Combined heat and power(CHP) units are affected by the output fluctuation of large-scale operating conditions

resulting in poor overall control quality of power generation load-extraction steam flow-throttle pressure in the turbine-boiler coordinated control system. To solve this problem

a multi-condition adaptive control method based on multi-agent deep deterministic policy gradient(MA-DDPG) is proposed. Firstly

according to the nonlinear dynamic mechanism of the unit

multiple operating condition sub-models considering the changes of state parameters are established

and the optimal operating condition sub-model is obtained by the integral function switching mechanism. For the multi-loop complex control requirements of the coordinated control system

a multiagent synchronous operation mechanism is proposed

and the reward function is designed with the coordinated control objectives of rapid response

stable heating and safe operation. Finally

by training the agent to interact with the environment

the multi-loop gains are continuously adjusted online to achieve multi-condition adaptive control.Simulation results show that

compared with the conventional control method

the proposed method can effectively improve the load response rate under wide range operating conditions

and ensure the stability of heating.