Modeling PV Production with Machine Learning

Background: 
For small photovoltaic systems (<30kWp), CKW only measures the energy fed into its grid. This means CKW lacks direct insight into how much energy is actually produced by the PV systems. Our challenge aims to solve this gap by developing a machine learning model that estimates the total energy production and self-consumption of these smaller PV systems.

Data: 
In the CKW grid, there are approximately 9'600 PV systems smaller than 30 kWp without production measurement. Participants will use time series energy data, weather data (from Meteomatics), and egid data to develop models for estimating energy production and self-consumption. The dataset is fully prepared with basic feature engineering, allowing participants to focus on creating the best possible model. Both traditional machine learning techniques and advanced neural networks (e.g., LSTMs) are encouraged. 

Why it matters for CKW: 
This challenge has significant practical implications. Starting next year, PV production will be limited to 70% of the panel’s peak output as per new guidelines. Accurate production estimates will help quantify these losses and support the development of innovative products to make solar energy more market-oriented and attractive. 

Skills: 

• Python Coding skills
• Experience with machine learning and neural networks
• Creativity and innovation