Research On Predictive Control For Ice Storage Air-conditioning Systems

With the development of economy, the city-size of Beijing expands rapidly. Numerous coal-fired power plants have been constructed to meet the increasing energy demand which leads to the increasing of generator capacity and the scale of grid load. They squander resources and destroy our environment. In order to save energy and improve environment, the government has introduced relevant policies to encourage applying energy storage device to reduce peak power. Under the government’s guidance, the ice-storage systems for air-conditioning applications have been widely installed and a series of problems appear. To name a few, operating according to personal experience may not achieve the designed performance. As a consequence, it is urgent to study the optimization control strategy of ice storage system.Lacking simulation platform to verify the strategy for optimal control in Ice storage system is a vital drawback. Designing controllers for traditional HVAC model lies in successful dynamic models as well as corresponding math equations to describe its behavior. E.g. ice-tank model of ice storage system is not suitable for control simulation. As a result, the tank model needs to be designed to simulate the ice-storage air conditioning system according to practical engineering. Also, the programs are not available for cross-platform simulation. The developing module function and external call function is employed to construct a co-simulation platform of TRNSYS and MATLAB. We have developed the dynamic models of the plate heat exchanger and the ice tank in the platform to set up the whole ice-storage air conditioning system and to verify the optimization control strategy.The ice-storage air conditioning system in project 4-2 was built based on the co-simulation platform. The double-effect chiller and the ice tank affect the chilled water temperature dropped from 13.2℃ to 2.2℃. In a practical project, the temperature of 2.2℃ for the chilled water can not only meet the requirements of the buildings’ load but also save the energy of pumps in the water system and the air system.A multi-layer forward neural network acted as the optimal feedback controller, which was trained with optimization algorithm based on the Hamilton-Jacobi-Bellman(HJB) and Euler-Lagrange(EL) equations as well as multi-step predictive performance function. After been trained, the neural network controller can approximate the optimal feedback solution of nonlinear systems without the complexities of computation and storage problems. The neural network predictive control algorithm, which is of high precision and strong robustness, works effectively with small amount of computation, has been programed and applied in the ice-storage air conditioning system model. Simulation results show that in the condition of large variation load, it is by employing the neural network predictive control algorithm that the controlled variables manage to achieve the pre-set values rapidly in 180 senconds, while the PID control algorithm needs 20 minutes to achieve stability. Hence, it proves that Neural network predictive control algorithm works better for guiding the practical operation of the project and reducing the energy consumption of ice-storage air conditioning system.