摘要: 在电力市场中,价格一直受到买卖双方的广泛关注。但是,电价影响因素的不确定性给电价的预测带来难度。针对该问题,提出一种通过结合人工神经网络和KNN算法来进行时间序列预测的模型,用KNN算法找出历史数据中相似的数据子序列集合(最近邻),并用人工神经网络来寻找这些最近邻的最优权重,得出预测的时间序列。以美国纽约州电力市场的电价数据进行实验分析,同时比较了利用ARIMA算法以及Naive I预测的结果,证明该方法简单、有效。
关键词:
电价预测,
人工神经网络,
KNN算法,
权重
Abstract: In a deregulated electricity market, electricity price is one of the main concerns for both suppliers and consumers. However, uncertain factors on electricity prices make it difficult to forecast them. This paper proposes a time series forecasting model which integrates K-Nearest Neighbor(KNN) and Artificial Neural Network(ANN) algorithms. The KNN algorithm is used to search similar time subseries (neighbors) in historical data and the ANN algorithm is used to search the optimal weights on these neighbors. The model is tested with the electricity price in New York electricity market. Compared with traditional ARIMA and Naive I model, the model can get more accurate forecasting results with smaller volatility.
Key words:
electricity price forecasting,
Artificial Neural Network(ANN),
K-Nearest Neighbor(KNN) algorithm,
weight
中图分类号:
周芳. 基于KNN-ANN算法的边际电价预测[J]. 计算机工程, 2010, 36(11): 188-189,194.
ZHOU Fang. Marginal Electricity Price Forecasting Based on KNN-ANN Algorithm[J]. Computer Engineering, 2010, 36(11): 188-189,194.