Online identification of agent-based multi-model system and its application to the control valve circuit

An agent-based online multi-model system is proposed in this paper, and the identification method of the model is described in detail. In the agent-based system, each agent can execute the task alone or collaborate with other agents to accomplish a fixed goal. The agent is structured by some cluster centres that can be obtained through the fuzzy subtractive clustering method. The behaviour of the agent can be described by a dynamic equation whose parameters can be identified online by the weighted recursive least squares method. In addition, an initial setting method is used to adjust the local model parameters when new data come to the system. The implementation of the mentioned system to the control valve in cooling water circuit is studied, and the simulation results illustrate the proposed strategy can obtain satisfactory performance.