Short-term Photovoltaic Power Forecasting Model Based on Improved BP Neural Network Method

Photovoltaic power generation is an important part of the global energy system and energy internet. Accurate photovoltaic power forecasting is beneficial to the optimization of power grid scheduling and the promotion of clean energy consumption. This paper analyzes the seasonal and daily characteristics of the photovoltaic power, uses self-organizing maps (SOM) to dig out the inherent disciplines and essential properties of data, clustering sequences with similar characteristics. Then the MEA-BP model is established which applies mind evolutionary algorithm (MEA) to improving BP neural network. MEA coordinates the local and global values by introducing the evolution mechanism of convergence and dissimilation to optimize the weight and threshold of BP. Next, neural network error correction model is proposed to fix forecasting error. Based on the data of a 50MW photovoltaic power station, the numerical results show that the model can improve the accuracy of photovoltaic power forecasting effectively.