Projects.

Data analytics for AMM data processing

ČEZ Distribuce, a.s.

Customer

ČEZ Distribuce a.s., is the biggest distribution system operator in Czech Republic. It operates 3.5 millions of consumption points, has 46.000 secondary substations and more than 100.000km of low voltage networks.

Challenge

Measurements of low voltage grid was interesting historicaly just for billing purposes. But history is passed. Solution for full utilisation of smart metering infrastructure is good step in DSOs strategy for managing the grid.

To build a smart metering infrastructure and substation monitoring infrastructure is big investment for DSO. How to get maximal value from these infrastructures? What are possible use-cases of using low voltage measurements for DSO processes optimization? How to effectively monitor such a large metering infrastructures?

Solution

Following areas were analysed:

  • Scope of measured data from smart meters and substations
  • Data validations methods
  • Missing meaurements estimation methods
  • Measurement processing methods
  • Monitoring of large metering infrastructures
  • Metrics for data value calculations

During the project:

  • More than 150 millions of smart metering measurements from field was analysed.
  • Various algorithms for predictions, estimations, anomaly detections, situation classification and segmentation was evaluated according their precision, performance and requirements on input data quality.
  • Several analytical applications on Qliksense platform were developed.
  • Lot of inspirative discussions with DSOs experts.

Result

New estimation algorithm was designed and tested – is has 15% better precision results than actually used by DSO.

Machine learning algorithm for anomaly detection was successfully tested on smart meter measurements and events concerning tariff and relay switching. Several problems of used meters and concentrators were discovered by using this method.

Smart metering data value metric was designed. It calculates value of collected data and can help to define which data to collect from smart metering.

Algorithm for estimation of the order in which phases are connected at individual connection points was successfully tested.