1.Key Laboratory of Control of Power Transmission and Conversion, Ministry of Education (Shanghai Jiao Tong University), Shanghai 200240, China;2.Shanghai Non-carbon Energy Conversion and Utilization Institute, Shanghai 200240, China;3.China Electric Power Research Institute (Nanjing), Nanjing 210003, China;4.East China Branch of State Grid Corporation of China, Shanghai 200120, China

The operation of photovoltaic-energy storage plants includes two parts: photovoltaic power forecasting and resource optimal scheduling. The photovoltaic power forecasting is usually carried out first with the goal of the highest forecasting accuracy, and then the optimal scheduling of the photovoltaic-energy storage plant is carried out based on the forecasting curve. However, there is a nonlinear and asymmetric relationship between the objective function value of optimal scheduling problem and the photovoltaic power forecasting error. At the approximate forecasting error level, the photovoltaic power forecasting result aiming at the highest accuracy does not necessarily lead to the maximum operation revenue of the photovoltaic-energy storage plant. In this regard, a value-oriented photovoltaic power forecasting method for improving the operation revenue of photovoltaic-energy storage plants is proposed. Firstly, a bi-level optimization problem is constructed, which includes photovoltaic power forecasting and power plant operation. The upper layer is the training problem of the day-ahead photovoltaic power forecasting model, and the lower layer is the two-stage optimal scheduling problem of day-ahead bidding and intra-day operation for the photovoltaic-energy storage plant under the conditions of given photovoltaic power forecasting model. Then, the upper forecasting problem is transformed into a combined forecasting form, and the forecasting model parameters are set as weight coefficients. A parameter solving method for the forecasting model based on iterative optimization is designed. Finally, using actual photovoltaic power plant data and electricity price data for case analysis, and compared with the photovoltaic power forecasting method that aims to achieve the highest forecasting accuracy, the effectiveness of the proposed method in improving the operation income of power plants is verified.