Short Term Load Forecasting Using Machine Learning
Abstract
Short term load forecasting is required for power system planning, operation and control. It is utilized by utilities, system operators, power marketers, generators etc. In this project, load forecasting has been done with Regression Analysis using ANN (Artificial Neural Network). The load profile is determined by factors including majorly weather, time of day, economic factors and random effects. Accordingly, forecasting is done using a neural network to determine the total hourly load for 3 towns in Panama. Hourly temperature, humidity, precipitation, wind speed, nature of holiday and the electrical load of the three towns has been taken from 2015 to 2020. The ANN model is trained and tested using hourly data randomly attained from the entire dataset. Performance of forecaster is calculated using MAPE, MSLE and correlation of predicted data with the actual data. The built ANN model is deployed via a Web App with an interface to be utilized by the utility or transmission company to determine the hourly load forecast and peak demand for that day. A CSV file with 24 hours pre-forecast weather hourly data is required is provided as the input, and the trained model with its weights will determine the 24-hour hourly load forecast as well as the peak load demand for that very day.