Challenge

Engie, a global energy provider, is committed to enhancing energy intelligence through data-driven insights. A key challenge in renewable energy planning is accurately identifying buildings with installed solar panels. Traditional methods for gathering this data rely on manual surveys, public databases, or incomplete utility records, which are;

  • Time-consuming and costly to update
  • Inconsistent, with varying data quality across regions
  • Difficult to scale for national and international coverage

To improve solar energy mapping and energy consumption predictions, Engie needed a solar panel detection system to identify solar-equipped buildings at scale.

Solution

Faktion developed an AI-driven solar panel detection model using computer vision to analyse aerial imagery and identify buildings with solar installations. The system combines high-resolution satellite data, geographic databases, and deep learning to automate large-scale solar panel detection.

Key components of the solution include:

  • Deep learning model: a convolutional neural network (CNN) trained to detect solar panels in aerial images
  • Geospatial data integration: the model is linked with address and building footprint databases, ensuring accurate location mapping
  • Automated image processing pipeline: the AI processes aerial images, assigns confidence scores, and updates a solar panel database
  • Batch processing for large-scale coverage: the system can analyse entire regions in a single pipeline, making it scalable for national expansion

By replacing manual inspections with AI-powered automation, Engie gains a cost-effective, scalable, and consistent method to track solar adoption and plan future energy initiatives.

Approach

We followed a structured development process, ensuring a robust and scalable AI model:

  1. Data collection and preprocessing: the team gathered high-resolution aerial images, spatial building data, and address records from Geopunt, CRAB, and GRB datasets
  2. Model training and fine-tuning: a supervised learning model was trained using labeled datasets containing images of buildings with and without solar panels
  3. Validation & accuracy testing: the model was evaluated on a test dataset to ensure high detection accuracy, with additional manual validation where needed
  4. Integration & deployment: the model was deployed in Engie’s cloud infrastructure, allowing seamless batch processing of large geographical areas
  5. Expansion strategy: after successful testing, the AI model was adapted for deployment beyond the initial region, expanding to other areas like the Netherlands

This systematic approach ensured that the solar panel detection model was scalable, accurate, and efficient for real-world applications.

Outcome

The solar panel detection system has delivered key advantages for Engie:

  • Automated large-scale analysis: the system scans thousands of building, significantly reducing manual workload
  • Improved data accuracy: AI-driven detection ensures higher precision compared to traditional survey methods
  • Scalability for future expansion: the model can be applied nationwide, with the potential for international expansion
  • Enhanced energy forecasting: accurate solar panel data supports better grid management and energy strategy planning

This project laid the foundation for AI-powered energy intelligence, enabling Engie to optimise solar adoption tracking and support sustainable energy initiatives.