Energy flexibility quantification of a tropical net-zero office building using physically consistent neural network-based model predictive control

Advances in Applied Energy, Journal Year: 2024, Volume and Issue: 14, P. 100167 - 100167

Published: Feb. 24, 2024

Building energy flexibility plays a critical role in demand-side management for reducing utility costs building owners and sustainable, reliable, smart grids. Realizing tropical regions requires solar photovoltaics storage systems. However, quantifying the of buildings utilizing such technologies has yet to be explored, robust control sequence is needed this scenario. Hence, work presents case study evaluate controls operations net-zero office Singapore. The utilizes data-driven quantification workflow employs novel model predictive (MPC) framework based on physically consistent neural network (PCNN) optimize flexibility. To best our knowledge, first instance that PCNN applied mathematical MPC setting, stability system formally proved. Three scenarios are evaluated compared: default regulated flat tariff, real-time pricing mechanism, an on-site battery (BESS). Our findings indicate incorporating into could more beneficial leverage decisions than flat-rate approach. Moreover, adding BESS PV generation improved self-sufficiency self-consumption by 17% 20%, respectively. This integration also addresses mismatch issues within framework, thus ensuring reliable local supply. Future research can proposed PCNN-MPC different types.

Language: Английский

---

Hydrogen jet and diffusion modeling by physics-informed graph neural network

Renewable and Sustainable Energy Reviews, Journal Year: 2024, Volume and Issue: 207, P. 114898 - 114898

Published: Sept. 6, 2024

Language: Английский

---

Biomass gasification modeling based on physics-informed neural network with constrained particle swarm optimization.

Energy, Journal Year: 2025, Volume and Issue: unknown, P. 135392 - 135392

Published: March 1, 2025

Language: Английский

---

Data-driven adaptive building thermal controller tuning with constraints: A primal–dual contextual Bayesian optimization approach

Applied Energy, Journal Year: 2024, Volume and Issue: 358, P. 122493 - 122493

Published: Jan. 9, 2024

We study the problem of tuning parameters a room temperature controller to minimize its energy consumption, subject constraint that daily cumulative thermal discomfort occupants is below given threshold. formulate it as an online constrained black-box optimization where, on each day, we observe some relevant environmental context and adaptively select parameters. In this paper, propose use data-driven Primal-Dual Contextual Bayesian Optimization (PDCBO) approach solve problem. simulation case single room, apply our algorithm tune Proportional Integral (PI) heating pre-heating time. Our results show PDCBO can save up 4.7% consumption compared other state-of-the-art optimization-based methods while keeping tolerable threshold average. Additionally, automatically track time-varying thresholds existing fail do so. then alternative where aim with budget. With formulation, reduces average by 63% safe required

Language: Английский

---

Long-term experimental evaluation and comparison of advanced controls for HVAC systems

Applied Energy, Journal Year: 2024, Volume and Issue: 371, P. 123706 - 123706

Published: June 22, 2024

Language: Английский

---

A review on hybrid physics and data-driven modeling methods applied in air source heat pump systems for energy efficiency improvement

Renewable and Sustainable Energy Reviews, Journal Year: 2024, Volume and Issue: 204, P. 114804 - 114804

Published: Aug. 14, 2024

Language: Английский

---

Disentangled Representation Aided Physics-Informed Neural Network for Predicting Syngas Compositions of Biomass Gasification

Energy & Fuels, Journal Year: 2024, Volume and Issue: 38(3), P. 2033 - 2045

Published: Jan. 9, 2024

Machine learning (ML) has been extensively studied and applied in the biomass gasification field currently. However, insufficient experimental data tends to cause a mismatch between ML model physical mechanism, particularly for feedstocks that do not appear training set, becoming significant challenge creating credible models gasification. Therefore, this study proposes disentangled representation-aided physics-informed neural network method, briefly called DR-PINN, predict syngas components. First, DR-PINN extracts latent variables represent feedstock properties through representation generates synthetic samples gasification-related variable space cover full range of types. Then, employs inequality constraints embed priori monotonic relationships into loss function. Finally, are simultaneously considered process realize synergy complementarity actual information existing knowledge using an evolutionary algorithm. As result, shows good prediction performance (the within set: R2 ≈ 0.96, root-mean-square error (RMSE) 1.7; outside 0.81, RMSE 3). Moreover, even with can strictly abide by prior relationships, consistency degree equal 1. Overall, proposed demonstrates superior generalization interpretability compared other methods, such as RF, GBR, SVM, ANN, PINN.

Language: Английский

---

Modularized neural network incorporating physical priors for future building energy modeling

Patterns, Journal Year: 2024, Volume and Issue: 5(8), P. 101029 - 101029

Published: July 19, 2024

Building energy modeling (BEM) is fundamental for achieving optimized control, resilient retrofit designs, and sustainable urbanization to mitigate climate change. However, traditional BEM requires detailed building information, expert knowledge, substantial efforts, customized case-by-case calibrations. This process must be repeated every building, thereby limiting its scalability. To address these limitations, we developed a modularized neural network incorporating physical priors (ModNN), which improved by model structure heat balance equations, physically consistent constraints, data-driven modular design that can allow multiple-building applications through sharing inheritance. We demonstrated scalability in four cases: load prediction, indoor environment modeling, retrofitting, optimization. approach provides guidance future into models without extensive paving the way large-scale BEM, management, buildings-to-grid integration.

Language: Английский

---

A review of physics-informed machine learning for building energy modeling

Applied Energy, Journal Year: 2024, Volume and Issue: 381, P. 125169 - 125169

Published: Dec. 21, 2024

Language: Английский

---

Opening the Black Box: Towards inherently interpretable energy data imputation models using building physics insight

Energy and Buildings, Journal Year: 2024, Volume and Issue: 310, P. 114071 - 114071

Published: March 15, 2024

Missing data are frequently observed by practitioners and researchers in the building energy modeling community. In this regard, advanced data-driven solutions, such as Deep Learning methods, typically required to reflect non-linear behavior of these anomalies. As an ongoing research question related Learning, a model's applicability limited settings can be explored introducing prior knowledge network. This same strategy also lead more interpretable predictions, hence facilitating field application approach. For that purpose, aim paper is propose use Physics-informed Denoising Autoencoders (PI-DAE) for missing imputation commercial buildings. particular, presented method enforces physics-inspired soft constraints loss function Autoencoder (DAE). order quantify benefits physical component, ablation study between different DAE configurations conducted. First, three univariate DAEs optimized separately on indoor air temperature, heating, cooling data. Then, two multivariate derived from previous configurations. Eventually, thermal balance equation coupled last configuration obtain PI-DAE. Additionally, commonly used benchmarks employed support findings. It shown how enhance inherent model interpretability through physics-based coefficients. While no significant improvement terms reconstruction error with proposed PI-DAE, its enhanced robustness varying rates valuable insights coefficients create opportunities wider applications within systems built environment.

Language: Английский

---