A phase change material-operation parameter integrated intelligent optimization method for low-carbon buildings
By using an integrated intelligent optimization method, the problem of disconnect between the selection of phase change materials and operating parameters in buildings has been solved, enabling rapid and accurate matching and dynamic adaptive control in different climate zones, thereby improving the robustness and optimization efficiency of building systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SOUTHEAST UNIV
- Filing Date
- 2026-01-22
- Publication Date
- 2026-06-09
AI Technical Summary
In existing technologies, the selection of phase change materials in buildings is disconnected from the optimization of operating parameters, resulting in insufficient trade-offs between objectives, inadequate generalization and robustness across climate zones, high computational costs, and difficulty in achieving rapid and accurate matching and coordinated optimization of operating strategies under multiple meteorological scenarios.
By constructing feature vectors, clustering algorithms, and Markov chains to generate representative meteorological sequences, a transient heat transfer physical model is established. Machine learning surrogate models and multi-objective optimization algorithms are used in conjunction with a digital twin system to achieve integrated intelligent optimization of phase change materials and operating parameters, enabling rapid and accurate parameter matching and strategy optimization. Dynamic adaptive control is achieved through model predictive control.
It enables precise and rapid application in different climate zones, reduces computing costs, improves optimization efficiency, enhances system robustness and real-time response capabilities, and ensures optimal operation of building systems in the face of weather changes and internal disturbances.
Smart Images
Figure CN122174309A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of building thermal technology, specifically to an integrated intelligent optimization method for phase change materials and operating parameters in low-carbon buildings. Background Technology
[0002] Buildings are a significant source of energy consumption and carbon emissions, accounting for approximately 30% of global final energy consumption and 26% of energy-related emissions. Improving the thermal performance of building envelopes is a key pathway to achieving building energy conservation. Solar energy has advantages such as wide distribution, easy access, and no pollution, making it an important renewable energy source for building heating, air conditioning, and other loads. However, its inherent intermittency and volatility lead to instability in the heat storage and release process of the building envelope, making it difficult to maintain a relatively constant indoor thermal environment over the long term.
[0003] To improve the thermal stability of building envelopes, researchers have introduced phase change materials (PCMs) into structures such as walls and roofs to passively store low- and medium-temperature thermal energy. However, the actual thermal performance of the building envelope is determined by the coupling of PCM properties and structural parameters (such as phase change temperature, latent heat, thermal conductivity, thickness, doping, and location), external meteorological parameters (solar radiation, outdoor temperature and humidity, wind speed, and sky conditions), and operational parameters (set temperature, ventilation, shading, and nighttime ventilation strategies).
[0004] Existing research generally suffers from the following limitations: (1) Limited selection methods: Phase change material parameters are mostly based on experience or single-scenario simulation trial and error, lacking synergistic optimization with operating parameters (such as set temperature, shading, ventilation / night ventilation); (2) Single objective: It is difficult to simultaneously consider energy consumption, thermal comfort, peak load, carbon emissions, and cost; (3) Insufficient generalization and robustness: Insufficient consideration of migration across multiple climate zones and meteorological / construction uncertainties; (4) High computational cost: A large number of high-fidelity simulations result in high costs and long cycles. Therefore, there is an urgent need for a general method that integrates and synergistically optimizes phase change materials and operating parameters under multiple meteorological scenarios, and possesses robustness and interpretability, so as to achieve rapid and accurate matching of phase change material parameters under different operating conditions and optimize them in conjunction with operating strategies. Summary of the Invention
[0005] Purpose of the invention: The purpose of this invention is to provide an integrated intelligent optimization method for phase change materials and operating parameters in low-carbon buildings, which solves the problems of disconnect between phase change material selection and operating parameters under different meteorological conditions, insufficient target trade-offs, and inadequate cross-climate generalization and robustness.
[0006] Technical solution: The present invention provides an integrated intelligent optimization method for phase change materials and operating parameters in low-carbon buildings, comprising the following steps:
[0007] (1) Obtain multi-year historical hourly meteorological data of the target building location, and perform cleaning and feature extraction to construct a feature vector to characterize the climate features;
[0008] (2) Based on feature vectors, clustering algorithms are used to classify complex meteorological data into multiple typical meteorological day categories, and representative meteorological sequences that retain typical day characteristics and conform to actual time series change patterns are generated based on Markov chain method;
[0009] (3) Establish a transient heat transfer physical model of the building envelope with integrated phase change material layer, and generate high-fidelity simulation samples covering design variables in batches through parametric scripts to form a training database;
[0010] (4) Define the physical properties and structural parameters of phase change materials and the operation control parameters of buildings as optimization variables, and construct a multi-objective optimization problem that simultaneously minimizes the building’s total annual energy consumption, peak load, thermal discomfort time, annual carbon emissions and life cycle cost.
[0011] (5) Use the training database to train the machine learning agent model and establish a fast mapping relationship from optimization variables to multiple performance indicators; adopt a multi-fidelity modeling strategy to integrate simulation data of different accuracies to improve model efficiency and accuracy; and introduce transfer learning or meta-learning techniques to realize the rapid generalization of the agent model to different climate zones.
[0012] (6) Couple the trained surrogate model with a multi-objective optimization algorithm, perform rapid optimization on representative meteorological sequences, search for Pareto optimal solution set; and conduct robustness test on candidate solutions by simulating input uncertainty, and select the optimal solution with robust performance.
[0013] (7) Analyze the contribution of various meteorological characteristics, phase change material parameters and operating parameters to various performance targets, and generate parameter recommendation intervals and operating strategy guidelines corresponding to meteorological categories;
[0014] (8) Deploy a digital twin system during the building operation period to collect data in real time to dynamically correct the proxy model; and based on the corrected model, use model predictive control algorithm to perform rolling optimization and adaptive control of the operating parameters in the future short time window.
[0015] Furthermore, in step (2), the clustering algorithm is the K-means algorithm, and the actual date closest to the cluster center in each cluster is selected as the typical date, while the dates with extremely high radiation and extremely low temperature are additionally retained as extreme working condition days.
[0016] Furthermore, in step (4), the phase change material parameters include phase change temperature, latent heat, thermal conductivity, thickness, doping amount and layer position; the operating parameters include indoor set temperature curve, nighttime ventilation start and stop threshold, solar radiation intensity threshold when shading is turned on and the number of air changes.
[0017] Furthermore, in step (5), the machine learning agent model is a Gaussian process regression model, and an active learning strategy is adopted to intelligently select sample points for high-fidelity simulation based on model uncertainty.
[0018] Furthermore, in step (6), the multi-objective optimization algorithm is a multi-objective particle swarm optimization algorithm, and the robustness test is implemented through Monte Carlo simulation to simulate the performance impact caused by meteorological fluctuations and construction deviations.
[0019] Furthermore, in step (7), grey relational analysis or SHAP value analysis based on entropy weight method is used to quantify the contribution of each parameter to the performance target.
[0020] Furthermore, in step (8), the model predictive control algorithm optimizes the shading, ventilation and air conditioning set temperature parameters within the future time window based on short-term weather forecast information.
[0021] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the methods described herein.
[0022] An electronic device according to the present invention includes a memory and a processor, wherein the memory stores a computer program, and when the program is executed by the processor, it implements any of the methods described herein.
[0023] Beneficial Effects: Compared with existing technologies, this invention has the following significant advantages: It forms a complete integrated collaborative decision-making chain from meteorological processing and modeling optimization to online control, greatly reducing manual trial and error and reliance on experience. Multi-objective Robustness: It uniformly considers energy consumption, thermal comfort, peak load, carbon emissions, and life-cycle cost (LCC), and ensures the robustness of the solution through uncertainty testing; the optimization efficiency is improved by several orders of magnitude through the core framework of "surrogate model + multi-objective optimization"; the combination of multi-fidelity learning and meta-learning ensures the accurate and rapid application of the method in different climate zones. Through the closed-loop design of digital twins and model predictive control, the building system is equipped with the ability to cope with real-time weather changes and internal disturbances, realizing the transformation from "static optimization design" to "dynamic optimal control". Attached Figure Description
[0024] Figure 1 This is a flowchart of the present invention. Detailed Implementation
[0025] The technical solution of the present invention will be further described below with reference to the accompanying drawings.
[0026] like Figure 1 As shown, this embodiment of the invention provides an integrated intelligent optimization method for phase change materials and operating parameters in low-carbon buildings, comprising the following steps:
[0027] S1 data collection and preprocessing includes:
[0028] S1.1 Data Collection
[0029] Collect meteorological parameters: Collect hourly dry-bulb temperature, relative humidity, horizontal / normal / diffuse irradiance, wind speed / direction, and cloud cover for the target area for at least 10 years.
[0030] Collect building information: Collect the building's geometric information (building orientation, window-to-wall ratio), the structural layering of the building envelope (material type, thickness) and thermal properties (thermal conductivity, density, specific heat capacity) as well as interior design parameters (set temperature, air exchange rate, equipment power).
[0031] S1.2 Data Cleaning and Repair
[0032] For parameters with diurnal periodicity, such as irradiance, a conservative interpolation method is used (without crossing the diurnal boundary); for temperature and humidity parameters, a weighted average of adjacent time windows is used and combined with a seasonal model for correction.
[0033] Statistical methods (such as the 3σ criterion) are used to identify and remove outlier data points that deviate significantly from physical laws.
[0034] Convert all data to the local standard time zone and ensure that all physical quantities are in the International System of Units (SI).
[0035] S1.3 Feature Engineering
[0036] Macroscopic features such as total annual irradiance, number of cooling / heating days, average diurnal temperature range, and cumulative residence time in each enthalpy range are extracted.
[0037] Construct a time series vector with hourly meteorological parameters (dry-bulb temperature, relative humidity, total irradiance, and wind speed) as the core, and derive derivative features such as "3-hour moving average value" and "slope of change".
[0038] The missing detection rate of the processed data is no more than 1%; the correlation coefficient between the extracted key features and the original time series data is no less than 0.98, ensuring the representativeness of the features.
[0039] S2 meteorological condition decomposition and typical day generation, including:
[0040] S2.1 Construction and Normalization of Cluster Feature Space
[0041] The diurnal variation profiles of (dry-bulb temperature, relative humidity, total irradiance, and wind speed) and their daily statistics (such as daily total irradiance and daily temperature difference) are selected as clustering features.
[0042] All features are Z-score normalized. For temporal contours, Euclidean distance is used for similarity measurement.
[0043] S2.2 Clustering Methods and Selection of Typical Days
[0044] K-means clustering was used, with the sum of squared errors as the objective function. A graph showing the relationship between the sum of squared errors and the number of clusters was plotted. The value of k could be determined based on the slope trend of the graph.
[0045] The actual date closest to the cluster center in each cluster is selected as the typical date, and dates with extremely high radiation and extremely low temperature are additionally reserved as extreme operating condition days.
[0046] S2.3 Representative Sequence Generation
[0047] Based on the clustering results, the daily transition probability between each category is calculated, and a first-order Markov chain model is constructed. This Markov chain is then used to generate representative weekly / monthly series with real-world meteorological time-series variation characteristics for subsequent optimization and robustness verification.
[0048] The output consists of a typical daily set and a representative sequence. The generated sequence must cover at least 90% of the original annual irradiance and temperature distribution (based on passing the Kolmogorov-Smirnov test).
[0049] S3 Envelope Structure - Multiphysics Modeling of Phase Change Materials, including:
[0050] S3.1 Heat Transfer Model Establishment
[0051] Considering outdoor solar radiation intensity and temperature and humidity, a one-dimensional unsteady multi-layer enclosure heat transfer model is established, in which the phase change material layer is modeled using the enthalpy method.
[0052] S3.2 Determine boundary conditions
[0053] For the building's exterior surfaces, convective heat transfer, absorption of shortwave solar radiation, and longwave radiation heat transfer with the sky and environment are considered. For the building's interior surfaces, convective heat transfer and longwave radiation heat transfer with other interior surfaces are considered. The exterior convective heat transfer coefficient h is... out and the internal surface convective heat transfer coefficient h in Determined based on empirical formulas. The effect of external shading is converted into the effective solar absorptivity of the outer surface.
[0054] S3.3 Numerical Simulation and Verification
[0055] Mesh refinement was performed in the phase change material region. The number of meshes and the time step were determined through mesh independence and time step independence analyses. The enthalpy-porosity method was used to solve the temperature field distribution characteristics of the building envelope-phase change material. The results of the numerical simulation were compared and validated with a one-dimensional unsteady multilayer building envelope heat transfer model.
[0056] S4 design variables and runtime parameters include:
[0057] S4.1 Phase Change Materials and Structural Variables (X)
[0058] To match the target meteorological and room temperature requirements, the phase change temperature Tm is set to 18–30 ℃; the phase change interval ΔT is set to 1–4 ℃; based on the heat storage capacity requirements, the latent heat L of the phase change material is set to 100–250 kJ / kg; based on the thermal conductivity of existing phase change materials, the thermal conductivity k is set to 0.2–0.6 W / m·K; considering construction requirements and costs, the thickness d is set to 10–50 mm; to adapt to microcapsule or composite panel processes, the dosage φ is set to 10–30%; layer position (insulation inner / outer side or plaster inner side), and coating method (microcapsule / panel, meeting fire resistance requirements).
[0059] S4.2 Operating Parameters (U)
[0060] Indoor temperature setting curve: Summer 24-27 ℃, Winter 19-21 ℃ (occupiable / non-occupiable time-sharing).
[0061] The nighttime ventilation threshold is the outdoor temperature T. out ≤20-24℃ and relative humidity RH out ≤65-70%.
[0062] The solar radiation intensity trigger threshold for external shading in summer is G≥300-500 W / m².
[0063] The air exchange rate is 0.5–2.0 h⁻¹.
[0064] Continuous variables are encoded using real numbers, while discrete variables, such as strata, are encoded using enumeration.
[0065] S5 multi-objective and constraint modeling includes:
[0066] S5.1 Determine the objective function
[0067] This includes minimizing: 1) total annual energy consumption E; 2) peak load Pmax; and 3) number of unsatisfactory hours I. A ; 4) Annual carbon emissions (Cemi); 5) Life cycle cost (LCC).
[0068] S5.2 Determine the constraints
[0069] Determine the upper limits for the thickness and surface area of the enclosure structure;
[0070] Ensure that the fire resistance rating and VOC limit of the materials meet the standards;
[0071] Room temperature quantile constraints: Tin ≤ upper limit in summer (P95), Tin ≥ lower limit in winter (P95).
[0072] Evaluation was conducted on typical daily sequences and the entire meteorological year (sequences were used for optimization, and the entire year was used for verification).
[0073] S6 agent model and multifidelity learning, including:
[0074] S6.1 Sample Generation
[0075] Establish a low-fidelity model of a one-dimensional thermal network and a high-fidelity model in step S3.
[0076] By adopting an active learning strategy, based on the uncertainty or information entropy of the surrogate model, the system intelligently selects the next batch of sample points that need to be simulated with high fidelity, thereby improving sampling efficiency.
[0077] S6.2 Proxy Model Training
[0078] The Gaussian process regression model is chosen for training. First, a large amount of rapid but somewhat coarse simulation data is used to allow the model to grasp its basic patterns. Then, a small amount of fine-grained simulation data is used to precisely correct the model's prediction bias, forming a high-precision "performance predictor." To ensure its reliability, "five-fold cross-validation" is used to evaluate the model's performance, requiring a prediction accuracy exceeding 90% in the vast majority of cases, with errors controlled within a very small range. If this target is not met, more training data is added until it is satisfactory.
[0079] S7 Multi-Objective and Robust Optimization: A multi-objective particle swarm optimization algorithm is used to select a Pareto optimal solution set that meets the multi-objective requirements. Subsequently, Monte Carlo simulations are used to "stress-test" these optimized solutions, simulating real-world uncertainties such as weather fluctuations and construction deviations. Finally, a "robust optimal solution" that maintains stable performance under various potential risks is selected, ensuring that the recommended solution is not only theoretically optimal but also sufficiently reliable in practical applications.
[0080] S8 interpretability analysis and parameter recommendations include:
[0081] S8.1 Importance and Interaction Effects
[0082] Using SHAP, calculate the global and local contributions to five objectives: annual total energy consumption (E), peak load (Pmax), unsatisfactory hours (IA), annual carbon emissions (Cemi), and life cycle cost (LCC), and draw a SHAP dependency graph to identify the interaction effects between key variables.
[0083] S8.2 Recommendation Rules
[0084] Based on the analysis results, parameter recommendations and operational strategies can be directly applied to engineering design (e.g., Tm=24-26℃, L≥160 kJ / kg, d=15-25 mm; nighttime ventilation: Tm). out ≤22 ℃ and RH≤70%; shading threshold ≥400 W / m²).
[0085] S9 deployment and online updates include:
[0086] S9.1 Digital Twin Deployment: Deploy a sensor network in a physical building to collect real-time data on indoor and outdoor temperature and humidity, solar irradiance, and equipment operating status.
[0087] Based on real-time data, the proxy model is subjected to online incremental learning and correction to ensure that its prediction accuracy is consistent with the actual building condition.
[0088] S9.2 Adaptive Model Predictive Control:
[0089] The model predictive control algorithm is adopted to continuously optimize operating parameters such as shading, ventilation and temperature setpoints within future time windows based on the corrected model and short-term weather forecasts.
[0090] The system has a preset safety rollback strategy that automatically switches to protection mode when an anomaly is predicted or detected.
[0091] S9.3 Operations and Continuous Optimization:
[0092] The system automatically triggers a reassessment process during seasonal changes to check and update operational strategies. All strategy changes are logged for easy tracking and analysis.
Claims
1. A method for integrated intelligent optimization of phase change materials and operating parameters in low-carbon buildings, characterized in that, Includes the following steps: (1) Obtain multi-year historical hourly meteorological data of the target building location, and perform cleaning and feature extraction to construct a feature vector to characterize the climate features; (2) Based on feature vectors, clustering algorithms are used to classify complex meteorological data into multiple typical meteorological day categories, and representative meteorological sequences that retain typical day characteristics and conform to actual time series change patterns are generated based on Markov chain method; (3) Establish a transient heat transfer physical model of the building envelope with integrated phase change material layer, and generate high-fidelity simulation samples covering design variables in batches through parametric scripts to form a training database; (4) Define the physical properties and structural parameters of phase change materials and the operation control parameters of buildings as optimization variables, and construct a multi-objective optimization problem that simultaneously minimizes the building’s total annual energy consumption, peak load, thermal discomfort time, annual carbon emissions and life cycle cost. (5) Use the training database to train the machine learning agent model and establish a fast mapping relationship from optimization variables to multiple performance indicators; adopt a multi-fidelity modeling strategy to fuse simulation data of different accuracies to improve model efficiency and accuracy; Furthermore, transfer learning or meta-learning techniques are introduced to enable the surrogate model to be rapidly generalized to different climate zones; (6) Couple the trained surrogate model with a multi-objective optimization algorithm, perform rapid optimization on representative meteorological sequences, search for Pareto optimal solution set; and conduct robustness test on candidate solutions by simulating input uncertainty, and select the optimal solution with robust performance. (7) Analyze the contribution of various meteorological characteristics, phase change material parameters and operating parameters to various performance targets, and generate parameter recommendation intervals and operating strategy guidelines corresponding to meteorological categories; (8) Deploy a digital twin system during the building operation period to collect data in real time to dynamically correct the proxy model; and based on the corrected model, use model predictive control algorithm to perform rolling optimization and adaptive control of the operating parameters in the future short time window.
2. The integrated intelligent optimization method for phase change materials and operating parameters in low-carbon buildings according to claim 1, characterized in that, In step (2), the clustering algorithm is the K-means algorithm, and the actual date closest to the cluster center in each cluster is selected as the typical date. At the same time, the dates with extremely high radiation and extremely low temperature are reserved as extreme working condition days.
3. The integrated intelligent optimization method for phase change materials and operating parameters in low-carbon buildings according to claim 1, characterized in that, In step (4), the phase change material parameters include phase change temperature, latent heat, thermal conductivity, thickness, doping amount and layer position; the operating parameters include indoor set temperature curve, nighttime ventilation start and stop threshold, solar radiation intensity threshold when shading is turned on and air exchange rate.
4. The integrated intelligent optimization method for phase change materials and operating parameters in low-carbon buildings according to claim 1, characterized in that, In step (5), the machine learning agent model is a Gaussian process regression model, and an active learning strategy is adopted to intelligently select sample points for high-fidelity simulation based on model uncertainty.
5. The integrated intelligent optimization method for phase change materials and operating parameters in low-carbon buildings according to claim 1, characterized in that, In step (6), the multi-objective optimization algorithm is a multi-objective particle swarm optimization algorithm. The robustness test is implemented through Monte Carlo simulation to simulate the performance impact caused by meteorological fluctuations and construction deviations.
6. The integrated intelligent optimization method for phase change materials and operating parameters in low-carbon buildings according to claim 1, characterized in that, In step (7), grey relational analysis or SHAP value analysis based on entropy weight method is used to quantify the contribution of each parameter to the performance target.
7. The integrated intelligent optimization method for phase change materials and operating parameters in low-carbon buildings according to claim 1, characterized in that, In step (8), the model predictive control algorithm optimizes the shading, ventilation and air conditioning set temperature parameters within the future time window based on short-term weather forecast information.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.
9. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, implements the method as described in any one of claims 1-7.