Learning predictive control for gas heat pump

A predictive control algorithm is effective for systems which have large time constants and dead times. The response models are necessary in order to predict the future responses. In the control of the gas heat pump which is an air-conditioning system, the model parameters of the transfer functions are frequently changed according to operating conditions. We propose a learning predictive control algorithm, which determines control input using the identified system models. These models are stored in a fuzzy map and reused to infer the system response in new operating conditions based on past experienced conditions