Studies on Highly accurate Short-term Electric Power Load Forecasting with Fuzzy Data Mining

模糊数据挖掘高精度短期电力负荷预测研究

基本信息

批准号：
13650319
负责人：
MORI Hiroyuki
金额：
$ 1.73万
依托单位：
Meiji University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
2001
资助国家：
日本
起止时间：
2001 至 2003
项目状态：
已结题

项目摘要

In this project, a new hybrid intelligent system has been proposed for short-term load forecasting in power systems. The proposed method is based on the regression tree of data mining, Simplified Fuzzy Inference and Tabu Search of Meta-heuristics. The regression tree works to extract rules from data base though the decision tree so that if-then rules are obtained. Simplified Fuzzy Inference (SFI) is a good nonlinear approximation technique for nonlinear systems that is equivalent to the multi-layered perceptron (MLP) of artificial neural network. The use of Tabu Search allows SFI to construct the globally optimal rules in terms of the number of fuzzy membership functions and their location. As a result, the proposed model is superior to MLP in terms of the prediction error. At the same time, SFI is applied to the regression tree to improve the boundary conditions of the splitting conditions. The fuzzy rules contributed to the classification of data on load forecasting.Also, the use of TS is easier to determine the forecasting model from a standpoint of minimizing the maximum errors of load forecasting model through the learning process due to the advantage without any constraints. Therefore, power system operators have flexibility to give priority to the maximum or the average squared errors.In addition, the developed model contributed to the reduction of the reserves of generation so that it plays an important role as the decision making system of selling and buying the electricity and make power system operation and control more effective.

在该项目中，提出了一种新的混合智能系统，用于电力系统的短期负荷预测。该方法基于数据挖掘的回归树、简化模糊推理和元启发式禁忌搜索。回归树的作用是通过决策树从数据库中提取规则，从而获得if-then规则。简化模糊推理（SFI）是一种很好的非线性系统非线性逼近技术，相当于人工神经网络的多层感知器（MLP）。使用禁忌搜索允许 SFI 根据模糊隶属函数的数量及其位置构建全局最优规则。因此，所提出的模型在预测误差方面优于 MLP。同时，将SFI应用于回归树，改善分裂条件的边界条件。模糊规则有助于对负荷预测数据进行分类。而且，从通过学习过程最小化负荷预测模型最大误差的角度来看，使用TS更容易确定预测模型，因为它具有不受任何约束的优点。因此，电力系统运营商可以灵活地优先考虑最大或平均平方误差。此外，所开发的模型有助于减少发电储备，从而在售电决策系统中发挥重要作用。使电力系统运行和控制更加有效。