Energy Systems Modeling

Excited to share our new review article, which synthesizes evidence across disciplines to understand the #psychological and #contextual determinants of household clean #energy adoption. 🏠 Our review focuses on high-impact technologies with major climate change mitigation potential: • Electric vehicles • Residential solar PV systems • Heat pumps • Battery storage systems Despite technical progress and policy support, adoption rates remain lower than expected in most climate scenarios. Our review discusses why and what can be done to accelerate uptake. 🔎 Some takeaways from the review: 1️⃣ Psychological and contextual factors jointly shape adoption decisions. Decisions to adopt EVs, PV systems, heat pumps, and battery storage are influenced by cognitive biases (e.g., loss aversion, temporal discounting), motivational factors (e.g., values, worldviews, identity), and social influences (e.g., norms, peer behavior, symbolic meaning). These operate alongside and interact with structural conditions, such as income, infrastructure, and policy. 2️⃣ Consumer behavior deviates systematically from techno-economic assumptions. Standard energy models often assume rational utility maximization. However, real-world adoption is more complex and influenced by a wealth of factors, including bounded rationality, information misperceptions, and emotional responses—all of which can distort cost–benefit assessments and delay adoption, even when technologies are financially advantageous. 3️⃣ Contextual heterogeneity is critical but often under-addressed. Socio-demographic characteristics, geographic setting, institutional design, and cultural values significantly impact the likelihood of adoption and the effectiveness of interventions. Yet, much of the empirical evidence remains concentrated in high-income contexts. Broader cross-cultural and field-based studies are urgently needed. 4️⃣ Effective interventions must be both targeted and integrated. Standalone approaches—whether economic (e.g., subsidies) or behavioral (e.g., social norm messaging)—often fall short because they address only a subset of the barriers to adoption. We need intervention portfolios that are strategically matched to specific psychological and contextual determinants. Read-only (free) link: https://rdcu.be/epJUS Published version (paywalled): https://lnkd.in/dKSJZABs (I'm happy to share the published version - just message me.) Huge thanks to Anne Günther and Lukas Engel for leading this collaborative effort and to co-authors Matthew Hornsey, Joyashree Roy, Linda Steg, Kim-Pong Tam, Anne van Valkengoed, Kim Wolske, Gabrielle Wong-Parodi, and Ulf Hahnel! Copenhagen Business School University of Basel Environmental Psychology Groningen Centre for Sustainability International Energy Agency (IEA) #climatechange #behaviorchange

Perspective: Large language models for building energy applications: Opportunities and challenges Large language models (LLMs) are gaining attention due to their potential to enhance efficiency and sustainability in the building domain, a critical area for reducing global carbon emissions. Built on transformer architectures, LLMs excel at text generation and data analysis, enabling applications such as automated energy model generation, energy management optimization, and fault detection and diagnosis. These models can potentially streamline complex workflows, enhance decision-making, and improve energy efficiency. However, integrating LLMs into building energy systems poses challenges, including high computational demands, data preparation costs, and the need for domain-specific customization. This perspective paper explores the role of LLMs in the building energy system sector, highlighting their potential applications and limitations. The authors propose a development roadmap built on in-context learning, domain-specific fine-tuning, retrieval augmented generation, and multimodal integration to enhance LLMs’ customization and practical use in this field. This paper aims to spark ideas for bridging the gap between LLMs capabilities and practical building applications, offering insights into the future of LLM-driven methods in building energy applications. Details of this Perspective can be found at https://lnkd.in/gbdMArEe, authored by Mingzhe Liu, Liang Zhang, Jianli Chen, Wei-An Chen, Zhiyao YANG, L. James Lo, Jin Wen & Zheng O'Neill #buildingsimulation #Perspective #LLM #FDD #Building #LLMasagent 

⚡ Hydrogen Production Facilities — A New Frontier in Power System Modelling As the world accelerates toward net-zero emissions, hydrogen production is emerging as a critical load segment within the power grid. 🌍 But unlike traditional industrial loads, electrolyzer-based hydrogen facilities introduce a new class of electrical behavior that power system planners and modelers must now capture with precision. 💧 1️⃣ The Process Behind the Load Hydrogen production facilities using electrolysis convert electrical energy directly into chemical energy by splitting purified water (H₂O) into hydrogen and oxygen. Before this, water is treated and deionized to remove dissolved solids, organic materials, and other impurities — ensuring the required purity for the electrolyzer stacks. Each electrolyzer unit typically operates in the 5–10 MW range, and large facilities can host dozens or even hundreds of such modules — aggregating into multi-gigawatt loads. This makes them one of the largest emerging controllable loads in modern power systems. 🏭 2️⃣ Load Composition in a Hydrogen Facility The electrical demand in these facilities is not uniform — studies show: 👉 Nearly 85% of total power is consumed by power electronic converters feeding the electrolyzer stacks. 👉 The remaining 15% comes from motor-driven systems for hydrogen compression, water treatment, cooling, and auxiliary equipment. This unique mix of converter-based and motor-based loads means that hydrogen facilities interact with the grid very differently from conventional industrial consumers. 🔌 3️⃣ Modelling in Power System Simulations From a simulation standpoint, electrolyzer-based hydrogen plants introduce several important considerations: ✅ Dynamic Response: The converter-fed electrolyzers can show fast and nonlinear power variations, depending on hydrogen demand and system voltage conditions. ✅ Voltage Sensitivity: Their performance is strongly tied to the DC bus stability and AC supply voltage, making voltage dip or frequency deviation studies critical. ✅ Grid Interaction: Large-scale hydrogen plants may act as demand response resources, ramping power up or down to balance renewable generation. ✅ Harmonics and Power Quality: Converter-dominated loads can influence harmonic distortion and reactive power flow, requiring detailed electromagnetic transient (EMT) or RMS-level models. 🧠 4️⃣ Representation in Load Models In simulations tools, hydrogen production facilities can be modelled as a hybrid load — combining: Converter-based static load models (for electrolyzers), and Induction motor dynamic models (for auxiliary drives and pumps). As facilities scale up, aggregated dynamic models become essential for evaluating grid stability, power quality, and control interactions — especially when multiple electrolyzer farms are connected near renewable hubs. Ref: NERC white paper on large loads

🔌 Modeling Inverter-Based Resources (IBRs) – The Key to a Reliable Grid: As renewable energy takes center stage, Inverter-Based Resources—including solar PV, wind turbines, and BESS—are transforming how power systems are planned and operated. But to ensure grid stability, reliability, and compliance, accurate IBR modeling is no longer optional—it’s essential. 📊 Why Modeling Matters: Simulates dynamic behavior during faults, frequency deviations, and voltage disturbances. Helps grid operators evaluate Fault Ride-Through (FRT), reactive power support, and grid-forming capabilities. Ensures compliance with grid codes (G99, IEEE 1547, NERC, etc.) before connection. 🛠 Key Modeling Inputs: OEM control models (PPC, PLL, AVR, Governor, LVRT/HVRT logics) Power plant layout & electrical parameters Protection schemes and ride-through curves Communication and SCADA integration details 🎯 Applications in Power Studies EMT simulations for transient stability Harmonic analysis and compliance checks Grid impact and hosting capacity assessments Black start and islanding performance evaluation 💡 Bottom Line: The better your IBR model, the more predictable your grid’s response—and the smoother your path to renewable integration. #RenewableEnergy #PowerSystemStudies #GridIntegration #BESS #WindEnergy #SolarEnergy #PowerSystems #IBR #EnergyTransition

Energy transition models can improve accuracy by up to 27% when they include social factors like public acceptance and investment risks. Energy system models have struggled to account for the human elements of transitioning to clean energy. While these models excel at technical and economic calculations, they often miss social dynamics that can make or break real-world implementation. A new study examines which societal factors matter most and how to include them effectively. By analyzing power system transitions across 31 European countries from 1990-2019, researchers found that incorporating societal factors improved model accuracy by up to 27% for predicting the installed capacity of individual technologies. Three factors emerged as particularly important: public acceptance of new energy infrastructure, investment risk considerations, and the tendency of existing infrastructure to create system lock-in. This research hints at new pathways for updating energy transition models. By systematically identifying which social factors matter most, modelers can better simulate how energy systems evolve - helping policymakers design more effective interventions. S/O to Vivien Fisch-Romito, Marc Jaxa-Rozen, Xin Wen , and Evelina Trutnevyte.

Enhancing occupant behavior representation for interoperability between building information modeling and building energy modeling. The Green Building XML (gbXML) schema facilitates data exchange between Building Information Modeling (BIM) and Building Energy Modeling (BEM) software tools. However, limited occupant behavior (OB) representation in BIM often leads to inconsistent and inaccurate energy simulation in BEM software. In this study, a critical set of enhancements have been identified and implemented in gbXML (a BIM schema) and obXML (an OB model schema) to improve representation of occupant behaviors for interoperability between BIM and BEM tools. Details are available at the free access article at https://rdcu.be/ewIpL

🏡 Have you ever come home early and your HVAC had no idea you were there? That gap between what occupants intend to do and what building systems assume is a major source of energy waste in residential buildings. Our new paper in Advances in Applied Energy addresses exactly this. We developed a framework that uses multimodal large language models to infer occupant intentions from multiple input signals and integrate that understanding directly into a model predictive controller for residential HVAC. The result: a system that doesn't just react to occupancy — it anticipates it. 📄 Energy-Efficient HVAC Control in Residential Buildings through Occupant Intention Inference with Multimodal Large Language Models. https://lnkd.in/gmAEA2XC This work reflects our group's ongoing mission to bring cutting-edge AI tools into practical building energy applications. Zixuan Qi, Zhiyao YANG, Mingzhe Liu #SmartBuildings #HVAC #OccupantIntentionInference #MultimodalAI #LLMs #EnergyEfficiency #BuildingControls #MPC #ResidentialEnergy J. Mike Walker '66 Department of Mechanical Engineering at Texas A&M University

Check out our new paper on agentic AI for energy operations! Shoutout to the coauthors Cong Chen and Ömer Karaduman Paper: https://lnkd.in/gw2MeQMi - The ideas is to use AI-powered LLM generative agents to simulate how "real" people manage home energy storage like batteries. We built a simulation framework (TARJ) with realistic prompts, memories modeling, and reactions. Instead of conventional simulations where agentic behaviors were abstracted into utility functions and Markov chains, in this case, we literally ask the agent what they want to do and why. - Our agents nail realistic human behavior mostly. In simple markets, they make near-optimal choices, but as complexity ramps up, their decisions get more varied; just like us. This type of scaling makes them viable for modeling real-world energy use / consumer behavior. - We gave agents distinct personas prompts —Thinker, Realist, Feeler. It turns out each makes unique decisions: Thinkers chase profits, Feelers prioritize safety, and Realists balance both. This diversity helps us design better, targeted energy programs: we can as, for instance, what could happen to a neighborhood after an outage in terms of usage patterns, given existing demographic data and past behavioral data? - But how do the agents react to sudden, tail events? When we threw blackouts into the mix, all agents adapted fast, but DIFFERENTLY. Feelers stockpiled energy for security, while Thinkers stayed profit-focused. These insights could shape emergency energy strategies. - Our agents don’t just "act"; they also explain "why". By analyzing their reasoning in the recorded memories, we uncovered distinct motivations, from profit-driven to emotionally guided. This is the type of interpretability that was until now entirely beyond the reach of conventional simulation models, and can really help craft policies that make sense with real people.

Model system performance and troubleshoot issues. If you run operations in energy, this is your daily reality.     Executives in service and maintenance don’t need more dashboards. We need systems that see the physics, predict the failure, and help us fix the right thing at the right time. The cost of getting this wrong is real: delays in licensing, spiraling maintenance expense, and plants that miss decarbonization targets.     Here’s the shift I see working: move from static models to a living, high‑fidelity digital twin. Not a diagram. A virtual reactor that evolves with the physical asset, predicts thermal behavior, and flags risks before they show up on the floor. In the material, you’ll see how reactor teams replaced one‑dimensional tools with 3D twins to resolve pebble‑level temperatures, validate turbulence models for liquid metal coolants, and cut validation time without building expensive demonstration units. That’s practical performance insight, not theory.     Why this matters for operations: a twin that captures real physics turns troubleshooting into a targeted plan. When hot spots and thermal striping are modeled accurately, you don’t over‑maintain or guess. You condition the system where it needs it, prove it to regulators with validated data, and keep output steady.     Try this today: pick one chronic issue your team sees repeatedly, temperature excursions, mixing anomalies, or vibration in assemblies. Build a minimal digital twin around that single behavior. Validate it with the highest‑fidelity data you have, then downgrade to faster models once accuracy is proven. Use it to guide maintenance windows and set thresholds you can defend.     If you’re accountable for uptime and safety, a living twin is the simplest way to see problems before they become events.  

Advancing BESS Safety Through Digital Twins: Modeling Thermal Runaway Propagation In the fast-growing battery energy storage system (BESS) sector, fire safety is a critical challenge for developers, manufacturers, insurers, and AHJs. A recent conversation spotlighted an innovative physics-based digital twin approach to model thermal runaway events and improve design and risk management. The core idea starts with detailed cell- and module-level data; temperature rise, pressure rise, and thermal runaway behavior from standard tests like UL 9540. This data becomes the foundation to simulate propagation from individual cells through modules, racks, and entire containers. Manufacturers can use the digital twin to virtually test thousands of scenarios, optimizing rack spacing, barrier thickness, vent areas, HVAC systems, flooring, and other parameters — reducing costly physical trial and error. Strong Validation Process Container-level predictions are rigorously validated against real thermocouple measurements, gas sensor data, and physical test results. Only after confirming accuracy at the container level does the model scale to site-level analysis. This is essential because full-scale testing of many containers is impractical. The model incorporates site-specific factors such as container spacing, fire-rated walls, wind direction, and historical wind speed data to evaluate propagation risk between units. Bridging Developers and Insurers A common industry tension exists around spacing. Developers want to minimize land use for better economics, while insurers often push for conservative distances (e.g., 25 feet) to limit property damage and liability, including potential off-site impacts. Digital twin modeling offers a science-based middle ground, providing quantitative, site-specific recommendations that can help balance safety, cost, and land efficiency. Insurance and risk engineers frequently ask detailed questions about model granularity, data inputs, validation methods, and plume movement. This approach delivers defensible insights for underwriting and project reviews. Path Forward Transparent validation and selective large-scale tests for adjacent containers will build confidence. By cutting the number of expensive physical iterations, digital twins can help manufacturers clear safety hurdles faster while giving insurers reliable, scenario-based evidence. Proactive use of this kind of analysis early in design, construction, or underwriting phases can lead to safer and more efficient BESS deployments across the industry. What are your thoughts? BESS manufacturers, developers, and risk engineering professionals; how are you currently addressing thermal runaway propagation and spacing decisions? Have you used digital twins or similar modeling tools?