Solar Wise Energy Forecast

Inspiration

The problem: Renewable energy sources like solar panels are highly dependent on weather conditions, making their energy output variable and therefore challenging to predict accurately. This variability can strain energy grids and affect the stability of power supply.

The Solution: The Solar Wise Energy Forecast, a cutting-edge machine learning algorithm harnessing the power of advanced AI to predict solar energy production based on real-time weather conditions, illuminating a more sustainable and efficient energy landscape. Accurate green energy forecasting can help energy grid operators efficiently manage the integration of renewable energy sources, reduce reliance on fossil fuels, and minimize energy waste. The program could contribute to a more sustainable energy ecosystem and reduce greenhouse gas emissions.

What it does

Solar Wise provides accurate solar energy production forecast in the next 14 days based on real-time weather conditions.

How we built it

Data Collection

We gathered historical data on weather conditions and corresponding energy production data from solar energy farms.

Data Processing & Features Engineering

We decided to build our algorithm based on the following relevant features from the weather data that could impact solar energy production: cloud cover duration, visibility, dew point, humidity, windspeed, and solar energy.

Weather Forecast

We leveraged weather predictions obtained from a Visual Crossing's API, which supplies detailed weather forecasts for the upcoming 14 days. Additionally, we explored the possibility of incorporating Google DeepMind's GraphCast, a state-of-the-art AI model just launched earlier this week that is capable of delivering exceptionally accurate 10-day weather predictions in under one minute. GraphCast offers advanced capabilities such as predicting the tracks of cyclones and identifying atmospheric rivers associated with flood risk, which could provide early warnings of extreme weather events. However, due to constraints in computing resources, we were unable to integrate GraphCast into our current model. This remains a promising avenue for future enhancements to our forecasting capabilities.

Prediction of Solar Energy Production

Armed with accurate weather forecasts, we leveraged machine learning models to predict solar energy production based on various weather conditions. Our model underwent rigorous testing with multiple algorithms, including Linear Regression, Long-Short Term Memory, and AdaBoosting. Remarkably, our AdaBoosting model achieved an impressive R-squared value of 96%, providing a reliable prediction of solar energy production, essential for informed energy management in Boston. For scalability, we plan to transition to Google Cloud's AutoML model training and deployment tool, VertexAI, as our dataset grows. Despite initial attempts, we encountered limitations due to the dataset's size.

Visualization and Reporting

To make our predictions accessible to users, we developed a user-friendly Streamlit application. This application allows users to input their location (address, city, zipcode, etc.) and receive a 14-day forecast of solar energy production. The prediction comes from the Visual Crossing API. Utilizing our prediction model and API calls, we retrieve real-time and forecasted weather data and calculate the anticipated solar energy production. Furthermore, our application (Power BI dashboard) calculates the potential carbon emission reductions and cost savings achieved by transitioning to solar energy, bringing the climate benefits of renewable energy adoption closer to home.

Challenges we ran into

We tested a couple of different prediction algorithms, from ADAboost to linear regression to a complex neural network, and had to play around with hyperparameters to result in an effective algorithm. After settling on ADAboost, we had some issues when our data was printing inaccurately on Streamlit. Again, we slightly tweaked our ADAboost based model to result in a realistic prediction.

Accomplishments that we're proud of

We take pride in several key accomplishments achieved through Solar Wise Energy Forecast. We're proud to have proposed a data-driven solution that take into consideration the most emerging AI technologies. Specifically, GraphCast is a new tool that was launched just a few days ago by Google DeepMindFirst and foremost, our AdaBoost model achieved an impressive R-squared value of 96%, showcasing the reliability of our solar energy production predictions. We are also proud of the development of a user-friendly Streamlit application, which makes our predictions accessible to users, allowing them to receive real-time forecasts and calculate potential carbon emission reductions and cost savings associated with transitioning to solar energy.

What we learned

Throughout the development process, we gained invaluable insights into the intricacies of green energy forecasting and its vital role in achieving a sustainable future. We also learned new algorithms and new tools individually as well as collectively such as Streamlit, Google Cloud's AutoML, Power BI, and Canva.

What's next for Solar Wise Energy Forecast

Looking ahead, our vision for Solar Wise Energy Forecast includes several exciting steps. Firstly, we plan to transition to Google Cloud's AutoML model training and deployment tool, VertexAI, to accommodate the growth of our dataset and further enhance the accuracy of our predictions. Additionally, we aim to explore opportunities to integrate GraphCast's advanced weather predictions, which could provide early warnings of extreme weather events and contribute to the climate resilience efforts. We also want to integrate the Power BI dashboard into Streamlit so that users can view the details of their energy consumption, production prediction, and how much financial and environmental impacts they make year-to-date. Ultimately, we aspire to expand our project's reach and impact by collaborating with institutions, local communities, energy providers, and policymakers to promote the adoption of renewable energy sources and a more sustainable future.