Anomaly ranking of District Heating Substations

Title Anomaly ranking of District Heating Substations
Summary Implementing anomaly ranking algorithm to monitor district heating substations.
Keywords anomaly detection, self monitoring, data mining, learnin-to-rank
References M. Goldstein and S. Uchida, "A Comparative Evaluation of Unsupervised Anomaly Detection Algorithms for Multivariate Data", PLOS ONE, vol. 11, no. 4, p. e0152173, 2016.

P. Arjunan, H. Khadilkar, T. Ganu, Z. Charbiwala, A. Singh and P. Singh, "Multi-User Energy Consumption Monitoring and Anomaly Detection with Partial Context Information", Proceedings of the 2nd ACM International Conference on Embedded Systems for Energy-Efficient Built Environments - BuildSys '15, 2015.

D. Araya, K. Grolinger, H. ElYamany, M. Capretz and G. Bitsuamlak, "An ensemble learning framework for anomaly detection in building energy consumption", Energy and Buildings, vol. 144, pp. 191-206, 2017.

S. Rayana and L. Akoglu, "Less is More", ACM Transactions on Knowledge Discovery from Data, vol. 10, no. 4, pp. 1-33, 2016.

Huang, Huaming, "Rank Based Anomaly Detection Algorithms" (2013). Electrical Engineering and Computer Science - Dissertations.Paper 331.

Prerequisites Artificial Intelligence and Learning Systems courses; good knowledge of data mining; programming skills for implementing machine learning algorithms
Supervisor Ece Calikus, Sławomir Nowaczyk
Level Master
Status Open

District heating system is a common way to distribute heat through underground pipelines for residential and commercial requirements. Faults are common in district energy systems due to the high number of substations and instrumentation components. Also, the standard energy-metering instrumentation is designed for low cost and billing, not for automated fault detection. Large variations in building dynamics, building subsystems, human behaviour and the environment make the system complex to model and analyse.

Anomaly detection refers to the process of detecting abnormal events that do not conform to expected patterns. However, it is hard to differentiate actual faults from the changes in energy consumption due to seasonal variations and changes in personal profiles such as holidays etc.

In practice, multiple anomaly detection tools are used to continuously raise alarms for different application domains. These alarms include both true positives and false alarms. Operators act on these alarms for diagnosis and deeper root cause analysis and take appropriate maintenance actions to mitigate the anomalous behaviours. Given the scale and scope of the district heating substations, the operators can be overwhelmed with the large number of alarms at any given instant. It is therefore necessary to prioritize and rank these alarms by their severity. In this project, we aim to propose a novel anomaly ranking algorithm in order to monitor district heating substations.

Examples of techniques to try are e.g. ensemble methods or learning-to-rank.

It is also interesting to look at ways to categorise different anomalies, based on different criteria.


-Applying and comparing state-of-the-art unsupervised anomaly detection methods

-Implementing novel anomaly ranking schema, for example based on aggregating results from different detectors

-Categorisation of anomalies

-Testing anomaly-ranking framework on district heating and heat pump datasets.

We have collaboration with 2 companies within this project, i.e., HEM and Öresundskraft.