基于AdaBoost的短期边际电价预测模型  

Short-term Marginal Price Forecasting Model Based on AdaBoost

作　　者：张俊玮 史文彬 欧家祥 丁超 ZHANG Junwei;SHI Wenbin;OU Jiaxiang;DING Chao(Guizhou Electric Power Research Institute,Guiyang 550002;Shanghai University of Electric Power,Shanghai 200090)

机构地区：[1]贵州电网有限责任公司电力科学研究院,贵阳550002 [2]上海电力学院,上海200090

出　　处：《计算机与数字工程》2020年第2期442-446,473,共6页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:61672337);基于深度学习的大客户负荷预测技术研究与应用(编号:066600KK52170002)资助。

摘　　要：在电力系统中,系统边际电价(SMP)反映了电力市场中电力商品短期供求关系,是电力市场的经济纽带,对电力的市场营销起重要作用。传统的单一预测算法模型对短期边际电价的预测存在误差大、泛化能力低等缺点,因此论文提出了基于AdaBoost的短期边际电价预测集成学习算法。论文分析了影响短期边际电价的主要因素,基于集成学习的AdaBoost算法对短期边际电价预测问题进行建模。通过系统边际电价测试数据验证模型,和SVM以及BP神经网络作对比,AdaBoost算法的准确率明显优于传统模型SVM和BP神经网络,在电力市场中具有实际的应用价值。In the power system,the system marginal price(SMP)reflects the short-term supply and demand of power commodities in the electricity market,and is the economic ties of the electricity market,playing an important role in the marketing of electricity. The traditional single prediction algorithm model has shortcomings such as large error and low generalization ability for short-term marginal price. Therefore,this paper proposes an integrated learning algorithm for short-term marginal price forecast based on AdaBoost algorithm. This paper analyzes the main factors affecting the short-term marginal price. The AdaBoost algorithm based on integrated learning models the short-term marginal price forecasting problem,and compares it with the SVM and BP neural network through the system marginal price test data validation model. The accuracy rate of the AdaBoost algorithm is obviously superior to the traditional model SVM and BP neural network,and has practical application value in the power market.

关 键 词：电力系统 电价预测 系统边际电价 ADABOOST 市场电价营销 

分 类 号：TM744[电气工程—电力系统及自动化] TP311[自动化与计算机技术—计算机软件与理论]