Energy optimization at UBS using artificial intelligence

We are delighted that you are interested in our white paper. We will send you the white paper by e-mail.

Thank you for your interest. Please check your inbox for the download link to the whitepaper.

09.07.2024

UBS is pursuing the ambitious goal of reducing its operating costs through improved energy efficiency and at the same time achieving a higher rating for sustainability certificates. This was achieved through an innovative project that relies on the use of artificial intelligence (AI) to optimize the energy consumption of a total of 14 of the bank's existing buildings.

Project description

Part 1: Monitoring

The first step is to closely monitor energy consumption in the various UBS buildings. This involves the use of various, often heterogeneous building management systems (BMS), which have grown organically over the years. These systems provide inconsistent and often unstructured data, which impairs the consistency and quality of the data. Monitoring is used to establish a clear baseline of current energy consumption and identify areas with potential for improvement.

Part 2: Optimization with artificial intelligence

At the heart of the project is the optimization of energy consumption using AI, which is executed in the cloud. Eliona and UBS are working together with partner Dabbel, which creates a digital twin for each building. This digital twin makes it possible to virtually model and optimize the building and its energy flows, taking into account additional data (such as historical and current weather data, people counting). Dabbel's AI analyses the data, identifies inefficient patterns and calculates optimized target values that are sent back to the building management systems.

Data harmonization and semantic order

A central element of this project is the harmonization of incoming data streams. Due to the heterogeneous nature of the existing building management systems, some of which consist of different versions, the data quality is often unstructured. Eliona takes on the task of data harmonization by semantically organizing, tagging and translating the data. This requires intensive research in the field of data modeling and the development of an ontological and semantic data model.

The challenge is that building management systems are designed as closed systems and often do not allow external intervention. In order to enable comprehensive data integration and harmonization, Eliona had to develop innovative approaches to securely connect these closed systems and make the data accessible.

Technical challenges

Data points and scaling

A central technical problem is the optimal number of data points per building. While some AI models can work effectively with less than 1,000 data points per industrial building, others require up to 15,000 data points. It has been found that a higher number of data points does not necessarily lead to better results. The challenge lies in finding a balance between data quantity and quality of results.

Another problem is scaling. The planned international rollout envisages expanding the system to over 100 buildings. The number of data points varies considerably, which can affect the performance of the systems. Strategies need to be developed to scale the amount of data efficiently and optimize performance.

Security architecture

A key issue is the store & forward procedure in the event of connection interruptions, which ensures that no data is lost and that all actions on the edge nodes remain traceable (audit trail). Every interaction with the buildings must maintain the existing security chains, which poses additional technical and security challenges.

Proof of Concept (PoC)

The first step was to carry out a proof of concept (PoC) to test the technical interfaces and prove the functionality of the systems. The PoC showed that energy savings of up to 25% are possible in a single building. The final results for the buildings as a whole are expected in October.

Rollout and ongoing operation

The rollout strategy is divided into three phases:

Rollout phase 1: Implementation in 7 buildings.
Rollout phase 2: Expansion to a further 7 buildings in Switzerland.
International rollout: Long-term planning for cross-corporate monitoring.

In the long term, UBS plans to roll out the system internationally and introduce cross-corporate monitoring in order to optimize energy efficiency in all its buildings.

In the first three months of operation (mid-May to mid-July), savings totaling 27% of energy consumption have already been achieved. We are already looking forward to the energy-intensive winter months and what results can be achieved during this time of year.