为了优化完全竞争市场中新能源光储电站的储能充放电与售电经济性决策,提出了一种基于长短时记忆(Long Short Term Memory,LSTM)神经网络及日电价极值研判的电能存储与销售的优化策略.采用LSTM神经网络挖掘电价信息的隐含特征,对日内实时电价进行滚动预测;构建电价系数指标,确立日电价极值时刻,并进行电价极值概率校验;采用动态规划算法将优化策略分解为初始状态到终值状态的多阶段优化决策,确立了售电优化策略.使用某竞争性电力市场数据进行了算例测试,结果表明,实施优化后的电站全年售电收益提升了 17.4％.相比于4种典型传统模型,所提策略的售电收益分别提高了 3.8％、11％、10.4％和7.0％.
Power Sale Strategy for Photovoltaic Plants with Energy Storage Based on Electricity Price Prediction and Extreme Value Judgement
An optimization strategy for power storage and sales based on long short-term memory(LSTM)neu-ral network and intraday extreme value judgement is proposed for the economic decision of photovoltaic plants with energy storage in a perfectly competitive market.An LSTM neural network is used to mine the implied fea-tures of electricity price information to make rolling forecasts of intraday real-time electricity prices.An electric-ity price coefficient index and a daily electricity price extreme moment are established respectively.Then the electricity price extreme probability calibration is conducted.The dynamic programming algorithm is used to de-compose the optimization strategy into a multi-stage optimization decision from the initial state to the final value state.The real electricity price data are used for example testing.The test results indicated that the annual power sale revenue increased by 17.4％following the implementation of the proposed optimization strategy.Compared with the four traditional models,the electricity sale revenue based on the proposed strategy increased by 3.8％,11％,10.4％,and 7.0％,respectively.
photovoltaic plant with energy storageelectricity price predictionextreme value judgementpower sale strategy
万方数据
维普
