Abstract: According to the actual water supply situation of a small and medium-sized water company in China, the influences of weather and other factors on daily water supply are analyzed by comparing the correlation coefficient, so as to determine the input parameters required for daily water consumption prediction. The application performance of three traditional water volume prediction methods is compared using the actual operating data. To solve the severe errors existing in the traditional methods, an improved method is proposed, which takes the water consumption of the previous day and 8 hours into consideration. The efficiency of the proposed algorithm is verified by the tests in the information system of the water supply company. According to the performance and implementation complexity of the algorithm, a water quantity prediction algorithm and its suitable implementation form for urban water affairs are proposed, which can help the water affair system improve the water quantity prediction accuracy, thus effectively improving the utilization rate of water resources.

Key words: smart water platform, water volume prediction, big data, neural network algorithm