DataCity entrepreneurs: #3 How to bring the electricity bill of a city 10% down ? - Datacity

DataCity entrepreneurs: #3 How to bring the electricity bill of a city 10% down ?


This serie presents the entrepreneurs that participated to DataCity Paris 2nd edition and their projects that were presented during the Demo Day on June 6, 2017

Clémence Fischer, Head of Smart City

Paul Moreau, Innovation Program Manager

 

Clémence, Paul, within Datacity, which projects were you working on?

 

“Making night lighting “smart” by adapting it to the actual presence of people”

 

Currently, Paris is equipped with 199,000 light points in its streets, gardens and squares to provide lighting at night to the city’s 2.2 million down town residents and many visitors.  We worked on ‘adaptive’ lighting in the city to reduce light pollution and save energy. The idea behind the challenge is to know if urban lighting can be modulated based on activity in a street, without using motion sensors in order to keep infrastructure costs down.

Just recently, French regulation changed to allow dimmed lighting during off-peak times at night in a moderate capacity that ensures the safety of pedestrians and vehicles. The challenge created by Bouygues Energies & Services and taken up by startups Quantmetry and Dataiku, aimed to assess the relevance of anonymised and aggregated data from mobile phones to determine street activity. If the data was found to be useful, the ultimate goal was to offer a new street lighting map for off-peak times.

We worked on this challenge with Alexandre StoraBenjamin Habert and Sacha Samama at Quantmetry, with help from Matthieu Scordia and Yannis Ghazouani at Dataiku. In terms of major firms, Jean-Bernard Sers at Bouygues Energies and ServicesFrédéric Gauzy at SFR, and Maxime Billot at EVESA (the lighting network operator of the City of Paris) assisted Quantmetry and Dataiku throughout the challenge.

Jean-Bernard Sers (Bougyes Energy and Services), Benjamin Habert (Quantmetry), Frédéric Gauzy (SFR), Yannis Ghazouani (Dataiku), Maxime Billot (Evesa) et Alexandre Stora (Quantmetry)

What was the most difficult moment and which result are you most proud of?

 

The first hurdle in this challenge was to identify startups that combine a strong data science expertise and a keen interest in the challenge. A limited number of startups applied, but those which did were of excellent quality.

The second hurdle was to process the data itself.

The goal was to measure variations in night-time urban traffic to adjust the brightness of Paris street lighting. Quantmetry and Dataiku analysed SFR data and were confronted with two major challenges: establishing geographic units and the accuracy of the data.

First: how do you incorporate the anonymous flow of data into the geographic map of street columns? In other words, once the urban travels at night have been modelized, how do you translate it into concrete action for the street lights ?
Division into IRIS units (groups of streets within a district) made it possible to achieve this.

Second: how can we ensure that traffic measurement is reliable? Data scientists compared SFR flow measurements with open data to calibrate a prediction model. Teams worked on dozens of data sets, combining state-of-the-art technologies in data analysis, including DSS from Dataiku, PostgreSQL, PostGIS, Python, and D3.js.

The quality of the work performed by data scientists at Quantmetry and Dataiku was a decisive factor in the project’s success.

Participants in the challenge are proud to have proven that mobile phone data can be used to save energy on lighting – the main hypothesis of the experimental phase.

If this project was replicated in all cities worldwide …

 

In terms of impact, rolling out such a solution in the city of Paris would translate into a 6% decrease in the city’s energy bill without any tangible investment.

Similarly, this solution can be rolled out in any city in the world equipped with LED lighting columns and street lights that can be set to at least two levels of brightness, at on- and off-peak times.

“Building local communities for energy production and consumption”

 

Our second project was to take up the challenge of collective self-supply electricity, or “local energy communities”. The challenge, to allow every consumer to purchase electricity produced by their neighbour and bypass the national network, represents a major disruptive force in the world of energy. This new market will progressively shake up today’s energy landscape, traditionally based on high-capacity power plants which distribute electricity on national and transnational networks.  In France, solar energy represented 1.4% of total production in 2015; the multi-year Energy Plan (PPE) of October 2016 aims to triple that figure by 2023, creating major potential for progress.

The challenge was taken up by Danish startup Linc, represented by its CEO Pranay Kirshen. In terms of major firms, Fabrice Casciani at EDF, Herbert Beck at Nexity and Jean-Bernard Sers at Bouygues Energies et Services (BYES) also rose to the challenge and helped the startup build a solution. We decided to work on this challenge for its cutting-edge potential and because French regulation is undergoing changes that will allow this type of market to emerge.  For example, following the experimental phase, an implementing decree was issued to permit the purchase of electricity.

Herbert Beck (Nexity), Pranay Krishen (Linc), Fabrice Casciani (EDF) and Jean-Bernard Sers (Bouygues Energy And Services)

What was the most difficult moment and which result are you most proud of?

 

On the technical side, four smart meters were tested and installed at the EDF Renardières Lab, with the back-end of Linc connected to the site’s meters to allow direct interaction between a home producing energy with solar panels and a consumer household. On the energy consumer side, we started from scratch due to the non-existence of a market in France, where changes in regulation were yet to take place. An important part of the challenge was to assess the value proposition with potential users of a local-level energy exchange solution. We established a profile: a standard individual who we thought might be interested in joining a local energy community to buy or sell electricity from/to a neighbour (“early adopter”). According to an Opinion Way survey, 47% of the French population would be willing to invest in self-supplied solar energy, meaning that the potential share of early adopters is significant.  We then wanted to define a specific example of how the technology could be used, to identify individuals for interviews and establish their needs. But how can a value proposition be tested when no reference points exist for a service or its pricing? Should we target solar energy producers or simply any resident whose neighbours produce solar energy?

We formulated several hypotheses in a short space of time to focus on one specific value proposition.  We tested them using 10 qualitative interviews of about one hour in length with people likely to adopt the solution and who matched the “early-adopter” profile identified at the outset. We showed them our user interface prototypes, accessible on a Linc web site, to test the value proposition and service price. Over 30 interfaces were developed for the test. This fast prototyping and testing method made it possible to quickly confirm or reject our initial hypotheses regarding the product and user needs. The ultimate goal is to align the collective self-supply offer with the real needs of people who are likely to enter this market.

If this project was replicated in all cities worldwide …

 

… then a large part of world production would be renewable and local! Given that 50% of the world’s population already lives in cities, that is a lot of people. Of course, solar power alone would not suffice (at night, for example); other means of energy production and storage would have to be added, or the national grid used. According to the European Commission, 264 million European citizens could produce their own electricity as early as 2050. The impact on greenhouse gas emissions will be enormous.

“Predicting malfunctions in city infrastructure”

 

The third challenge we took up focused on the Predictive maintenance of city infrastructure. The ultimate goal is to offer an approach that predicts malfunction in lighting columns and street lights in the city of Paris. Data used to address this issue included voltage curves, amperage, energy, power and power factors, measured on 20,000 control units connected to street lights operated by the City of Paris, with data transmitted on an hourly basis. Lights present a different energy profile according to how they are used and their surrounding environment. The hypothesis tested during the experiment asserted that the study of a large volume of energy data could identify and predict malfunction in urban lighting systems.

The challenge was taken up by Saagie, a company specialised in processing big data and constructing machine learning algorithms. For Saagie, the goal of the experiment was to offer a new way to use its technology and build a new offer for urban infrastructure maintenance.

We worked with Stéphanie Brion, Sébastien Follet and Karine Goutorbe at Saagie. Maxime Billot and Alexis Abdelmoula were our main contacts at EVESA (a Bouygues Energies et Services subsidiary (BYES), responsible for urban lighting systems in Paris), along with Jean-Bernard Sers (BYES).

The data record provided by EVESA allowed Saagie to first identify electrical signals which led to malfunction in the field. After identifying the curves, we were able to group together malfunction causes and create a classification of potential failures. For example, data analysed by Saagie was used to identify arcing and capacitor faults, as well as cases of contact resistance with a permanent, occasional drop in tension.  Saagie’s work identified 1,276 failures in total on 648 units over a three-month study period on average (failures which led to visible malfunction were repaired by EVESA on site).

The next phase involved correlating this classification system with contextual data from the field, with the ultimate goal of predicting a failure at specific, precise intervals. Comparing the data record with real failures made it possible to validate the approach and the prediction results. Saagie’s work identified a list of 2,806 potential failures, each with its own confidence rating (likelihood of the failure in the field), calculated using a statistical distribution of the distance between the anomaly and the failure across the data set. This data was compared with field conditions and will continue to be over the next few months.

Jerome Tredan (Saagie), Alexis Abdelmoula (Evesa), Karine Goutorbe (Saagie), Maxime Billot (Evesa), Remi Coulon et Jean-Bernard Sers (Bougues Energy and Services)

What was the most difficult moment and which result are you most proud of?

 

The challenge was difficult in that the extent and depth of the data was only just enough: the units used to transmit electrical data were installed by EVESA at the end of 2016. We had to wait for further data to arrive from the field over time, which was only a short-term problem: the more new data was added to the algorithm, the more accurate the predictions became.

Saagie also offers a visual interface in the form of a map which can quickly identify street lights likely to fail. Technicians can use it directly to plan predictive maintenance tasks.

The direct impact for the urban lighting system operator is a decrease in the time needed to identify failures and a decrease in the number of actual failures. For citizens, the system improves city living by reducing the number of faulty street lights.

If this project was replicated in all cities worldwide …

 

The offer developed by Saagie could be replicated in most urban areas at both the national and international level. The impact for public lighting managers represents significant savings. The predictive maintenance solution developed by Saagie enables managers to prevent failures before they become too serious, thereby reducing spending on repairs and spending less time identifying faulty street lights.

 

Written by:

Clémence Fischer, Head of Smart City

Paul Moreau, Innovation Program Manager

 

Read the other article of this serie here!

Get all the information on the DataCity program and its edition in Paris

Datacity in 2 minutes (video)

To join us or get to know more about the program:  [email protected] !