The Application Of Spiking Neural Network In Wind Speed Forecasting Of The Wind Plant

In recent years,with the rapid development of clean energy,wind energy is widely concerned due to its no pollution,renewable and wide distribution,wind power is a major means of wind energy development and utilization,with the large-scale wind farms and industrialization and large-scale wind power grid,the inherent incontrollable,intermittency and randomness of wind power have dreadful effect on the safe and stable operation of power system and the quality of power.Wind speed is the most important factor affecting the output power of wind farms.Therefore,wind speed forecasting is an important part of the research of power system operation and can provide important reference for power load forecasting.This paper focuses on the research of wind speed forecasting,summarizes the development status and the practical significance of the wind speed forecasting research.With the history of wind speed measurement data as the research object,according to the wind data inspection standard to check the bad data and find the missing data then optimized and completed these data by sliding average method and interpolation method.The neuron model,network structure,time coding method and network learning algorithm of Spiking neural network are introduced in detail.Based on Spiking neural networks(SNN)as the core,combined with improved SpikeProp learning algorithm to establish wind speed prediction model,based on the basic SpikeProp network learning algorithm,to use the adaptive learning rate with momentum term to speed up convergence and improve the dynamic performance of the wind speed training process.In order to prove the effectiveness of the SNN wind speed forecasting model,using the BP neural network and RBF neural network for comparison,using the optimized wind data as the experimental sample for simulation.The simulation results were evaluated by the mean absolute error(MAE),the mean absolute percentage error(MAPE)and the root mean square error(RMSE)three kinds of evaluation indicators,the results show that the SNN prediction model has higher accuracy,but the convergence speed is slower.To demonstrate the validity of the improved SpikeProp algorithm,the simulation results are compared with the SpikeProp algorithm simulation.Simulation results show that in the condition that the number of neurons in the hidden layer is the same,the improve prediction model converges faster and the prediction accuracy is further improved,which proves the feasibility and effectiveness of the SNN wind speed prediction model.