Data Validation for Digitally Enabled Operation & Maintenance of District Heating Systems

Digitization provides many opportunities and benefits for the district heating sector to increase equipment up-time, availability, increase in the lifetime, and overall effectiveness. Additionally, digitalization promises a reduction in overall maintenance costs and premature asset replacement. However, for this to be possible, the first step must be ensuring that the data meets a sufficient quality. For this reason, we employ a set of data validation methods to data from a real district heating system. The methods are a frozen test, ranged test, Shewhart chart, spike test, sliding window, and principal component analysis coupled with Hotelling's T2 and Q statistic. For validation of the methods, we investigate the detection rate of four known anomalies. We also show that combining different methods can reduce false positive detections. Lastly, we demonstrate a method to identify poorly performing sensors based on the combined detections of the univariate data validation methods.