中长期负荷预测型及其稳健性评价.pdfVIP

中长期电力负荷预测模型及其稳健性评价 摘 要 本文针对最小二乘估计的缺陷 提出了三个改进方法减弱异常数 据对参数估计的影响 首先 用加权最小二乘法进行参数估计 权系 数函数是预测残差的反函数 其次 通过对回归系数的模糊化建构模 糊线性回归模型预测负荷的变化范围 减弱模型对原始数据的依赖程 度 最后 采用稳健估计减小目标函数各项的乘方次数估算模型参数 实例证明加权最小二乘法 模糊化回归系数和稳健估计三个改进 方法正确而有效 具有重要的理论意义和工程实用价值 特别是在历 史数据资料不够完全 准确的情况下有相当的指导价值 本文还结合线性相关性检验和函数试探法提出了非线性相关性 检验和模糊相关性检验 为回归模型中相关因素选择提供科学参考 实例验证线性 非线性和模糊相关性检验的方法简单而有效 在考察 电力负荷与GDP年最高温度等因素的相关关系方面具有重要的实用 价值 关键词 中长期负荷预测 最小二乘估计 稳健估计 加权最小二乘 估计 模糊 统计检验 MID-LONG TERM ELECTRIC LOAD FORECASTING (1.1)MODEL AND ITS ROBUSTNESS EVALUATION ABSTRACT This paper presents three new methods to weaken outliers’ impact on parameters estimation, namely weighed least square estimation, fuzzy linear regression and robust estimation. Weighed least square estimation tune each item’s coefficient of LSE’s object function. The weight function is inverse against forecasting errors. Fuzzy linear regression model is built up by fuzzying the parameters; therefore, weaken model’s dependence on the raw data. Robust estimation is applied to minish the power of LSE’s object function. All of these are aimed to improve the forecasting model’s robustness against outliers. Case study results reveal the validity of these three new methods. Especially when the uncertainty and fuzzy characteristic are found in the raw data, those three methods have much more advantages. Besides, this paper put forward nonlinear and fuzzy relativity test, combined linear relativity test with function trial method. Relativity test provides scientific reference for choosing relative factors. The practical examples validate the linear, nonlinear and fuzzy relativity tests. Particularly in the i