Research On Photovoltaic Power Prediction Model Based On Echo State Neural Network

With the acceleration of global energy Internet construction and large-scale development of clean energy, the newly-emerging energy technology is receiving growing attention from the whole world. Photovoltaic as a kind of clean energy in the smart grid construction, has been widely applied and developed. Photovoltaic power generation system of the smart grid, however, due to the influence of various factors, can easily suffer instability in terms of its powers, which results in an impact on the grid after massive parallel operations, thus failing to support the grid to perform safely and steadily.Therefore, the accurate and efficient prediction of photovoltaic power is of profound and great significance for advancing the construction of smart grid.This paper elaborates on the technologies related to photovoltaic and conducts a deep discussion on their development in the smart grid, pointing out the high necessity of the accurate prediction of photovoltaic power. After comprehensive analyses on the influence factors of photovoltaic power generation, it proves that the full use of those influence factors bearing stronger correction is but an effective way to predict the photovoltaic power precisely.On account of the instability of photovoltaic power, this paper presents a corresponding forecasting model based on echo state neural network. State of the echo of the neural network hidden layer is a kind of dynamic reserve pool structure, with the echo state attributes, not only enhance the stability of network prediction, and only use the weights of the network output can be obtained by linear algorithm, simplifies the process of training, and overcome the slow convergence speed and traditional neural network into the problem of local minimum. Using actual of the historical data and meteorological data of the photovoltaic power station simulation, simulation results show that based on echo state photovoltaic power prediction model of neural network has good prediction accuracy and prediction stability.Due to the closed-loop autonomy of this neural network prediction model based on echo state, the predicted results can easily sustain the error accumulation. In order to further improve the precision and stability of the photovoltaic power prediction, this paper also presents the prediction model of photovoltaic power based on modular echo state neutral network. Firstly, this model reconstructs the state space of reserve pool, and then rank of reserve pool structure so as to make each layer contain equal area modules,among which each bears a different function, thus greatly reducing the coupling between neurons and improving the dynamic handling ability of the reserve pool. The simulation shows that the prediction model of photovoltaic power based on echo state neural network, compared with that built on modular echo state neutral network, has favorable prediction performance.