<![CDATA[The advancement of smart building technologies requires energy management systems that are both efficient and capable of adapting to dynamic operational conditions. A key component of such systems is reliable electrical load forecasting, as building energy demand is affected by environmental conditions, occupancy behavior, and operational activities that exhibit nonlinear and time-dependent characteristics. This study explores the use of the Long Short-Term Memory (LSTM) approach for forecasting smart building electricity consumption based on multivariate time-series data. The input dataset incorporates temporal features, ambient temperature, humidity levels, occupancy-related patterns, and major electrical load components within the building. The research workflow consists of data preprocessing, normalization, time-series construction using a sliding window strategy, LSTM model training, and evaluation of forecasting performance. The findings indicate that the building’s electricity demand varies approximately between 6 kW and 17 kW, with an average load ranging from 11 to 12 kW. Performance assessment yields an RMSE of about 3 kW and a MAPE of roughly 25%. The largely symmetric error distribution around zero suggests minimal systematic bias in the predictions, although the model’s accuracy during peak demand periods remains limited.]]>
Copyrights © 2026
