Prediction of Air-Conditioning Cooling Load Using PSO-ESN

Accurately predicting the air-conditioning cooling load (ACCL) is crucial for energy-saving control of the air-conditioning system and plays a key role in enhancing the green economy of public buildings. To address the persistent challenges of the mismatched ACCL and terminal cooling demand, this study proposes a CL prediction model based on the echo state network (ESN) improved by particle swarm optimization (PSO), denoted as PSO-ESN. To overcome the issues of complex structure and low efficiency in traditional neural network models, the ESN model is employed to establish a prediction model. Since the prediction accuracy of ESN is significantly influenced by the reservoir parameters, PSO is proposed to optimize these parameters. The results show that compared to the other models, PSO-ESN reduces the mean absolute error (MAE) by 63.81%, 59.78%, and 53.75%, and the root mean square error (RMSE) by 72.54%, 71.25%, and 62.54%, respectively. Therefore, PSO-ESN has great validity and superiority for the prediction of ACCL.