Load demand forecasting in air conditioning a rotor Hopfield neural network approach optimized by a new optimization algorithm.

Load demand forecasting is crucial for optimal energy management and sustaining comfortable indoor environments for air conditioning systems. The current research provides load demand prediction by a new modified rotor Hopfield neural network (RHNN) integrated with a fractional order of seasons optimization algorithm (FO-SOA) to overcome the challenge of predicting load demand. The RHNN extracts historical data patterning and predicts load demand prediction for future time using past data, and the FO-SOA includes infinitesimal calculus in its process to optimize its solution by considering repeating operation of honeybee agent and also extracting long-term memory operation without requiring additional memory access in the process to make it best at exploration/exploitation among optimization process. The model includes an incorporation model of key factors including ambient temperature, humidity, occupancy pattern, etc., for enhancing the reliability and the prediction accuracy. A case study validated the proposed RHNN/FO-SOA model and allowed for a comparison with several state-of-the-art methods, such as LSTM-based hybrid ensemble learning (LSTM/HEL), LSTM/RNN, deep neural networks (DNN), and deep learning models (DLM). The results showcase optimal performance, yielding an R value of 0.95, along with the lowest MSE, RMSE, and MAE values when compared to the other tested models. A correction coefficient increased the goodness of fit from 0.77 to 0.85. The RHNN/FO-SOA method may contribute to improve energy performance and reduce costs in air conditioners, shown by the findings.