Data for: Optimal Planning and Forecasting of Active Distribution Networks Using a Multi-stage Deep Learning Based Technique
Description
This data is associated with the paper titled: Optimal Planning and Forecasting of Active Distribution Networks Using a Multi-stage Deep Learning Based Technique. The data includes results from the the models and key input data. This research introduces a comprehensive methodology that integrates long-term forecasting and optimal sizing and siting techniques to streamline the planning of distribution networks. The primary goal is to jointly forecast solar irradiation, wind speed, and electric load demand using advanced AI-based models. The methodology consists of two main stages: forecasting and optimal sizing and siting. In the forecasting stage, load demand, PV irradiance, and wind speed are forecasted, and these outputs serve as initial inputs for the subsequent optimal sizing and siting stage. The latter stage formulates an optimization problem to determine the best size and location for PVs, WTs, and BESSs based on the forecasted data. The forecasting process involves five key steps: data type selection, load bus selection (for load forecasting only), data reading, data preprocessing, and calling, training, testing, and evaluating the forecasting model. The Load dataset covers hourly data from 3-11/2017 to 31/05/2020, comprising 22,552 data points sourced from reference [1]. Further details regarding data features and data split can be found in subsequent subsections. The dataset has been meticulously pre-processed and prepared for one-year long-term forecasting. Additionally, a three-year historical meteorological dataset from a specific UK location spans from 1/1/2017 to 31/05/2020. Sourced from [2], this dataset contains comprehensive meteorological parameters, including solar radiation, temperature, humidity, pressure, wind speed, and wind direction, which are utilized for forecasting solar irradiance and wind speed. The network demand was forecasted for all 12 load buses across all seasons in the 16-bus network. Likewise, solar irradiance and wind speed data were forecasted for all seasons. [1] E. Borghini, C. Giannetti, J. Flynn, and G. Todeschini, “Data-Driven Energy Storage Scheduling to Minimise Peak Demand on Distribution Systems with PV Generation,” Energies, vol. 14, no. 12, p. 3453, 2021. [2] Nasa, “The Power project,” The Power project, 2023. https://power.larc.nasa.gov/.
Files
Steps to reproduce
This work was supported by Innovate UK GCRF Energy Catalyst Pi-CREST project under Grant number 41358