【摘要】 为了解决中长期负荷预测存在时间跨度大和广域分布广等难题,提出具有"厚近薄远"特性的指数平滑法。指数平滑法是电力负荷预测中一种重要的方法,该方法中平滑系数α为预测精确度的关键。在传统的指数平滑法预测电力负荷模型的基础上,考虑到模型各期负荷数据误差的区别,引入拟合误差权重系数,提出具有"厚近薄远"特性的指数平滑法预测模型,并通过等分法搜索出最优平滑系数α。通过统计学实验,对负荷模型添加随机扰动产生符合实际要求的多组负荷时间序列,统计在不同负荷模型下"厚近薄远"的规律,即拟合误差权重系数β的取值规律。最后将统计实验结果与仿真实例进行对比得到一致结果,从而证明指数平滑法"厚近薄远"规律的正确性。更多还原

【Abstract】 In order to solve the problems of medium-long-term load forecasting,such as large time span and wide distribution,etc.this research proposes an exponential smoothing method with "thickness near thin"characteristics.Exponential smoothing method is an important method in power load forecasting.In this method,the smoothing coefficientαis the key to the prediction accuracy.In this paper,based on the traditional exponential smoothing method for predicting the power load model,taking into account the difference of the model load data error,the fitting error weight coefficient is introduced and an exponential smoothing prediction model with"Valuing Near Errors and Belittling Far Errors"characteristics is proposed.The optimal smoothing coefficientαis searched by a bisection method.Based on statistical tests,random disturbances are added to the load model to generate multiple sets of load time series that meet actual requirements,and the law of"Valuing Near Errors and Belittling Far Errors"under different load models is calculated,that is,the law of the value of the fitting error weight coefficientβ.Finally,the statistical test results are compared with the simulation results to obtain a consistent result,which proves the correctness of the exponential smoothing method"Valuing Near Errors and Belittling Far Errors".更多还原