Currents: AI & Energy Insights - April 2024
With today’s issue I’m excited to introduce Currents, a monthly column from Reimagine Energy dedicated to the latest news at the intersection of AI & Energy. Every last Sunday of the month you’ll be receiving an expert-curated summary of the most relevant updates from the sector. I’ll be focusing on major industry news, published scientific articles, a recap of the month’s posts from Reimagine Energy, and a dedicated job board.
1. Industry news
WattCarbon is launching the OpenEAC Alliance, designed to foster collaboration and standardization in the measurement and verification (M&V) of clean energy technologies. The alliance focuses on enhancing the quality and transparency of evaluating the climate benefits of decarbonizing buildings and addresses structural barriers that have undermined the proper valuation of these benefits. The kickoff call will be hosted on May 16, 2024 and you can find more info in the official post below.
What I’m thinking: Open source, standardized methodologies for clean energy M&V have the potential to unlock substantial investments in the sector, as I discuss in detail here. I’m looking forward to personally participating in the kickoff call!
Ento entered the AI-driven HVAC management market with our latest product, Ento Control. The solution leverages Model Predictive Control (MPC) to dynamically adjust HVAC systems based on predictive analytics—integrating weather forecasts, occupancy, and soon energy pricing, to optimize energy use and enhance comfort in real-time. Additional info here.
What I’m thinking: Within a few months this project went from inception to market release. I’m amazed by what can be achieved by a couple driven engineers and momentum from the market. Now it’s time to get the consumption of as many buildings as possible under control (pun intended).
BrainBox AI launched ARIA, an AI assistant that utilizes Large Language Models and proprietary algorithms to provide real-time operational insights and recommend strategic actions. The tool is tailored for facility managers and real estate professionals, requiring the installation of BrainBox AI technology to maximize its capabilities. Key features of ARIA include real-time insights, proactive alerts, energy efficiency recommendations, and advanced diagnostics, all aimed at enhancing building efficiency and sustainability. Read the full release notes here.
What I’m thinking: I’m excited to finally see LLM applications appearing in the energy and buildings sector. At the same time, I’m curious to see the rate of adoption in a field that has historically lagged behind in terms of technological advances.
During the first week of April, most of the European electricity markets experienced negative electricity prices. The negative prices were largely attributed to decreased average gas prices, and increased solar energy production. Temperature increases also contributed to these trends, in the form of reduced electricity demand. On April 7th, more than 6 GW of wind and solar energy were being curtailed in Spain, while the country was importing 3 GW from France due to even more negative energy prices in the rest of Europe. The analysis from Modo Energy, below, highlights the negative prices trend in Great Britain, with a significant increase in 2024.
What I’m thinking: The infamous California “duck curve” is coming to Europe as well. This opens up incredible value for storage and flexibility solutions. AI will be a key enabler in storage optimization and demand-side flexibility deployment, using real-time data and forecasts to balance supply and demand.
2. Scientific publications
It’s hard to keep track of everything that is released in the academic sector, but I always try to reserve some time to look at the most interesting publications of the month. Two articles from academia caught my attention in April:
Interpretable domain-informed and domain-agnostic features for supervised and unsupervised learning on building energy demand data by Canaydin et al.
Closing the energy flexibility gap: Enriching flexibility performance rating of buildings with monitored data by de-Borja-Torrejon et al.
Interpretable features for building energy demand
The authors compiled a large meta-dataset of energy demand in over 13,500 buildings from different countries in the world. They then applied feature extraction techniques transforming the high-resolution consumption time series from the buildings to a lower dimensional feature space. The methodology relied on interpretable feature extraction techniques from the Python packages Nixtla and IFEEL. The final feature matrix includes 130 features, representing a compression ratio of multiple orders of magnitude compared to the initial dataset. The extracted features were then used to classify buildings based on their geographical characteristics and forecastability.
What I’m thinking: This work opens up many interesting possibilities in the domain of building classification and data imputation. With a sufficiently diverse dataset, the extracted features could predict a range of building characteristics beyond geographic location—such as usage type, energy efficiency labels, and geometric profiles. Additionally, the inclusion of metadata such as the presence of solar installations or EV charging stations could further refine these predictions, adding layers of utility and applicability to the classification models. Another intriguing possibility is employing these features in transfer learning strategies, particularly for imputing missing data in time series. With higher smart meter data availability, building metadata will become increasingly more valuable over the next years.
Data-driven energy flexibility evaluation
This article provides a good overview of current state-of-the-art methodologies to quantify the Energy Flexibility of buildings. The authors compare a model-based methodology and a purely data-driven methodology to estimate the Flexibility Performance Indicator (FPI) in two different households that participated to a Demand Response program in the UK. The study shows that most of the Energy Flexibility metrics within the FPI rating method can be calculated using monitored data, without the need of creating an energy model of the building.
What I’m thinking: The study is definitely a step in the right direction. Quantifying flexibility using only smart meter data—eliminating the requirement for indoor temperature readings or complex energy models—will be a key enabler for the deployment of energy flexibility across thousands of buildings.
3. Reimagine Energy publications
The two Reimagine Energy articles of this month featured a case study and a code tutorial to verify energy efficiency savings in buildings with Machine Learning. Read them below!
Case Study: Using AI to Verify Savings Across 9,000 Buildings
Code Tutorial: Building a Counterfactual Energy Model for Savings Verification - Part 1
4. AI in Energy job board
This space is dedicated to job posts in the sector that caught my attention during the last month. I have no affiliation with any of them, I’m just looking to help readers connect with relevant jobs in the market.
Data Scientist Forecasting and Data Engineer Forecasting at Electricity Maps
Senior Data Scientist at BrainBox AI
Doctoral scholarship holder artificial intelligence for renewable energy at University of Antwerp
Conclusion
With so much going on in the sector it’s not easy to follow everything. If you’re aware of anything that seems relevant and should be included in Currents (job posts, scientific articles, relevant industry events, etc.) please reach out to me on LinkedIn and I’ll be happy to consider them for inclusion!