Time-of-use tariff based control of energy consumption of households in a smart grid environment

Budapest University of Technology and Economics Department of Control Engineering and

Information Technology

Time-of-use Tariff Based Control of Energy Consumption of Households

in a Smart Grid Environment

Doctoral Thesis

Author Salma Taik

Supervisor Dr. Bálint Kiss

June 16, 2022

A la mémoire de ma chère mère, Paix à son âme, pour ton amour et tes sacrifices durant mon enfance, tu es toujours dans nos cœurs, Que Dieu te garde dans ses paradis éternels.

A mon père, pour son amour et son soutien inestimable, Aucune dédicace ne saurait être assez éloquente pour exprimer ce que vous méritez pour tous les sacrifices que vous n’avez cessé de me donner depuis ma naissance, durant mon enfance et même à l’âge adulte.

Salma

Acknowledgements

Firstly, I am thankful to God for this opportunity I was blessed with to expand my knowl- edge, challenge myself, and meet wonderful people. I would like to express my gratitude to the Tempus Foundation for granting me the Stipendium Hungaricum to start this amazing journey.

I would like to thank my supervisor Dr. Kiss Bálint for accepting me as one of his stu- dents, for his patience, for supporting and guiding me throughout the years of my curriculum.

I will always carry with me the knowledge I gained from him into my carrier.

I would like to thank my department colleagues and my professors for their generous support and kindness. I would like to offer my sincere thanks and gratitude to all those who directly or indirectly helped complete this thesis work.

I would like to thank the NRDI Fund (TKP2020 IES, Grant No. BME-IE-MISC) based on the charter of bolster issued by the NRDI Office under the auspices of the Ministry for Innovation and Technology for supporting this research work.

Lastly, I am grateful to my family and friends for their constant love, endless support, and prayers that kept me strong throughout this journey.

Remerciement

Je tiens tout d’abord à remercier Dieu le tout puissant et miséricordieux, qui m’a donné la force et la patience d’accomplir ce modeste travail. Je remercie aussi la fondation Tem- pus pour m’avoir accordé la bourse de Stipendium Hungaricum pour commencer ce voyage incroyable.

Au second lieu, je tiens à remercier mon encadrant, le Dr Kiss Bálint, de m’avoir accepté comme l’un de ses étudiants, pour sa patience, pour m’avoir soutenu et guidé tout au long des années de mon cursus. J’emporterai toujours avec moi les connaissances que j’ai acquises grâce à lui dans ma carrière.

Je tiens à remercier mes collègues du département et mes professeurs pour leur généreux soutien et leur gentillesse. Je tiens à exprimer mes sincères remerciements et ma gratitude à tous ceux qui ont contribué de près ou de loin à la réalisation de ce travail de thèse. Je tiens à remercier le Fonds NRDI (TKP2020 IES, Grant No. BME-IE-MISC) basé sur la charte de renforcement émise par le Bureau NRDI sous l’égide du Ministère de l’Innovation et de la Technologie pour avoir soutenu ce travail de recherche.

Je ne saurais terminer sans remercier mon père qui était toujours à mes côtés pour m’encourager et me soutenir inestimablement. Je remercie également ma soeur et mon frère pour leur amour et leurs soutiens. Je tiens à remercier mon mari pour sa croyance en moi. Finalement, je tiens à remercier mes amis pour tous les divertissements et le soutien émotionnel.

Abstract

Forecasting the demand in the residential households results in minimizing the environmen- tal impact of energy production since the related infrastructure will be used efficiently (e.g., scheduling the renewable energy resources production). On the other hand, the utility com- pany (UC) seeks to match the supply with the demand. The supply is subject to prediction errors due to the real consumption deviation from the predicted one. Consequently, the UC must purchase energy with a spot market price to fulfill consumers’ demands when it exceeds the supply.

Dynamical pricing is a simple yet effective way to influence consumption as price-sensitive consumers may reschedule the operation of some of the appliances in the household. Opti- mal pricing strategies allow both the UC and the consumers to benefit from a steady and predictable overall energy consumption since the UC needs a smaller amount of extra en- ergy purchase thanks to the reduced fluctuations around the predicted consumption. This often involves consumer-side demand management in residential areas using dynamic pricing structures. Such strategies work if the consumer-side response is at least partly automated using real-time optimization strategies.

My research focuses on managing, controlling, and monitoring the consumption of a population of residential electricity consumers by designing demand management strategies for the UC to be implemented on the consumer side. The thesis aims to introduce methods and algorithms that create closed-loop control between consumers and UC and contribute to the existing literature with new optimal results.

The proposed control loop comprises two components. Firstly, a consumer-side optimiza- tion to automate the control of the household electrical appliances (scheduling) responding to a dynamic pricing structure and controlling thermal appliances to preserve a given ther- mal comfort indicated by the household’s indoor temperature. The two components are in continuous interaction to efficiently plan the electric energy consumption in the household.

The first optimization component schedules the home appliances based on a Mixed Integer Programming approach. Various distributed energy resources are considered, such as the electric vehicle (EV), with energy storage capability to provide energy to the households’

appliances and the neighboring households.

The second optimization component is a Model Predictive Control (MPC) strategy dedi- cated for the control of an electric heating system. Due to outside temperature variations, the input constraints may impede the MPC to maintain the required thermal comfort, which triggers a rescheduling event for the first component. The efficiency of the framework is presented in multiple simulations for scenarios with different consumer behaviors. Another MPC strategy is proposed where the heating system is considered to be mainly controlled by a binary state thermostat.

Secondly, a selective and dynamic Time-of-Use (ToU) electricity pricing strategy applica- ble by a UC for residential electricity consumers is proposed. The method includes consumer clustering and cluster categorization based on the likelihood of appliance rescheduling. The parameters of the ToU tariff, namely prices and tariff periods for each eligible consumer category, are obtained by minimizing a cost function.

The underlying consumer behavior model is based on price elasticity. The profitability

robustness analysis shows that the scheme remains profitable to both the consumers and the UC even for uncertain price elasticity parameters. Moreover, a simple method to identify these parameters while respecting the consumers’ privacy is suggested. The elements of the strategy are illustrated on real consumption data obtained for two towns in Hungary.

Résumé

L’estimation de la demande dans les ménages résidentiels permet de minimiser l’impact envi- ronnemental de la production d’énergie puisque l’infrastructure correspondante sera utilisée efficacement (par exemple, en programmant la production de ressources d’énergie renouve- lable). D’autre part, la compagnie utilitaire (CU) cherche à égaliser l’offre à la demande prévue. L’offre est sujet à des erreurs de prédiction en raison de l’écart de la consommation réelle par rapport à celle prévue. Par conséquent, la CU doit acheter de l’énergie au prix du marché pour répondre aux demandes des consommateurs lorsque la dernière dépasse l’offre.

La tarification dynamique est un moyen simple mais efficace pour influencer la consomma- tion, vu que les consommateurs sensibles aux prix peuvent reprogrammer le fonctionnement de certains des appareils électroménagers. Des stratégies de tarification optimales permettent à la fois à la CU et aux consommateurs de bénéficier d’une consommation d’énergie globale stable et prévisible, car la CU a besoin d’une plus petite quantité d’achat d’énergie supplé- mentaire grâce aux fluctuations réduites autour de la consommation prévue. Cela implique souvent une gestion de la demande du côté des consommateurs dans les zones résidentielles en utilisant des structures de prix dynamiques. Ces stratégies fonctionnent si la réponse au côté consommateur est partiellement automatisée à l’aide d’une méthode d’optimisation en temps réel.

Cette thèse traite la gestion, le contrôle et la supervision de la consommation d’une population de consommateurs d’électricité résidentiels et de la conception de stratégies de gestion de la demande pour les CU à mettre en œuvre du côté des consommateurs. Le but de la thèse est d’introduire des méthodes et des algorithmes qui créent une boucle de contrôle fermée entre les consommateurs et les CU pour améliorer la littérature existante pour des résultats préférables.

La boucle de régulation proposée consiste de deux parts. Premièrement, une optimisation côté consommateur pour automatiser le contrôle des appareils électroménagers (ordonnance- ment) répondant à une tarification dynamique et contrôlant les appareils thermiques pour préserver un confort thermique donné indiqué par la température intérieure de la maison. Les deux composants sont en interaction continue pour planifier efficacement la consommation d’énergie électrique dans la maison.

Le premier composant d’optimisation planifie les appareils électroménagers sur la base d’une approche de programmation mixte en nombres entiers. Diverses ressources énergé- tiques distribuées sont incorporés, telles que le véhicule électrique (VE), qui possèdent une capacité de stockage d’énergie pour fournir de l’énergie aux appareils au sein de la maison et aux voisins.

Le deuxième composant d’optimisation est une stratégie de commande predictive (MPC) dédiée au contrôle d’un système de chauffage électrique. En raison des variations de tempéra- ture extérieure, les contraintes d’entrée peuvent empêcher le MPC de maintenir le confort thermique requis, ce qui déclenche un événement de reprogrammation par le premier com- posant. L’efficacité du cadre est illustrée par plusieurs scénarios de simulations pour des différents comportements de consommation. Une a utre stratégie MPC est proposée où le système de chauffage est considéré comme étant principalement contrôlé par un thermostat à état binaire.

Deuxièmement, une stratégie de tarification de l’électricité en fonction du temps d’utilisation (ToU) sélective et dynamique applicable par une CU pour les consommateurs d’électricité résidentiels est proposée. La méthode comprend le regroupement de consommateurs et leurs catégorisations en se basant sur la probabilité de reprogrammation des appareils. Les paramètres du tarif ToU, à savoir les prix et les périodes tarifaires pour chaque catégorie de consommateurs éligibles, sont obtenus en minimisant une fonction de coût.

Le modèle de comportement des consommateurs utilisé est basé sur l’élasticité du prix.

L’analyse de robustesse de la rentabilité montre que le système reste rentable à la fois pour les consommateurs et la CU même pour des paramètres d’élasticité du prix incertains. De plus, une méthode simple est suggérée pour identifier ces paramètres tout en respectant la confidentialité des consommateurs. Les éléments de la stratégie sont illustrés sur des données de consommation réelles obtenues pour deux villes de la Hongrie.

Contents

List of Symbols xiv

1 Introduction 1

1.1 Definitions and Fundamentals of Smart Grid Concept . . . 3

1.1.1 Energy Management on UC Side . . . 4

1.1.2 Energy Management on Residential Consumer Side . . . 6

1.2 Problem Statement and State-of-The-Art . . . 7

1.2.1 Consumer side . . . 7

1.2.2 Utility company side . . . 9

1.3 Research Methodology . . . 10

1.4 Thesis Outline . . . 11

2 Consumer side Management and Control Algorithms 13 2.1 Overview of The Control Framework . . . 14

2.2 Framework Setup . . . 16

2.2.1 Thermal Model of The Household . . . 16

2.2.2 Appliance Classification and Operation . . . 19

2.3 Appliances Scheduling Algorithm . . . 21

2.3.1 Decision Variables and Constraints . . . 22

2.3.2 Cost Function . . . 23

2.3.3 Simulation Results . . . 23

2.4 Indoor Temperature Control . . . 25

2.4.1 Inconvenience-based Compensation Cost . . . 27

2.4.2 MPC formulation . . . 29

2.4.3 Variation Effects of The Inconvenience Weighting Factor on The In- door Temperature . . . 33

2.4.4 MPC Strategy For Thermostat’s Setpoint Control . . . 36

2.5 Components Interaction . . . 42

2.5.1 Rescheduling Events Scenarios . . . 44

2.5.2 The Plant Model Mismatch Analysis . . . 47

2.6 Community-Level Optimization Framework . . . 47

2.6.1 Framework Description . . . 49

2.6.2 Energy Exchange Model At The Community-Level . . . 52

2.6.3 Appliances Scheduling Algorithm . . . 54

2.7 Conclusions . . . 61

3 Design and Analysis of Dynamic Pricing Strategies 63

3.1 Description of The Structure of The Proposed Methodology . . . 64

3.2 Consumption analysis . . . 65

3.2.1 Consumption Data Clustering . . . 66

3.2.2 Clusters-based Categorization . . . 66

3.3 Clustering and categorization based on consumption data . . . 67

3.4 Consumer Behavior Modeling . . . 73

3.5 Design of The Time-of-Use Electricity Tariffs for a Single Consumer Category 74 3.5.1 Decision Variables . . . 76

3.5.2 Optimization Constraints . . . 76

3.5.3 Cost Function . . . 77

3.5.4 Optimal Time-of-Use Electricity Tariff Structure For One Consumer Category . . . 78

3.6 Design of The Time-of-Use Electricity Tariffs For a Population of Residential Consumers . . . 81

3.6.1 ToU Tariff Periods . . . 81

3.6.2 ToU Tariff Variables . . . 82

3.6.3 Cost-based Objective Function . . . 83

3.7 Optimal Time-of-Use Tariffs Based on Consumption Data for City A . . . . 84

3.8 ToU Electricity Tariffs Analysis for City B . . . 87

3.9 Profitability Robustness Analysis . . . 89

3.9.1 Price-elasticity uncertainties . . . 89

3.9.2 Cost-function Components Weighting Analysis . . . 91

3.10 Price Elasticity Parameter Identification . . . 94

3.11 Conclusions . . . 96

4 Conclusions 98

List of Figures

1.1 Smart grid architecture. . . 3

1.2 Thesis framework. . . 11

2.1 Components of the proposed optimization framework and their interactions. 14 2.2 Heating process in the household. . . 17

2.3 Shiftable appliancesA_sschedule with the nonshiftable appliancesA_nswithout considering the consumer’s time preferences. . . 24

2.4 Appliances operation times considering the earliest defined time preferences by the consumer. . . 26

2.5 Appliances schedule considering the defined time preferences by the consumer. 26 2.6 Closed-loop control of T_in^k. . . 27

2.7 Example of the indoor temperature. . . 28

2.8 Available power margin for the heating system operation before and after the addition of the stored power in the EV’s battery. . . 31

2.9 Indoor temperature variation T_in with the weather changes (top) and the heating system power consumptionP OW_hvac(bottom) with the assigned ToU tariffs. . . 32

2.10 Total power consumption of the household in 24 hours. . . 32

2.11 Indoor temperature without and with the implementation of the inconvenience- based algorithm in the MPC setup. . . 33

2.12 Indoor temperature at different weighting factors. . . 35

2.13 Thermostat behavior. . . 37

2.14 Control loops setup L₁ and L₂. . . 38

2.15 Indoor temperature changes using thermostat control only (top) and the out- door temperature changes (bottom). . . 39

2.16 Power consumed by the heating system controlled by the thermostat only. . 40

2.17 Indoor temperature changes with the MPC strategy to control the thermostat control. . . 40

2.18 Optimal temperature setpoints resulting from the MPC strategy optimization. 41 2.19 Power consumed by the heating system when implementing the MPC strategy. 41 2.20 Rescheduling scheme. . . 43

2.21 Rescheduling event when MPC’s horizon exceeds the scheduling horizon. . . 44

2.22 Indoor temperature without (top) and with(third) the rescheduling with the corresponding power margin remaining from the first component without (sec- ond) and with (bottom) the rescheduling. . . 45 2.23 Appliances schedule in the new schedule horizon with the assigned ToU tariff. 45

2.24 Indoor temperature without (top) and with (third) the rescheduling with the corresponding power margin remaining from the first component without (sec- ond) and with (bottom) the rescheduling. . . 46 2.25 Indoor temperature when the window is open without (top) and with (bottom)

the rescheduling. . . 48 2.26 Total power consumption of the household in 24-hours without (top) and with

(bottom) the rescheduling. . . 48 2.27 Community-level framework consisting of multiple smart households. . . 49 2.28 Power generated due to the photovoltaic panel. . . 52 2.29 Optimal consumption schedule forh=1 (top) and the energy sources supply-

ing it (bottom). . . 59 2.30 Optimal consumption schedule forh=2 (top) and the energy sources supply-

ing it (bottom). . . 59 2.31 Optimal consumption schedule forh=3 (top) and the energy sources supply-

ing it (bottom). . . 60 3.1 Workflow of the proposed methodology. . . 64 3.2 Segmentation of the categories interval. . . 67 3.3 City A: weekly load profiles of clusters K1-K5 (winter season); solid black line:

average. . . 68 3.4 City B: weekly load profiles of clusters K1-K5 (winter season); solid black line:

average. . . 68 3.5 City A: weekly load profiles of clusters K1-K5 (summer season); solid black

line: average. . . 69 3.6 City B: weekly load profiles of clusters K1-K5 (summer season); solid black

line: average. . . 69 3.7 Clusters’ share in the total consumption (winter season; left: City A; right:

City B). . . 70 3.8 Average consumption weekly load profiles representing five clusters, each be-

longing to a different category (City A, winter). . . 71 3.9 Approximation of the operation of shiftable appliances (green bars) to the

average consumption of each category (blue) in one weekday in City A. . . . 72 3.10 Approximation of the operation of shiftable appliances (green bars) to the

average consumption of each category (blue) in one weekday in City B. . . . 72 3.11 Load demand curve. . . 73 3.12 Example of a daily consumption pattern of one category partitioned to 3 periods. 75 3.13 Total consumption of category 3 before and after the implementation of the

optimal ToU tariffs in winter season for City A. . . 79 3.14 Consumption in each period using flat rate versus the optimal ToU tariffs in

winter season for City A. . . 79 3.15 Average consumption of category 3 in winter season based on flat-rate versus

the optimal ToU tariffs. . . 80 3.16 Labelled tariff periods (Category 3, City A, winter season). . . 82 3.17 Average Monday consumption of all consumers in the cluster belonging to

each eligible categories and their contribution to each period in City A. . . . 83

3.18 Total consumption of the eligible categories average: 3, medium-high: 2, and high: 1 and the consumption distribution before (left) and after (right) im- plementing the optimal ToU tariffs for City A in winter season. . . 85 3.19 Average consumption curve for the three categories after and before the im-

plementation of the optimal ToU electricity tariffs in City A in winter season.

. . . 86 3.20 Contribution of eligible clusters to the overall power peak in City B in winter

season. . . 87 3.21 The ToU tariffs structures for each consumer category of City B in winter

season. . . 88 3.22 Average consumption curve for the three categories after and before the im-

plementation of the optimal ToU electricity tariffs in City B in winter season. 90 3.23 Total consumption of the eligible categories medium-low: 4, average: 3,

medium-high: 2, and high: 1 and the consumption distribution before (left) and after (right) implementing the optimal ToU tariffs for City B in winter season. . . 90 3.24 Dependence of UC gains on price sensibility parameter variations. . . 92 3.25 Dependence of consumer gains on price sensibility parameter variations. . . . 92

List of Tables

1.1 DR types with the corresponding adjusted load shape. . . 5

2.1 Available temperature in the framework. . . 17

2.2 Household model’s new notation. . . 19

2.3 The power profile and length of operation of some appliances in A_s (WM - washing machine; DW - dishwasher; OV - Oven). . . 20

2.4 EV’s Characteristics (Nissan Leaf). . . 24

2.5 Consumer’s Time Preferences. . . 25

2.6 Temperature inconvenience compensation rates. . . 29

2.7 Results of inconvenience factor without and with the implementation of the inconvenience-based algorithm. . . 33

2.8 Inconvenience and compensation at different weighting factors. . . 34

2.9 Results of inconvenience factor without and with the implementation of the inconvenience-based algorithm. . . 47

2.10 Household states corresponding to the binary code and decimal code. . . 53

2.11 Time Preferences of each consumer h=1,2,3. . . 59

3.1 Categorization change of clusters based on c^σ_k. . . 67

3.2 Examples of consumers’ categories in the seasons in city A. . . 71

3.3 ToU Tariffs of Category 3. . . 78

3.4 Profits of the UC and consumers before and after implementing the DR program. 80 3.5 Labelling rules in consumer category iand for the tariff period j (j=1,. . .,C_i). 82 3.6 Tariffs in City A (normalized to T_F, rounded to two decimals). . . 85

3.7 Average profits per consumer before and after implementing the DSM program (City A, winter). . . 86

3.8 Tariffs in City B (normalized to T_F, rounded to two decimals). . . 88

3.9 Average profits per consumer before and after implementing the DSM program (City B, winter). . . 89

3.10 UC gain calculated by changing the values of w₁ and w₂. . . 93

3.11 Consumer gain calculated by changing the values of w₁ and w₂. . . 93

3.12 UC gain calculated by changing a and b. . . 94

3.13 Consumer gain calculated by changing a and b. . . 94

List of Symbols

A,A_s,A_ns set of nonthermal/shiftable/nonshiftable appliances

α weighting factor for the inconvenience costs

β constant to limit the variation of peak tariffs

C number of daily periods for ToU structure

C_hh calculated compensation [currency/kWh]

c rate of compensation

c_a specific heat capacity of the household envelope

[J/kg/^◦C]

¯

c_hr,d normalized hourly consumption

c_in specific heat capacity of the air [J/kg/^◦C]

¯

c_k,c¯^σ_k,c_k,c^σ_k average daily comsmption, its standard deviation, clus- ter centroid and its standard devialtion

d_h,d^∗_h decision and optimal decision vectors

E_EV EV’s battery capacity [kWh]

E_h,ESS ESS’s energy capacity

E_c energy needed to eliminate the error calculated by the household thermal model [kWh]

E_day total energy that could be drawn from the grid in one day [kWh]

EV_h,nd,EV_h,dst,EV_h,mrg EV’s number of travelled days, distance traveled and maximum distance range

ϵ thermostat switching range

h_a2ext convection heat transfer coefficient between the house- hold envelope and the outside air [W/m²/^◦C]

h_in2a convection heat transfer coefficient between the house- hold envelope and the inside air [W/m²/^◦C]

hin2ext convection heat transfer coefficient between the inside and outside air for open surfaces [W/m²/^◦C]

I_{M P C} inconvenience factor

Ir,Ir_{ST C} irradiance and Standard Test Conditions λ₁,λ₂ parameters to determine ToU labels

m_a mass of the house envelope [kg]

min mass of the air inside the household [kg]

N number of consumer categories

N_app number of shiftable appliances

N_prd prediction horizon

N_slot number of slots in the scheduling horizon n_h,s,n_h,p PV’s number of modules in series and parallel nld_i number of load phases for the appliance i

np_i number of operational programs of appliance i

η_s slack variable penalty belonging to the inequality con- straints on the desired temperature

θ acceptable width of thermal comfort

on_h,EV,on_h,ESS variable to prevent EV and ESS charging and discharg- ing overlapping

P_j⁰,P¯_j initial and new demand consumption in period j Pⁱ_j total baseline of category i at period j

P OW_app power schedule of the shiftable and nonshiftable appli- ances in the scheduling horizon [kW]

P OW_av^k available power at time slot k provided by the utility company [kW]

P OW_EV charging power of the EV’s battery [kW]

P OW_EV,rate EV’S maximum charging rate

P OW_h,EV,ch,P OW_h,EV,dis EV’s charging and discharging powers P OW_h,ESS,ch,P OW_h,ESS,dis ESS’s charging and discharging powers

P OW_i,prg power profile of the appliance i running on prg program [kW]

P OW_hvac power to actuate the heating system [kW]

P OW_{h,P V} PV’S generated power

P OW_h,lc,init initial available power at householdh P OW_h,lc(d_h) available power after the decisiond^∗_h

P OW_max power limitation of the heating system [kW]

P OWmrg power margin for the heating system [kW]

P OW_s optimal power schedule of the smart appliances [kW]

P OW_ns estimation of the nonshiftable appliances power con- sumption [kW]

P OW_T total power consumption of the appliances and heating operation [kW]

P R^k time-of-use tariff at time slot k [currency/kW]

prof it_initial,prof it_{f inal} UC profits before and after implementing the ToU strat-

∆P,∆T egydemand and price changes

|∆¯P_j|,|∆P_j| remaining and shifted loads in/to periodj

q_d,a2in thermal heat flowing from the interior to the house-

hold’s envelope by conduction [W]

q_d,ext2a thermal heat flowing from the household’s envelope to the exterior by conduction [W]

q_hvac thermal heat flowing from the heating system to the interior [W]

¯

q_m mass flow rate of the air passing through the heating system [kg/s]

q_max maximum heat generated by the heating system

q_v,a2ext thermal heat flowing from the household’s envelope to the exterior by convection [W]

q_v,in2a thermal heat flowing from the interior to the house-

hold’s envelope by convection [W]

qv,in2ext thermal heat flowing from the interior to the exterior through open spaces by convection [W]

R_a2in thermal resistance between the inside space and the en- velope [^◦C/W]

R_ext2a thermal resistance between the outside space and the

envelope [^◦C/W]

rp_{P V} PV’s rated power

S_i,j^k decision variable indicating the state of the appliance i during the load phase j at time slot k

S_EV,j decision variable indicating the state of the EV during the load phase j

S_h,lc,S_h,gd,N S_h,lc,N S_h,gd decision variables for shiftable appliances and non- shiftable appliances locally supplied and from the grid S_h,N,N S_h,N decision variable for shiftable and nonshiftable appli-

ances supplied by the neighborhood

SF_a2ext surface of convection between the envelope and the out- side [m²]

SFin2a surface of convection between the envelope and the in- side [m²]

SF_in2ext surface of convection between the inside and the outside [m²]

SoC_min(max),SoE_EV,min(max) EV minimum and maximum SoC and SoE [%]

SoC₁₀₀ power to fully charge the EV [%]

SoE_h,EV,init EV’s intial SoE in household h

SoE_h,ESS,init ESS’s initial SoE

SoE_h,ESS measured ESS’s SoE

s^k slack variable

sw_h,j,sw_dec variable indicating the switch state in binary and deci- mal code at household h

Ta household’s envelope temperature [^◦C]

T_c temperature error [^◦C]

T_d desired temperature by the consumer [^◦C]

Text outdoor temperature [^◦C]

T_F,T_j flat and ToU tariffs in period j

T_F^{U C},T_H^{U C} day-ahead and spot market prices (UC)

T_h temperature of the heating system [^◦C]

T_in measured indoor temperature [^◦C]

T_{M P C} optimal temperature setpo-points

T_o initial temperature of the household [^◦C]

T_s sampling time

t_ar arrival period of the EV

t_dp departure period of the EV

t_{OF F,i} switching off time slot of the appliance i t_ON,i switching on time slot of the appliance i

t100 time required to fully charge the EV

th threshold of the UC profits

∆T_{M P C},∆T_min,∆T_max rate of control change, its minimum and maximum

∆t duration of one time slot

U_a thermal energy stored in the household’s envelope [J]

U_in Thermal energy stored inside the household [J]

u^k,u^k_{T T} control actions at time slot k

∆u rate of change of the control actions

U_h(d_h) profit function

U_min negative constant to prevent actions overlapping w₁,w₂ weighting factors on UC cost function

ξ,ξii,ξij price elasticity, self and cross elasticities

List of Acronyms

AC Air conditioner

CPP Critical peak pricing DR Demand response

DSM Demand side management DW Dishwasher

EMPC Economic Model Predictive Control ESS Energy storage system

ESS2H Energy storage system to home EV Electric vehicle

GA Genetic Algorithm

GUI Graphical User Interface H2H Home to home

HEMS Home energy management system HVAC Heating, ventilating, air condition-

ing

IBDR Incentive-based demand response

MIP Mixed Integer Programming MPC Model Predictive Control OV Oven

PBDR Price-based demand response PV Photovoltaic

PV2H Photovoltaic to home RTP Real-time pricing SG Smart grid

SoC State of charge SoE State of energy ToU Time-of-Use UC Utility company V2H Vehicle to home WM Washing machine

Chapter 1 Introduction

In the last two decades, electricity consumption in the residential sector has known a rapid surge worldwide, especially in developed countries [1]. The main reasons are the exponential growth of the population and the electrification of transport due to the growing number of electric vehicles (EV) [2]. Large-scale adoption of EVs may result in serious problems such as line congestion and voltage limit violations [3].

The Utility company (UC) plays an essential role as an intermediate between the power generators (purchases energy from) and the consumers (distributes energy to). Thus, the consumers consider the UC the official energy distributor for their direct interaction. Despite the planning of the energy production and the demand prediction, the utility company (UC) still faces the challenge of robustly and cost-effectively supplying the consumers. Hence, the UC may resort to open-loop solutions to avoid unwanted surges and electric power quality degradation. Such solutions are the enlargement of the production capacity and the enlargement of the existing infrastructure (integrating new power plants and distribution lines) to meet higher demands [4]. Such infrastructure expansion represents a significant impact to the environment [5], such as greenhouse gas emissions, where only Europe is estimated to be emitting 6.7 tonnes of CO2per person in 2019 [6]. The UC may also schedule a rolling blackout when the power system is congested. These possibilities are expensive, time-consuming, represent undesired environmental impacts, and cause inconveniences to the consumers.

Integrating renewable energy resources into the power grid system is considered an effi- cient and environment-friendly opportunity to fulfill the energy needs without enlarging the grid infrastructure. However, considering such a solution as additional energy sources to the grid, other challenges are still to be overcome. The latter is presented in the fact that the consumers still lack the knowledge on how to consume the energy efficiently to match the UC’s objectives. In other words, the consumers cannot react to the information available from the UC in very granular time intervals. Thus, it can lead to unwanted surges in the grid and degradation of the supply quality.

Monitoring and controlling residential consumers’ energy consumption can be considered alternative solutions to match, or at least limit, the deviation of demand from the supply continuously [7, 8]. Consumer behavior undergoes large instantaneous changes that make the prediction of the next move complicated. Therefore, the UC must often purchase ad- ditional electric energy with a spot price market when the demand exceeds the supply [9].

Consequently, the UC calculates the electricity prices depending on the costs of supplying and production costs. Hence, the consumers will be charged high prices when the electricity demand is high and/or production shortages occur. However, in a traditional power grid, the consumers are exposed to flat rates that implicitly consider the price fluctuations in the elec- tricity market throughout the consumption horizon. Therefore, the existing pricing strategy represents no economic incentive for the consumer to change their consumption behavior.

The alternative solutions have been investigated, and new strategies are being introduced:

developing software functions and integrating and/or replacing minor hardware components to the power grid. This solution aims to create a closed loop between the consumers and the UC to regulate the demand and supply imbalance. Such strategies are presented under the smart grid concepts and known as demand-side management (DSM) [10,11].

Due to the duplex communication infrastructure introduced in smart grids [12], a DSM program dedicated to the consumer side is implemented to decrease consumption in peak hours, namely the demand response (DR) program [13]. The latter enables the closed- loop control of consumer-side consumption by reading the real-time consumption, thanks to the intelligent meter technology [14], and applying motivating strategies proposed by the UC [15] to modify their consumption profile. As a result, the consumers can be motivated to participate in DSM programs to save on their bills. The UC, on the other hand, will improve its energy planning (purchase and supply) due to the continuous consumption data fed back from the consumers and avoid extra generation costs at power peaks times.

The DSM programs remain an attraction to researchers due to their advantages to the UC and consumers and, most importantly, the environment. They have been implemented in real-time and tested for industrial consumers; however, few attempts are recorded for the implementations for residential consumers. For example, the UK and France have aggre- gation companies in charge of making the optimal decisions for the industrial consumers’

DR (France: Energy Pool [16] and UK: Flextricity [17]). Their customers’ DR has known a pattern change that significantly affects the electricity system as a whole.

This research aims to investigate, propose and design control strategies that ensure the balancing of the demand and supply for residential consumers population. The proposed algorithms are implemented on the consumer side and are adopted by the UC to effectively decrease the power peaks and maintain the sustainability and stability of the power system.

The proposed strategy guarantees a profitable situation for the consumers and the UC.

Motivated by the difficulties mentioned above, I proposed a closed-loop control problem that relates the consumers and the UC. I decoupled the proposed strategy into two main tasks: designing DR programs based on real consumption data to be implemented by the UC and developing algorithms on the consumer side for appliances’ management and control in response to the implemented DR programs.

The results presented in this thesis are based on simulation scenarios and real consump- tion data and demonstrate the efficiency and robustness of the proposed framework. How- ever, real-time implementation can be carried out easily thanks to the utilization and veri- fication of the proposed strategies elements with the real consumption data.

1.1 Definitions and Fundamentals of Smart Grid Con- cept

The traditional power grid architecture comprises power production plants, utility compa- nies, transmission and distribution grids [18]. The sole role that the power grid played for years is a linear functioning where it transports and distributes energy from the power generators to the consumers with the best possible performance and robustness.

With the problems coming in hand with the rise of population, an upgrade to the so- called smart grid (SG) [19–22] is necessary. The SG is a modern concept bringing a new future of energy that focuses on clean and green energy. Unlike the traditional grid, the SG is characterized by its communicative features between the connected entities. Thanks to technological advancement, the SG allows the bi-directional way of communication of energy and information, increases decentralized production and storage units, and advanced demand control [23, 24]. The SG objective is to optimize energy planning and consump- tion by balancing the supply with the predicted demand at all times to enhance the grid’s performance.

The SG concept introduces new technologies and strategies to energy systems, espe- cially in the consumption, generation, distribution, and management sectors, bringing more accuracy and efficiency to the whole power system. The general architecture of the SG, illustrating the essential aspects from high voltage to low voltage grids, is summarized in Fig. 1.1.

Figure 1.1: Smart grid architecture.

The essential entities can be deduced from the SG general architecture are the following:

• Independent System Operator (ISO): A nonprofit and independent entity from the other market players. Its responsibilities are to orchestrate the electricity grid and its development. The ISO accommodates the competitive market of biding the generated energy. It ensures a fair and transparent market by controlling any manipulation among the generators. The ISO seeks that demand is always met at any time of the day.

• Generators: Their main role is to generate the electric energy and sell it on the market before the next transmitting horizon. The generators have long-term contract with the retailers so that their income is secured for the next generation operation. Multiple sources can be combined with the conventional plants to generate the electric energy necessary to supply the consumers population (industrial, residential, and commer- cial). With the thriving popularity that renewable energy gained for its impact on the environment and the cheap generation, renewable energy accommodation became inevitable. Its role can be summarized as a wholesaler.

• UC: Operators or service providers are the intermediates between the low voltage consumers and the generators. They provide direct services to customers such as energy distribution, calculating and charging, energy planning and management, and troubleshooting. Their objectives are to provide energy smoothly in real-time, maintain the distribution system’s stability and efficiency, and ensure the privacy and security of their subscribers. In the thesis, the UC is considered as a retailer.

• Consumers: In the SG, the consumers will actively participate in the energy market and make decisions concerning their consumption profiles. Therefore, the SG will enable the consumers to be better informed about the market fluctuations by sending price signals. Moreover, The SG supports the distributed generation at or near the consumption areas. Hence, the concept of prosumers appears where the consumers can produce and consume energy [25,26,S8]. It will further allow the consumers to share energy with other consumers and the grid.

The electricity market is a unique market with multiple actors subject to several con- straints, including the production and transmission infrastructure limitations and regulatory measures. Moreover, this market is currently undergoing disruptive changes. The detailed analysis of these changes and their economic consequences is beyond the scope of the current thesis. My research focuses on two related optimal control problems: one at the consumer (household) level and one at a retailer level. The solution of both control problems is pre- sented in Chapter 2 and 3, respectively.

1.1.1 Energy Management on UC Side

Thanks to the intelligent sensors and the two ways of communication, the UC can track and monitor their customers’ energy consumption. Hence, it gives an opportunity to develop automated strategies to enhance the direct interaction with the customers in a reliable and secure way [27].

The DSM is an advanced control tool presented under the SG concept to match the demand with the supply instead of the usual way around [28]. The DSM ensures energy efficiency by planning ahead the purchased energy to avoid scenarios where the UC must expand its capacity [29]. As a consequence, the DSM results in limiting the supply costs. Its concept is to motivate consumers to change their consumption patterns voluntarily in peak hours in exchange for a specific incentive. Several incentives can be used (e.g., free energy, etc.), but the consumers’ most tempting and accepted method is the monetary incentives.

The DSM strategy also allows higher penetration of renewable energy resources [29, 30].

Thus, the objectives of the UC from the DSM implementation are reducing the costs of electricity, improving environmental and social development, and limiting the power grid issues.

Among the DSM strategies, the DR method has been investigated in recent years. The DR programs are economic strategies that are implemented on the demand-side. Its goal is to manage the consumer’s power consumption profile (by controlling the electric appliances) in response to proposed strategies by the UC to reduce electricity costs. Different techniques exist to modify the shape of electricity demand; hence the common techniques of DSM [31,32]

can be summarized in Table 1.1.

Table 1.1: DR types with the corresponding adjusted load shape.

Peak clipping Valley filling Load shifting

The DR programs encouraging the consumers can be broadly classified into two main strategies [33] namely the incentive-based and price-based strategies.

Incentive-Based Demand Response

The UC uses the incentive-based demand response to lower the electricity consumption at critical times (e.g., high energy market price and/or power system overload) by motivating the consumer through an incentive payment [34]. In the literature, the incentive-based pro- grams are divided into the following [35]: direct load control, interruptible load, emergency demand response, capacity resource, spinning/responsive reserves, non-spinning reserves, regulation service, and demand bidding and buyback. The common goal of these strategies is that the UC has the permission to intervene and control the consumption instantaneously, due to the installed control equipment on the consumer side, to maintain the grid stability.

Price-Based Demand Response

The idea behind the price-based strategy is that the consumer rearranges his electricity consumption in response to a price signal that changes over time [36]. The price signals are designed to reflect the costs of power generation, transmission and distribution during periods of time. Thus, consumers can be charged with higher prices in peak times (i.e., when all consumers consume simultaneously) to motivate them to consume less, where they are charged with relatively lower prices in non-peak times. The existing price schemes [37–

39] that can be implemented are the Real-Time Pricing (RTP) [40–42], Time-of-Use tariff (ToU) [43–45], and Critical Peak Pricing (CPP) [46,47].

The RTP and ToU are identical strategies except for the difference in the time span of the changes in the prices. The granularity of the RTP changes can vary from some seconds to minutes, where the ToU tariffs cover more extended periods ranging from several minutes to hours. Moreover, the ToU tariffs are set prior to the actual period implying that the prices do not reflect any changes in the energy conditions in the actual period. Hence, the ToU tariff can be a monthly or longer-based strategy. In contrast, the RTP prices capture the actual energy conditions and may change to adjust accordingly. Therefore, the RTP strategy can be an hour-ahead or a day-ahead program. The CPP is characterized with higher rates than RTP and ToU tariffs and is identified based on historical data.

1.1.2 Energy Management on Residential Consumer Side

On the other end of the communication network is the consumers that can receive information from the UC. Consumers can react to this information to avoid overconsuming on power peak times. Automated and control software-based strategies are implemented to manage energy consumption on the consumer side. The purpose of implementing the DR program is to change the shape of the loads on the consumer side. From the residential consumer perspective, the operation of different appliances can be rearranged to respond to the DR program implemented by the UC.

Two major appliances category can be found in a residential household: thermal and nonthermal loads. These appliances can be monitored and controlled by the home energy management system (HEMS) via home networks. The HEMS is a system capable of commu- nicating with several devices inside the household and can also communicate with external entities such as the UC and neighboring interconnected households [48]. Different com- panies have commercialized HEMS for residential consumers for device-level control (e.g., Ecobee [49, 50] and EnergyHub [50] for thermostat control and intelligent plugs), and cen- tralized and on-board control (e.g., Control4 [51], Google home [52], and Whirlpool [53] to control the smart appliances and to communicate with multiple devices in the household).

Thermal Appliances

Thermal appliances such as HVAC systems can also be controlled to consume energy ef- ficiently. The thermal appliances should have the appropriate tools and support commu- nication protocols for responding to control inquiries such as Bluetooth, Zigbee, Wi-fi, and Li-fi [54]. The thermal appliances in this thesis are considered smart. They can communicate with other household appliances and not only be programmed to reach specific temperatures at different times.

Nonthermal Appliances

Not all household appliances can be operated at different times than usual since it might cause inconveniences to the consumers. However, a type of appliance called shiftable appliance [S7]

can operate flexibly as many times as desired at any time of the day (e.g., washing machine, dishwasher), preferably when the electricity prices are low. The unmanageable appliances are called hereinafter by nonshiftable appliances [S7], and they could be operated manually

by the consumer (e.g., lights, laptops, etc.) or operate in a batch-like fashion depending on their internal state (e.g., refrigerator). However, the electric energy source to operate the nonshiftable appliances can be controlled, especially with the presence of local distributed energy resources.

The EVs are high-power consumers and flexible loads at the same time. They represent an opportunity to regulate energy consumption if their charging time is managed as suggested in [55–57]. Hence, they are considered an essential factor in the modernization of the grid.

With the EV’s ability to charge and store the energy, one can view the EV as a distributed energy resource if a certain charging level is achieved [57]. Therefore, when the power system is overwhelmed, and the power peaks appear, the EV may be discharged to supply locally interconnected consumers.

Along with the EV, integrating an energy storage system (ESS) as a distributed energy resource adds flexibility to the grid to deal with unpredicted power shortages. The ESS is a battery storage system that can be implemented at a household level as distributed storage (i.e., supply with a small amount of power over short periods of time) or commercial level as mass storage (supply with a larger amount of power over more extended periods) [58]. It can be used to store the generated power from renewable energy resources if it is not used, or it can be charged from the grid when the electricity prices are low.

1.2 Problem Statement and State-of-The-Art

Monitoring and controlling the consumption profile of residential consumers and coordinating the available distributed energy resources in addition to the grid may result in an optimal balance to protect the whole power system. This task can be solved by applying optimal scheduling algorithms and consumption controllers on the consumer side.

Considering the fact that the UC’s ultimate goal from the DSM is to match the electricity supply with the predicted demand at all times, an appropriate strategy should be proposed and applied by the UC to encourage the consumers to participate (by shifting loads and controlling the appliances’ consumption level) in the DSM program guaranteeing their con- tinuous profits. This second task consists of designing an incentive-based and/or price-based program to which the consumer will react. Therefore, a closed-loop between the consumers and the UC will be created where solving one task is necessary to carry on with the second task and vice versa.

1.2.1 Consumer side

Many strategies presented in the literature solve the underlying scheduling task to shift the appliances’ operations and reduce the electricity costs. The optimization-based techniques have been investigated and proven efficient for peak load reduction and consumers’ profits maximization. Such techniques can be exact or approximate methods depending on the problems’ complexity [59].

Considering the implementation of a price-based demand response program on consumer- side, automated algorithms can be developed to be responsible for appliances management.

Several approaches can be established to assign the operation times of the household appli- ances to when the electricity price is low, which will result in reducing the costs [60–62]. For example, considering a combination of two or more dynamic prices may result in reaching an optimum appliances schedule [63, 64]. More solutions consist of reducing the costs and PAR by shifting the loads and smoothing out the demand curve [65,66] and prioritizing the operation of certain appliances [67]. Moreover, forecasting loads and prices to correlate with the actual loads and prices can be used in the presence of training data [68]. A multi-agent system framework results in minimizing the household electricity costs by managing the shiftable appliances [69].

Regulating the consumption of aggregated households is another approach to increase the DR margins of the participants. For instance, managing the distribution transformers at the neighborhood level can reduce the overall peak [70,71]. Participants in the DR program can also be involved in a game to select the best strategy for them to shift their appliances [72].

Their strategy is visible to all participants so that an optimal one is selected. Exchanging electric energy is also possible [73] if appropriate infrastructure is available in each household.

Dynamic pricing also requires optimizing the continuously operated electric systems such as heating, ventilation, and air conditioning (HVAC) systems. Multi-agent decision making and control framework is used as a solution to optimally control the HVAC systems [74].

This solution results in significant energy savings, however, the issue related to the hardware implementation for this approach is undeniable. On the other hand, the model predictive control (MPC) [75] approach is often used for HVAC systems [76]. This popularity is due to its flexibility and ability to take into account constraints and external dynamics that might be complicated to resolve with other approaches [77].

There exist in the literature two approaches of how the MPC can control the HVAC systems: continuously ramping up and down its consumption or to send binary signals so that the HVAC system is strictly on either the on or off mode. The occupancy information inside of a room can be employed to control the indoor temperature using the MPC to potentially save energy [78]. A different MPC strategy can be used, namely the economic MPC (EMPC) (i.e., nonlinear MPC), also for the same purpose to directly optimize the economic objectives in real time [79, 80]. As for the second approach, the MPC in this strategy can turn on and off the HVAC system where the latter is operating with its full capacity [81]. Similarly, the thermostat setpoint can be manipulated between upper and lower limits predicted from the historical data [82].

The (E)MPC strategy can also be used to schedule the appliances and control the tem- perature [83,84] or a binary-based logic algorithm responsible for scheduling the appliances as well as the HVAC system can be deployed [85,86]. Despite the efficiency of other strate- gies in reducing the energy consumption and achieving a specific thermal comfort, the MPC strategy is validated to ensure a 100% thermal comfort and higher energy savings [87].

Despite the richness of the presented literature, the work carried out does not consider the following important points:

• The time allocated for a given appliance may be inconvenient to the consumer, which may lead to limited DR. That is due to no time preferences are considered for operating the appliances.

• The constraints considered in some references models the appliances to be interrupted,

which may not be the case in a realistic scenario, since some appliances should run uninterruptedly to ensure the consumer’s comfort.

• The EV is scheduled as the rest of the appliances without considering the battery’s sustainability.

• The proposed MPC in the literature may be complicated and includes many infor- mation exchanges that should be avoided due to the existing HVAC communication protocols [88].

• The consumer preferences to achieve thermal comfort are not considered when schedul- ing shiftable appliances and the EV.

• The MPC controls the HVAC system to be on and run entirely with its nominal power for long periods or off, which consumes a large amount of energy since the two possible states are on or off. Hence, minimizing the periods when the HVAC system is on can be optimal.

• In most of the existing literature, the MPC usually does not consider the remaining appliances’ consumption patterns.

• The alternative approaches may not result in significant energy saving to reduce the peak power.

• The unallowed integration of distributed energy resources such as EV-to-home (V2H) [89]

and home-to-home (H2H) (also known as peer-to-peer) [90,91] technologies.

In summary, the proposed approaches may result in balancing the demand with sup- ply; however, their imperfections leave plenty of room for developing new strategies seeking further improvements to fill in the gaps.

1.2.2 Utility company side

For fruitful DR programs implementation, the UC must plan a strategy that reflects the amount of energy ready to be supplied and its costs. Moreover, this strategy should award the consumers for participating without minimizing the UC’s income.

The incentive-based DR has been studied in the literature to design the incentives as well as to manage the household appliances [92–94]. However, the costs of enrolling the consumers into incentive-based programs are expensive due to the installation costs of control equipment and the need to make regular incentive payments to customers [95]. Moreover, incentive- based strategies are not preferred by consumers due to concerns related to the involvement of consumers in making decisions and associated with consumers’ privacy and data security in the grid [96]. Hence, the focus is mainly on the price-based demand response that exposes the consumers directly to wholesale market prices and activates the role of the consumers in the power system [97].

Optimal ToU tariffs have been intensively studied recently. Authors of [98] surveyed optimal ToU tariff structures. Based on evolutionary game-theoretic approaches, in [99,100]

the optimal ToU tariffs for different energy markets can be obtained. The ToU tariff periods are predefined for all consumers. The ToU tariffs are bounded at each time period so that the consumers will not be charged with higher costs. In [101], the UC designs the ToU tariffs starting from the flat tariff and charges the consumers with both to compare the effectiveness of the ToU tariffs. [55] proposed four different tariff schemes for a residential sector suitable to low-income consumers. They found out that the best strategy for better results is a combination of an inclining block and the traditional ToU pricing scheme. Authors of [102] analyzed how to design the parameters influencing tariff structure and how these parameters may affect UC’s profit based on the consumers’ responses. [103] design a ToU tariff for residential electricity consumers starting from a flat rate while the earnings of the UC are kept unchanged. In [104], a pricing algorithm is proposed to reduce the peak- to-average ratio of the aggregated load demand if the UC is uncertain about consumer responsiveness. The electricity price model in [105] calculates flat and dynamic prices where the extra energy generation and purchase costs are divided equally as an award among the consumers. The consumer reaction is addressed in [106] to maximize the suppliers’ profits.

A different methodology is presented in [107], where a semi-vectorial bilevel programming approach is developed. The aim is to find the best pricing decisions for the UC by modeling the interaction between the latter and consumers to optimize electricity time-of-use retail pricing.

Unlike in the existing literature, in a realistic scenario, optimal pricing must guarantee that both the UC and the consumers robustly benefit from an improved overall consumption pattern even if the predicted consumers’ reaction to the tariffs is subject to uncertainty.

Moreover, the consumers with different consumption characteristics usually coexist in an area supplied by the same retailer; hence consumer classification and analysis of the ability to shift some of their loads are necessary. Each consumer class contributes differently to the overall power peak that the UC seeks to smoothen. This implies that distinguishing which groups contribute the most to the power peak and applying different tariffs to the appropriate groups may result in a better optimum. Furthermore, The challenge is to determine the shape of the designed ToU tariff within the optimization to capture the fluctuation of the demand curve of each type of consumer. Hence, adopting fixed time blocks cannot be applicable to all kinds of consumers. Predefining constraints on possible minimum and maximum ToU tariffs may limit the calculation range, and the optimal ToU tariffs may not be reached.

Despite the uncertainty, consumer behavior model parameters should be identifiable based on consumption data obtained after introducing the ToU tariff while respecting the consumers’

privacy, which allows reducing the uncertainty and periodically recalculating optimal prices.

1.3 Research Methodology

Due to several factors, such as the modernization of the power grid, the rapid advancing of technologies, and the future towards clean energy, the idea behind my research was born to extend the existing literature by proposing and designing novel methods. The focus of my thesis is mainly on creating control opportunities and developing algorithms responsible for that.

In this research work, I started by studying the state-of-art methods regarding DSM

strategies. An overview of their advantages and disadvantages concerning the consumers’

acceptance of DR programs and the efficiency of achieving the required goals on consumer and UC sides is presented in the previous sections.

My focus is on the price-based DR strategies since they are more attractive to the con- sumers and can be easily implemented. This solution will allow the consumers to have an active role in the power system to maximize their profits by receiving incentives from the UC after their participation. The DR program presented here consists of automated algorithms that control the thermal and nonthermal appliances of multiple households and manage the procurement of electric energy from various energy sources.

For the completeness of my work, dynamical tariffs are designed based on the consump- tion data of residential consumers to motivate them to rearrange their consumption profile.

An analysis of consumers’ behavior in the same area is first carried out to obtain a better optimum by implementing dynamic tariffs. The consumption data are real data provided by Hungarian utility companies.

My goal is to construct a closed-loop control strategy consists of different consumers responding to dynamical electricity tariffs and UC designing appropriate dynamical prices based on the recorded consumption data and the predicted consumption of the following supply horizon for the consumers population. Figure 1.2 summarizes the proposed thesis framework.

Figure 1.2: Thesis framework.

The designed algorithms are implemented in the Matlab/Simulink environment, and their efficiency is demonstrated by using real data through different simulation scenarios. Some parameters properties of the presented algorithms are also analyzed.

1.4 Thesis Outline

The rest of the thesis is organized as follows:

Chapter 2 introduces the algorithms developed for consumers to control their consump- tion profile in response to a ToU electricity tariff. The EV charging and discharging manage- ment are also developed through a novel algorithm for one household. Furthermore, a novel method is introduced to robustify the control components to guarantee consumer satisfaction at all times.

Multiple households with different consumption behaviors interact to minimize the amount of electric energy procured from the grid is also presented. This algorithm assumes the pres- ence of distributed energy resources such as photovoltaic panels, energy storage systems, and neighbors. The algorithm seeks to create a virtual energy market where each household can sell and purchase energy from the neighbors.

The design of dynamic pricing strategy on UC side is elaborated in Chapter 3. The latter consists of algorithms that analyze consumers’ behavior, select the eligible consumers that can participate in the DR program, and design the various ToU electricity tariffs structures compatible with the consumers’ behaviors. For that purpose, a consumer behavior model is also introduced where identification of its parameters is also presented. The proposed identification method is consumer privacy-friendly. A robust analysis of the designed ToU tariffs is studied against the uncertainties presented in the consumer behavior model.

Finally, Chapter 4 summarizes the main points of the proposed work and the results obtained as well as future research activities are provided.