基于神经网络的新能源功率预测

New Energy Power Prediction Based on Neural Networks

随着新能源发电向电网的渗透,电力系统运行中的模糊性增加,因此,为使电力系统网络稳定运行,准确有效地预测新能源发电量至关重要。本文采用多层前馈人工神经网络(FF-ANN)模型对新能源发电预测数据集进行训练,研究内容涉及两个步骤,即训练和预测。在训练过程中,为了优化FF-ANN的参数,使用了长短记忆学习算法。为了预测新能源功率,提出了一种新的预测技术,该算法被称为加权最小二乘误差相关法(WLSEC),该方法已在C++平台上实现。该模型的性能已经在实际运行中进行了测试,考虑了每小时分辨率的一年的历史数据。本文预测的新能源小时平均绝对百分比误差(MAPE)为7.32%,而反向传播神经网络(BPNN)为9%,这清楚地表明了本文所提出的预测新能源发电模型的有效性。

With the infiltration of new energy into the power grid,the ambiguity in the operation of the power system increases.Therefore,in order to ensure the stable operation of the power system network,accurate and effective prediction of the new energy generation is crucial.This paper uses a multi-layer feedforward artificial neural network(FF-ANN)model to train the new energy generation prediction dataset.The research involves two steps,namely training and prediction.During the training process,a long and short memory learning algorithm is used to optimize the parameters of the FF-ANN.In order to predict the power of the new energy sources,a new prediction technique is proposed,which is called the weighted least squares error correlation(WLSEC)and has been implemented on the C++platform.The performance of this model has been tested in actual operation by taking into account one year of historical data with hourly resolution.The hourly average absolute percentage error(MAPE)of predicting the new energy is 7.32%,while the backpropagation neural network(BPNN)is 9%,this clearly demonstrates the effectiveness of the proposed model in predicting the new energy generation.