Our partner #EnTune optimizes the internal climate of commercial buildings - significantly saving energy and cost while reducing #CO2 emission. The software for this “EnTune BuildingAI” uses machine learning algorithms and physics models to optimally predict energy usage and automatically adjust settings. We built the #AI solution that powers this software and saves up to 30% in energy costs. 

Azure Functions: fast, easy, and self-healing

Every building is unique in terms of thermal characteristics and heating/cooling dynamics. It can use many different systems, and those develop over time; so scalability and modularity is key. 
Therefore, we designed a "service bus"-like architecture, such that we can decouple the different parts of the solution and adjust any part without impacting the whole architecture.

We built the processing core with Azure Functions, providing processing and orchestration capability. This Azure building block supports event-driven triggers, to connect to other services without having to write extra code. It is a serverless solution, allowing code to be deployed and executed without the need for a server. 

As gathered data, e.g. ambient temperature, weather data, personal preferences, or energy prices, comes in bulk and needs continuous analysis, we chose Azure Blob Storage combined with Apache Parquet files which we partition to allow for easy data management.
All data is analyzed in AI-powered software and processed into an energy plan that minimizes cost within the boundaries that define a comfortable indoor climate. 

Dynamic data is handled by machine learning algorithms to predict future values. For example, the number of people in the building on a typical day.
More static data, for example the surface area of windows, is processed using physics models - to predict how much energy will be needed to keep the internal atmosphere comfortable.

Before the software triggers an energy system to heat or cool, the AI model checks the latest weather forecast and energy prices and then can decide to postpone or bring forward any adjustments. 
To constantly assess the general health of the solution, we used Azure Application Insights including the integrated alerting functionality. And by configuring retry-policies and dead-letter queues to store failed messages, we made the software more resilient.

An AI-assistant that reduces cost and optimizes your indoor climate 

This AI-powered solution was built with several #Azure building blocks, that integrate easily and match the desired functionality perfectly. This reduced the need for coding and allowed us to integrate the solution within a few months.