Application of multi-agent EADRC in flexible operation of combined heat and power plant considering carbon emission and economy

Abstract

Flexibility enhancement of combined heat and power (CHP) plant is extremely urgent to accommodate renewable energy and maintain the stability of power grid. In this paper, a new control strategy seamlessly integrates the error-based active disturbance rejection control (EADRC), Kalman filter and reinforcement learning algorithm is proposed and applied to CHP plant for operational flexibility promotion. Firstly, in EADRC, load fluctuation, uncertainties and heating supply interference are regarded as a total disturbance precisely estimated and compensated in real-time. Secondly, the optimal parameters of controller are obtained via the twin delayed deep deterministic policy gradient algorithm based multi-agent system. Under the demand of flexible operation, a comprehensive optimization objective including general control objectives and carbon reduction requirements is constructed for CHP plant. By setting the multi-objective function as the reward function of multi-agent system, plant is guided to the Utopia point which emblems the flexible, economic and low-carbon operation. The new controller effectively fuses the features from each component, namely state estimation, dynamic optimization, and continuous learning. Thirdly, the stability analysis of proposed controller based on singular perturbation theory is provided. Finally, the flexible and low–carbon operation capacity of CHP plant with the new control scheme under wide-load operating conditions is successfully verified through extensive simulations.