一种基于深度强化学习算法的新能源风光场站参与未来放松管制情况下的零售电力市场的运维及交易策略  

Deep reinforcement learning based operation and bidding strategies for renewable energy sources projects in the future deregulated retail electricity market

作　　者：梁哲铭 王力军 赵建勇 刘庆伏 叶林 王建国 LIANG Zheming;WANG Lijun;ZHAO Jianyong;LIU Qingfu;YE Lin;WANG Jianguo(Huaneng Renewables Co.,Ltd.,Huaneng Build.,Haidian 100036 Beijing,China)

机构地区：[1]华能新能源股份有限公司,北京海淀100036

出　　处：《电力大数据》2021年第9期17-24,共8页Power Systems and Big Data

摘　　要：新能源风光场站参与未来放松管制情况下的零售电力市场面临如下问题:(1)上网电量结算方式由原有的含补贴的固定电价转为无补贴的实时电价;(2)新能源风光场站在运行和维护过程中会遇到大量的不确定因素,包括风速、太阳辐照度、限功率运行情况以及升压站内部用电等;(3)新能源风光场站需要在最小化运维成本的情况下满足较高运维水平的要求。本文引入了一种无监督的深度强化学习算法,利用新能源风光场站内部可控电力设备解决运维及交易过程中遇到的不确定因素,并使售电电量与运维满意度最大化。大量基于实际数据的仿真结果表明,本文所提出的基于深度强化学习算法的运维及交易策略既能最大限度地降低新能源风光场站运维成本,又能最大限度地向零售电力市场售电并维持较高的综合运行水平。The following problems will be faced when the new energy scenic spot station participates in the future retail electricity market under the condition of deregulation:(1)the settlement method of electricity is changed from the original fixed electricity price with subsidy to the real-time electricity price without subsidy;(2)a large number of uncertainties may be encountered during the operation and maintenance process of the renewable energy resources projects,including wind speed,solar irradiance,power limiting operation and internal power consumption of the substations;(3)the renewable energy resources projects needs to meet the requirements of improving the comprehensive operation and maintenance satisfaction levels while minimizing the operation and maintenance costs.In this paper,an unsupervised deep reinforcement learning algorithm is introduced to solve the uncertain factors in the process of operation and transaction by using the controllable electric power equipment in the new energy scenic spot station.A large number of simulation results based on actual data show that the operation and transaction strategy based on the deep reinforcement learning algorithm proposed in this paper can not only reduce the operation and maintenance cost of the new energy scenic spot station to the maximum,at the same time,it can sell electricity to the retail electricity market to the maximum extent and maintain a high level of comprehensive operation.

关 键 词：深度强化学习算法 零售电力市场 新能源风光场站 不确定因素 运维水平 

分 类 号：TM732[电气工程—电力系统及自动化]