JunctionX KAUST 2019 : Predicting the Future of Solar Power Generation at NEOM

Ayachips

September 15, 2019

Solar power generation forecasts will be a critical need if SOs are to balance NEOM’s grid. We built an ML model to forecast solar power generation that takes into account NEOM’s weather patterns.

Devpost link

ml_prediction dashboard

Inspiration

Solar power generation forecasts will be a critical need if SOs are to balance NEOM’s smart electricity grid with nearly 100% renewables. Even though NEOM is blessed with plenty of solar radiation, NEOM also experiences substantial fluctuations in temperature, wind, and dust and these factors can all have a substantial impact on solar power generation. We built an ML model capable of accurately forecasting solar power generation that takes into account NEOM’s unique weather patterns and created a few prototype interactive dashboards to display the data.

What it does

Our model uses ML techniques to predict future solar power generation at NEOM as a function of the weather data as well as the history solar power generation. After training our model on historical data, we can generate a new forecast for next day’s solar power generation. Once the next day’s actual values of solar power generation are observed, our model can be automatically re-trained and improved. Model can easily be retrained with weekly or monthly forecast horizons if longer forecasts are required by the SO.

Our interactive dashboard allows the user to interrogate the historical weather and solar power generation data for NEOM as well as to display forecasts of future solar power generation. Dashboard can generate user notifications via the Telegram mobile app to indicate significantly changes to either weather or solar energy forecasts.

Our web and mobile app prototype will allow users (SOs, residential, and industrial prosumers) at NEOM to interact with weather and solar energy forecasts and to receive alerts about significant upcoming changes to either.

How we built it

We built the ML model using widely used open-source tools: Python, Jupyter, Scikit-learn, Keras, and Tensorflow. Our interactive dashboard leveraged another open-source tool called Grafana. We used Adobe XD to prototype our web and mobile apps.

Challenges we ran into

Accomplishments that we’re proud of

What we learned

What’s next for Forecasting Solar Power Generation for NEOM using ML?

Predicting the future or electricity demand at NEOM!

Built With

Try it out