Research On Prediction Of Photovoltaic Power Generation Based On Intelligent Water Drops Algorithm And Neural Network

With the development of industrialization and the increase of population, human's demand for energy continues to grow, the traditional fossil energy is facing the problem of depletion, develop and utilize clean energy, goes a green, low-carbon, clean alternative route is the main theme of the future. Solar energy is a kind of clean energy, photovoltaic power generation is the main way to develop and use solar energy, with the advantages of safe, reliable, flexible application, simple installation and maintenance, it has broad prospects for development. But the generation of photovoltaic are affected by weather type, temperature, solar radiation and other factors, so its output has obvious fluctuation, intermittent and uncontrollability, large-scale photovoltaic power plants connect to grid will cause serious influence and economic operation to the power system, so the accurate prediction of photovoltaic power generation has important significance.Firstly, a new swarm intelligent optimization algorithm named Intelligent Water Drops algorithm is introcuded in this paper. The algorithm is formed by obversing the interaction between river and bed, it has the advantages of strong robust and good global optimization. However, the standard intelligent water drop algorithm is prone to stagnation and obstruction in the choice of nodes, and then the convergence speed is slow and easy to fall into local optimal solution. Two strategies are proposed to improve the algorithm, then the function of Rosenbrock, Rastrigin, Griewank and Schaffer F6 are optimized by the algorithm respectively. The experimental results show that the improved algorithm can effectively improve the efficiency of the water drop, and it is suitable for solving optimization problems.Secondly, for the drawbacks of static feedforward BP neural network has low accuracy and no memory problems in sequence prediction. A power forecasting model based on Elman neural network with dynamic and local memory function is proposed. The weights and thresholds of Elman neural network are optimized by using the improved intelligent water drop algorithm, which improves the training efficiency of the network.Finally, the influence factors of photovoltaic power generation are analyzed in detail, and a model named intelligent water drops algorithm optimized the Elman neural network is established to forecast output of photovoltaic plant. According to the idea of modularization, the model be divided into three sub models, which are sunny, cloudy and rainy day. IWD-Elman model, the traditional Elman neural network model and BP neural network model are compared with the actual output data of photovoltaic power station. The results show that the proposed IWD-Elman neural network model has higher prediction accuracy.