Machine learning–assisted prediction of heat fluxes through thermally anisotropic building envelopes

Building and Environment, Год журнала: 2023, Номер 234, С. 110157 - 110157

Опубликована: Март 1, 2023

Язык: Английский

---

Prediction of Energy Consumption of an Administrative Building using Machine Learning and Statistical Methods

Civil Engineering Journal, Год журнала: 2023, Номер 9(5), С. 1007 - 1022

Опубликована: Май 1, 2023

Energy management is now essential in light of the current energy issues, particularly building industry, which accounts for a sizable amount global use. Predicting consumption great interest developing an effective strategy. This study aims to prove outperformance machine learning models over SARIMA predicting heating usage administrative Chefchaouen City, Morocco. It also highlights effectiveness with limited data size training phase. The prediction carried out using (artificial neural networks, bagging trees, boosting and support vector machines) statistical methods (14 models). To build models, external temperature, internal solar radiation, factor time are selected as model inputs. Building simulation conducted TRNSYS environment generate database validation models. models' performances compared based on three indicators: normalized root mean square error (nRMSE), average (MAE), correlation coefficient (R). results show that all studied have good accuracy, 0.90 < R 0.97. artificial network outperforms other (R=0.97, nRMSE=12.60%, MAE= 0.19 kWh). Although methods, general terms, seemingly outperform it worth noting reached accuracy without requiring too much Doi: 10.28991/CEJ-2023-09-05-01 Full Text: PDF

Язык: Английский

---

BIM Integration with XAI Using LIME and MOO for Automated Green Building Energy Performance Analysis

Energies, Год журнала: 2024, Номер 17(13), С. 3295 - 3295

Опубликована: Июль 4, 2024

Achieving sustainable green building design is essential to reducing our environmental impact and enhancing energy efficiency. Traditional methods often depend heavily on expert knowledge subjective decisions, posing significant challenges. This research addresses these issues by introducing an innovative framework that integrates information modeling (BIM), explainable artificial intelligence (AI), multi-objective optimization. The includes three main components: data generation through DesignBuilder simulation, a BO-LGBM (Bayesian optimization–LightGBM) predictive model with LIME (Local Interpretable Model-agnostic Explanations) for prediction interpretation, the optimization technique AGE-MOEA address uncertainties. A case study demonstrates framework’s effectiveness, achieving high accuracy (R-squared > 93.4%, MAPE < 2.13%) identifying HVAC system features. resulted in 13.43% improvement consumption, CO2 emissions, thermal comfort, additional 4.0% gain when incorporating enhances transparency of machine learning predictions efficiently identifies optimal passive active solutions, contributing significantly construction practices. Future should focus validating its real-world applicability, assessing generalizability across various types, integrating generative capabilities automated

Язык: Английский

---

Progress in artificial intelligence-based visual servoing of autonomous unmanned aerial vehicles (UAVs)

International Journal of Thermofluids, Год журнала: 2024, Номер 21, С. 100590 - 100590

Опубликована: Янв. 28, 2024

Unmanned aerial vehicles (UAVs) have attracted massive attention in many engineering and practical applications the last years for their characteristics operation flexibility. For UAV system, suitable control systems are required to operate appropriately efficiently. An emerging technique is visual servoing utilizing onboard camera inspecting UAV's environment autonomously controlling operation. Artificial intelligence (AI) techniques widely deployed of autonomous applications. Despite increasing research field AI-based systems, comprehensive review articles that showcase general trends future directions this limited. This work comprehensively examines application advancements AI-enhanced covering critical tasks offering insights into enhancing performance applicability which limited current literature. The paper first reviews intelligent executing various tasks, including 3D positioning, ground object following, obstacle avoidance, landing. Second, progresses applying AI discussed analyzed. Finally, gaps further improving included.

Язык: Английский

---

Hybrid forecasting model of building cooling load based on combined neural network

Energy, Год журнала: 2024, Номер 297, С. 131317 - 131317

Опубликована: Апрель 13, 2024

Язык: Английский

---

A protocol for developing and evaluating neural network-based surrogate models and its application to building energy prediction

Renewable and Sustainable Energy Reviews, Год журнала: 2024, Номер 193, С. 114283 - 114283

Опубликована: Янв. 9, 2024

Because of their low computational costs, surrogate models (SMs), also known as meta-models, have attracted attention simplified approximations detailed simulations. Besides conventional statistical approaches, machine-learning techniques, such neural networks (NNs), been used to develop models. However, based on NNs are currently not developed in a consistent manner. The development process the is adequately described most studies. There may be some doubt regarding abilities due lack documented validation. In order address these issues, this paper presents protocol for systematic NN-based and how procedure should reported justified. covers model sample generation, data processing, SM training validation, report implementation, justify modeling choices. critically review quality SMs prediction building energy consumption. Sixty-eight papers reviewed, details summarized. developing procedures were evaluated using criteria proposed protocol. results show that selection number neurons best-implemented step with justification, followed by determination architecture, mostly justified discussion way. While greater focus given dataset especially input variables selection, considering independence check clear validation test data. Also, preprocessing strongly recommended.

Язык: Английский

---

Challenges and opportunities in European smart buildings energy management: A critical review

Renewable and Sustainable Energy Reviews, Год журнала: 2024, Номер 199, С. 114472 - 114472

Опубликована: Май 15, 2024

The substantial stock of European buildings, accounting for more than 40% energy consumption, has prompted member states to establish a renovation standard with stringent performance criteria. As advancing into the era digital transformation, concept smart buildings emerges as solution create sustainable, efficient, resilient, active, and comfortable living working spaces. This is achieved through intelligent resource use optimization, including management production, storage distribution systems. Smart operate by harnessing monitoring data leveraging artificial intelligence algorithms big techniques. integration contextual information, such building information modelling, physics, or simulation models, enhances resources. Moreover, incorporation metrics like readiness indicator promotes adoption buildings. study delves significance these techniques, expanding on existing research in field It integrates concepts enrichment, smartness, user-centric approaches. Key findings provide insights future opportunities within sector, emphasizing need user awareness strategies, development new algorithms, services that incorporate indicator. also advocates widespread twins.

Язык: Английский

---

Optimizing building energy performance predictions: A comparative study of artificial intelligence models

Journal of Building Engineering, Год журнала: 2024, Номер 88, С. 109247 - 109247

Опубликована: Апрель 4, 2024

Язык: Английский

---

Novel design of recurrent neural network for the dynamical of nonlinear piezoelectric cantilever mass–beam model

The European Physical Journal Plus, Год журнала: 2024, Номер 139(1)

Опубликована: Янв. 3, 2024

Язык: Английский

---

Future energy insights: Time-series and deep learning models for city load forecasting

Applied Energy, Год журнала: 2024, Номер 374, С. 124067 - 124067

Опубликована: Авг. 3, 2024

Most of the utility meters in Sweden are now integrated with Internet Things (IoT) technology. This modern approach significantly enhances our understanding energy consumption patterns and empowers consumers detailed insights into their power usage. Additionally, it provides companies grid owners critical data to facilitate future production planning. However, having at disposal is only half battle won. The method employed forecast equally important due complex interplay between long-term trends, seasonal fluctuations, other unpredictable factors. To optimally utilize this data, we analyzed several robust time-series forecasting models: Random Forest, XGBoost, SARIMAX, FB Prophet, a Convolutional Neural Network (CNN). Each these models was chosen for its unique strengths capturing trends short-term variations, making them appropriate candidates predicting consumption. We showcase models' performance on from commercial property 2021 evaluate based key metrics such as mean squared error (MSE), root (RMSE), absolute (MAE), percentage (MAPE), Relative Root Mean Square Error (RRMSE), Coefficient determination (R2), Standard Deviation (SD). Our results demonstrate that while ability effectively factor external parameters price temperature, fared well aggregated consumption, outperformed by CNN classifier. model demonstrated exceptional prediction capabilities flexibility adding additional features model. For example, highest accuracy showed lowest MSE compared Prophet reductions 75.70%, 69.48%, 49.45%, 30.62%, respectively. superior R2 values, indicating better fit data. Specifically, value 0.93% training set 0.60% testing set, outperforming terms explained variance. also utilized AutoML analyze 4-year dataset (2021–2023) generalizability models. Using AutoML, increased 47% 83% an expanded dataset, will achieve results. From qualitative perspective, contrary prevailing notion deep learning demand substantial resources, experience revealed did not pose greater challenges than traditional reinforces untapped potential forecasting, highlighting problems like electricity forecasts may benefit advanced solutions CNN.

Язык: Английский

---