Merge branch 'master' of github.com:sanderploegsma/ti3706

chapters/conclusion.tex

@@ -2,9 +2,11 @@ \section{Conclusion}\label{conclusion}
 \acresetall
 In this paper we addressed the question: ``What are the current challenges in the design of smart grids when utilising game theoretic approaches?'' In order to give a well formed answer to this question we have looked at some problems that might occur when using smart grids, and how game theoretic methods can help solve them. When we looked at challenges using smart grids several key issues popped up. 
 
-The most important task of the smart grid is evidently supplying all of its users with sufficient energy. This is not trivial because energy demand cannot always be met by energy production. Additionally the amount of energy that can be produced by renewable energy sources varies from time to time due to weather circumstances for instance, and the amount of energy that is requested by users fluctuates during the day. There are several games that target user demands and aim to spread energy demand equally over the day. Energy demand can be managed with adjustments to the energy tariff. A \ac{rtp} method based on the \ac{vcg} mechanism can ensure that users oblige to predict their own energy consumption and penalising them for (grave) inaccuracy in their prediction. This way users are encouraged to report their estimated energy consumption truthfully and accurately. 
+The most important task of the smart grid is evidently supplying all of its users with sufficient energy. This is not trivial because energy demand cannot always be met by energy production. Additionally the amount of energy that can be produced by renewable energy sources varies from time to time due to weather circumstances for instance, and the amount of energy that is requested by users fluctuates during the day. There are several games that target user demands and aim to spread energy demand equally over the day. Energy demand can be managed with adjustments to the energy tariff. 
 
-Another \ac{rtp} method based on a Stackelberg game can be modelled. The load on the energy grid can be spread out with changes to the energy tariff. When all prosumers install smart meters in their homes, the meter will negotiate when energy will be used, and how much. For instance, this way the program of a washing machine can be scheduled to run at a beneficial time. When \acp{ev} are incorporated into a home, the smart meter, and by extent the smart grid, can decide to use energy from the battery of an \ac{ev} in order to save expenses when tariffs are high, or even to sell at a good price to other users that need energy. The \ac{ev} can then later be charged again when the tariffs are lower. 
+\ac{rtp} methods based on the \ac{vcg} and \ac{agv} mechanisms can ensure that users have to predict their own energy consumption accurately and announce it truthfully. The \ac{rtp} method based on \ac{vcg} shows that there is no better payoff when inaccurate and false information is announced. Another method based on a \ac{agv} mechanism penalises users that provide inaccurate or false information with a fee.
+
+Another \ac{rtp} method based on a Stackelberg game can be modelled. The load on the energy grid can be spread out with changes to the energy tariff. When all prosumers install smart meters in their homes, the meter will negotiate when energy will be used, and how much. For instance, this way the program of a washing machine can be scheduled to run at a beneficial time. When \acp{ev} are incorporated into a home, the smart meter can decide to use energy from the battery of an \ac{ev} in order to save expenses when tariffs are high, or even to sell at a good price to other users that need energy. The \ac{ev} can then later be charged again when the tariffs are lower. 
 We think especially dynamic energy pricing and energy storage are important factors that contribute to a steady power supply in the smart grid.
 
 The smart grid has a lot of \acp{der}, consumers and prosumers connected to it. This can be a challenge when trying to solve computational challenges, such as predicting energy consumption patterns or trying to reach an equilibrium with \ac{rtp}. When there are too many users the problem will be too complex and cannot be solved quickly or accurately enough. 

chapters/demand_side_management.tex

@@ -1,6 +1,6 @@
 \section{Demand side management}\label{dsm}
 \acresetall
-Another challenge in the design of smart grids is managing the demand of the consumer. \ac{dsm} focuses on two main problems: reducing the total consumption of energy and shifting demand during peak hours \cite{keypaper}. Additionally, the power grid is currently specifically designed to be able to cope with peak rather than average demand to achieve high levels of reliability. As a result, it causes underutilisation of the power system \cite{MaDengSongEtAl2014}. Utilising game theory, solutions to implementing an efficient scheme for these problems, have been proposed in \cite{SamadiMohsenian-RadSchoberEtAl2012, SamadiSchoberWong2011, MaDengSongEtAl2014, MaharjanZhuZhangEtAl2013, ChenKishoreSnyder2011, ChenLiLowEtAl2010, Mohsenian-RadWongJatskevichEtAl2010a, SalinasLiLi2013, CaronKesidis2010, DepuruWangDevabhaktuni2011a}. These solutions range from devising smart pricing schemes, shaping behaviour of consumers and scheduling schemes for energy consumption. In this chapter we will discuss the proposed solutions respectively. Out of many proposed solutions we consider these to be showing very promising results for future research and integration with the smart grid.
+Another challenge in the design of smart grids is managing the demand of the consumer. \ac{dsm} focuses on two main problems: reducing the total consumption of energy and shifting demand during peak hours \cite{keypaper}. Additionally, the power grid is currently specifically designed to be able to cope with peak rather than average demand to achieve high levels of reliability. As a result, it causes underutilisation of the power system \cite{MaDengSongEtAl2014}. Utilising game theory solutions to implement an efficient scheme for these problems have been proposed in \cite{SamadiMohsenian-RadSchoberEtAl2012, SamadiSchoberWong2011, MaDengSongEtAl2014, MaharjanZhuZhangEtAl2013, ChenKishoreSnyder2011, ChenLiLowEtAl2010, Mohsenian-RadWongJatskevichEtAl2010a, SalinasLiLi2013, CaronKesidis2010, DepuruWangDevabhaktuni2011a}. These solutions range from devising smart pricing schemes, shaping behaviour of consumers and scheduling schemes for energy consumption. In this chapter we will discuss the proposed solutions respectively. Out of many we consider these to be showing very promising results for future research and integration with the smart grid.
 
 \subsection{Smart pricing}
 One way to manage the demand side is by designing smart tariffs. Designing these smart tariffs can be done through pricing methods such as \ac{rtp} (i.e. a higher load correlates to a higher price and vice versa), adaptive pricing and peak load pricing \cite{SamadiMohsenian-RadSchoberEtAl2012}. A new \ac{rtp} method based on the \ac{vcg} mechanism is proposed in \cite{SamadiMohsenian-RadSchoberEtAl2012}. In the proposed method tariffs are determined based on the local information obtained from users reporting their expected demand during certain time slots. The method aims to reduce total energy consumption. This is achieved by the supplier calculating the optimal consumption level for the user based on the reported energy consumption and assigning a new tariff. Because users can anticipate the impact on the tariffs and therefore possibly lie about their true energy consumption to get a lower tariff, the VCG mechanism will help to ensure that users will not have any incentive to do so by showing that the user cannot do better than reporting their demand truthfully. As a result, the proposed system is deemed efficient and shows that it lowers total energy consumption, because the utilities of the users and cost on the supplier is respectively maximised and minimised. Another \ac{vcg}-based mechanism is proposed in \cite{SamadiSchoberWong2011}. It tries to shift the energy consumption to off-peak hours. Similar to \cite{SamadiMohsenian-RadSchoberEtAl2012}, it requires users to report their demand to the supplier. However, the supplier computes the optimal consumption schedule for the user rather than the optimal level of consumption. This schedule aims to balance the total level of consumption.
@@ -13,9 +13,9 @@ \subsection{Smart pricing}
 We have revealed that applications of smart pricing can both reduce the total level of energy consumption and shift peak load. However, we believe smart pricing may cause large price swings, when energy is either superfluous or scarce. Consumers may find the instability of the tariffs undesirable, because the more rational choice to either consume more energy when tariffs are low or less when tariffs are high may interfere with their daily life. 
 
 \subsection{Energy consumption scheduling}
-A lot of papers have presented research in energy consumption scheduling. A good example is where two models are given to match the supply and shape the demand of energy in order to get a market equilibrium \cite{ChenLiLowEtAl2010}. An algorithm is given where both companies and customers jointly run the market with an iterative bidding scheme to find an equilibrium price and allocation before the actual action of demand reduction \cite{ChenLiLowEtAl2010}. This is a way to achieve a market-clearing price (i.e. a price where supply equals demand).
+A lot of papers have presented research in energy consumption scheduling. A good example is where two models are given to match the supply and shape the demand of energy in order to get a market equilibrium \cite{ChenLiLowEtAl2010}. An algorithm is given where both companies and customers jointly run the market with an iterative bidding scheme to find an equilibrium price and allocation before the actual action of demand reduction \cite{ChenLiLowEtAl2010}. This is a way to achieve a market-clearing price, i.e. a price where supply equals demand.
 
-The previous schemes about direct load control and smart pricing are all focused on the interaction between energy suppliers and individual end-users. Other studies show that it can be beneficial to model the energy usage of end-users all together in an aggregate load \cite{Mohsenian-RadWongJatskevichEtAl2010a, SalinasLiLi2013, ZhuTangLambotharanEtAl2011}. The algorithm to optimise the schedule for all users however is proven to be non-polynomial \cite{CaronKesidis2010}. To minimise the cost for all the users would take too long and would not be able to be computed in time. Instead of this, the problem can be distributed over all users. This can be done by letting the users communicate whenever it can be beneficial to coordinate their usage. That idea is modelled as a non-cooperative congestion game. The goal of the game is of course to minimise the total cost of the energy used in a smart grid. In the game all schedules are known and in turn each user chooses its schedule such that it has the lowest cost possible. It then sends this new schedule to the rest of the users, who will in turn, if possible with this new information, choose a new schedule with a lower cost than before. The mechanic of this game is based on the fact that if the total load increases the cost is higher and that the cost functions are convex. Because the users profit from scheduling their energy consumption at times with the lowest load a Nash equilibrium can be established such that no single user can increase its profit any further \cite{Mohsenian-RadWongJatskevichEtAl2010a, ZhuTangLambotharanEtAl2011, IbarsNavarroGiupponi2010}.
+The previous schemes about direct load control and smart pricing are all focused on the interaction between energy suppliers and individual end-users. Other studies show that it can be beneficial to model the energy usage of end-users all together in an aggregate load \cite{Mohsenian-RadWongJatskevichEtAl2010a, SalinasLiLi2013, ZhuTangLambotharanEtAl2011}. The algorithm to optimise the schedule for all users however is proven to be non-polynomial \cite{CaronKesidis2010}. To minimise the cost for all the users would take too long and would not be able to be computed in time. Instead of this, the problem can be distributed over all users. This can be done by letting the users communicate whenever it can be beneficial to coordinate their usage. That idea is modelled as a non-cooperative congestion game. The goal of the game is of course to minimise the total cost of the energy used in a smart grid. In the game all schedules are known and in turn each user chooses its schedule in such a way that it has the lowest cost possible. It then sends this new schedule to the rest of the users, who will in turn, if possible with this new information, choose a new schedule with a lower cost than before. The mechanic of this game is based on the fact that, if the total load increases, the cost is higher and that the cost functions are convex. Because the users profit from scheduling their energy consumption at times with the lowest load a Nash equilibrium can be established in such a way that no single user can increase its profit any further \cite{Mohsenian-RadWongJatskevichEtAl2010a, ZhuTangLambotharanEtAl2011, IbarsNavarroGiupponi2010}.
 
 The \emph{players} of this game are not the actual users of the homes but rather the smart meters they have installed. If the equilibrium is reached and the meter has an optimal schedule it can autonomously enforce this schedule. It of course takes into account what appliances can and cannot be scheduled at different times. For example a fridge always needs to be turned on and a washing machine can be scheduled to run at a different time \cite{DepuruWangDevabhaktuni2011a}.