Abstract: To improve the prediction accuracy of energy consumption of air conditioning system in subway station, it is an effective method to establish energy consumption forecasting model by Least Squares Support Vector Machines (LS-SVM). However, LS-SVM is difficult to determine the optimal model parameter value in dealing with regression problem for large datasets, which affects the fitting precision and generalization ability of the model to a large extent. An Improved Seeker Optimization Algorithm (ISOA) is proposed to optimize the model parameters in the LS-SVM modeling process by introducing an improved algorithm from search step and search direction. The energy consumption forecasting model based on ISOA-LS-SVM is applied to the training platform of subway in a school of Beijing. The results show that the model can accurately predict the energy consumption of the system. Compared with the grid search method, the particle swarm algorithm and the traditional population search algorithm, the LS-SVM is improved in speed and precision.

Key words: air conditioning system, energy consumption prediction model, least squares support vector machines, seeker optimization algorithm