根据Supervisory Control and Data Acquisition （SCADA）系统异常监测值能及时预警供水管网爆管，但由于用水周期变化、随机波动及监测误差，导致实际中很难预警较小爆管。针对该难题，开展了基于自适应卡尔曼滤波的供水管网爆管信号识别研究，提出将历史监测数据按用水周期分解，采用自适应卡尔曼滤波结合平均低通滤波对管网供水量进行实时估计，根据监测值与估计值的差，预警爆管、估算爆管流量。分别采用仿真数据与实测数据验证所提出方法，结果表明，所提出方法可用于实际供水管网爆管预警；对所采用实测数据，爆管预警精度约为最大时水量的9%；此外，实际爆管预警精度主要取决于用水量本身的随机波动，同时与监测数据采样频率相关。
The pipe burst of water distribution systems (WDSs) can be alarmed on-line according to the abnormal change of measurements of Supervisory Control and Data Acquisition(SCADA) system. However, due to the periodic variation and random fluctuation of the water consumption, as well as the measurement errors, it is often hard to alarm the burst with small discharge in practice. To solve this problem, this paper proposes a method for burst alarming of WDSs based on adaptive Kalman filter, by which the historical data are first decomposed according to periodic variation of water consumption, and then the adaptive Kalman filter combined with the average low pass filter are used to estimate the water supply real-timely, finally the burst is alarmed and its discharge is estimated based on the difference between the measurements and the estimates. The test results from simulation data and real measurements indicate the proposed method is valid for burst alarming in practical application, and the accuracy of burst alarming is about 9% of the maximum hour water consumption. The accuracy of the burst alarming depends mainly on the random fluctuation of the water consumption itself, and is also related to the sampling frequency of measurements.