Abstract
In order to reduce the adverse impact of photovoltaic generation instability on power grid, a model for predicting photovoltaic output value based on Echo State Networks with weather type index was proposed. The weather type index was established by the ratio relationship between the average power generation of different weather types, and then it was used as one of the input variables of the prediction model. Compared with traditional neural network, Echo State Networks has great improvement in stability, global optimality, local minimum problem, convergence speed and complexity of training process. Furthermore, leaky integrator neuron was used to optimize the classical neuron model. It enhanced the short-term memory ability of the reserve pool and improved the prediction ability and accuracy for time series. In the simulation experiment, using MATLAB it was verified that compared with BP neural network and classical ESN, the proposed method had higher prediction accuracy and showed better prediction effect in predicting photovoltaic power output.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call