Project purpose and unfolding, clients.
The project of smart metering has been developed and deployed principally between 2015 and 2018.
At the beginning the project was launched in Engie start-up company “Homni”, situated in La Defense quarter in the west part of Paris. With the time the project has been integrated in the Smart metering department inside Engie itself.
The initial purpose of the project was calculation of gaz / electricity consumption (in meters cub / kilowatts) and it’s visualization on the smartphone application, as it’s shown on abovementioned pictures. The application is supposed to show in mobile application (f.e., in Android) the curve of gaz consumption with the possibility to control and program the consumption in advance.
Clients of these projects are the physical persons, Engie clients, consumers of gaz and electricity.
Technical realization of project.
The calculation of gaz consumption was organized with the help of sensor situated in front of the traditional gaz/electricity counters with number of wheels as shown at the pictures above. Sensor emitted regularly pulses in blue spectrum and it received the reflected signal. In function of figure (1,2,3,.,0) on the wheel, the level of received pulses is being changed. But every gaz meter had its own output for the cycle between 0 and 0 figures. This signature, pattern was especially repetitive Then the special digital signals processing algorithm was supposed to transform the signal wave to number wheel cycles, that is converted to the consumptions of gaz/energy. The information from the abovementioned captor is transmitted via the wireless protocol to freebox. Then this information arrives to date base server to be stocked and processed.
My participation in project and planning.
The patterns detection rate (or PSR, packet success rate) was supposed to be at least 90% according to the requirements. At the beginning PSR on all the test set of signals (gaz night light, strong consumption-shower, moderate consumption-oven) reached 50%. After the first month of new algorithm design the rate went to 80%, after the second: 90% (for some cases it reached 100%).
My participation included: planning development, planning of tests, debug organization, algorithm development, fixed point C code porting, team organization, contacts with clients, and contacts with providers.
The planning included consequently the following stages of development:
- Problem analysis
- Theoretical signal processing model design
- Matlab GUI analysing tool development (curves, framework, PSR calculation, signal forms visualization , tuning buttons, etc)
- First model tested on the real test vectors since different gaz and electrical counters present in France (Itron, SagemCom, Elster, Magnol,etc).
- Improved model with machine learning methods with better results, profiling.
- Portage of floating point Matlab code to fixed point, C/asm STM 32.
- Taking into account the target restrictions (microcontroller): RAM, stack, MIPS, parallelization. Finding the gold middle between the algorithm quality and performance.
- Test in laboratory. Correction the algorithm.
- Real tests at private homes using industrial server
Environment and technics used in project.
Software: Matlab, Visual C++
Languages: C, Matlab, asm
Methods: portage floating point to fixed point, machine learning, digital filtering, GUI tool development, autocorrelation, synchronization, resampling, demodulation, DC offset, amplitude detection, signal form recognition.
Hardware: gaz/electricity counters: Itron, SagemCom, Elster, Magnol; gaz/electricity measuring test bench, ST Microcontroller, bluewave emitter-captor, data base server, mcController evaluation board.
Main results.
- Pattern detection rate has been improved from 50% to 90..95%
- Matlab code has been successfully ported from Matlab floating point to STM32 real time embedded fixed point C.
- Successful tests in laboratory, confirmed algorithm functioning as well as the correct Matlab to C porting.
- Successful clients tests for principal types of situations.
- Excellent team ambiance