【作者】 李媛媛；

【导师】 牛东晓；

【作者基本信息】 华北电力大学（河北） ， 管理科学与工程， 2005， 硕士

【摘要】 中长期负荷预测对电力系统规划和运行是非常重要的,本文的目的是探讨如何准确的进行中长期负荷预测。　本文首次提出以联合数据挖掘技术对历史数据进行预处理,提取出规律强化的数据序列,并在此基础上建立电力负荷优选组合预测模型。　　月度负荷具有增长和波动二重趋势,本文提出以纵向历史数据为原始序列,用GM (1,1)进行增长趋势预测,以横向历史数据为原始序列,用ANN 进行波动趋势预测,并引入最优可信度的概念,建立综合最优预测模型。　将多层递阶回归分析方法引入了中长期负荷预测领域,该方法较好的考虑了相关因素影响,还对电力系统自身的时变特性具有较强的适用性。　实例验证了上述方法的先进性和适用性。　最后,针对辽宁电网进行了中长期负荷预测软件的开发。更多还原

【Abstract】 Mid-long term power load forecasting is fundamental to power system planning and operation. The purpose is to discuss how we can forecast electrical demand accurately. The united data mining technology for historical data preprocessing is proposed in the thesis. The data sequence which can boost up rules is composed through the method. The optimum combined model is built based on that. For the monthly load with double trends of increasing and fluctuating, the integrated optimum gray neural network model of monthly load forecasting is proposed in the thesis for the first time. In the model, we regard vertical historical data as the primitive array to forecast increasing trend by the gray model, and regard horizontal historical data as the primitive array to forecast fluctuating trend by the ANN . Based on that, the concept of the optimum credibility is introduced, and the integrated optimum model is build in the thesis. In this thesis, multi-degree recursive regression analysis is applied to forecast the mid-long term load for the first time. The method can perfectly reflect the import function of the related factors as well as have strong adaptability to the power system with sequential variance. The advantage and applicability of the three methods mentioned in the front have been verified by instances. Lastly, mid-long term load forecasting software is developed for LiaoNing power grid in this thesis.更多还原