日前交易边际电价的预测方法  

Forecasting Methods of Market Clearing Price in Day-ahead Electricity Market

作　　者：杨波[1] 赵遵廉[2] 陈允平[1] 韩启业[3] 

机构地区：[1]武汉大学电气工程学院,武汉430072 [2]国家电网公司,北京100031 [3]华中电网有限公司,武汉430077

出　　处：《高电压技术》2007年第7期144-150,共7页High Voltage Engineering

基　　金：华中电网有限公司科技基金(KJ2006-0604-21)~~

摘　　要：在电力市场日前交易中,边际电价对独立发电商、输配电服务提供者、电力零售商和电力客户等市场成员的经济利益影响重大。针对电力市场研究的热点问题即边际电价预测问题,分析了边际电价的微观经济机理,指出供求均衡规律是影响边际电价的主要因素,以及边际电价具有周期性、波动性和分时均值回复等特征;对边际电价预测方法从ARIMA模型、GARCH模型、动态回归模型、传递函数模型、灰色系统模型、混沌相空间重构、人工神经网络、委员会机器、支持向量机、市场模拟等方面进行了评述;提出了选择边际电价预测方法的建议,并指出由于市场力、博弈、串谋、容量持留等因素会影响预测精度,因此组合预测模型是提高预测精度的一种可行方法。In day-ahead electricity market, market clearing price （MCP） has a great effect on economic benefit of market participants including independent power producer, transmission and distribution service provider, retailer, and customer. Based on the analysis of economic principle of MCP, it is concluded that demand-supply equilibrium rule, which is the main reason affecting MCP and MCP, has characteristics of periodicity, fluctuation, and mean-reversion. Next, MCP forecasting methods are reviewed from different aspects such as ARIMA, GARCH, dynamic regression model, transfer function model, grey system, chaos, artificial neural network, committee machine, support vector machine, and market simulation. Finally, some suggestions are given in the selection of MCP forecasting method. Due to influence of market power, game strategy, collusion, and capacity withholding, combination of different forecasting models is regarded as feasible way to improve MCP forecasting accuracy.

关 键 词：边际电价 ARIMA GARCH 灰色系统 混沌 人工神经网络 委员会机器 支持向量机 

分 类 号：TM731[电气工程—电力系统及自动化] F407.61[经济管理—产业经济]