能源供给调度优化方法数据集

本数据集基于某钢铁企业电力消耗数据与自主生成的分布式调度算例，采用深度学习与强化学习方法构建，旨在为能源供给调度优化研究提供基准数据。数据集核心内容包括：（1）基于互信息法进行特征提取，并利用CNN-BiLSTM-Attention混合神经网络模型进行短期电力负荷预测的完整实验数据，涵盖模型代码、参数、训练过程及性能对比结果；（2）针对分布式能源场景，通过遵循泊松分布的随机工件到达模式生成不同规模的测试算例，并采用田口实验优化参数的强化学习算法进行实时调度优化，输出包括反世代距离与纯度在内的多目标性能指标。数据产生于2025年3月，实验在受控的实验室环境下完成，计算在配置NVIDIA RTX 3080 GPU的工作站上依托TensorFlow框架执行。数据集包含1个文件夹，名称为能源供给调度优化方法数据集，容量约为6.6MB，内含1个预测模型说明文档、1个调度规则库代码文件、3个数据表格、8个图片文档以及2篇论文和1项专利，数据内容包括模型参数、神经网络结构、测试数据集以及实验结果。并包含2个子文件夹分别命名为Instances和在线调度代码，其中包含测试算例数据和在线调度算法源代码。

This dataset is constructed based on the power consumption data of a steel plant and independently generated distributed scheduling test cases, using deep learning and reinforcement learning methods, aiming to provide benchmark data for energy supply scheduling optimization research.
The core contents of the dataset are as follows:
(1) Complete experimental data for short-term electric load forecasting: feature extraction is performed via the mutual information method, and a CNN-BiLSTM-Attention hybrid neural network model is utilized for prediction. The dataset covers model codes, hyperparameters, training processes and performance comparison results;
(2) For distributed energy scenarios, test cases of different scales are generated with random job arrival patterns following the Poisson distribution. A reinforcement learning algorithm optimized via Taguchi experiments is adopted for real-time scheduling optimization, and the outputs include multi-objective performance metrics such as inverted generational distance and purity.
The dataset was generated in March 2025. All experiments were conducted in a controlled laboratory environment, and all computations were executed on a workstation equipped with an NVIDIA RTX 3080 GPU using the TensorFlow framework.
The dataset consists of one main folder named *Energy Supply Scheduling Optimization Method Dataset*, with a total size of approximately 6.6 MB. The folder contains 1 prediction model specification document, 1 scheduling rule base code file, 3 data tables, 8 image documents, 2 academic papers and 1 patent. The dataset contents include model parameters, neural network architectures, test datasets and experimental results. Additionally, there are two subfolders named *Instances* and *Online Scheduling Code*, which store test case data and the source codes of the online scheduling algorithm respectively.

搜集汇总

背景与挑战

背景概述

该数据集包含基于钢铁企业电力消耗数据的短期电力负荷预测实验数据和分布式能源场景的调度优化算例，采用深度学习与强化学习方法构建，提供模型代码、参数、训练过程及性能对比结果，适用于能源供给调度优化研究。