Forecasting Integrated Smart Storage Solutions

• 

  We used fish-eye camera, irradiance sensor and Raspberry Pi to create The Cheapest Solar Forecasting Device in the world!

• 

  We can forecast the solar power in the next 15 minutes to 24 hours with a average accuracy of 93%

• 

  For some buildings, solar alone is not enough to avoid expensive demand charge

• 

  We utilize forecast data and battery control algorithm to shave the peak and reduce demand charge

• 

  We would like to work with the city to put our forecasting system on every libraries in San Diego and provide forecast data for all!

• 

  We use machine learning program to identify the clouds and forecast the solar irradiance

Inspiration

Solar is an expensive investment by far, and the only thing that encourages people to install more solar panels, is bigger savings. However, for commercial solar users in San Diego who have unregulated load, solar energy can't really help to reduce their demand charges(San Diego has the highest demand charge rate throughout the nation). A well managed ,forecasting integrated micro-grid system with storage can effectively shave the peak and cut the demand charges with a relatively low cost compare to current micro grid-model.

What it does

We developed an micro-grid optimization algorithm that can foresee the next demand peak by forecasting the solar generation and building energy demand, and reduce the demand by utilizing the storage. The users of the algorithm may even use their EVs as the storage systems to save their budget.

How we built it

The algorithm contains three parts, solar forecasting, load forecasting, and battery control. We used machine learning method in our forecasting part so that the result can self-involve and become more precise over time. We used Matlab to write all the scripts.

Challenges we ran into

There are not enough data to show the solar panel number on a certain building, so we need to estimate a number. Matlab runs too slowly, we have to simplify the code for this Hackathon beause otherwise it will take forever to run the result out.

Accomplishments that we're proud of

We have developed the cheapest solar forecasting device. It makes solar forecasting becomes a subscribe data service instead of an expensive scientific measurement. Base on the forecasting and machine learning technology, we figured out a new way for non-residential users to save on their electricity bills.Our machines learning algorithm allows the users have flexibility on choosing any kinds of storage system that fit in their budget(Li-ion, Used EV battery, Lead-Acid or EV)

What we learned

Each energy user has a unique energy bill, we need to evaluate carefully before we decide what type of energy saving actions they need. And it is really important to collaborate with the city.

What's next for Forecast Integrated Smart Storage Solutions

We are going to work with the city to build our forecasting network in San Diego. And become a important data provider in solar industry. We are going to design a special EV charging station that runs on our Micro-grid control algorithm. And initiate a pioneer project to install the new charging station in buildings of San Diego.

Built With

• matlab
• python