Short-Term Residential Load Forecasting for Microgrid: Enhancing Performance through PCA and Clustering Techniques.

Short-term load forecasting (STLF) is indispensable in electrical energy systems, providing real-time insights into energy demand patterns and fluctuations. STLF is vital for the stability and efficient functioning of microgrid. It empowers microgrid operators to optimize energy generation, distribution, and grid management; enabling improved responses to changes in load, integration of renewables, and cost-effective energy generation. Residential load consumption-based STLF analyzes consumer behavior and power consumption patterns, enabling individuals to optimize energy consumption along with identifying opportunities for energy savings. Neural network models are effective in STLF as they can outperform conventional load forecasting models by managing complicated relationships between many input variables and adapting to changing patterns and trends. In this work, we present our recent development of an artificial neural network (ANN) model, and a long short-term memory (LSTM) model using K-means clustering as well as Principal Component Analysis (PCA). PCA reduces the dimensionality of the input data, allowing it to identify underlying patterns and relationships in the data. Different load profiles or consumption patterns among households are identified by K-means clustering, thereby aiding in dataset categorization. A dataset capturing the electricity usage within a single residence, at a frequency of one sample per minute is employed to assess the model’s performance with root mean squared error (RMSE), mean absolute percent error (MAPE), and mean square error (MSE). The LSTM model exhibits better accuracy in prediction as it is better at capturing temporal dependencies in the data, with an MSE value of 0.0097 and the value of RMSE is 0.374 (cluster-1).