Agent Based Reinforcement Learning Approach towards Smart Generator Scheduling under Deregulated Electricity Market

Abstract

Electric Power System had undergone remarkable changes in the last few years in order to achieve the aim of being a Smart Grid. In India the change is being visible after the implementation of the Electricity Act 2003. This mainly includes the articles related to trading of electricity and promoting competition in this field. One of the main features of this act is delicensing of generation and hence captive generation is being freely permitted. There is a provision for unbundling of services provided by state owned power entities. This action helps in the development of deregulated electricity market. All the entities in the power market are under the control of Independent System newlineOperator (ISO). The objective of each entity in power market is to accomplish high net newlineearnings. Every Generation Company bids in market and the winners will clear the newlinemarket at a particular Market Clearing Price. This process will continue until congestion newlinehappens in the transmission line. The marginal pricing scheme implemented is Locational Marginal Pricing (LMP) once the power transfer exceeds the limit. In order to achieve the objective in less time a layered modelling structure is required, for that all the entities is considered as agents having the learning capability. The learning method used is reinforcement learning technique. The strategic trading in electricity market and hence the achievement of attaining high net earnings is implemented through Variant Roth- newlineErev (VRE) Reinforcement algorithm.The research work mainly concentrates on formulating a mathematical model for executing the agent based computation and hence to implement an efficient electricity market structure considering the entities as agents. The results obtained considering various IEEE test systems and real time systems show the importance of agent based newlineapproach in a deregulated electricity market. Here GenCo will report higher than true cost newlineand will act as an agent having the capability of interactive learning. Such agents exhibit economic capacity ..