Data for : An Enhanced CNN-LSTM Based Multi-Stage Framework for PV and Load Short-Term Forecasting: DSO Scenarios

The provided data is linked to the paper titled "An Enhanced CNN-LSTM Based Multi-Stage Framework for PV and Load Short-Term Forecasting: DSO Scenarios." In this research, a novel multi-stage framework for PV and load Short-Term Forecasting (STF) is introduced, incorporating feature generation, feature selection, and optimal hyperparameter tuning preprocessing techniques. The final stage of the proposed framework presents an enhanced hybrid CNN-LSTM deep learning model architecture.

The effectiveness of this framework is evaluated and compared against other state-of-the-art approaches across various DSO scenarios, encompassing multiple single-phase residential loads, three-phase feeders, and secondary substations. Remarkably, the proposed framework exhibits significant reductions in forecasting errors.

The provided time series data serves the purpose of testing the proposed short-term forecasting methodology. It features a 5-minute resolution for one month (July 2021, a summer month) and consists of 8,920 data points for each data profile.

This data can be categorized into two main categories: load data and PV data. The load data includes three sub-categories representing different DSO scenarios: Multiple residential loads (with 100 and 400 residential loads in separate datasets), three-phase feeder load demands, and three-phase substation load demands. On the other hand, the PV data folder contains PV data for one month (July), along with corresponding weather data.

本数据集关联于题为《基于改进型CNN-LSTM的光伏与负荷短期预测多阶段框架：配网系统运营商场景》的论文。本研究提出了一种全新的光伏（Photovoltaic, PV）与负荷短期预测（Short-Term Forecasting, STF）多阶段框架，整合了特征生成、特征选择与最优超参数调优三类预处理技术。所提框架的最终阶段采用了改进型混合CNN-LSTM深度学习模型架构。

该框架的有效性通过各类配网系统运营商（Distribution System Operator, DSO）场景下的前沿基准方法进行评估与对比，涵盖多组单相居民负荷、三相馈线与二次变电站场景。值得注意的是，所提框架的预测误差实现了显著降低。

本公开的时序数据用于验证所提出的短期预测方法。该数据集的时间分辨率为5分钟，覆盖时长为1个月（2021年7月，夏季月份），每条数据序列包含8920个数据点。

该数据可分为两大类别：负荷数据与光伏数据。负荷数据包含三类对应不同DSO场景的子类别：多居民负荷（分为分别包含100户与400户居民负荷的独立数据集）、三相馈线负荷需求以及三相变电站负荷需求。另一方面，光伏数据文件夹包含1个月（7月）的光伏数据以及对应的气象数据。