Building carbon emission prediction method and system based on BO-LGBM

Abstract

The invention belongs to the technical field of green buildings, and particularly discloses a building carbon emission prediction method and system based on BO-LGBM. Comprising the following steps of: analyzing correlation among building carbon emission influence factors, selecting characteristics which have obvious correlation with building carbon emission so as to construct an index system of city-level building full-life-cycle carbon emission, and constructing a data set according to the index system; bayesian is adopted to optimize hyper-parameters of the LGBM model, based on the optimal hyper-parameters, the LGBM model is set and adjusted, and a building full life cycle carbon emission prediction model is established; and carrying out optimization training on a building full-life-cycle carbon emission prediction model by adopting the data set so as to construct a nonlinear relationship between the influence factors and building full-life-cycle carbon emission. According to the method, the prediction precision of the carbon emission in the whole life cycle of the building is improved by combining the LGBM model and the Bayesian optimization method, and scientific decision support is provided for energy conservation and emission reduction of the building industry.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of green building, more specifically, relates to a building carbon emission prediction method and system based on BO-LGBM. BACKGROUND

[0002] Building carbon emissions are a key to addressing global climate change, as they account for more than 40% of global carbon emissions and consume more than 20% of China's total energy use. According to the "2022 China Building Energy Consumption and Carbon Emission Research Report" released by the China Building Energy Efficiency Association (CABEE) in 2022

[0003] It can be seen that building energy consumption and building material manufacturing are the main sources of building carbon emissions.

[0004] With urbanization and economic growth, building energy consumption in China is increasing year by year. In the process of urbanization, the increase of newly built buildings and the energy consumption of existing buildings have an impact on carbon emissions. The construction industry in China has great potential in reducing carbon emissions as it is the industry with the highest resource consumption and environmental load in today's society. However, the construction sector in China faces severe challenges in terms of emission reduction. First, building energy consumption continues to rise throughout the entire process, and the growth rate of building energy consumption still exceeds the overall average level, reflecting the existence of a series of obstacles in energy efficiency in the current construction industry. Second, the total amount of building carbon emissions continues to rise. Third, the challenge faced by the construction industry comes from the difficulty of energy saving and reconstruction of existing buildings. Many old buildings have low energy efficiency and are difficult to effectively renovate. At the same time, the lack of unified standards and supervision has led some new buildings to fail to fully consider energy saving and emission reduction requirements during the design and construction phase, which has made the existing building stock a bottleneck for emission reduction and hindered the green transformation of the entire industry. Traditional building carbon emission research is mostly based on regression statistical models. These methods are simple and fast, but have the problems of insufficient data information mining and low precision. SUMMARY

[0005] In view of the above defects or improvement needs of the prior art, the present application provides a building carbon emission prediction method and system based on BO-LGBM, which integrates LGBM model and Bayesian optimization. This method not only improves the prediction accuracy of building life cycle carbon emissions, but also automates the hyperparameter tuning process, thereby improving the generalization ability of the model. This method can accurately capture the complex nonlinear relationship between influencing factors and carbon emissions and quantify the importance of each factor, providing data-driven decision support for energy saving and emission reduction strategies in the construction industry. This helps to optimize the building design, construction and operation process, achieve more effective energy management and carbon emission control, and thus promote the realization of sustainable development and environmental goals.

[0006] To achieve the above object, according to one aspect of the present application, a building carbon emission prediction method based on BO-LGBM is provided, comprising the following steps:

[0007] Step one, the correlation between the influencing factors of building carbon emission is analyzed, the characteristics with significant correlation with building carbon emission are selected to construct the index system of city-level building full life cycle carbon emission, and the data set is constructed according to the index system;

[0008] Step two, the hyperparameters of LGBM model are optimized by using Bayesian, and the LGBM model is set and adjusted based on the optimal hyperparameters to establish the building full life cycle carbon emission prediction model;

[0009] Step three, the building full life cycle carbon emission prediction model is optimized and trained by using the data set to construct the nonlinear relationship between the influencing factors and the building full life cycle carbon emission;

[0010] Step four, the reliability of the building full life cycle carbon emission prediction model is verified;

[0011] Step five, the trained prediction model is used to predict the building full life cycle carbon emission, and the importance of each influencing factor is analyzed.

[0012] As a further optimization, in step one, the index system comprises:

[0013] Unit area building material production carbon emission X1, construction area X2, building area X3, power consumption X4, power grid factor X5, per capita regional total output value X6, proportion of tertiary industry X7, total population X8, urbanization rate X9, energy structure X 10 , climate zone X 11 ;

[0014] As a further optimization, in step one, the calculation model of building full life cycle carbon emission comprises:

[0015] E build (t)=exp(α·ln(η(t)·ρ(t)))+β·(1+γ·ξ(t)) δ ·(1+κ′·E e (t) μ ·U(t) v )

[0016] Wherein, E build(t) is the carbon emission index of the building at time t, a is the first adjustment factor of the building carbon emission calculation model, η(t) is the energy efficiency coefficient of the building at time t, ρ(t) is the personnel density of the building at time t, β is the second adjustment factor of the building carbon emission calculation model, γ is the third adjustment factor of the building carbon emission calculation model, ξ(t) is the energy consumption of the building at time t, which is used to describe the energy consumption of the building affected by the season, δ is the fourth adjustment factor of the building carbon emission calculation model, κ' is the fifth adjustment factor of the building carbon emission calculation model, E e (t) is the influence index of the environment on carbon emission at time t, μ is the sixth adjustment factor of the building carbon emission calculation model, U(t) is the actual energy demand of the building at time t, v is the seventh adjustment factor of the building carbon emission calculation model.

[0017] As a further preferred, step two comprises the following steps:

[0018] (21) Determine the hyperparameters that need to be optimized in the LGBM model and their possible value ranges;

[0019] (22) Take the regression error of the LGBM model as the objective function, use the Bayesian optimization library to initialize the optimization process, define the objective function, the search space of hyperparameters and the optimization algorithm;

[0020] (23) Run the Bayesian optimization algorithm, specify the number of iterations, and in each iteration, the Bayesian optimization algorithm will select a set of hyperparameters and evaluate its performance using the objective function;

[0021] (24) After completing the optimization, obtain the optimal hyperparameter combination from the Bayesian optimizer;

[0022] (25) Based on the optimal hyperparameters, set and adjust the LGBM model and establish a building life cycle carbon emission prediction model.

[0023] As a further preferred, step (25) comprises the following steps:

[0024] The LGBM model uses a pre-ordered decision tree, uses parallel learning of parallel voting DT in the training process, and uses the Leaf-wise method to find suitable leaves in the optimization process, wherein the objective function of the LGBM model is:

[0025] Obj(t) = L(t) + Ω(t) + c

[0026] In the formula, L(t) and Ω(t) represent the regularization function and the loss function respectively, and c and t represent the additional parameter and the sampling time respectively;

[0027] The loss function of the LGBM model includes:

[0028]

[0029] wherein y i is a model fitting degree, is a predicted output;

[0030] The objective function is further represented as:

[0031]

[0032] wherein f i is the i-th parameter or parameter vector in the model.

[0033] As a further preferred, the analysis of the importance of each influencing factor comprises: using the LGBM algorithm to evaluate the importance of the input influencing factors, calculating the importance score of the influencing factors, and the importance of feature j in a tree is:

[0034]

[0035] wherein k is the number of leaf nodes, G is the total return, v t represents the feature associated with the node, and when the number of trees is M, the global importance of the feature is:

[0036]

[0037] wherein Tm represents the importance of the m-th tree.

[0038] As a further preferred, the fitting degree R 2 , the root mean square error RMSE and the mean absolute error MAE are used to verify the reliability of the building life cycle carbon emission prediction model:

[0039]

[0040]

[0041] wherein y obs is the corresponding data in the data set, y pred is the prediction model result, is the average value of the data, and n is the sample size.

[0042] According to another aspect of the present application, a building carbon emission prediction system based on BO-LGBM is also provided, comprising:

[0043] a first master control module, configured to analyze the correlation between the influencing factors of building carbon emissions, select features with significant correlation with building carbon emissions, construct an index system of municipal building life cycle carbon emissions, and construct a data set according to the index system.

[0044] The second master module is configured to optimize the hyperparameters of the LGBM model using Bayesian optimization, and to set and adjust the LGBM model based on the optimal hyperparameters to establish a building life cycle carbon emission prediction model.

[0045] The third master module is configured to optimize and train the building life cycle carbon emission prediction model using the data set to build a nonlinear relationship between the influencing factors and the building life cycle carbon emission.

[0046] The fourth master module is configured to verify the reliability of the building life cycle carbon emission prediction model.

[0047] The fifth master module is configured to use the trained prediction model to predict the building life cycle carbon emission and analyze the importance of each influencing factor.

[0048] According to another aspect of the present application, an electronic device is also provided, which comprises at least one processor, at least one memory and a communication interface connected in communication with the processor, wherein the processor, the memory and the communication interface communicate with each other, the memory stores program instructions executable by the processor, and the processor invokes the program instructions to execute the BO-LGBM-based building carbon emission prediction method of any of the above embodiments or a combination of multiple embodiments.

[0049] According to another aspect of the present application, a readable storage medium is also provided, which stores computer instructions, and the computer instructions enable a computer to execute the BO-LGBM-based building carbon emission prediction method of any of the above embodiments or a combination of multiple embodiments.

[0050] Overall, compared with the prior art, the above technical solutions conceived by the present application mainly have the following technical advantages:

[0051] 1. By using the LGBM model, we can effectively capture the nonlinear relationship between the building life cycle carbon emission and multiple influencing factors. As an integrated learning method based on gradient boosting, the LGBM model can improve the prediction accuracy by constructing multiple decision trees. In addition, combining Bayesian optimization to adjust the hyperparameters of LGBM can further optimize the model performance, find the best parameter combination to minimize the prediction error such as mean square error (MSE) or root mean square error (RMSE), and thus improve the prediction accuracy of the model.

[0052] 2.The present application predicts the objective function (such as the MSE of LGBM) by establishing a proxy model, and uses the acquisition function (such as the expected improvement EI) to balance exploration and utilization, automatically finding the optimal combination of hyperparameters. This method not only saves the time and effort of manual parameter tuning, but also can find better parameter settings, improving the generalization ability and prediction performance of the model.

[0053] 3.Through the feature importance analysis of the LGBM model, we can identify and quantify the contribution of each influencing factor to the carbon emissions of the building life cycle, which provides valuable insights for policymakers and architects. Understanding which factors have the greatest impact on carbon emissions can help them develop more effective energy-saving and emission-reduction strategies, such as optimizing the production process of building materials, improving energy efficiency, adjusting the energy structure, etc. In addition, accurate carbon emission prediction is of great significance for evaluating the impact of climate change policies and promoting sustainable development. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 is a flowchart of a building carbon emission prediction method based on BO-LGBM according to an embodiment of the present application;

[0055] Figure 2 is a correlation coefficient diagram of influencing factors according to an embodiment of the present application;

[0056] Figure 3 is a schematic diagram of the hyperparameter optimization process according to an embodiment of the present application;

[0057] Figure 4 is a schematic diagram of the building life cycle carbon emission test set results according to an embodiment of the present application;

[0058] Figure 5 is a schematic diagram of the building life cycle carbon emission influencing factor importance score according to an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0060] As shown in Figure 1 , the building carbon emission prediction method based on BO-LGBM provided by an embodiment of the present application comprises the following steps:

[0061] Step one, the correlation between the influencing factors of building carbon emissions is analyzed, the characteristics with significant correlation with building carbon emissions are selected to construct the index system of city-level building full life cycle carbon emissions, and the data set is constructed according to the index system; the index system includes:

[0062] Carbon emissions per unit area of building materials production X1, construction area X2, building area X3, power consumption X4, power grid factor X5, per capita regional total output value X6, proportion of the tertiary industry X7, total population X8, urbanization rate X9, energy structure X 10 , climate zone X 11 .

[0063] The analysis of the correlation between the influencing factors in the data includes: the Pearson correlation coefficient is used to calculate the correlation coefficient between the influencing factors, and the absolute value of the correlation coefficient is sorted, and the variables with significant correlation are removed, and the calculation formula of the Pearson correlation coefficient is as follows:

[0064]

[0065] In the formula, cov(X, Y) is the covariance between variables X and Y, μ X is the mean of variable X, μ Y is the mean of variable Y, σ X is the standard deviation of variable X, σ Y is the standard deviation of variable Y.

[0066] More specifically, in the present application, the calculation model of building full life cycle carbon emissions includes:

[0067] E build (t)=exp(α·ln(η(t)·ρ(t)))+β·(1+γ·ξ(t)) δ ·(1+κ′·E e (t) μ ·U(t) v )

[0068] Wherein, E build(t) is the carbon emission index of the building at time t, a is the first adjustment factor of the building carbon emission calculation model, which mainly adjusts the comprehensive influence degree of energy efficiency coefficient and personnel density on building carbon emission, embodies the comprehensive adjustment effect of building on carbon emission in energy efficiency utilization and personnel density distribution, the higher the energy efficiency and the more reasonable the personnel density distribution, the lower the carbon emission is generally, a plays a quantitative adjustment function on this relationship, η(t) is the energy efficiency coefficient of the building at time t, p(t) is the personnel density of the building at time t, β is the second adjustment factor of the building carbon emission calculation model, which is used to adjust the influence degree of building carbon emission under the joint action of many factors such as seasonal influenced energy consumption, environmental influence index on carbon emission and actual energy demand of building. It comprehensively considers the complex influence of environmental factors (such as energy consumption change caused by seasonal change of temperature, humidity, etc.) and energy demand on building carbon emission, quantifies and adjusts the comprehensive influence of these factors on carbon emission by combining with other parameters, so that the building carbon emission calculation model is closer to the actual carbon emission situation caused by the change of many factors. γ is the third adjustment factor of the building carbon emission calculation model, which mainly adjusts the influence intensity of seasonal influenced energy consumption of building on carbon emission, which can reflect the contribution degree of seasonal energy consumption change to carbon emission, so as to make the model more accurately consider the influence of seasonal factors on building carbon emission and make the calculation result more accurate. ξ(t) is the energy consumption of the building at time t, which is used to describe the seasonal influenced energy consumption of the building, δ is the fourth adjustment factor of the building carbon emission calculation model, which further adjusts and corrects the carbon emission under the cross influence of many factors such as seasonal influenced energy consumption, environmental influence index on carbon emission and actual energy demand of building, κ' is the fifth adjustment factor of the building carbon emission calculation model, which is used to adjust the comprehensive influence intensity of environmental influence index on carbon emission and actual energy demand of building in carbon emission calculation, which reflects the comprehensive influence degree of the interaction between environmental factors (such as wind direction, sunshine, etc.) and building energy demand on carbon emission, E e (t) is the environmental influence index on carbon emission at time t, μ is the sixth adjustment factor of the building carbon emission calculation model, which mainly adjusts the weight of environmental influence index on carbon emission in carbon emission calculation. Environmental factors have direct or indirect influence on building carbon emission, such as air quality, climate conditions, etc., which will affect the energy utilization efficiency and carbon emission, U(t) is the actual energy demand of the building at time t, v is the seventh adjustment factor of the building carbon emission calculation model, which is used to adjust the influence of actual energy demand of building in carbon emission calculation.

[0069] The energy efficiency coefficient η(t) of the building at time t includes:

[0070]

[0071] wherein η0is an initial energy efficiency coefficient, a T is a weight of the ambient temperature, T(t) is the ambient temperature at time t, T base is a reference temperature, λ T is an adjustment factor of the ambient temperature, β H is a weight of the ambient humidity, H(t) is the ambient humidity at time t, H base is a reference humidity, γ' is an adjustment factor of the ambient humidity, a W is a weight of the ambient wind speed, W(t) is the ambient wind speed at time t, a L is a weight of the light intensity, L(t) is the light intensity at time t, L max is a maximum value of the light intensity.

[0072] The energy consumption ξ(t) of the building at time t comprises:

[0073]

[0074] wherein ζ0is a reference energy consumption, λ κ is a weight of the thermal conductivity, κ(t) is the thermal conductivity of the building at time t, a κ is an adjustment factor of the thermal conductivity, δ C is a weight of the air permeability, C(t) is the air permeability of the building at time t, β C is an adjustment factor of the air permeability, p a (t) is the air density inside the building at time t, γ ρ is an adjustment factor of the air density, V is the volume of the building, δ V is an adjustment factor of the volume of the building, μ F is a weight of the wind speed through the interior of the building, F(t) is the wind speed through the interior of the building at time t, a F is an adjustment factor of the wind speed through the interior of the building.

[0075] The environmental impact index E e (t) on carbon emissions at time t comprises:

[0076]

[0077] wherein a1is a first adjustment factor of the environmental impact index on carbon emissions, b1is a second adjustment factor of the environmental impact index on carbon emissions, g1is a third adjustment factor of the environmental impact index on carbon emissions, A veg (t) is the vegetation coverage of the region where the building is located at time t, d1is a fourth adjustment factor of the environmental impact index on carbon emissions, z1is a fifth adjustment factor of the environmental impact index on carbon emissions, C abs(t) is the CO2 absorption efficiency of the region where the building is located at time t.

[0078] The actual energy demand U(t) of the building at time t includes:

[0079]

[0080] wherein a2 is the first adjustment factor of the actual energy demand, b2 is the second adjustment factor of the actual energy demand, T in (t) is the temperature in the building at time t, g2 is the third adjustment factor of the actual energy demand, d2 is the fourth adjustment factor of the actual energy demand, H in (t) is the humidity in the building at time t, k" is the fifth adjustment factor of the actual energy demand.

[0081] In the present application, the concept of top-down method and the weighted average strategy based on DN value are adopted to obtain the carbon emission of the municipal building in the whole life cycle. The definition is as follows:

[0082]

[0083] In the formula, CT z = [CT1, CT2, …, CT k ] T is the carbon emission data of different cities in the zth province, DN = [DN1, DN2, …, DN k ] T is the DN value of different cities in the zth province.

[0084] Step two, the hyperparameters of the LGBM model are optimized by using Bayesian, and based on the optimal hyperparameters, the LGBM model is set and adjusted to establish the building life cycle carbon emission prediction model. It specifically includes the following steps:

[0085] (21) Determine the hyperparameters to be optimized in the LGBM model and their possible value range; the hyperparameters include learning rate (learning_rate), minimum child sample number (min_child_samples), maximum leaf number (num_leaves), L1 regularization (reg_alpha) and L2 regularization (reg_lambda).

[0086] (22) Take the regression error of the LGBM model as the objective function, use the Bayesian optimization library to initialize the optimization process, define the objective function, hyperparameter search space and optimization algorithm;

[0087] (23) Run the Bayesian optimization algorithm, specify the number of iterations, and in each iteration, the Bayesian optimization algorithm will select a set of hyperparameters and evaluate its performance using the objective function;

[0088] (24) After the optimization is completed, the optimal hyperparameter combination is obtained from the Bayesian optimizer;

[0089] (25) Based on the optimal hyperparameters, the LGBM model is set and adjusted to establish a building full life cycle carbon emission prediction model.

[0090] Step (25) includes the following steps:

[0091] The LGBM model adopts a decision tree based on preordering. In the training process, parallel learning of parallel voting DT is used. In the optimization process, the LGBM model adopts a Leaf-wise method to find suitable leaves. The objective function of the LGBM model is:

[0092] Obj(t) = L(t) + Omega(t) + c

[0093] Where L(t) and Omega(t) represent the regularization function and the loss function, respectively, and c and t represent the additional parameter and the sampling time, respectively.

[0094] The loss function of the LGBM model includes:

[0095]

[0096] Where y i is the model fitting degree, is the predicted output.

[0097] The objective function is further expressed as:

[0098]

[0099] Where f i is the i-th parameter or parameter vector in the model.

[0100] In this embodiment, the predicted output of the LGBM model is defined as:

[0101]

[0102] Where, is the predicted building full life cycle carbon emission. x is the feature vector of the influencing factors, including unit area building material production carbon emission, construction area, building area, power consumption, power grid factor, etc. Theta is the hyperparameter set of the LGBM model, f(x; theta) is the prediction function of the LGBM model, alpha i is the weight of the i-th decision tree, and n is the total number of decision trees.

[0103] The objective function of Bayesian optimization is defined as the mean square error of the LGBM model:

[0104]

[0105] Wherein, yi is the actual building life cycle carbon emissions. f(xi; theta) is the prediction value of the LGBM model for the i th sample. n is the number of samples.

[0106] The proxy model of Bayesian optimization is constructed.

[0107] In the present application, the Bayesian algorithm adopts a Gaussian process (GP) proxy model. Gaussian process is an extension of multivariate Gaussian distribution to infinite-dimensional random process, which considers a limited number of observation combinations and follows a multivariate Gaussian distribution characterized by mean and covariance. Therefore, Gaussian process can be completely defined by its mean and covariance. Multivariate Gaussian model:

[0108]

[0109] Mu is the mean, and cov is the covariance.

[0110] In the present application, the acquisition function is used to obtain the optimal solution of the objective function. Exploration and exploitation are two functions of the acquisition function, which should be balanced during each iteration to ultimately find the parameter with the highest prediction accuracy and determine the final parameter. Common acquisition functions include expected improvement (EI), improvement probability (PI), upper confidence bound (UCB) and lower confidence bound (LCB). EI is used as the acquisition function:

[0111]

[0112] Phi (·) and Respectively represent the cumulative distribution probability and probability distribution function of the standard Gaussian distribution; y best Is the optimal value in the current sample space.

[0113] Step three, using the data set to optimize and train the building life cycle carbon emission prediction model, to construct the nonlinear relationship between the influencing factors and the building life cycle carbon emission.

[0114] Step four, verify the reliability of the building life cycle carbon emission prediction model;

[0115] The goodness of fit R 2 , root mean square error RMSE and mean absolute error MAE are used to verify the reliability of the building life cycle carbon emission prediction model:

[0116]

[0117] Wherein, y obs Is the corresponding data in the data set, y pred Is the prediction model result, is the data average value, n is the sample size.

[0118] Step five, using the trained prediction model to predict the building full life cycle carbon emissions, and analyzing the importance of each influencing factor.

[0119] The analysis of the importance of each influencing factor includes: using the LGBM algorithm to evaluate the importance of the input influencing factors, calculating the importance score of the influencing factors, and the importance of feature j in a tree is:

[0120]

[0121] In the formula, k is the number of leaf nodes, G is the total return, v t is the feature associated with the node, and when the number of trees is M, the global importance of the feature is:

[0122]

[0123] In the formula, Tm represents the importance of the mth tree.

[0124] Embodiment 2

[0125] In this embodiment, a building carbon emission prediction method based on BO-LGBM includes the following steps:

[0126] 1. Construction of index system

[0127] Since the influence of urban building carbon emissions is affected by economic level, urban development, energy consumption and other factors, based on a large number of literature summaries and the availability of municipal data, 11 important influencing factors are selected from five aspects to construct the index system of municipal building full life cycle carbon emissions, as shown in the following table:

[0128] Table 1 Index system of building full life cycle carbon emissions at municipal scale

[0129]

[0130] The above different building full life cycle carbon emission influencing factors may have complex coupling relationship, and the high correlation between parameters not only increases the complexity of the training model, but also reduces the accuracy of the model. In order to analyze the relationship between influencing factors, Pearson correlation coefficient is used to calculate the correlation coefficient between influencing factors. According to literature research, the correlation coefficient is between-1 and 1, the greater the absolute value, the stronger the correlation between variables, and the closer to 0, the weaker the correlation between variables. Generally, when the absolute value of the correlation coefficient between two variables is greater than 0.9, it indicates that there is significant correlation between variables, and the variables need to be removed. The calculation formula of Person correlation coefficient is as follows:

[0131]

[0132] In this embodiment, in order to further ensure the data quality, the correlation coefficient between the indicators is calculated, and the correlation matrix of the Pearson correlation coefficient between the indicators is obtained as shown in Figure 2 It can be seen from the figure that the correlation coefficient between the indicators is between-0.47 and 0.75, and all the coefficients are less than 0.9, which verifies that there is no high correlation data in the data set that will affect the model result, and the indicator system and the data set are reliable.

[0133] 2. Construction of municipal building carbon emission prediction model based on BO-LGBM

[0134] (1) Bayesian optimization of hyperparameters

[0135] For the LGBM algorithm, the selection of hyperparameters will directly affect the final regression prediction result. In order to ensure the effectiveness of the prediction result, when using the LGBM prediction algorithm to train the sample, it is necessary to first adjust the important parameters in the prediction model. The LGBM prediction model needs to adjust two main parameters, the first parameter is num_leave, that is, the number of leaves of each decision tree, and its main function is to control the complexity of the tree model. When the number of leaves is the same, because the tree that adopts the leaf-by-leaf growth strategy is much deeper than the tree that adopts the horizontal leaf growth strategy, so overfitting is common. The second parameter to be adjusted is the minimum number of leaf node samples min_child_samples, which is used to prevent overfitting. The setting range of these parameters will affect the establishment of the tree model and its regression fitting performance. In addition to the above two hyperparameters, four other hyperparameters such as learning rate are optimized in this embodiment, as shown in Table 2.

[0136] Table 2 Introduction of LGBM hyperparameters

[0137]

[0138] After obtaining the indicator system data, the data set of the prediction model is constructed, and the data set is divided according to the ratio of 4:1, the first 80% is used as the training set to train the prediction model, and the last 20% is used as the test set to evaluate the model precision. In order to make the LGBM training model obtain excellent prediction effect, this embodiment optimizes the six hyperparameters of the model based on the Python platform using Bayesian, the number of iterations of optimization is set to 200, and the mean square error MSE is used as the basis to verify the precision of the LGBM model. The Bayesian hyperparameter search process is shown in Figure 3 The optimal hyperparameter combination is shown in the following table. From Figure 3The super parameter optimization process can be seen that the loss of the prediction model decreases with the increase of the iteration number, which shows that the super parameter of the Bayesian optimization model is effective, and the prediction model realizes the minimum MSE at the 36th generation, which shows that the prediction accuracy of the LGBM model is the best at this time.

[0139]

[0140] (2) BO-LGBM prediction model construction

[0141] LGBM adopts the decision tree (DT) technology based on preordering, and uses parallel learning of parallel voting DT in the training process, which allows parallel learning of the model. In the optimization process, LGBM adopts the Leaf-wise method to find suitable leaves. The objective function of the LGBM model is given below:

[0142] Obj(t) = L(t) + Omega(t) + c

[0143] In the formula, L(t) and Omega(t) represent the regular function and the loss function respectively, and c and t represent the additional parameter and the sampling time respectively. The additional parameter c can prevent overfitting and optimize the depth of the tree, and the regular function L(t) reflects the complexity of the model. The loss function represents the fitness of the model obtained by comparing the actual value and the predicted output of N samples, which can be defined as:

[0144]

[0145] In the formula, y i is the model fitting degree, is the predicted output;

[0146] The objective function is further represented as:

[0147]

[0148] In the formula, f i is the i-th parameter or parameter vector in the model

[0149] After the model is optimized by BO, the optimal super parameter combination is input into the LGBM model, and the BO-LGBM model continuously learns and trains the nonlinear relationship between the input variables and the output variables in the training set, thereby establishing a stable performance municipal building full life cycle prediction model.

[0150] 3. Model accuracy evaluation

[0151] After establishing the BO-LGBM prediction model, in order to verify the reliability of the BO-LGBM model, the prediction accuracy needs to be evaluated, and in this embodiment, the goodness of fit (R 2), root mean square error (RMSE) and mean absolute error (MAE). R 2 The closer to 1, the better the model fitting effect. The lower the RMSE and MAE, the smaller the error of the model, and the higher the performance of the model.

[0152] After determining the optimal hyperparameters by Bayesian optimization, the model is set adjusted and the building life cycle carbon emission prediction model is established. According to R 2 , RMSE and MAE, the prediction accuracy of the model is evaluated. The building life cycle carbon emission prediction results obtained based on the BO-LGBM model are shown in Figure 4 .

[0153] It can be seen from Figure 4 that the predicted value and the simulated value of the building life cycle carbon emission at the city level on the training set are basically consistent, R 2 , RMSE and MAE are 0.981, 4.339 and 2.182 respectively, the model fitting effect is good, which shows that the BO-LGBM model well establishes the nonlinear relationship between the influencing factors and the building life cycle carbon emission.

[0154] 4. Importance analysis of influencing factors

[0155] In order to better understand how each index affects the city-level building life cycle carbon emission, so as to better develop building energy saving and emission reduction strategies, the embodiment uses LGBM algorithm to evaluate the importance of input indexes, and calculates the importance score of the indexes.

[0156] In order to further explore the internal relationship between the influencing factors and the city-level building life cycle carbon emission, and improve the transparency and interpretability of the prediction model, the embodiment uses the advantage of LGBM model in interpretability to quantitatively analyze the influence intensity of 11 influencing factors. According to the global importance calculation formula, the importance ranking of 11 influencing factors is obtained, as shown in Figure 5 .

[0157] As can be seen from the figure, the top three factors affecting the city building life cycle carbon emission are unit area building material production carbon emission, building area and power grid factor, and their importance values are significantly higher than those of other factors. The importance ranking of other factors is in turn construction area, energy structure, climate zone, power consumption, total population, urbanization rate, per capita regional GDP, and the proportion of the third industry.

[0158] In summary, the present application estimates the building carbon emissions at the municipal scale from the perspective of the whole life cycle by using IPCC, LCA and night light data, solves the problem of difficult to obtain statistical data at small scale, and provides a data basis for establishing a building whole life cycle carbon emission prediction model at the municipal scale. The BO-LGBM building whole life cycle carbon emission prediction model at the municipal scale established in this paper can well learn the complex nonlinear relationship between the input indicators and the prediction target, and the R 2 respectively, the RMSE is 4.339 and 9.067, and the MAE is 2.182 and 2.124, respectively, which realizes the fine prediction of the building whole life cycle carbon emission. According to the LGBM importance score, the importance of the building carbon emission influencing factors at the municipal scale is in the order of unit area building material production carbon emission, building area, power grid factor, construction area, energy structure, climate zone, power consumption, total population, urbanization rate, per capita regional GDP, and the proportion of the tertiary industry. For urban building carbon emission reduction, the carbon emission of building material production stage, the total area of existing buildings and the cleanliness of power grid need to be controlled.

[0159] According to another aspect of the present application, a BO-LGBM-based building carbon emission prediction system is also provided for implementing the method related to any of the above embodiments, comprising:

[0160] The first main control module is used for analyzing the correlation between the building carbon emission influencing factors, selecting the features with significant correlation with the building carbon emission, constructing an index system of the municipal building whole life cycle carbon emission, and constructing a data set according to the index system;

[0161] The second main control module is used for optimizing the hyperparameters of the LGBM model by using Bayesian, setting and adjusting the LGBM model based on the optimal hyperparameters, and establishing a building whole life cycle carbon emission prediction model;

[0162] The third main control module is used for optimizing and training the building whole life cycle carbon emission prediction model by using the data set, so as to construct the nonlinear relationship between the influencing factors and the building whole life cycle carbon emission;

[0163] The fourth main control module is used for verifying the reliability of the building whole life cycle carbon emission prediction model;

[0164] The fifth main control module is used for predicting the building whole life cycle carbon emission by using the trained prediction model, and analyzing the importance of each influencing factor.

[0165] The method of the embodiment of the present application is realized by relying on an electronic device, so it is necessary to introduce the related electronic device. For this purpose, the embodiment of the present application provides an electronic device, such asFigure 3 As shown in the figure, the electronic device includes at least one processor, a communications interface, at least one memory, and a communications bus, wherein the at least one processor, the communications interface, and the at least one memory complete mutual communication through the communications bus. The at least one processor can invoke a logical instruction in the at least one memory to execute all or part of the steps of the method provided by the foregoing various method embodiments.

[0166] In addition, the logical instruction in the at least one memory described above can be implemented in the form of a software functional unit and sold or used as an independent product, and can be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various method embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0167] The device embodiments described above are only schematic, wherein the units shown as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0168] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware. Based on such understanding, the technical solutions described above essentially or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in the various embodiments or some parts of the embodiments.

[0169] The computer program product of the present application can be a computer program product comprising a computer-readable medium bearing computer program code embodied therein for use with a computer. In operation, the computer program code causes a computer to perform the steps of the present application. The computer program code can be written in any form suitable for use with the computer for example, in the form of source code, object code, interpretative code, etc. The computer program code can be supplied to a user on a computer-readable medium such as a floppy disk, CD ROM, or over a computer network such as the Internet. The computer-readable medium can be magnetic, optical, or other computer storage medium. The computer-readable medium can be a distributed computer-readable medium that is remotely accessible via the Internet or other computer network.

[0170] Those skilled in the art will readily understand that the above description is only the preferred embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall fall within the scope of the present application.

Claims

1. A building carbon emission prediction method based on BO-LGBM, characterized in that: The following steps are involved: Step 1: Analyze the correlation between factors affecting building carbon emissions, select features with significant correlation with building carbon emissions, and construct an indicator system for the life cycle carbon emissions of municipal buildings. Then, construct a data set based on this indicator system. Step 2: Use Bayesian optimization to optimize the hyperparameters of the LGBM model. Based on the optimal hyperparameters, adjust the settings of the LGBM model and establish a building life cycle carbon emission prediction model. Step 3: Using the dataset to optimize and train a building life cycle carbon emission prediction model to construct a nonlinear relationship between influencing factors and building life cycle carbon emissions; Step 4: Verify the reliability of the building life cycle carbon emission prediction model; Step 5: Use the trained prediction model to predict the carbon emissions of the building throughout its life cycle.

2. The building carbon emission prediction method based on BO-LGBM according to claim 1 is characterized in that: In step 1, the indicator system includes: Carbon emissions per unit area of ​​building materials production X1, construction area X2, building area X3, electricity consumption X4, grid factor X5, per capita regional gross output value X6, proportion of tertiary industry X7, total population X8, urbanization rate X9, energy structure X 10 、Climate zone X 11 .

3. The building carbon emission prediction method based on BO-LGBM according to claim 1 is characterized in that: In step 1, the calculation model for building life cycle carbon emissions includes: E build (t)=exp(α·ln(η(t)·ρ(t)))+β·(1+γ·ξ(t)) δ ·(1+κ′·E e (t) μ ·U(t) v ) Among them, E build (t) is the carbon emission index of the building at time t, α is the first adjustment factor of the building carbon emission calculation model, η(t) is the energy efficiency coefficient of the building at time t, ρ(t) is the occupant density of the building at time t, β is the second adjustment factor of the building carbon emission calculation model, γ is the third adjustment factor of the building carbon emission calculation model, ξ(t) is the energy consumption of the building at time t, which is used to describe the energy consumption of the building affected by the season, δ is the fourth adjustment factor of the building carbon emission calculation model, κ′ is the fifth adjustment factor of the building carbon emission calculation model, E e (t) is the environmental impact index on carbon emissions at time t, μ is the sixth adjustment factor of the building carbon emission calculation model, U(t) is the actual energy demand of the building at time t, and v is the seventh adjustment factor of the building carbon emission calculation model.

4. The building carbon emission prediction method based on BO-LGBM according to claim 1 is characterized in that: Step 2 includes the following steps: (21) Determine the hyperparameters that need to be optimized in the LGBM model and their possible value ranges; (22) The regression error of the LGBM model is used as the objective function, and the Bayesian optimization library is used to initialize the optimization process, define the objective function, hyperparameter search space and optimization algorithm; (23) Run the Bayesian optimization algorithm and specify the number of iterations. In each iteration, the Bayesian optimization algorithm will select a set of hyperparameters and evaluate its performance using the objective function. (24) After the optimization is completed, the optimal hyperparameter combination is obtained from the Bayesian optimizer; (25) Based on the optimal hyperparameters, the LGBM model is adjusted and a building life cycle carbon emission prediction model is established.

5. The building carbon emission prediction method based on BO-LGBM according to claim 4 is characterized in that: Step (25) comprises the following steps: The LGBM model uses a pre-sorted decision tree. During the training process, parallel learning with parallel voting DT is used. During the optimization process, the LGBM model uses the Leaf-wise method to find suitable leaves. The objective function of the LGBM model is: Obj(t)=L(t)+Ω(t)+c Where L(t) and Ω(t) represent the regularization function and loss function, respectively, c and t represent additional parameters and sampling time, respectively; The loss function of the LGBM model includes: Where y i is the model fit, is the predicted output; The objective function is further expressed as: Where, f i is the i-th parameter or parameter vector in the model.

6. The building carbon emission prediction method based on BO-LGBM according to claim 1 is characterized in that: The analysis of the importance of each influencing factor includes: using the LGBM algorithm to evaluate the importance of the input influencing factors and calculating the importance score of the influencing factors. The importance of feature j in a tree is: Where k is the number of leaf nodes, G is the total revenue, and v is t Represents the features associated with the node. When the number of trees is M, the global importance of the feature is: Where Tm represents the importance of a tree when it is the mth tree.

7. The building carbon emission prediction method based on BO-LGBM according to claim 1 is characterized in that: Goodness of fit R 2 , root mean square error (RMSE) and mean absolute error (MAE) are used to verify the reliability of the building life cycle carbon emission prediction model: Among them, y obs is the corresponding data in the dataset, y pred To predict the model results, is the data mean, and n is the sample size.

8. A building carbon emission prediction system based on BO-LGBM, characterized by: include: The first main control module is used to analyze the correlation between factors affecting building carbon emissions, select features with significant correlation with building carbon emissions, and build an indicator system for the entire life cycle carbon emissions of municipal buildings. The data set is then constructed based on this indicator system. The second main control module is used to optimize the hyperparameters of the LGBM model using Bayesian methods. Based on the optimal hyperparameters, the LGBM model is adjusted and a building life cycle carbon emission prediction model is established. The third main control module is used to optimize and train the building life cycle carbon emission prediction model using the data set to establish a nonlinear relationship between influencing factors and building life cycle carbon emissions; The fourth main control module is used to verify the reliability of the building's full life cycle carbon emission prediction model; The fifth main control module is used to predict the carbon emissions of the building throughout its life cycle using the trained prediction model.

9. An electronic device, characterized in that: include: At least one processor, at least one memory and a communication interface communicatively connected to the processor; wherein the processor, memory and communication interface communicate with each other; the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute a building carbon emission prediction method based on BO-LGBM as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores computer instructions, and the computer instructions enable a computer to execute the building carbon emission prediction method based on BO-LGBM according to any one of claims 1 to 7.

Images