How to optimise and predict street furniture maintenance by analysing data generated by the furniture or outside databases?
Technicians servicing street lights, traffic lights and other street furniture typically follow predetermined schedules. They work on street furniture at standard intervals, which are based on estimates of each piece of equipment’s state of wear.
Data that recently became available from systems connected to this furniture, however, has opened the door to detecting and anticipating maintenance work requirements. Doing this serves two purposes: optimising work (and possibly predicting breakdowns) and enhancing quality of life in cities by reducing the number of unlit areas.
Use case and experimentation field
The experiment’s success will be measured in terms of the breakdowns identified by analysing data generated by street furniture compared to the anomalies recorded by teams tasked with repair work on the ground.
Over time, the goal is to shorten repair timeframes, reduce the number of malfunctions reported by residents, and reduce the number of unforeseen breakdowns.
The field available for this experiment encompasses all the street lighting in Paris. For the purposes of the experiment, however, only a restricted (but representative) sample of streets and areas will be used.
The data sets provided by the partners are:
– Hourly data from 16,000 units controlling street lights
– Georeferenced data on street lighting in the experiment area
– Outside data such as weather information
Evesa, a company related to Bouygues Energies & Services, is tasked with operating and servicing public lighting in Paris and has access to the data gathered by 16,000 units controlling those lights. Its goal is to reduce operating costs while offering new services.
Paris City Council is keen on providing its people with a high-quality and efficient public service.
From a more general perspective, any operator handling electric infrastructure – including EDF as well as many other players outside France – may be interested in optimising predictive maintenance on its facilities. The opportunities to use this experiment’s outcomes, in other words, are not limited to street furniture but encompass any device connected to a control unit and requiring maintenance.
Based on a set of historical data, Saagie analyzed each electrical curve associated with street lights in order to:
· create a typology and classify these anomalies in clusters depending on their cause,
· identify the context for each malfunction.
2806 dysfunctions have been identified using Saagie’s solution. A prediction model was developed, including a time range associated with a probability of lights malfunction, and mock-ups of a visual interface to be used by Evesa’s technicians.
- Solution immediately scalable to the 199 000 street lights in Paris
- Enhanced quality of service for citizens with lower rate of lights break-down
- Solution is scalable to any street furniture consuming electricity