基于改进卡尔曼滤波的供水管网爆管检测研究

An Improved Kalman Filter Method for Burst Detection in Water Supply Network

根据SCADA系统监测的流量及压力信号检测爆管时,监测误差及用水量随机波动形成的背景噪音会掩盖较小爆管信号。利用卡尔曼滤波能有效从背景噪音中识别爆管信号,但存在爆管检测精度偏低、误报率较高的缺点。改进了传统卡尔曼滤波法,对卡尔曼滤波残差进行累积和运算,选取0.5倍标准差为累积和阈值、3倍标准差为检测阈值进行爆管检测。应用数值仿真验证算法,结果表明:普通卡尔曼滤波检测信号不稳定,爆管检出信号断断续续且持续时长小于实际爆管时长,非爆管时段出现11次误报;改进卡尔曼滤波法检测信号稳定,爆管检测信号持续时间大于实际爆管发生时间,非爆管时段仅出现1次误报。相较于普通卡尔曼滤波法,改进卡尔曼滤波法将非爆管时段理论误报率由16%降至2.8%,同时放大了爆管信号强度、持续时间,提高了爆管时段爆管检出成功率。

In the detection of the burst in water supply network,when using flow and pressure signals from SCADA system,the burst signal can be masked by background noises such as monitoring errors and fluctuations of water consumptions.Studies have shown that the Kalman filter method is effective in identifying the burst signal from background noises,but has limitations of low detection accuracy and high false alarm rates.The traditional Kalman filter method was improved,in which the cumulative sum method was applied to Kalman filter residuals,half-time standard deviation was used as the cumulative threshold,and three-time standard deviation was used as the threshold for burst detection.Numerical simulations were used to demonstrate the proposed method.Results indicated large instability in the detection signal of ordinary Kalman filter.The ordinary Kalman filter showed intermittent detection signal and its duration was shorter than the actual duration of the burst;additionally,there were eleven false detections during the non-burst period.The improved Kalman filter showed stable detection signal and its duration was longer than the actual duration of burst.There was only one false detection during the non-burst period.In conclusion,compared with the ordinary Kalman filter method,the improved Kalman filter method reduced the theoretical false detection rate from 16%to 2.8%during non-burst period,and amplified the intensity and duration of the burst signal,which increased the success rate of burst detection during the burst period.