基于最小二乘及分类向量机的空气调节器故障检测

Fault detection of air conditioner based on least squares and support vector machine classification

为了对空气调节器的故障进行检测和诊断,提高建筑物管理系统的能源利用率,提出一种基于递归最小二乘的故障检测和诊断方法.方法包含特征选择、递归最小二乘和支持向量机分类三个部分,在特征选择中,将空气调节器的故障分为11个类型,并基于Relief F算法选取三个最显著的特征变量.在递归最小二乘中,通过最小化真实值与预测值之差的平方和对模型的参数进行估计,并基于二叉决策树思想采用支持向量机对11个故障状态和1个正常状态进行分类.结果表明,所提方法可以更好地对空气调节器中的故障进行检测,并对故障类型进行分类.

In order to detect and diagnose the faults of air conditioner and improve the energy utilization rate of building management system, a fault detection and diagnosis method based on recursive least squares was proposed. The method was composed of such three parts as feature selection, recursive least squares and support vector machine （SVM） classification. In the feature selection, the faults of air conditioner were divided into 11 types, and three variables with the most notable feature were selected based on ReliefF algorithm. In the recursive least squares, the parameters for the model was estimated through minimizing the quadratic sum of the differentials between the actual values and predicted values. Based on binary decision tree idea, the 11 fault states and a normal state were classified by SVM. The results show that the proposed method can better detect the faults of air conditioner, and can classify the fault categories.