基于贝叶斯理论的日用水量概率预测

为解决城市日用水量的概率预测问题,提出了基于贝叶斯理论的日用水量预测法.引入贝叶斯理论,建立了日用水量概率预测系统.在系统中,利用支持向量机建立日用水量预测模型、似然函数和先验密度,并采用自适应马尔可夫链蒙特卡罗模拟方法求解日用水量的后验密度,得到日用水量的概率预测值.实例表明,本文提出的预测方法不仅显著提高了日用水量的预测精度,而且通过定量给出预测值的置信区间,为城市供水系统的调度提供了更科学、可靠的决策依据.

Probabilistic daily water consumption forecasting usingBayesian theory

To probabilistically forecast the daily water consumption, a forecasting method based on Bayesian theory was proposed. The Bayesian theory was introduced to form the probabilistic forecasting system of daily water consumption. In the system, the daily water consumption forecasting model, prior density and likelihood function based on support vector machine (SVM) were built. An adaptive Markov chain Monte Carlo method was used to solve the posterior density of daily water consumption to obtain the probabilistic forecasts. Case study shows the proposed method has better estimating performance and it provides the confidence interval of the forecasts for the scientific and reliable dispatch of water distribution network.