Abstract:
Short-Term Load Forecasting (STLF) plays a crucial role in power system planning and
operation, as it helps utilities to efficiently allocate their resources and ensure reliable service
to customer. In this project the performance of different forecasting algorithms such as Long
Short-Term Memory (LSTM), Particle Swarm Optimization-Gated Recurrent Unit (PSO GRU), Multivariate LSTM, and 1-Dimensional Convolution Neural Network-Long Short Term Memory (1-D CNN LSTM) are evaluated. A widely used benchmark dataset namely
Global Energy Forecasting Competition (GEFCOM) dataset is used in this work for training
and performance validation. The performance of different load forecasting models is compared
using performance indices like accuracy and Mean Absolute Percentage Error (MAPE).
Among the different models used for short-term load forecasting, Multivariate LSTM model is
found to be more accurate than other models. The results indicate that Multivariate LSTM is a
promising approach for STLF, and its superior performance is attributed to its ability to handle
multiple input variables. The study highlights the importance of model selection in accurate
load forecasting and demonstrates the potential of Multivariate LSTM for STLF. The findings
can help power system planners and operators to choose an appropriate STLF algorithm based
on their specific needs and requirements
