Ultra-short-term photovoltaic cluster power prediction based on improved reinforcement learning
Photovoltaic energy is vulnerable to external interference,and its power fluctuates greatly.In or-der to ensure the operation safety of photovoltaic substation,an ultra-short-term photovoltaic cluster power forecasting method based on improved reinforcement learning is proposed.The key power data is extracted by mutual information analysis of the correlation between photovoltaic power data,the dimension of photo-voltaic power data is reduced by principal component analysis,and the predicted value of ultra-short-term cluster power is obtained by using LSTM neural network.By improving reinforcement learning algorithm,a large number of predicted values are continuously updated and trained,and the real-time power predicted value is obtained,so as to realize the ultra-short-term photovoltaic cluster power prediction.The experiment results show that the proposed method has good power prediction performance and shorter running time.
万方数据
维普
