基于功率谱分解和实时气象因素的短期负荷预测

提出了基于功率谱分解和实时气象因素的短期负荷预测方法，采用快速傅里叶变换(fast Fourier transform，FFT)对负荷序列进行变换得到功率谱，依据变换结果分析功率谱得出负荷基频、低频和高频分量的频率范围，采用有限脉冲响应(finite impulse response，FIR)滤波器从负荷中分离出各 个负荷分量。分析各个负荷分量的特点，针对各个负荷分量分别设计预测模型，对基频分量采用Elman回归神经网络进行预测，这部分较好地反映出基频分量的时间序列特性；对低频和高频分量分别采用自适应线性回归神经网络进行预测，在对这部分分量的预测中重点引入实时气象因素，以利用最新的气象信息提高预测精度。通过在某地区的实际应用证明了所提出方法的有效性。

Short-Term Load Forecasting Based on Power Spectrum Decomposition and Hourly Weather Factors

A short-term load forecasting method based on power spectrum decomposition and hourly weather factors is proposed,in which the fast Fourier transform is applied to power load series to obtain power spectrum and by means of analyzing the obtained power spectrum the frequency ranges of fundamental frequency component,low frequency component and high frequency component of power load are obtained,then by use of finite impulse response(FIR) filters each several load components are separated from the load.After analyzing the feature of each several load components,the forecasting model for each several load components is designed respectively,fundamental frequency component of the load is forecasted by Elman regressive neural network and the time series characteristic of fundamental frequency component can be reflected well;the low frequency components and high frequency components are forecasted by adaptive linear regressive neural network,in the forecasting of these components the hourly weather factors are introduced to improve the accuracy of forecasting by the latest weather information.The effectiveness of the proposed method is validated by the results of practical application of this method in a certain regional power system.