Research On Short-term Prediction Algorithm Of Wind Power Based On Weight Agnostic Neural Networks

As a kind of green and clean energy,wind power starts to change from supplementary energy to strategic alternative energy because of its advantages of low cost and easy exploitation and utilization.About 20%of China's land area is rich in wind power resources,and both the scale and level of development have made great progress and improvement.Wind power has huge development potential in China.However,due to the randomness and interm ittency of wind energy,its power output is not stable,and it also brings challenges to the normal and stable operation of power system.Therefore,only by doing a good job in wind power forecasting can we effectively manage the operation of wind farms and ensure the safety and power quality of power systems.Because of the nonlinear fitting advantage of neural network algorithm,the neural network algorithm is widely used in wind power prediction,but most of these studies have raw data that has not been preprocessed,and model input parameter selection problems such as too little or too much parameters been selected lead to poor prediction accuracy of the models,at the same time,the wind power short-term predicting algorithm based on neural network is basically no interpretability,this caused some difficulties to our understanding of the model.Based on this background,this paper takes the short-term power prediction method of fans as the research content.Through the means of neural network prediction,it studies and discusses the wind farm prediction method.The main work of this paper includes the following aspects:Firstly,the SCADA system data used in the short-term prediction of wind power is introduced,the influence factors of fan power were analyzed.The analysis results were combined with the parameter types recorded by SCADA system to determine the input parameters of the prediction model.Then the overall process of wind power prediction was established and the model evaluation indexes were studied.Secondly,Introduce the data preprocessing and analyze the process of data preprocessing required in this paper.Then,analyzing the characteristics of the data,the algorithm of processing wind speed data and wind power data outliers was studied according to the traditional data pretreatment method,and the data of SCADA system for wind farm in operation was used for verification.Finally,In order to facilitate the analysis and comparison of algorithm results,the prediction model based on BP neural network and the prediction model based on improved weight agnostic neural networkss are compared in this paper.BP neural network and weight agnostic neural networks are introduced,and the short-term fan power prediction algorithm based on BP neural network is analyzed experimentally with actual data.Then,the problems existing in the application of weight agnostic neural networks are analyzed,and an improved weight agnostic neural networks is proposed to solve the problems existing in the actual operation of weight agnostic neural networks,and then the short-term wind power prediction algorithm based on the improved weight agnostic neural networks is used for experimental analysis.The advantages and disadvantages of BP neural network and improved weight agnostic neural networks are compared to illustrate the feasibility and superiority of the method.