Short-term customer-centric electric load forecasting for low carbon microgrids using a hybrid model

The shift towards sustainable energy management, with a focus on demand-side flexibility which refers to the strategic adjustment of consumer power usage to match electricity supply variability, requires precise load forecasting that captures consumer behavior and consumption patterns to harmonize electricity supply and minimize costs. Traditional forecasting methodologies often fall short in reflecting the dynamic nature of customer behaviors, which are subject to regional, temporal, and individual variations. To address this gap, our study presents a groundbreaking electric load forecasting model that integrates data decomposition, advanced deep learning, and customer clustering techniques. Employing ensemble empirical mode decomposition, the model disaggregates load data into intrinsic mode functions, subsequently forecasting future consumption with a convolutional long-short-term memory network. The optimization of the model’s hyperparameters is achieved through the grey wolf optimization algorithm, coupled with K-Means clustering to consider behavioral diversity among customers. When evaluated on two comprehensive datasets, including 200 and 300 households, our method demonstrates a substantial improvement in forecasting accuracy, outstripping four analogous models by 20–80%. This enhancement has practical ramifications: it enables more precise grid management, facilitates the incorporation of renewable energy by aligning supply more closely with actual consumption, and empowers utilities to craft consumer-centric services such as dynamic pricing schemes and energy-saving initiatives. Ultimately, these advancements underpin our commitment to bolstering the reliability and sustainability of energy supply systems while accommodating the modern electricity landscape’s ever-evolving demands.