Research And Application Of Data Mining In Short-term Air-conditioning Load Prediction

Reliable Air-conditioning system load prediction is an important foundation to realize high efficiency and energy saving of the Air-conditioning system which is based on the actual circumstances of the Air-conditioning load characteristics, by the use of certain mathematical theory and calculation method, and of appropriate model to have an accurate load forecast results. However,Air-conditioning load historical data collected in practical work usually contains many problems such as the noise, the randomness, ect, in addition, the selection of the independent variables of the prediction model is often lack of basis. These problems will certainly affect the generalization of the predition model, and then affect the accuracy of load prediction. To improve the generalization of the predition model, Air-conditioning load historical data should be preprocessed.Aimed at the problem of bad data from the process of information gathering of the Air-conditioning system, this paper firstly builds a bad data cleaning model, extracting the characteristic curve of Air-conditioning load by using Kohonen neural network clustering, adopting the super circle covering artificial neural network positioning and identifying bad data, which effectively improves the quality of the Air-conditioning load historical data after data adjustment. As to problems of high independent variable dimension of load forecast and the difficulty of selection,correlation analysis and principal component analysis are used to analyze the relevant factors of independent variables, which reasonably reduces the input dimension of the predictive model, and so as to solve the fitting problem in the model prediction. Finally, in response to the issue of local convergence and low accuracy occurred easily in the traditional neural network model prediction,support vector regression model is established, and related parameters of support vector machine is optimized with the help of the heuristic algorithm, this paper finally predicts theShort-term Air-conditioning Load Prediction by using the optimized support vector machine(SVM)model. According to using examples and comparison, this paper proves that the method put forward by the author can effectively improve the generalization of short-term Air-conditioning load prediction.