基于LSTM模型的甘肃电力市场现货价格预测与最佳交易决策研究  

Research on Spot Price Forecasting and Best Trading Decision in Gansu Province Electricity Market Based on LSTM Modeling

作　　者：杨春祥 马寅 张晓斌 韩杰 张妙静 郝如海 

机构地区：[1]国网甘肃省电力公司,甘肃 兰州

出　　处：《数据挖掘》2025年第1期26-39,共14页Hans Journal of Data Mining

基　　金：本文由国网甘肃省电力公司科技项目“适应高比例新能源的省级双边现货市场监测关键技术研究及应用”(522722230062)资助。

摘　　要：甘肃电力市场由于新能源接入比例的提升增加了市场的不确定性和波动性。本文首先从数据预处理与特征工程入手,对2024年上半年的现货价格数据进行平滑处理,捕捉短期和长期的价格趋势。随后建立了高精度的长短期记忆网络模型(Long Short-term Memory, LSTM)对甘肃电力市场未来7天和15天的平均实时和日前现货价格进行了预测,结果表明LSTM模型在训练集和测试集上的预测准确度较高。其次,基于预测结果和风险溢价的对数收益分析,为市场交易建立了最佳交易决策模型;以案例分析的方式展示了如何利用预测价格与实际价格的差异来评估市场风险和制定相应的交易策略。研究表明,在高比例新能源电网背景下,利用先进的机器学习技术可以对电力现货市场价格进行准确预测,本文的最佳交易决策模型能够提升市场交易决策的科学性及与监测预警支持能力。The increase in the proportion of new energy access in the Gansu electricity market has increased market uncertainty and volatility. Starting with data preprocessing and feature engineering, this paper smoothed the spot price data for the first half of 2024 to capture short-term and long-term price trends. Subsequently, a high-precision Long Short-Term Memory (LSTM) network model was established to predict the average real-time and intraday spot prices of the Gansu electricity market for the next 7 and 15 days. The results showed that the LSTM model had high prediction accuracy on both the training and testing sets. Secondly, based on the prediction results and logarithmic return analysis of risk premium, an optimal trading decision model was established for market trading;This case study demonstrates how to use the difference between predicted prices and actual prices to assess market risks and develop corresponding trading strategies. Research has shown that in the context of a high proportion of new energy grids, advanced machine learning t

关 键 词：LSTM模型 电力市场 现货价格预测 最佳交易决策 风险溢价 

分 类 号：F42[经济管理—产业经济]