模型准确性对系统管网泄漏诊断的影响分析

摘要

This paper proposes an evaluating method for the effectiveness of leakage diagnosis by using quantitative index of model accuracy,aiming at the pressure sensitivity analysis-based leakage diagnosis method of the pipe network brought by model mismatch. FlowMaster software is used to establish the pipe network models with different accuracy of a real refrigerant system. Regarding the calibrated model as the benchmark of the pipe network,the improved Grey Relational Analysis(IGR2)is proposed to calculate model mismatch by estimating the similarity of pressure measurements between each model and the bench⁃mark model. Thus,the indexes of quantitative model accuracy are achieved. Furthermore,Fault Diagnosis Success Percent(FDSP)is defined to calculate the effectiveness of leakage diagnosis. The leakage is diag⁃nosed based on the models with different accuracy,and the effects of model accuracy on leakage diagnosis are obtained. Results illustrate that the effectiveness of this leakage diagnosis method is improved when the model becomes more accurate. When the refrigerant pipe network model accuracy is higher than 0.75,the FDSP is basically over 90%which shows that this leakage diagnosis is robust to model mismatch.%针对基于压力敏感性的管网泄漏诊断方法应用中由管网模型误差带来的鲁棒性问题，提出根据量化的模型准确度指标来分析诊断方法的有效性。利用FlowMaster流体仿真软件建立某实际冷却系统管网不同准确度下的水力模型。以该管网的标定模型为基准，采用改进的灰色关联度分析（IGR2）法对各模型与基准模型间各测点压力值的相似性进行评估来计算误差，得到模型准确度量化指标。采用故障诊断成功率（FDSP）计算泄漏诊断的有效性。基于不同准确度的管网模型进行泄漏诊断分析，获得不同准确度模型对泄漏诊断有效性的影响。结果表明，该泄漏诊断法的有效性随模型准确度的增大而提高，当管网模型准确度大于0.75时，故障诊断成功率均处于90%以上，说明该故障诊断方法对模型误差具有一定的鲁棒性。