Study On Optimal Control In Ice-storage System Based On Load Prediction

Ice-storage technology is a new technology with positive social and economic benefits; it has been effectively supported by the government because of its characteristics of “peak load shifting”. In order to popularize the application of ice-storage technology, the problem of current ice-storage system optimal control was further studied in this paper. The content of the study mainly included three aspects: establishing a general and simple load prediction model, establishing an optimal control model and its solution, and discussing the economy.In the establishment of neural network load prediction model, the choice of BP neural network played an important role. Therefore, the author firstly analyzed the computing method and data sources of selected parameters in the network(such as the outdoor temperature, the solar radiation intensity and other important parameters); summarized the characteristic of the neural network model and discussed the design and improvement of the neural network model. Secondly, to set up an integrated and concise neural network model about the public buildings, there were three parts should be done. Besides, taking the actual engineering data as the basis of the load prediction model, the MATLAB was used to establish the neural network model, so that the neural network could achieve higher precision.1. Based on the principal component analysis method(PCA), the original data was reduced via the software, and the input parameters were reduced from 9 to 4, and the new parameters of the network were obtained.2. By using the software, the number of hidden layer neurons and the length of the sample were iterated repeatedly, and a neural network model with high precision and high speed was realized eventually. The network structure is 4-10/11-1, and the best interval of sample length is [26~30].3. To testify the established model had higher precision and speed, the model was compared to the other three kinds of neural network model {routine algorithm model(without PCA method and L-M algorithm), the model based on PCA method, the model based on L-M algorithm}. The simulation results turned out that the comprehensive model was superior to other network models in precision and speed.For the application of the ice-storage system, load prediction wasn't the ultimate goal, but to realize the optimization of system control and make the operation cost the most economical. Hence, the author founded an economic mathematical model about ice-storage system based on optimal control strategy. Historical records of a practical engineering case were used as a data source to seek the optimum ice packing factor(IPF), which is 30.93%. The relationship between IPF and the system invest is discussed, and the effect of optimal control on the economy of ice-storage system was qualitatively analyzed.Finally, comparing the ice-storage air-conditioning system to the conventional air-conditioning system in the initial investment and operation cost, the static economic evaluation method has been used to prove that the ice-storage air-conditioning system has a good economic benefit. Analyzing the case, the payback period of the investment is 3.74 years while the program with the ice-storage air-conditioning system is applied.The optimal control strategy is the most economical among the several strategies in ice-storage system. Because of its high demands on the accuracy of construction load, the whole load prediction model was researched in detail and depth, so that the integrated neural network model can be established eventually and can provide theoretical basis for the load prediction in the future. Meanwhile, the economy of the ice-storage system has been discussed, which can provide reference for the clients on the application of the ice-storage system.