基于电力交易平台大数据的多区域电价预测研究

Multi region electricity price prediction based on big data of electric power trading platform

随着电力市场的发展,电力交易的实时数据通常来自于电力交易平台,但随着电力交易量的增加,电力交易平台用电成本逐渐增加。为此,基于多区域电力大数据,采用机器学习的方法,研究了电力价格建模和预测方法,通过分别采用线性回归、决策树和M5P三种机器学习算法,对同一组数据集进行训练、预测和比对,得出最佳学习算法,最后利用预测数据与实际数据对比计算误差,并讨论了电价预测准确性及其影响的主要因素。研究结果表明M5P算法具有最为准确的预测性能,建立的预测模型能够准确预测最低电价以及电价变化趋势,但是对短期电价波动的预测存在一定误差,误差主要来源于电价自身具有的时间序列性质。

With the development of power market,the real-time data of power transaction usually comes from the support of electric power trading platform,but with the increase of power transaction volume,the power cost of power data center increases gradually.Therefore,based on multi regional big data of electric power trading platform,this paper studies the modeling and forecasting methods of electric power price by using machine learning method.By using three machine learning algorithms of linear regression,decision tree and M5P respectively,the same data set is trained,predicted and compared,and the best learning algorithm is obtained.Finally,the error is calculated by comparing the predicted data with the actual data,the main factors affecting the accuracy of electricity price prediction are discussed.The research results show that M5P algorithm has the most accurate prediction performance,and the established prediction model can accurately predict the minimum price and the price change trend,but there are some errors in the short-term price fluctuation prediction.