基于GCN-LSTM的日前市场边际电价预测

Day Ahead Market Marginal Price Forecasting Based on GCN-LSTM

传统电价预测往往采用基于时间序列的时域预测方法,未能充分利用电力市场的地域信息,忽略了跨区域输电条件下影响区内电价的域外因素,为进一步提升电价预测精度提出一种基于图卷积神经网络与长短时记忆网络(graph convolution network-long short term memory,GCN-LSTM)的时空预测算法。该算法首先通过建立图模型,描述地域分布的电力市场数据,并使用图卷积神经网络,提取所研究区域和周围地区传导到域内的域外信息;其次,将不同时刻图卷积神经网络提取到的信息构成时间序列,输入长短时循环网络,从而对日前市场边际电价进行预测。利用北欧电力交易所Nord Pool的运营数据进行算例分析,通过与对照算法对比,该算法具有更好的预测精准度和普适性。

The commonly used traditional time-domain forecasting method based on time series often, fails to make full use of the regional information of the power market and ignores the extraterritorial influence factors of regional electricity price under cross regional transmission conditions.In order to further improve the accuracy of electricity price forecasting, a spatiotemporal forecasting algorithm(graph convolution network-long short term memory, GCN-LSTM)based on graph convolution network-long short term memory network was proposed. Firstly, the algorithm described the regional distribution of electricity market data by building a graph model, and the graph convolution neural network was used to extract the electricity market data around the studied area to import the outside information transmitted to the region.Secondly, the information extracted from the graph convolution neural network at different times constituted a time series and was input into the long short term memory network, so as to forecast the day ahead market marginal price. The operation data of Nord Pool was used for example analysis. Compared with the control algorithm, the algorithm has better prediction accuracy and universal applicability.