Research On Spot Electricity Market Decision-Making Models Based On Different Perspectives And Subject Characteristics

In the electricity market systems in the United States and Europe,the spot electricity market has played an important role in finding prices,balancing power and maintaining the security and stability of the power system.Therefore,in recent years,the spot electricity market has increasingly become an important sign of the relatively perfect construction of electricity market system in a country or a region.In September 2017,the National Energy Administration(NEA)issued Notice on The Pilot Project for The Construction of Electricity Spot Market which indicates that the construction of spot electricity market in China has been officially launched.In the spot electricity market environment,the participants(seller and buyer)involved in the transaction,as well as the relevant agency responsible for organizing the transactions,are all important economic behavior decision-making subjects.On one hand,the transaction participants can make the transaction declaration decision to improve their own profits,the transaction organizer can make the spot market clearing decision to ensure the dual realization of the dispatching operation and the market transaction;On the other hand,the whole market can achieve optimal allocation of electric resources under the action of all subjects’reasonable decisions.Thus,it can be seen that reasonably conducting or guiding various economic decision-makings are important means to ensure the effective operation of the spot electricity market,whether from the perspective of transaction participant or organizer,or from the perspective of the whole market.In this study,on the basis of the transaction participation of new energy sellers,distribution companies(buyers)and power retailers(buyers),we construct,improve and simulate various spot market economic behavior decision-making models from both perspectives of the single-and multi-subject,using the techniques of uncertainty optimization and multi-agent decision-making.The purpose of this study is to form a scientific and feasible spot market decision-making model system which is suitable for different subjects and market scales.The specific research content and results are as follows:(1)Construction and simulations of single-subject new energy seller’s decision-making model in spot market based on stochastic-robust hybrid optimization method.On the premise of large scale spot market,taking the wind farm-energy storage system(WF-ESS)as the representative for new energy seller,we construct and simulate the WF-ESS’s single-subject decision-making model.In view of the uncertainties faced by the WF-ESS in spot market,such as the day-ahead,balancing prices and the real-time natural wind power,as well as the flexible power regulation capability of the energy storage equipment,a single-subject decision-making model system based on the stochastic-robust hybrid optimization technology is constructed.This model system is composed of a day-ahead market model and several balancing market models,which can not only reasonably realize the functions of energy storage equipment to offset power deviation and to promote the strategic joint power offering,but also combine the characteristics and advantages of the stochastic and robust optimization methods for improving the effect of decision-making,while generating decision-making outputs such as day-ahead and balancing market joint power offerings during a rolling process.The simulations are carried out in the Matlab software by running the model related programs and based on the actual historical data.On one hand,the scientificity and feasibility of this model system are investigated through trial calculation and solution;On the other hand,by comparing with other methods,the decision-making effect of this model system in improving WF-ESS’s profit is verified.(2)Construction and simulations of single-subject buyer’s decision-making model in spot market based on random integer mixed optimization and two-level decision-making method.On the premise of large scale spot market,taking the power retailer as the representative for buyer,we construct and simulate the power retailer’s single-subject decision-making model.In view of the uncertainties faced by the power retailer in spot market,such as the day-ahead,balancing prices and the real-time load,a single-subject two-level decision-making model system based on the random integer mixed optimization technology is constructed.This model system can make decisions on both the optimal number of piecewise segments and optimal bidding price of every piecewise segment in the power retailer’s day-ahead bidding curve through the internal and external two-level model structure.Hence,compared with the methods which can only decide optimal bidding prices,this model system theoretically improves the effect of decision-making.The simulations are carried out in the Matlab software by running the model related programs and based on the actual historical data.On one hand,the scientificity and feasibility of this model system are investigated through trial calculation and solution;On the other hand,by comparing with other methods,the decision-making effect of this model system in improving power retailer’s profit is verified.(3)Construction and simulations of spot market multi-subject decision-making model based on GDCAC algorithm.Under the premise that the scale of spot market is small or relatively limited,taking the day-ahead market without new energy sellers as the main representative,we construct and simulate the day-ahead market multi-subject(bidding)decision-making model.On the basis of regarding all the conventional generators and power retailers as agents,it is introduced the gradient descent continuous actor-critic(GDCAC)reinforcement learning algorithm to simulate the decision-making mechanism of every agent in the bidding process to learn the optimal bidding strategy,thereby the day-ahead market multi-subject decision-making model based on GDCAC algorithm is constructed.Through the GDCAC algorithm,every seller or buyer can learn its optimal bidding strategy from the corresponding continuous value space,without causing the "curse of dimensionality".Hence,compared with the models which are constructed based on discrete space reinforcement learning algorithms,this model theoretically improves the effect of decision-making.The simulations are carried out in the Matlab software by running the model related programs and based on the market case(on IEEE30-bus test system)which contains 6 conventional generators and 20 power retailers bidding together.On one hand,the scientificity and feasibility of this model are investigated through trial calculation and solution;On the other hand,by comparing with other methods,the decision-making effects of this model such as improving agents’ profits and reducing the market operation cost are verified.(4)Spot market clearing model improvement and multi-subject decision-making model construction under new energy sellers bidding based on robust optimization and L SC AC algorithm.Under the premise that the scale of spot market is small or relatively limited,taking the day-ahead market with multi-wind farm bidding as the main representative,we improve,construct and simulate the day-ahead market clearing and multi-subject decision-making models.In view of the real-time natural wind power uncertainties in the market,based on the method of reflecting robustness in the constraints,the day-ahead market clearing model is transformed as a robust optimization one which can significantly improve wind power accommodation.Based on the proposed robust market clearing model,it is introduced the least square continuous actor-critic(LSCAC)reinforcement learning algorithm to simulate the decision-making mechanism of every wind farm agent in the bidding process to learn the optimal bidding strategy,thereby the day-ahead market multi-subject decision-making model based on LSCAC algorithm is constructed.Through the LSCAC algorithm,every wind farm agent also can learn its optimal bidding strategy from the corresponding continuous value space without causing the "curse of dimensionality".The simulations are carried out in the Matlab software by running the model related programs and based on the market case(on IEEE30-bus test system)which contains 5 wind farms bidding together.On one hand,the scientific and feasibility of the proposed market clearing and multi-subject decision-making models are investigated through trial calculation and solution;On the other hand,by comparing with other methods,the decision-making effects of the proposed market clearing and multi-subject models such as improving wind power accommodation,improving wind farm agents’ profits and reducing the market operation cost are verified.In summary,we propose a spot market decision-making model system which is suitable for different subjects and market scales.In this system,the proposed single-subject models can provide scientific quantitative decision-making tools for "price taker" transaction participants in spot market;The improved market clearing model can significantly improve wind power accommodation without introducing the capacity market;The proposed multi-subject models can,on one hand,provide scientific quantitative decision-making tools for "price taker" transaction participants in spot market,on the other hand,provide quantitative analysis tool and simulation platform for market economy operation simulation,market planning and design,and policy effectiveness evaluation etc..