A Short-Term Commercial Buildings Load Forecasting Framework using the Firefly Algorithm and Artificial Neural Network

This paper presents short-term commercial building load forecasting of daily peak loads. The operation of the building can be managed properly for the next hour load consumption daily forecasting data set instantly to collect the load usage in buildings. In this paper, we proposed a load-shedding process, and load forecasting framework hourly base data set using an artificial neural network (ANN) model. The Levenberg Marquardt Learning algorithm is a multilayer feed-forward neural network and also has good properties with robustness in prediction. The historical daily peak load data and corresponding daily peak temperature data are used as input to the network. The input data is to be trained by using the load requirement based on prediction. The proposed ANN technique is illustrated using an actual dataset from the Vels Institute of Science, Technology, and Advanced Studies (VISTAS) campus for short-term load forecasting challenges. The experimental data set was analyzed by using MATLAB 2021a. The artificial neural network, curve fitting method, and the firefly algorithm were also used to analyze the data collection of the VISTAS campus based on results evaluated daily peak load curves. The results are compared with proposed and conventional methods, which give more accurate predictions with the optimal number of neurons in the hidden layers of 0.0472 and forecasted error values are -0.0115 which gives better accuracy.