Coupled Building and System Simulations for Detection and Diagnosis of High District Heating Return Temperatures

Abstract

High return temperatures are a frequent issue leading to inefficiencies in district heating networks. The causes for high return temperatures usually lie on the secondary side, within the building heating system. However, the district heating operator will in most cases only have access to primary side data through the heat meter. This makes it difficult for the operator to identify and remedy these causes. This contribution uses coupled building and system simulations to investigate issues leading to high return temperatures. The resulting synthetic data replace inaccessible secondary side data for the training of supervised classification algorithms allowing these issues to be diagnosed based on temperature and flow measurements in district heating substations. These classification algorithms are tested with three cases differing in the assumed availability of secondary side data. Fault detection and diagnosis can be performed with primary data only, with a modest degree of accuracy. Temperature measurements on the secondary side of the substation are shown to considerably improve the quality of predictions, from 78% to 96% classification accuracy.

Intelligent Approaches to Fault Detection and Diagnosis in District Heating: Current Trends, Challenges, and Opportunities

“…It must be determined whether the data is appropriate for automatic FDD and what kind of additional information is needed to improve the FDD accuracy. For example, only a single study reports FDD accuracy gains when including secondary side information [36]. While secondary side data is sometimes available, it is not consistently available due to, e.g., privacy and practical issues.…”

“…Brès et al [36] present an FDD approach using a Binary Decision Tree (BDT) and building simulations to discover fault signatures and identify four issues that cause high return temperatures. The authors calculate the correlation coefficient and quotient of average values for each pair of variables.…”

“…Few studies focus on deep learning or logical models, while none utilize probabilistic methods for outlier detection. [71] TPOT Geometric Wang et al [66], Palasz and Przysowa [68] SVM Geometric Theusch et al [70], Farouq et al [62][63][64] kNN Geometric Lee et al [69], Theusch et al [70], Sandin et al [72] kM Geometric Farouq et al [63] CC Geometric Sandin et al [72] Limit-checking Geometric Palasz and Przysowa [68] GBDT Logical Brès et al [36] BDT Logical Brès et al [36] CART Logical Farouq et al [64] IF Logical Wang et al [60], Palasz and Przysowa [68] MLP Deep learning Wang et al [60], Zhang and Fleyeh [67] LSTM Deep learning Zhang and Fleyeh [67] AE Deep learning Zhang and Fleyeh [67] VAE Deep learning Johansson and Wernstedt [73] Visualisation Statistical Gadd and Werner [6] Manual analysis Statistical…”

Grey-Box Modelling and Virtual Sensors Enabling Continuous Commissioning of Hydronic Floor Heating

“…Fault detection in DH networks have mainly focused on DH substations using heat metering data [23,30,[55][56][57][58][59]. These methods detect temperature faults in DH substations but rarely identify their origins, which may occur in the end-users' heating systems [60]. In a simulation study, Brès et al [60] developed a fault detection algorithm using temperature and flowrate measurements from DH substations.…”

“…These methods detect temperature faults in DH substations but rarely identify their origins, which may occur in the end-users' heating systems [60]. In a simulation study, Brès et al [60] developed a fault detection algorithm using temperature and flowrate measurements from DH substations. Measuring the supply and return temperatures on the end-user's side significantly improved the algorithm's accuracy.…”