Smart Energy Conservation in Homes using Fully Residual Convolutional Networks for Solar Power Forecasting

DG is a power production strategy that enhances efficiency by reducing carbon demand peaks, emissions, and transmission losses through the deployment of multiple smaller on-site energy sources located within individual buildings. Although net metering aids in balancing energy supply and demand, scaling distributed generation across several houses to support intermittent renewable energy sources remains challenging. Furthermore, distributed generation is not economically attractive for widespread implementation given the existing energy price framework. To maximise Smart Energy Conservation and Solar Power Forecasting, meticulous data preparation is crucial for addressing these challenges. This paper reviews the methodologies and procedures employed in data preparation in prior studies. Our approach employs a FCN architecture, utilising ResNet101 as the backbone, with integrated upsampling skip links to enhance prediction accuracy. The proposed method achieves an accuracy of 96.56%, outperforming leading models and techniques in solar power forecasting. These findings underscore the importance of robust machine learning algorithms and meticulous data preparation for smart homes to optimise the use of renewable energy. Efficient distributed generation utilisation in residential applications facilitates smart energy conservation, hence enhancing energy sustainability and improving the accuracy of solar power forecasts.