Our first ever research paper

At Qubit Energy, we work to achieve our vision; sustainbilitu for humanty. To reach that goal, we've set our mission to be accelreting electrification of sociaty. Today, with the help AI, we can achivie many of our goals faster. Thogh in crosssection of energy, AI and building, there are many challenges, unkown paths and lack of data. Our work is therefore research based. In our lab, we are working on some the must cutting edge and pioniering solutions world wide. One them is Vertical AI Agents.

Multi-Modal Vertical Energy AI Agents

AI agents are proving to be great assistense in many sectors. We wanted to research on how they can provide value in energy optimalisation for buildings. We are still very early on in our work, but in order to achieve our vision and mission, we have chosen to be transparent and share our findings and knowledge with world. We are therefor excited to publish our ever first research paper. You can access it for free here:

https://drive.google.com/file/d/1u5VRHlt4iTEwRT9L7M05EL29uJ1k2HM9/view?usp=sharing

——-

Abstract—Energy optimization in buildings is a critical challenge in advancing sustainable urban development. This research introduces a novel AI agent system designed to optimize energy consumption in buildings by leveraging historic energy data, external variables such as weather, pricing, and sentimental data. The AI agent assists building managers in achieving energy efficiency by forecasting conditions and adjusting energy usage dynamically. Central to this system is a core module built on our Virtual Power Plant (VPP) platform, integrating memory modules, external resource connections, and a robust planning module. This paper outlines the architecture, methodologies, and initial results, demonstrating significant potential in reducing energy consumption and enhancing building energy management systems (BEMS).

INTRODUCTION

Energy consumption in buildings accounts for approximately 40% of global energy usage, with heating, ventilation, and air conditioning (HVAC) systems as primary contributors. Traditional energy management systems often fail to account for dynamic external and internal factors, leading to inefficiencies. This research explores the deployment of specialized vertical AI agents capable of analyzing historic data, detecting patterns, and integrating external variables to forecast and optimize energy usage in real time. These agents are designed for specific building operations, such as HVAC optimization, lighting control, and energy storage management, providing a modular and targeted approach to energy efficiency. We propose a modular AI system designed to function as an intelligent assistant for building managers. This system runs on Qubit OS, our operating system infrastructure platform with big data capabilities for LLMs and deep learning neural networks. Qubit OS is a pioneer model because its LLM is not only text-based but also more intelligent due to connected sensors from different types of IoT devices. Additionally, we incorporate Retrieval-Augmented Generation (RAG) into the model.

Retrieval-Augmented Generation (RAG): RAG is a hybrid approach combining pre-trained language models (LLMs) with external knowledge retrieval systems. Instead of relying solely on the internal knowledge of the LLM, RAG retrieves relevant data from historic records or external databases, improving the accuracy and relevance of generated outputs. In our system, RAG allows the AI to pull real-time insights from IoT data, weather updates, and historic energy consumption patterns. By integrating this technology, the model can make context-aware predictions and suggest optimized planning actions dynamically, based on the latest available data. This capability is particularly valuable in energy flexibility modules, where precise recommendations require real-time information. This system leverages the foundational capabilities of a Virtual Power Plant (VPP) platform, extending its utility to energy optimization within buildings.

II. RELATEDWORK

2.1 Building Energy Management Systems (BEMS) Existing BEMS primarily focus on static rule-based control systems. However, these systems lack adaptability to dynamic external variables such as weather or energy pricing.

2.2 AI in Energy Optimization Recent advancements in AI have introduced machine learning models for predicting energy consumption patterns. Yet, integration with external factors such as real- time weather data and market conditions remains underdeveloped.

2.3 Virtual Power Plants (VPPs) VPPs aggregate distributed energy resources to optimize energy distribution. Our research extends this concept, using VPPs as an operational system for localized energy optimization in buildings.