Hi everyone. My name is Manuel Jesús Rodríguez. I’m currently one of the software developers in Liquid Studio for SAP Leonardo from Accenture Madrid, Spain. We’ve been working on some very interesting new projects and would like to share a bit of them with you all.
I’m going to talk you about our Intelligent Climate project, a solution for self-controlling HVAC (heating, ventilation and air conditioning) systems with pioneering technologies such as machine learning and IoT communication. It’s not intended to be a deep break-down of the solution, but hopefully you’ll have a sharp image of what we’ve been doing.
In this project we have three main parts: Java software, SCP Integration and User Experience. All of them are connected but are also independent, making up a structure of microservices that helps scalability to take part.
Our Java software is connected directly to the provider’s gateway, which uses standard BACnet protocol to communicate and take control over all HVAC devices. It’s also connected to our SAP Hana database and SAP Internet of Things Foundation in SAP Cloud Platform, which allows it to manage all commands and rules coming from the different sources and write them into the correspondent HVAC device.
Our Hana environment has a R Virtual Machine, which runs machine learning algorithms to make predictions on how the HVAC devices should behave. All these algorithms and other Hana features and procedures are run by XSJS services, set up by jobs when necessary.
The User Experience is developed to bring the user a pleasant and easy journey when checking and managing the system. We've taken advantage of the SAPUI5 framework, building a responsive web app that offers comprehensive information to the user and also includes a dashboard where the user can directly command some operations to the devices. Those command operations can be done by a conversational assistant as well, thanks to the newly incorporated Conversational Artificial Intelligence (Recast-ai) and even be sent by Telegram, thanks to its integration with Recast-ai. Commands arrive to our Java software thanks to the connection to the IoT foundation.
Most of HVAC systems follow a very traditional approach when ruling the environmental management programs, based on hard-coded rules that command the whole system in an inflexible fashion.
In our project, we have developed a machine learning model powered by R engine that takes real-time data ingestion, such as weather forecast, current office temperature, people influx and more, to predict how the HVAC devices should behave.
This algorithm is based on three main features:
Model Predictive Control algorithms are an effective method to improve building energy efficiency. This new “how-to” is the real innovation and the key feature that really makes it when it comes to save money.
All commands and other communications between the HVAC devices and the different parts of the project are possible thanks to IoT Foundation and its services.
The main feature of the IoT foundation service, in terms of communications, is to send and/or receive commands, and send measures to its IoT cloud storage system. Once the service is properly set up in its cockpit (including the broker in the software that is going to receive), it is possible and easy to send commands calling a simple API that the foundation kindly provides, as well as writing measures in the virtual devices created in the IoT cockpit.
These services allow us to grant the HVAC devices with rapid communications, that makes any changes made from any source almost real-time, without any delay or lag.
We’ve developed a front-end application that offers a real-time status for every HVAC device registered in the system. In its main page a total percentage of the savings in time is shown, as well as a heat map where they can be easily located and have a quick first-sight of the temperature in the place.
Also, in subsequent sections of the web app, users can access a bunch of graphical charts with real-time statistics and history data, so they can compare the correct work of the R algorithm.
There’s a dashboard to change any devices’ property, such as set point, work mode, status and so on. The user have the option of talking to a brand new conversation assistant, powered by Recast-ai (recently acquired by SAP, and named “Conversational Artificial Intelligence") and make quick changes based on the location and other information given to the assistant.
It’s even possible to have a conversation with the assistant by Telegram, which we’ve implemented and works perfectly.
All industrial spaces with implemented HVAC systems or enterprises that need to manage smart buildings/spaces can improve their energy efficiency and save in this unavoidable cost. Out of all the concepts and advantages that this solution brings, three key ideas can be extracted:
Our solution was presented in the 10th annual event for R users, which took place in Murcia (Spain), obtaining a wide approval from the attendance. Additionally, this project was nominated to the SAP Innovation Awards and the "El Español" Spanish newspaper, and is receiving huge support within Accenture global.
That's all folks! Please let me know in the comments what you think about this project. If you have any questions, I'll try to answer asap 😉
Thank you so much for your time!
I’m going to talk you about our Intelligent Climate project, a solution for self-controlling HVAC (heating, ventilation and air conditioning) systems with pioneering technologies such as machine learning and IoT communication. It’s not intended to be a deep break-down of the solution, but hopefully you’ll have a sharp image of what we’ve been doing.
Architecture: main solution and key features
In this project we have three main parts: Java software, SCP Integration and User Experience. All of them are connected but are also independent, making up a structure of microservices that helps scalability to take part.
Our Java software is connected directly to the provider’s gateway, which uses standard BACnet protocol to communicate and take control over all HVAC devices. It’s also connected to our SAP Hana database and SAP Internet of Things Foundation in SAP Cloud Platform, which allows it to manage all commands and rules coming from the different sources and write them into the correspondent HVAC device.
Our Hana environment has a R Virtual Machine, which runs machine learning algorithms to make predictions on how the HVAC devices should behave. All these algorithms and other Hana features and procedures are run by XSJS services, set up by jobs when necessary.
The User Experience is developed to bring the user a pleasant and easy journey when checking and managing the system. We've taken advantage of the SAPUI5 framework, building a responsive web app that offers comprehensive information to the user and also includes a dashboard where the user can directly command some operations to the devices. Those command operations can be done by a conversational assistant as well, thanks to the newly incorporated Conversational Artificial Intelligence (Recast-ai) and even be sent by Telegram, thanks to its integration with Recast-ai. Commands arrive to our Java software thanks to the connection to the IoT foundation.
Innovation: machine learning – data prediction in R
Most of HVAC systems follow a very traditional approach when ruling the environmental management programs, based on hard-coded rules that command the whole system in an inflexible fashion.
In our project, we have developed a machine learning model powered by R engine that takes real-time data ingestion, such as weather forecast, current office temperature, people influx and more, to predict how the HVAC devices should behave.
This algorithm is based on three main features:
- Model-based reinforcement learning: supervised learning model and
optimization algorithms / model-free linear regression. - Environment: learning from history to predict indoor temperatures and energy consumption using Gradient Boosting and LSTM (Long short-term memory networks).
- Optimization: three different algorithms are being used to optimize the climate conditions: Ant colony optimization; Fuzzy logic and Q-learning.
Model Predictive Control algorithms are an effective method to improve building energy efficiency. This new “how-to” is the real innovation and the key feature that really makes it when it comes to save money.
IoT: boosted communications
All commands and other communications between the HVAC devices and the different parts of the project are possible thanks to IoT Foundation and its services.
The main feature of the IoT foundation service, in terms of communications, is to send and/or receive commands, and send measures to its IoT cloud storage system. Once the service is properly set up in its cockpit (including the broker in the software that is going to receive), it is possible and easy to send commands calling a simple API that the foundation kindly provides, as well as writing measures in the virtual devices created in the IoT cockpit.
These services allow us to grant the HVAC devices with rapid communications, that makes any changes made from any source almost real-time, without any delay or lag.
User Experience: integrating services
We’ve developed a front-end application that offers a real-time status for every HVAC device registered in the system. In its main page a total percentage of the savings in time is shown, as well as a heat map where they can be easily located and have a quick first-sight of the temperature in the place.
Also, in subsequent sections of the web app, users can access a bunch of graphical charts with real-time statistics and history data, so they can compare the correct work of the R algorithm.
There’s a dashboard to change any devices’ property, such as set point, work mode, status and so on. The user have the option of talking to a brand new conversation assistant, powered by Recast-ai (recently acquired by SAP, and named “Conversational Artificial Intelligence") and make quick changes based on the location and other information given to the assistant.
It’s even possible to have a conversation with the assistant by Telegram, which we’ve implemented and works perfectly.
Conclusion
All industrial spaces with implemented HVAC systems or enterprises that need to manage smart buildings/spaces can improve their energy efficiency and save in this unavoidable cost. Out of all the concepts and advantages that this solution brings, three key ideas can be extracted:
- What makes this solution different from others is the integration of SAP Cloud Platform and Leonardo as a response to the immediate need for solutions that reduce the environmental impact without compromising comfort.
- It’s been proven that an energy efficiency concept solution can be integrated in SAP package with cutting-edge technology, such as machine learning and IoT.
- Preliminary results show implementing this system will represent an energy consumption saving between 20%-40% and ROIs in less than one year depending on the hardware physical installation, without lowering comfort standards.
Our solution was presented in the 10th annual event for R users, which took place in Murcia (Spain), obtaining a wide approval from the attendance. Additionally, this project was nominated to the SAP Innovation Awards and the "El Español" Spanish newspaper, and is receiving huge support within Accenture global.
That's all folks! Please let me know in the comments what you think about this project. If you have any questions, I'll try to answer asap 😉
Thank you so much for your time!
- SAP Managed Tags:
6 Comments
