A building thermal insulation external wall thermal insulation effect prediction method and system

Abstract

This invention relates to the field of building exterior wall insulation prediction technology, specifically disclosing a method and system for predicting the thermal insulation effect of building exterior walls. The method first acquires energy consumption segment data and corresponding environmental action time-series data of the exterior wall during the operation and maintenance phase, providing an accurate data foundation for subsequent model training and parameter identification. Then, a data-physics dual-driven thermal insulation effect attenuation prediction model is constructed, and an equivalent thermal resistance attenuation memory function is introduced into the model. This approach aims to constrain the solution boundary of the data-driven model with physical equations, while dynamically correcting the time-varying parameters of the physical model using data. Finally, the current equivalent thermal resistance attenuation rate is extrapolated to obtain the thermal insulation effect value and the corresponding building energy consumption increment, achieving quantitative assessment of the thermal insulation performance throughout the entire life cycle of the exterior wall and advanced early warning of energy consumption risks.

Description

Technical Field

[0001] This invention relates to the field of building exterior wall insulation prediction technology, specifically to a method and system for predicting the insulation effect of building exterior walls. Background Technology

[0002] In energy conservation management throughout the entire building lifecycle, the long-term performance maintenance of external wall insulation is a key factor affecting building energy consumption and indoor environmental quality. With engineering companies increasingly involved in investment, construction, and operation of integrated super high-rise buildings, large public buildings, and resettlement housing projects, more and more buildings have entered the long-term operation and maintenance phase. In actual service, external wall insulation is not static. Insulation materials (such as polystyrene boards, rock wool boards, and vacuum insulation boards) will experience irreversible performance degradation during long-term service, such as polystyrene board powdering and rock wool board moisture absorption and settlement. Environmental loads such as temperature and humidity cycles, wind pressure fatigue, and seismic forces can cause interface peeling or local debonding between the insulation layer and the base wall. In addition, external wall modifications and pipeline openings during the operation and maintenance phase may also cause cumulative damage to the insulation layer. All of the above factors together lead to a non-linear and non-stable decline in the effectiveness of external wall insulation over service time, resulting in a year-on-year increase in building energy consumption, a decrease in indoor thermal comfort, and even durability problems such as condensation and mold.

[0003] To address the aforementioned issues, some existing technologies have methods for evaluating or predicting the thermal insulation performance of exterior walls. For example, patent application CN119167198A discloses a method and system for evaluating the thermal insulation performance of building exterior walls. The method includes: acquiring temperature and humidity data of the building exterior walls based on data acquisition frequency; preprocessing the temperature and humidity data to obtain target temperature and humidity data; extracting features from the target temperature and humidity data to obtain target features; calculating the temperature and humidity change trend using a preset time series analysis method based on the target features, environmental factors, material properties, temperature data, and humidity data; calculating predicted thermal insulation performance information using a preset prediction method based on the target features and target temperature and humidity data; calculating thermal insulation performance index values ​​based on the target temperature and humidity data, insulation material thickness, exterior wall area, and internal and external temperature difference; and evaluating the thermal insulation performance based on the temperature and humidity change trend, predicted thermal insulation performance information, and thermal insulation performance index values ​​to obtain the thermal insulation performance evaluation result.

[0004] For example, the invention patent application with publication number CN120296561A discloses a method and system for predicting the risk of thermal insulation detachment from external walls based on a deep learning model. First, it collects multi-dimensional state data of the target building's external walls, covering insulation layer material properties, environmental exposure history, structural connection strength, surface deformation monitoring, and construction process records. Next, it extracts material aging feature vectors from the insulation layer material property data and extracts spatiotemporal distribution features from the environmental exposure history data to generate an environmental impact feature tensor. It then fuses the structural connection strength and surface deformation monitoring data into a structural deformation correlation map, converts the construction process record data into a probability distribution of process defects, and inputs these into a pre-trained multimodal risk prediction network. The network outputs a thermal map of the external wall insulation layer detachment risk level and risk area. Finally, based on the results, it generates a dynamic monitoring strategy that includes sensor deployment optimization schemes and detection cycle adjustment parameters.

[0005] In summary, the following two core technical problems still exist in the prediction of building exterior wall insulation effects: Firstly, current data analysis methods either employ a purely data-driven approach or a purely physical model approach. This approach struggles to balance physical interpretability with prediction accuracy under small sample conditions. Furthermore, when using energy consumption or environmental data for inversion, only single or current batch data is typically utilized, making it difficult to suppress noise interference from single measurements. Secondly, existing methods usually employ fixed-frequency continuous data acquisition or passively wait for data to arrive. When new data arrives, the model cannot be proactively updated based on the deviation between the new data and the predicted value. In addition, current outputs are mostly insulation performance levels, risk heat maps, or monitoring parameter adjustment suggestions. Ultimately, this prevents the direct generation of quantifiable long-term prediction data on insulation performance degradation for engineering decisions during the operation and maintenance phase. It also makes it difficult to provide early warnings of energy consumption exceeding limits or thermal and humidity failure risks caused by nonlinear degradation of insulation performance, thus hindering the development of accurate insulation repair or replacement plans within the optimal time window. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for predicting the thermal insulation effect of building exterior walls, which can effectively solve the problems mentioned in the background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of the present invention provides a method for predicting the thermal insulation effect of building exterior walls, comprising: S1. acquiring energy consumption segment data of the target exterior wall during the operation and maintenance phase and corresponding environmental action time-series data; S2. constructing a data-physical dual-driven thermal insulation effect attenuation prediction model, wherein the thermal insulation effect attenuation prediction model includes a physical driving part, a data driving part, and an equivalent thermal resistance attenuation memory function, and outputting the current equivalent thermal resistance attenuation rate; S3. extrapolating the current equivalent thermal resistance attenuation rate through an extrapolation model to predict the equivalent thermal resistance attenuation rate at future times, and converting it into a thermal insulation effect value and the corresponding building energy consumption increment.

[0008] As a further method, the initial value of the equivalent thermal resistance attenuation memory function is preset and offset by the equivalent initial attenuation amount of construction defects. Specifically, the construction acceptance data of the target exterior wall during the completion and acceptance stage is obtained. The mapping relationship between each defect type and the local thermal resistance weakening coefficient is established through the construction acceptance data. After weighted superposition, it is converted into the global initial attenuation amount. The global initial attenuation amount is superimposed with the nonlinear attenuation component to jointly constitute the complete equivalent thermal resistance attenuation process.

[0009] As a further method, self-checking of prediction results and triggering re-prediction are also included, the specific process of which is as follows: After completing the extrapolation prediction, the next energy consumption segment data is compared with the predicted energy consumption value of the insulation effect attenuation prediction model under the same environmental conditions. If the comparison deviation exceeds the acceptable threshold but is lower than the failure threshold, the fine-tuning operation of the insulation effect attenuation prediction model is triggered, and only the data-driven part is updated. If the comparison deviation is not lower than the failure threshold, a full re-prediction is triggered, and the inversion and extrapolation are re-executed. The acceptable threshold and the failure threshold are dynamically adjusted according to the current uncertainty propagation amount.

[0010] A second aspect of the present invention provides a system for predicting the thermal insulation effect of building exterior walls, comprising: a data acquisition module for acquiring energy consumption segment data of the target exterior wall during the operation and maintenance phase and corresponding environmental action time series data; a model building module for constructing a data-physical dual-driven thermal insulation effect attenuation prediction model, wherein the thermal insulation effect attenuation prediction model includes a physical driving part, a data driving part, and an equivalent thermal resistance attenuation memory function, and outputs the current equivalent thermal resistance attenuation rate; and a thermal insulation effect prediction module for extrapolating the current equivalent thermal resistance attenuation rate through an extrapolation model to predict the equivalent thermal resistance attenuation rate at future times, and converting it into a thermal insulation effect value and the corresponding building energy consumption increment.

[0011] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) This invention provides a method and system for predicting the thermal insulation effect of building exterior walls. First, it acquires the energy consumption data of the exterior wall during the operation and maintenance phase and the corresponding environmental action time series data, which can provide an accurate data foundation for subsequent model training and parameter identification. Then, it constructs a data-physical dual-driven thermal insulation effect attenuation prediction model and introduces an equivalent thermal resistance attenuation memory function into the model. The purpose of this approach is to constrain the solution boundary of the data-driven part with physical equations, and at the same time, use data to dynamically correct the time-varying parameters of the physical model, accumulate and transmit the historical impact of the attenuation effect in the time dimension, which helps to reduce or even avoid the defects of the pure data model in the present when the data is sparse, such as overfitting and lack of physical meaning, or the pure physical model being unable to adapt to the time-varying nature of complex working conditions. Finally, it extrapolates and extends the current equivalent thermal resistance attenuation rate, and finally obtains the thermal insulation effect value and the corresponding building energy consumption increment, so as to realize the quantitative assessment of the thermal insulation performance of the exterior wall throughout the entire life cycle and the early warning of energy consumption risk.

[0012] (2) The present invention also includes steps for identifying and eliminating abnormal energy consumption segments and self-checking and triggering re-prediction of prediction results. By processing different types of abnormalities in different ways, the effective signal-to-noise ratio of input data can be improved, avoiding misreading intermittent energy consumption abnormalities or sensor jump errors as real changes in insulation performance, thereby improving the accuracy of attenuation prediction. After completing the extrapolation prediction, the data comparison is used to determine whether to trigger the re-prediction operation, which helps to refresh the prediction trajectory in time after environmental changes or sudden damage, prevent unacceptable deviations between historical patterns and current physical reality, and maintain the timeliness of operation and maintenance decisions.

[0013] (3) The present invention presets the initial value of the equivalent thermal resistance attenuation memory function, that is, it converts the construction acceptance data into the global initial attenuation amount. This conversion enables the model to carry the initial defect information introduced by the actual construction quality from the zero moment of the operation and maintenance stage, eliminates the systematic underestimation starting from the ideal design value, and superimposes this initial attenuation amount with the nonlinear attenuation component to form a complete equivalent thermal resistance attenuation process, so as to internalize the difference in construction quality into the initial state of the model and make the subsequent attenuation prediction based on a more realistic physical starting point. Attached Figure Description

[0014] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0016] Figure 2This is a schematic diagram of the system module connections of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] The first embodiment, referred to Figure 1 The diagram shown is a flowchart of the method steps of the present invention. The first aspect of the present invention provides a method for predicting the thermal insulation effect of building exterior walls, including: S1. Obtaining energy consumption segment data and corresponding environmental action time series data of the target exterior wall during the operation and maintenance phase.

[0019] It should be noted that the environmental impact time series data comes from meteorological data or simple on-site sensors at the building location, and the data mainly includes one or more of outdoor temperature, relative humidity, solar irradiance, wind speed and rainfall.

[0020] It also needs to be explained that the discontinuous, short-term energy consumption data employs an adaptive sampling strategy based on weather type stratification: a weather classification system including sunny, cloudy, rainy, cold, and hot types is pre-established; based on weather forecasts, future weather types are identified, and an energy consumption segment is collected for each uncollected weather type; when extreme weather is predicted, additional collection is automatically triggered, unrestricted by conventional sampling rules, to obtain energy consumption segment data under extreme conditions; this data can be one or more of the following: exterior wall heat flux density, inverted value of heat transfer coefficient under indoor-outdoor temperature difference, or unit area air conditioning energy consumption increment; the energy consumption segment data can be obtained by deploying heat flux meters and temperature sensors at typical locations on the inner surface of the target exterior wall for short-term, high-frequency data acquisition. The purpose of using a discontinuous, short-term acquisition method is to reduce the cost of long-term continuous monitoring while selectively acquiring the most representative thermal response characteristics and avoiding redundant data accumulation.

[0021] Specifically, the aforementioned target exterior wall refers to a specific external wall cladding of a building whose thermal insulation performance degradation is to be evaluated, and the operation and maintenance phase refers to the entire operation and maintenance cycle after the building is completed and delivered and put into daily use and management.

[0022] S2. Construct a data-physical dual-driven prediction model for the attenuation of thermal insulation effect, and introduce an equivalent thermal resistance attenuation memory function into the model.

[0023] In this embodiment, the above-mentioned thermal insulation effect attenuation prediction model not only includes two dual-drive components, data-driven and physical-driven, but also introduces an equivalent thermal resistance attenuation memory function. That is, the relationship between this model and this function is that the equivalent thermal resistance attenuation memory function serves as the post-processing and smoothing module of the thermal insulation effect attenuation prediction model. It receives the single-time correction attenuation rate output by the physical drive and data drive and performs time-series fusion on it. This relationship can suppress the jump in prediction results caused by the fluctuation of data quality in a single energy consumption segment, making the output of the thermal insulation effect attenuation prediction model more stable and continuous on the time axis.

[0024] The data-physical dual-drive approach represents both data-driven and physical-driven approaches. The physical-driven approach takes the environmental action time series data as input, establishes the theoretical decay function of the material's equivalent thermal resistance over service time through the thermal aging mechanism of the insulation material, and outputs the theoretical decay rate.

[0025] The data driver takes the energy consumption segment data as input, uses the energy consumption segment data to perform inverse correction on the theoretical decay function, and outputs the corrected decay rate.

[0026] The equivalent thermal resistance attenuation memory function takes the corrected attenuation rate and the inversion attenuation results at several historical time points as input, and outputs the current equivalent thermal resistance attenuation rate through weighted fusion to suppress the interference of noise and uncertainties in single energy consumption segment data on the attenuation degree estimation.

[0027] It should be explained that the inversion attenuation results at several historical time points mentioned above represent the estimated values ​​of the effective equivalent thermal resistance attenuation rate inverted and recorded by the data-physics dual-drive model before the current time. The number of time points is determined by the size of the preset memory depth window, usually selecting the first 5 to 10 effective sampling points.

[0028] Specifically, the equivalent thermal resistance decay memory function adopts a time decay weighted mechanism, and the specific process is as follows: The time distances between several historical time points and the current time point are arranged in ascending order from smallest to largest. The inversion decay result corresponding to the first ranked result is assigned a first weight coefficient, and the remaining inversion decay results are assigned corresponding weight coefficients in ascending order. The weight coefficients decay exponentially.

[0029] For example, the time decay weighting mechanism can be expressed as:

[0030] Among them, R current This represents the current equivalent thermal resistance attenuation rate output after time-weighted fusion, where i is the number of the historical time point, i=1,2,3,...,m, and m is the total number of historical time points, r iw represents the inversion decay result at the i-th historical time point. i Let w represent the weighting coefficient corresponding to the inversion decay result at the i-th historical time point. That is, if i is 1, then w is the weighting coefficient at that time. i The first weighting coefficient is obtained by multiplying a preset forgetting factor by the time distance between the historical time point and the current time point, taking the reciprocal, and then comparing it with the corresponding posterior uncertainty. Finally, an exponential process is performed to obtain its weighting coefficient. The weighting coefficient obtained here takes into account both time decay and data confidence. The data confidence term is directly derived from and related to the uncertainty propagation calculation of the output of the equivalent thermal resistance decay memory function, which is described later. This uncertainty propagation is formed by the time decay weighted propagation of the posterior uncertainty of each historical inversion result.

[0031] In this embodiment, the smaller the time distance between a certain historical time point and the current time point, the closer the historical time point is to the current time point, and the more timely it is to reflect the current true decay state. Therefore, the inversion decay result corresponding to the historical time point with the smallest distance or the closest time is given the highest weight, and so on. The inversion decay result corresponding to the historical time point with the larger time distance is given the lowest weight, that is, the value of the first weight coefficient mentioned above is the largest.

[0032] The weighted fusion also outputs uncertainty propagation. If the uncertainty propagation exceeds a preset propagation threshold, a prompt is automatically triggered to re-collect energy consumption data.

[0033] It should be noted that the uncertainty propagation quantity of the above output is in parallel with the current equivalent thermal resistance attenuation rate of the above output. The uncertainty propagation quantity is calculated by superimposing the posterior uncertainty of each historical inversion attenuation result in the weighted fusion process through time attenuation weighted propagation. It is used to quantify the confidence interval width of the current estimate caused by the accumulation of noise in historical data. This uncertainty propagation quantity increases monotonically with the extension of the prediction period and is presented in the form of an interval in the final prediction result. Its presentation method can be the current equivalent thermal resistance attenuation rate ± uncertainty propagation quantity.

[0034] Preferably, within the equivalent thermal resistance attenuation memory function, a dynamically updated attenuation state matrix is ​​also maintained. This attenuation state matrix not only records the historical inversion attenuation rate, but also records the confidence level, dominant environmental factors, and data quality labels used in each inversion. The dynamic update of the attenuation state matrix means that each time a new inversion is completed and the current equivalent thermal resistance attenuation rate is output, the inversion attenuation rate, confidence level, dominant environmental factors, and data quality labels of this inversion are added as a new column to the attenuation state matrix, while the earliest record exceeding the preset memory duration is removed to maintain a constant matrix size.

[0035] It should be explained that the aforementioned historical inversion attenuation rate represents the equivalent thermal resistance attenuation rate output after Bayesian inversion correction at each historical time point; the confidence level of the inversion represents the maximum posterior probability density value of the posterior distribution of the attenuation rate output by the Bayesian inversion framework, used to quantify the reliability of the inversion result itself; the dominant environmental factor represents the code of the single environmental factor that contributes the most to the variance of the environmental impact time series data in the previous time window, obtained by principal component analysis, such as labels for high temperature and high humidity, freeze-thaw cycle, etc.; the data quality label used for the inversion can be the quality level assigned to the energy consumption segment data used in the historical inversion after the abnormal energy consumption segment identification and removal process, such as high quality, corrected, low weight, etc.

[0036] The weighted fusion process, based on the recorded information in the aforementioned decay state matrix, can adaptively adjust the fusion weights of the inversion decay results at each historical time point. The adjustment method is that when the dominant environmental factor at a certain historical time point is less than the threshold in terms of similarity measurement with the environmental characteristics of the current time window, its original weight is reduced by an environmental mismatch penalty factor less than 1, even if the time distance is relatively close. At the same time, when the data quality label of a certain historical time point is corrected or low-weight, its weight will be multiplied by a discount coefficient preset by the label, thereby suppressing the influence of low-quality data in time series fusion.

[0037] It should also be explained that the aforementioned environmental mismatch penalty factor is calculated by subtracting the average outdoor temperature and relative humidity of the current time window from the average temperature and humidity of each historical time point recorded in the decay state matrix. If the absolute values ​​of the two differences are both within their respective preset allowable deviation ranges (e.g., temperature deviation not exceeding 5 degrees Celsius, humidity deviation not exceeding 15%), then the environmental mismatch penalty factor corresponding to that historical time point is set to 1. If either difference exceeds the allowable deviation range, then the value is set to a preset constant less than 1 (e.g., 0.5), thereby reducing the weight of historical data with environmental mismatch in the weighted fusion.

[0038] Furthermore, the physical-driven theoretical decay function is based on the temperature and humidity coupled accelerated aging model of the Arrhenius equation, which expresses the decay rate of the equivalent thermal resistance as the product of the temperature excitation function and the humidity excitation function.

[0039] Preferably, the theoretical decay function mentioned above is a temperature and humidity coupled accelerated aging model based on the Arrhenius equation. In fact, it is equivalent to establishing a theoretical decay function of the material's equivalent thermal resistance with service time through the thermal aging mechanism of the insulation material. Specifically, the aging process of the material is regarded as a chemical reaction process driven by thermal activation and moisture degradation. The rate constant of the exponential decay of the equivalent thermal resistance with time is described by the product of the temperature exponential term and the humidity power law term. The final output theoretical decay rate is the percentage decrease of the material's equivalent thermal resistance calculated here based on the current cumulative service time.

[0040] In addition, the above-mentioned temperature and humidity coupled accelerated aging model refers to a model that uses outdoor temperature and relative humidity as the main environmental stress variables, and multiplies the temperature excitation function with the humidity excitation function to amplify or reduce the baseline aging reaction rate. The temperature excitation function is represented by an exponential term based on the Arrhenius formula, reflecting the accelerating effect of temperature increase on the molecular degradation reaction rate; the humidity excitation function is represented by a power function or exponential function with relative humidity as the independent variable, reflecting the accelerating effect of moisture penetration on the hydrolysis or plasticization of materials.

[0041] A nonlinear coupling term is introduced between the temperature excitation function and the humidity excitation function. This nonlinear coupling term characterizes the synergistic accelerated aging effect produced when temperature and humidity act on the material within a specific range. Specifically, an exponential enhancement factor is added to the product of the temperature excitation function and the humidity excitation function, with the interaction term of temperature and relative humidity as the variable. This is achieved by multiplying the coupling strength coefficient with the bilinear interaction function of temperature and relative humidity, then applying an exponential function to obtain the exponential enhancement factor. The coupling strength coefficient is determined by the material type. The bilinear interaction function of temperature and relative humidity increases significantly when both temperature and relative humidity enter their respective high-level ranges, causing the overall decay rate constant to exceed that of temperature alone. The level predicted by the product of the effect and the individual humidity effect captures the synergistic effect of the rapidly accelerating hydrolytic breakage rate of polymer molecular chains under high temperature and high humidity conditions. However, when only a single factor is at a high level, the contribution of this coupling term degenerates into a weak enhancement effect close to 1. Here, close to 1 means that it changes continuously from 1.0 and converges to the range of 1.05. The 5% slight fluctuation in the range indicates that the coupling is only a secondary perturbation at this time. The specific range represents the high temperature and high humidity combined range where the temperature exceeds 40 degrees Celsius and the relative humidity exceeds 70%, as well as the freeze-thaw-high humidity coupled range where the temperature fluctuates periodically near the freezing point and the relative humidity exceeds 85%. In these ranges, the microporous structure inside the insulation material is accelerated by the alternating effects of ice crystal expansion and moisture plasticization, resulting in irreversible microcrack propagation and a nonlinear drop in the equivalent thermal resistance.

[0042] Each type of thermal insulation material corresponds to a set of model parameters. For example, extruded polystyrene board corresponds to its pre-exponential factor, activation energy, and humidity sensitivity index, while rock wool corresponds to its fiber adhesive hydrolysis activation energy. The set of model parameters is obtained by collecting data on the change of equivalent thermal resistance of thermal insulation materials with aging time under different temperature and humidity combinations, and then fitting it through nonlinear regression.

[0043] It should be explained that the above model parameter set was calibrated in advance through laboratory accelerated aging experiments. The experiment set multiple temperature levels (such as 25 degrees Celsius, 40 degrees Celsius, 55 degrees Celsius, and 70 degrees Celsius) and multiple humidity levels (such as 30% relative humidity, 50% relative humidity, 70% relative humidity, and 90% relative humidity). The thermal insulation material was continuously aged in constant temperature and humidity chambers with different temperature and humidity combinations. The thermal conductivity was measured at fixed time intervals and converted into equivalent thermal resistance. After collecting the data on the change of equivalent thermal resistance with aging time, the optimal estimated value of each model parameter was obtained by using the temperature and humidity coupled accelerated aging model as the regression equation and nonlinear regression fitting methods such as least squares method or maximum likelihood estimation.

[0044] The initial values ​​of the above model parameter set depend on the initial thermal insulation performance parameters of the target exterior wall, namely the initial equivalent thermal resistance value calculated from the initial thermal conductivity and initial insulation layer thickness recorded during the completion and acceptance phase. This value serves as the normalization benchmark for the theoretical attenuation function at time zero. Obtaining the initial thermal insulation performance parameters during the completion and acceptance phase helps to provide an accurate starting point for material properties in the thermal insulation effect attenuation prediction model and eliminates the model baseline deviation caused by differences in construction quality.

[0045] The advantage of setting initial values ​​for the model parameter set is that it allows for the calibration of material parameters using actual acceptance data, ensuring that the starting point for calculating the theoretical attenuation rate is precisely matched with the actual initial state of the target exterior wall, thereby improving the absolute accuracy of short- and medium-term attenuation prediction.

[0046] Specifically, data-driven reverse correction is performed using a Bayesian inversion framework, and the specific execution process is as follows: The unknown parameters in the theoretical decay function are treated as random variables and assigned a prior distribution. These unknown parameters are field correction factors that are difficult to determine completely through laboratory experiments. They mainly include the field deviation values ​​of the actual activation energy adjustment coefficient and the humidity sensitivity index. The field deviation value refers to the difference between the activation energy and humidity sensitivity index obtained by inverting the energy consumption fragment data of the actual building exterior wall and the corresponding parameter standard values ​​calibrated by the laboratory accelerated aging experiment. This difference is calculated by subtracting the laboratory calibration standard value from the expected value of the posterior distribution in the Bayesian inversion framework, and is used to quantify the degree of deviation of aging behavior between the field service environment and the laboratory accelerated environment. The energy consumption fragment data is updated to a posterior distribution using the Bayesian formula, and the expected value of the posterior distribution is used as the corrected decay rate. The expected value of the posterior distribution can be obtained from the maximum a posteriori estimate or the mean of the posterior distribution probability density function.

[0047] The Bayesian inversion framework also outputs posterior uncertainty, which takes the standard deviation of the posterior distribution or the half-width of the confidence interval as input. The posterior uncertainty is defined as the ratio of the standard deviation to the mean of the posterior distribution. The posterior uncertainty is used in conjunction with the uncertainty propagation in the equivalent thermal resistance attenuation memory function. Specifically, the posterior uncertainty is used as a confidence weight for a single inversion and substituted into the weighted fusion process of the subsequent memory function. That is, the inversion attenuation result at each historical time point is divided by the square of its corresponding posterior uncertainty to obtain a normalized confidence weight value. This weight value is then used in the exponential weight allocation in the time attenuation weighting mechanism to control the contribution of the inversion result to the time series smoothing.

[0048] In this embodiment, the setting of the above-mentioned prior distribution also depends on the initial thermal insulation performance parameters of the target exterior wall. The specific setting process is to set the prior mean of the activation energy adjustment coefficient to 1, and introduce an uncertainty proportional to the initial equivalent thermal resistance measurement uncertainty into its standard deviation; the prior of the humidity sensitivity index deviation value is centered at 0, and its standard deviation is set according to the moisture absorption rate fluctuation range of the initial thermal insulation material batch test report.

[0049] S3. The current equivalent thermal resistance attenuation rate is extrapolated and extended using a preset extrapolation model. The extension method is to extend forward along the time axis, that is, using the current equivalent thermal resistance attenuation rate as the initial value, and using the temperature and humidity coupled aging trend term in the physical drive to recursively calculate the future attenuation; to predict the equivalent thermal resistance attenuation rate at several predicted maintenance time points, where several predicted maintenance time points refer to multiple predicted maintenance time points within a predicted maintenance cycle. Of course, if a full life cycle maintenance plan needs to be formulated, it can also be set as multiple predicted maintenance cycles. This cycle is not fixed, but dynamically determined by the expected remaining design service life of the building and the set evaluation frequency; and the equivalent thermal resistance attenuation rate is converted into the insulation effect value and the corresponding building energy consumption increment through a preset conversion table. The conversion method is as follows: First, based on the preset conversion relationship between equivalent thermal resistance and insulation effect value, the insulation effect value is converted into several predicted operation and maintenance time points; then, based on the preset mapping relationship between insulation effect value and building energy consumption, the building energy consumption increment corresponding to the insulation effect value is calculated.

[0050] Finally, the insulation performance value and the building energy consumption increment will be updated to the building operation and maintenance management platform.

[0051] In a specific embodiment, the energy consumption data of the exterior wall during the operation and maintenance phase, as well as the corresponding environmental impact time series data, are first acquired. This provides an accurate data foundation for subsequent model training and parameter identification. Then, a data-physical dual-driven insulation effect attenuation prediction model is constructed, and an equivalent thermal resistance attenuation memory function is introduced into the model. The purpose of this approach is to constrain the solution boundary of the data-driven part with physical equations, while dynamically correcting the time-varying parameters of the physical model with data. This accumulates and transmits the historical impact of the attenuation effect over time, which helps to reduce or even avoid the defects of existing pure data models in the case of data sparsity, such as overfitting and loss of physical meaning, or the inability of pure physical models to adapt to the time-varying nature of complex working conditions. Finally, the current equivalent thermal resistance attenuation rate is extrapolated to obtain the insulation effect value and the corresponding building energy consumption increment, thereby realizing a quantitative assessment of the insulation performance of the exterior wall throughout its entire life cycle and an early warning of energy consumption risks.

[0052] Reference Figure 2 The diagram shown is a schematic diagram of the system module connection of the present invention. The second aspect of the present invention provides a building insulation external wall insulation effect prediction system, including: a data acquisition module, a model construction module and an insulation effect prediction module.

[0053] The second aspect of this invention provides a building insulation and exterior wall insulation effect prediction system, which also includes a building operation and maintenance management platform for storing all preset values ​​involved in this invention, including but not limited to preset deviation thresholds, model parameter sets, and baseline values ​​for acceptable thresholds and failure thresholds. The specific storage method is illustrated using the acceptable threshold as an example, and the storage process is as follows: First, the building operation and maintenance management platform aggregates historical energy consumption data of similar segments in the climate zone where the building is located and compares them with the predicted energy consumption values ​​of the insulation effect attenuation prediction model. It then statistically analyzes the distribution of the comparison results corresponding to the first occurrence of condensation risk or energy consumption exceeding limits, taking the 10th percentile of this distribution as the initial benchmark value for the acceptable threshold. During system operation, the platform continuously records user feedback scores on the reasonableness of the threshold after each operation and maintenance decision. These feedback scores are the scores given by operation and maintenance personnel after implementing the maintenance plan, indicating whether the threshold is too high, too low, or reasonable, ranging from 1 to 5 points. A Bayesian online update algorithm is used to iteratively optimize the threshold. Specifically, the current threshold is used as the mean of the prior distribution, a likelihood function is constructed using the newly collected feedback scores, the posterior distribution is calculated using the Bayesian formula, and the expected value of the posterior distribution is used as the updated threshold. The optimized value is written into the platform configuration database in structured field form and managed through version numbers.

[0054] Other values ​​can be obtained through similar methods, such as the iterative optimization based on the statistical distribution of historical data and operational feedback, as described above. This embodiment does not impose any special limitations on this.

[0055] Preferably, the aforementioned building operation and maintenance management platform can refer to an existing building information model operation and maintenance management (BIM-FM) platform, building energy management system (BEMS), or building carbon emission intelligent monitoring cloud website.

[0056] The data acquisition module is connected to the model building module, the model building module is connected to the thermal insulation effect prediction module, and the data acquisition module, model building module, and thermal insulation effect prediction module are all connected to the building operation and maintenance management platform.

[0057] The data acquisition module is used to acquire energy consumption data segments of the target exterior wall during the operation and maintenance phase, as well as corresponding environmental impact time-series data.

[0058] The model building module is used to construct a data-physical dual-driven prediction model for the attenuation of thermal insulation effect, and introduces an equivalent thermal resistance attenuation memory function into the model.

[0059] The thermal insulation effect prediction module is used to extrapolate the current equivalent thermal resistance attenuation rate through a preset extrapolation model to predict the equivalent thermal resistance attenuation rate at several predicted operation and maintenance time points. The equivalent thermal resistance attenuation rate is converted into thermal insulation effect value and corresponding building energy consumption increment through a preset conversion table. Finally, the thermal insulation effect value and building energy consumption increment are updated to the building operation and maintenance management platform.

[0060] Specifically, this also includes the identification and removal of abnormal energy consumption segments. The specific identification process is as follows: The theoretical baseline value for zero energy consumption is generated by using initial thermal insulation performance parameters and concurrent environmental data. The generation process is as follows: Using the initial thermal conductivity and thickness of the insulation layer measured during the completion and acceptance phase, the initial equivalent thermal resistance is calculated. Then, combined with the same environmental data such as the outdoor temperature and indoor set temperature during the same period as the energy consumption segment collection time, the theoretical heat flow through the exterior wall during this period is calculated using the steady-state heat transfer formula, which serves as the theoretical zero-attenuation energy consumption benchmark value.

[0061] Each energy consumption segment is compared with the theoretical zero-attenuation energy consumption benchmark value. Specifically, the heat transfer coefficient or heat flux value calculated from the actual energy consumption segment is compared with the benchmark value to calculate the relative deviation. If the deviation result exceeds the preset deviation threshold, it is marked as abnormal, and the abnormality type is distinguished. It can be screened by the persistence and directionality of the deviation. Specifically, if the deviation remains in the same direction and the value is similar in multiple consecutive sampling periods, it is determined to be a correctable abnormality caused by sensor drift; if the deviation only appears in a single sampling period and has no directional pattern, it is determined to be temporary interference; if the deviation persists but the direction or value has no stable pattern, it is classified as other abnormality types.

[0062] If the anomaly is temporary interference, it is directly removed. If the anomaly is a correctable anomaly caused by sensor drift, data correction is performed. The specific correction process involves comparing the normal data segments before and after the anomaly, calculating the linear or nonlinear trend of the drift, obtaining the trend line equation using least squares fitting, subtracting the drift calculated by the trend line equation from the data during the anomaly period, thereby compensating for sensor zero drift, and performing baseline compensation for the data during that period. If it belongs to other anomaly types, the likelihood weight of the anomaly data point in the subsequent Bayesian inversion is reduced, that is, by expanding the standard deviation corresponding to the data item in the likelihood function to several times its original value to reduce its weight.

[0063] The initial thermal insulation performance parameters include one or more of the following: the thermal conductivity of each layer of wall material, the thickness of the insulation layer, and the defect rate detected by infrared thermography, as recorded in the completion acceptance record. These parameters can be obtained by reviewing the completion documents or conducting on-site specialized testing.

[0064] The concurrent environmental data includes one or more of the outdoor air temperature, solar irradiance, and wind speed corresponding to the time of energy consumption segment collection, which can be obtained through the historical records of the building's own weather station or a nearby weather station.

[0065] Preferably, before the above-mentioned anomaly identification, the energy consumption segment data needs to be preprocessed. This involves acquiring high spatial resolution temperature field data through a distributed fiber optic temperature sensing network deployed on the inner surface of the target exterior wall. The temperature field data is then used to locate thermal anomalies in the energy consumption segment data in a spatial dimension. Specifically, the distributed fiber optic temperature sensing network is divided into multiple spatial grid cells along the inner surface of the exterior wall, and the time-series temperature data of the inner surface of each grid cell is collected in real time. After a certain energy consumption segment is initially identified as an anomaly, the spatial temperature field distribution corresponding to that time period is retrieved, and the temperature of each grid cell is calculated relative to the set indoor temperature during the same period. For deviation values, the platform automatically draws a spatial distribution cloud map of temperature deviation values; it clusters grid cells whose temperature deviation values ​​exceed the preset spatial anomaly threshold, identifies spatially continuous anomaly areas with consistent temperature deviation directions, and outputs the spatial coordinates, area, and temperature deviation amplitude of the anomaly area as the thermal anomaly location result. Combined with the orientation of the building facade and the distribution map of known construction defects, it helps to determine whether the root cause of the anomaly is the deterioration of local insulation layer defects, water erosion in a specific area, or a global systematic error caused by sensor drift, thereby improving the accuracy of anomaly type determination and assisting in the determination of the above anomaly types.

[0066] Furthermore, it also includes self-checking of prediction results and triggering re-prediction, the specific process of which is as follows: After completing the extrapolation prediction, the next energy consumption segment data is compared with the predicted energy consumption value of the insulation effect attenuation prediction model under the same environmental conditions. The same environmental conditions refer to inputting parameters such as temperature and humidity from the environmental action time series data that are concurrent with the newly collected energy consumption segment into the prediction model to make the comparison benchmark consistent. The predicted energy consumption value here is the corresponding energy consumption value obtained by substituting the equivalent thermal resistance attenuation rate obtained by extrapolation into the building energy consumption calculation model.

[0067] If the comparison deviation (i.e., the absolute value of the relative error between the measured energy consumption value and the predicted energy consumption value) exceeds the acceptable threshold but is lower than the failure threshold, a fine-tuning operation of the insulation effect decay prediction model is triggered. Only the data-driven part is updated. Specifically, the posterior distribution parameters of the field correction factor in the Bayesian inversion framework are updated. That is, the energy consumption segment data at the extrapolated prediction time is used as the new observation data, and a Bayesian update is performed again to refresh the posterior distribution of the field correction factor to an updated posterior distribution containing new information.

[0068] If the comparison deviation is not lower than the failure threshold, a full re-prediction is triggered, and the inversion and extrapolation are re-executed.

[0069] The acceptable threshold and the failure threshold are dynamically adjusted based on the current uncertainty propagation amount. The relationship between the two is that the acceptable threshold is less than the failure threshold. The adjustment process involves multiplying the baseline threshold by an expansion coefficient that is positively correlated with the uncertainty propagation amount. This expansion coefficient is defined as 1 plus the normalized uncertainty propagation amount multiplied by a preset sensitivity factor. The normalized uncertainty propagation amount is obtained by dividing the current uncertainty propagation amount by the uncertainty propagation amount of the previous prediction period. When the uncertainty accumulation is high, both thresholds are relaxed simultaneously to tolerate greater fluctuations. When the uncertainty is low, the thresholds are tightened to improve sensitivity to anomalies.

[0070] By identifying and eliminating abnormal energy consumption segments and implementing self-checking and triggering re-prediction of prediction results, different processing methods are applied to different anomaly types. This improves the effective signal-to-noise ratio of input data, avoids misinterpreting intermittent energy consumption anomalies or sensor jump errors as real abrupt changes in insulation performance, and thus improves the accuracy of attenuation prediction. Furthermore, after completing the extrapolation prediction, data comparison is used to determine whether to trigger the re-prediction operation. This helps to refresh the prediction trajectory in a timely manner after environmental changes or sudden damage, prevents unacceptable deviations between historical patterns and current physical reality, and maintains the timeliness of operation and maintenance decisions.

[0071] In the second embodiment, under the premise that other contents remain unchanged, when executing the equivalent thermal resistance attenuation memory function, the initial zero value may lead to insensitivity to early damage. Therefore, it is necessary to assign a preset offset value to the initial value of the equivalent thermal resistance attenuation memory function based on construction defects to avoid the negative situation that the attenuation degree is seriously underestimated in the early stage of operation because the initial value cannot reflect the existing defects. Preferably, the initial value of the equivalent thermal resistance attenuation memory function is preset offset by the equivalent initial attenuation amount of construction defects. The specific process is as follows: Acquire construction acceptance data for the target exterior wall during the completion and acceptance phase, which refers to the quality verification stage after the main building structure and external envelope have been completed but not yet handed over for use. The construction acceptance data includes the insulation board joint width, the proportion of hollow areas, and anchor penetration information. The insulation board joint width represents the gap size between adjacent insulation boards, which can be collected by measuring multiple points on-site with a feeler gauge and taking the average value. The hollow area proportion represents the percentage of the total area of ​​cavities caused by bonding failure between the exterior wall insulation layer and the base layer to the total area of ​​the exterior wall, which can be collected by scanning and statistically analyzing each area using infrared thermography combined with tapping. Anchor penetration information represents the number of anchors per square meter of insulation layer and the density of thermal bridge points caused by penetration of the insulation layer, which can be collected by reviewing construction records and conducting on-site sampling verification.

[0072] The mapping relationship between each defect type and the local thermal resistance reduction coefficient was established using construction and acceptance data. The specific establishment process is as follows: The joint area ratio is calculated based on the joint width of the insulation board, and multiplied by the thermal conductivity enhancement factor of the thermal bridge at the joint to obtain the local thermal resistance reduction coefficient corresponding to the joint defect; the hollow area ratio is directly used as the reduction factor of the equivalent thermal resistance of the area; based on the anchor penetration information, the equivalent area influence of the point thermal bridge heat transfer model of a single metal anchor is calculated, and then multiplied by the anchor density to obtain the local thermal resistance reduction coefficient corresponding to the anchor thermal bridge.

[0073] After weighted summation, it is converted into the global initial attenuation amount. Specifically, the local thermal resistance reduction coefficient of each defect type is multiplied by its corresponding external wall area weight and then summed to obtain the global initial attenuation percentage representing the overall thermal defect level of the entire external wall. The global initial attenuation amount is a fixed offset component that does not change with time.

[0074] The equivalent thermal resistance decay memory function also integrates a saturation indicator function for irreversible damage accumulation. When the cumulative change of the corrected decay rate within a preset time window is lower than the preset minimum change threshold and the duration exceeds the preset stagnation judgment time, the saturation indicator function is triggered to output a saturation warning signal, and the saturation warning signal is used as a constraint condition to constrain the upper boundary of the extrapolation model.

[0075] Specifically, the above saturation indicator function means that it outputs 1 under the above conditions and outputs 0 when the above conditions are not met. The cumulative change refers to the absolute value of the difference between the decay rate at the end and the beginning of the window. The duration refers to the duration of the cumulative change being lower than the preset minimum change threshold. When the saturation indicator function outputs 1, the extrapolation model fixes the current decay rate to the upper limit of saturation and stops extrapolating.

[0076] The global initial attenuation amount is superimposed with the nonlinear attenuation component. The nonlinear attenuation component refers to the partial attenuation rate that increases over time, jointly output by the temperature and humidity coupled accelerated aging model and Bayesian inversion correction in the first embodiment, i.e., the component that increases over time driven by subsequent environmental effects. Together, they constitute a complete equivalent thermal resistance attenuation history, that is, the initial value of the equivalent thermal resistance attenuation memory function is shifted from zero to the global initial attenuation amount, so that the memory function has a non-zero initial attenuation level that reflects construction defects during the first calculation. If subsequent operation and maintenance repair data is input, the fixed offset component of the corresponding area is deducted. The deduction method is based on the proportion of the repair area to the entire exterior wall area, removing the partial local thermal resistance weakening coefficient corresponding to the repair area from the weighted summation term of the global initial attenuation amount, and recalculating the global initial attenuation amount after deduction to reflect the thermal resistance recovery effect.

[0077] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A method for predicting the thermal insulation effect of building exterior walls, characterized in that, include: S1. Obtain energy consumption data of the target exterior wall during the operation and maintenance phase, as well as the corresponding environmental impact time series data; S2. Construct a data-physical dual-driven prediction model for the attenuation of thermal insulation effect. The prediction model for the attenuation of thermal insulation effect includes a physical driving part, a data driving part, and an equivalent thermal resistance attenuation memory function, and outputs the current equivalent thermal resistance attenuation rate. S3. Extrapolate the current equivalent thermal resistance attenuation rate using an extrapolation model to predict the equivalent thermal resistance attenuation rate at future times, and convert it into a thermal insulation effect value and the corresponding building energy consumption increment.

2. The method for predicting the thermal insulation effect of building exterior walls according to claim 1, characterized in that: The physical drive section takes the environmental action time series data as input, establishes the theoretical decay function of equivalent thermal resistance with service time through the thermal aging mechanism of thermal insulation material, and outputs the theoretical decay rate. The data-driven part takes the energy consumption segment data as input, performs inverse correction on the theoretical attenuation function, and outputs the corrected attenuation rate.

3. The method for predicting the thermal insulation effect of building exterior walls according to claim 2, characterized in that: The theoretical decay function is a temperature and humidity coupled accelerated aging model based on the Arrhenius equation. It expresses the decay rate of the equivalent thermal resistance as the product of the temperature excitation function and the humidity excitation function, and introduces a nonlinear coupling term between the temperature excitation function and the humidity excitation function.

4. The method for predicting the thermal insulation effect of building exterior walls according to claim 3, characterized in that: The nonlinear coupling term is an exponential enhancement factor with the interaction term of temperature and relative humidity as variables. When both temperature and relative humidity enter their respective high-level ranges, the enhancement factor increases significantly. When only a single factor is at a high level, the enhancement factor degenerates into a weak enhancement effect close to 1.

5. The method for predicting the thermal insulation effect of building exterior walls according to claim 4, characterized in that: The inverse correction of the data-driven part is performed through a Bayesian inversion framework. The unknown parameters in the theoretical decay function are treated as random variables and assigned a prior distribution. The energy consumption fragment data is updated to a posterior distribution using the Bayesian formula, and the expected value of the posterior distribution is used as the corrected decay rate.

6. The method for predicting the thermal insulation effect of building exterior walls according to claim 5, characterized in that: The equivalent thermal resistance decay memory function takes the corrected decay rate and the historical inversion decay result as input, and outputs the current equivalent thermal resistance decay rate through time decay weighted fusion. In the time decay weighted fusion, the smaller the time distance between the historical time point and the current time point, the higher the weight is assigned, and the weight coefficient decays exponentially.

7. The method for predicting the thermal insulation effect of building exterior walls according to claim 6, characterized in that: The initial value of the equivalent thermal resistance attenuation memory function is preset and offset by the equivalent initial attenuation amount of construction defects. Specifically, the construction acceptance data of the target exterior wall during the completion and acceptance stage is obtained. The mapping relationship between each defect type and the local thermal resistance weakening coefficient is established through the construction acceptance data, and the weighted superposition is converted into the global initial attenuation amount. The equivalent thermal resistance decay memory function also integrates a saturation indicator function for irreversible damage accumulation. When the cumulative change of the corrected decay rate within a preset time window is lower than the preset minimum change threshold and the duration exceeds the preset stagnation judgment time, the saturation indicator function is triggered to output a saturation warning signal, and the saturation warning signal is used as a constraint condition to constrain the upper boundary of the extrapolation model. The global initial attenuation and the nonlinear attenuation component are superimposed to form a complete equivalent thermal resistance attenuation process.

8. The method for predicting the thermal insulation effect of building exterior walls according to claim 1, characterized in that: This also includes the identification and removal of abnormal energy consumption segments. The specific identification process is as follows: A theoretical baseline value for energy consumption without attenuation is generated by using initial thermal insulation performance parameters and concurrent environmental data. Each energy consumption segment is compared with the theoretical zero-attenuation energy consumption benchmark value. If the deviation result of the comparison exceeds the preset deviation threshold, it is marked as abnormal. If the anomaly type is temporary interference, it is directly removed; if the anomaly type is correctable, data correction is performed; if it belongs to other anomaly types, the likelihood weight in Bayesian inversion is reduced.

9. The method for predicting the thermal insulation effect of building exterior walls according to claim 1, characterized in that: It also includes self-checking of prediction results and triggering re-prediction, the specific process of which is as follows: After completing the extrapolation prediction, the data of the next energy consumption segment is compared with the predicted energy consumption value of the insulation effect decay prediction model under the same environmental conditions. If the comparison deviation exceeds the acceptable threshold but is below the failure threshold, a fine-tuning operation is triggered in the thermal insulation effect decay prediction model, updating only the data-driven part. If the comparison deviation is not lower than the failure threshold, a full re-prediction is triggered, and the inversion and extrapolation are re-executed. The acceptable threshold and failure threshold are dynamically adjusted based on the current uncertainty propagation.

10. A system for predicting the thermal insulation effect of building exterior walls, characterized in that: include: The data acquisition module is used to acquire energy consumption data segments of the target exterior wall during the operation and maintenance phase, as well as corresponding environmental impact time-series data. The model building module is used to build a data-physical dual-driven thermal insulation effect attenuation prediction model. The thermal insulation effect attenuation prediction model includes a physical driving part, a data driving part, and an equivalent thermal resistance attenuation memory function, and outputs the current equivalent thermal resistance attenuation rate. The thermal insulation effect prediction module is used to extrapolate the current equivalent thermal resistance attenuation rate through an extrapolation model, predict the equivalent thermal resistance attenuation rate at future times, and convert it into a thermal insulation effect value and the corresponding building energy consumption increment.

Images