3 - GridPulseMT

The objective of this project is to develop a day-ahead load forecasting model for the high-voltage (HV)/medium-voltage (MV) grid supply area. 

The supply areas are inherited by the physical structure of the energy distribution grid. Retail user consumption is powered through the medium voltage MV and each substation converting high voltage (HV) to MV has smart meters measuring power with a 15-minute granularity. Each HV/MV substation is not connected physically to another, but they are connected to the retail user of their area. The areas are not in contact between themselves. This means that the power at each substation is a measure of the load consumed by the retail users. 

Romande Energie provides electrical power to the majority of Canton Vaud, which means a wide surface and a large typology of clients. As an example, different areas of the canton might have different PV panels installed, thus having different net usage depending on the daily sunshine. Temperature and wind might play a role as well. In addition, the difference in local weather between two opposites of our service area might change the net charge of the distribution areas. 

Dynamic pricing requires an accurate forecast of the grid net charge for the future, possibly splitting the clients geographically. Using historical data from the substation power and the historical data on a grid for weather data, the objective is to develop the day-ahead load forecast: the forecasted load for the next full day using the data available at 17h.

This project sits on the development of dynamic pricing. Dynamic pricing is a methodology that grid operators can use to shift retail user consumption towards optimized grid usage.

The price per kWh billed under the “grid usage” part of the bill can be dynamical and adapt to the grid load. This method is interesting now that the consumption of most users is measured with a 15 minutes timescale. Dynamic pricing requires an accurate forecast of the grid net charge for the future, splitting the clients geographically.

• Load and coordinates for 21 substation, with a 15-minutes granularity, going from 01-01-2021 to 31-12-2025 
• Weather measurements for almost 200 latitude/longitude couples , with a 15-minutes & hourly granularity. 

The data will include radiation (direct and indirect), wind, temperature, relative humidity, cloud cover, rainfall among others.

Each hacker who wishes to participate in the challenge will need to sign an NDA.

• Data science
• Machine learning
• Timeseries forecasting