Ice-storage Air Conditioning Supervisory Based On SVM

Air-conditioning systems are the major energy consumer and it is also the main cause of the difference of on-peak laod and off-peak load. Ice-storage system is one of the effective ways of solving these problems, which produces ice during off-peak period and melt it in on-peak period to provide cooling for the buiding. So it has become one of the most important techniques to shift peak load and reduce operating cost and energy consumption.Ice storage system revived in 1990 s because it can move the power used from day to night. As the power price of day is higher than that of night, ice storage system can help to save power charge. Chiller priority is the most common control strategy for existing ice storage systems. But only optimal control can significantly reduce the operating cost to the minimum. The accuracy of the load prediction is a key for optimizing the system control.Excessive pursuit of traditional air-conditioning system load forecasting accurate loading data, require large amounts of historical data as a basis, but it has high error rates, a large number of research results in practical applications are unable to project site personnel play a practical role in guiding.Discussed in this paper which is suitable for load forecasting criteria for project use, namely load percentages to predict the future and then press load is calculated as a percentage of the hourly load distribution method to guide the engineering of the operator’s work, and integrate this into the cloud application platform software, on-line prediction of common scenarios.Firstly, this paper introduces an ice storage air conditioning system, collecting the operation of outdoor temperature and cooling load data in 1 year on this project, these data will be used as a basis for building load forecasting model.Secondly, introduced separately the most commonly used for load forecasting of BP neural network and RBF Neural network, analyzes their strengths and weaknesses.Introduces the features of SVM load forecasting, as opposed to BP network and RBF networks to summarize the advantages of SVM systemMoreover, using SVM binary tree multi-class classification method based on load prediction system, the load is divided into100%-80%, 80%-60%, 60%-40%, 40%-20%, 20%-0% five-term load forecast interval, comparing it with the traditionalforecasting error rate as well as its actual guidance to field staff.