Machine Learning-Based Dispatch Optimization of a Molten-Salt Concentrating Solar Power Plant Using Day-Ahead Electricity Price Forecasting

This dataset contains the electricity-market, gas-market, power-system and economic time series used to develop and benchmark the end-to-end dispatch-optimization algorithm presented in the associated article, which couples a deep neural network for day-ahead electricity price forecasting with a genetic algorithm for revenue-maximizing thermal energy storage (TES) dispatch in a molten-salt parabolic-trough concentrating solar power plant operating in the Spanish day-ahead market. All series cover 1 January 2016 – 31 December 2021 and comprise: hourly day-ahead marginal prices published by OMIE (marginalpdbc series, one file per calendar day, provided both in the original non-standard format and as converted CSV); the daily natural-gas day price and traded volume from MIBGAS; the daily generation structure by technology from Red Eléctrica de España (ESIOS); daily Brent crude-oil prices; the IBEX 35 and IBEX 35 Energy indices; and meteorological variables retrieved from the AEMET OpenData API. These variables were screened as candidate predictors through a Pearson-correlation analysis against the day-ahead price, which retained the natural-gas day price and a categorical day-of-week indicator as the final predictor set. The dataset also includes the consolidated modelling file actually supplied to the forecasting model, the preprocessing and dataset-composition scripts used to convert, clean and merge the raw files, and the benchmark results of the genetic-algorithm dispatch module across the 32 test scenarios (season × gas-price regime × day type × irradiance pattern). The direct normal irradiance profiles and engineering parameters of the reference commercial plant are not included, as they were provided under a confidentiality agreement.本数据集包含用于开发与基准测试相关论文中提出的端到端调度优化算法的电力市场、天然气市场、电力系统及经济时间序列数据。该算法将用于日前电价预测的深度神经网络，与用于在西班牙日前市场运营的熔盐槽式聚光太阳能发电厂中实现收益最大化的热能存储（TES, Thermal Energy Storage）调度遗传算法进行耦合。 所有序列的时间覆盖周期为2016年1月1日至2021年12月31日，涵盖以下数据：由OMIE发布的小时级日前边际电价（marginalpdbc序列，按自然日拆分为单个文件，同时提供原始非标准格式与转换后的CSV格式文件）；MIBGAS发布的日度天然气现货价格与交易量；西班牙电网公司Red Eléctrica de España（ESIOS）发布的按技术分类的日度发电结构数据；日度布伦特原油价格；IBEX 35指数与IBEX 35能源指数；以及从西班牙国家气象局（AEMET）开放数据API获取的气象变量。研究人员以日前电价为基准，通过皮尔逊相关性分析对上述候选预测变量进行筛选，最终保留天然气现货价格与分类式星期几指示器作为最终预测变量集。 本数据集还包含实际供给至预测模型的整合建模文件、用于转换、清洗与合并原始文件的预处理及数据集构建脚本，以及32种测试场景（季节×天然气价格态势×日期类型×辐照度模式）下遗传算法调度模块的基准测试结果。 参考商业电站的直接法向辐照度曲线与工程参数未纳入本数据集，因其需在保密协议约束下提供。