STEM, INC. IMPROVES ENERGY OPTIMIZATION USING PREDICTIVE AI
In today’s global energy market, saving energy, reducing costs for customers, and utilising renewables as much as possible are keys to a successful energy strategy. Energy strategies can vary market to market.
For example, Ontario, Canada is a specific market with a “global adjustment” model requiring energy providers to predict on what hour of the day peak electricity demand occurs.
Knowledge about how the demand will regularly fluctuate is therefore crucial. However, it is challenging to predict how hundreds of millions of consumers will behave and how their energy consumption will affect the aggregate electricity demand on a national level.
Generally, electricity consumption has to be evenly matched with electricity generation at all times to ensure a safe, stable supply. Still, the demand for and production of electricity fluctuates heavily depending on different hours, days of the week, months, etc., with supply not always coinciding with the electricity grid demand.
Stem Inc., a global leader in AI-driven energy storage, approached Endava to deliver a study on the Global Adjustment model and a pilot project creating a program that would predict peak electricity demands. This program would be meant to lessen energy costs by predicting the most cost-effective moment to buy electricity and enable efficient usage during more expensive peak hours.
Together with Stem, we made an artificial intelligence (AI) roadmap so we could create a proof of concept and ultimately deliver a custom-made solution for forecasting electricity demand.
Endava’s domain experts team closely collaborated with Stem and to evaluate the right approach that would allow us to recommend the most suitable data sources, algorithm, and enhancements.
Thorough testing techniques were implemented throughout the whole project. Endava wrote unit tests to ensure all code met quality standards before deployments, component tests to verify the input and output behaviour of the forecasting solution, and integration tests to eliminate any faults in the interaction of integrated units. Moreover, Endava estimated the developed models’ predictive power through carefully selected and designed AI validation procedures.
Our world-class expertise in the fields of predictive AI, machine learning (ML), and data analytics were a big factor for this project, as well as the following:
■ High-performance cloud development
■ Building scalable microservice architecture running in the cloud and on Internet of Things (IoT) devices
■ Developing application programme interfaces (APIs) and RESTful endpoints (i.e., the end locations or touchpoints of API ‘requests’ to receive information via ‘responses’ of transferrable data or resources that help APIs carry out their functions)
“With this Global Adjustment peak prediction project, we wanted to create a pilot that would enable us to continuously innovate, lead in multiple marketplaces, fulfil our promise to save energy, reduce costs for our customers, and support integration of renewables on the grid. In Endava and its innovative and resourceful AI team, we collaborated with a partner who supported the data engineering within our AI-driven smart energy storage services.” – Larsh Johnson, CTO, Stem, Inc.
The solution for forecasting electricity demand was made to improve Stem’s fixed assets efficiency by charging the batteries during off-peak time and using the stored energy during peak time when the difference in electricity price is highest.
By using batteries for electricity storage effectively, wed created a model that could substantially lower the electricity cost for Stem’s customers. In some cases, overall electricity costs could lower by 60%.
Through accurate forecasting, we could cut down the battery charge cycles and charging time, improving their lifespan, thus reducing servicing costs and potentially improving Stem’s profit margins.
Balancing the pressure on the transmission and distribution network during peak time could help reduce the costs of electricity production during this period and increase the infrastructure longevity.
We also created more opportunity to optimize the grid storage technology, paving the way to a greater adoption of renewable electricity sources such as solar power plants, tidal energy, and wind.