A construction process carbon emission prediction correction method

Abstract

The application provides a construction process carbon emission prediction correction method, and belongs to the technical field of carbon emission prediction.The method comprises the following steps: real-time monitoring of the total error value between the carbon emission prediction value and the actual carbon emission value during the construction process.When the total error value is greater than a first set value, calibration is performed, which comprises the following steps: sequentially performing weighted modeling, nonlinear correction and local optimization on the total error value.The genetic algorithm is used to optimize the model parameters of the carbon emission prediction model and the correction parameters introduced by the nonlinear correction, and the optimized model parameters and correction parameters are substituted into the carbon emission prediction model.The prediction error of the optimized carbon emission prediction model is compared with the first set value again.If the prediction error is still greater than the first set value, the calibration process is repeated until the prediction error is less than the first set value.Based on the above scheme, the first carbon emission prediction model can be adjusted in real time according to the actual situation encountered during the construction process, and the accuracy of the prediction result when facing unexpected situations is improved.

Description

Technical Field

[0001] The present invention generally relates to the technical field of carbon emission prediction, and in particular to a method for predicting and correcting carbon emissions during a construction process. Background Art

[0002] In existing technology, the total carbon emissions during construction are predicted using a multivariate regression model. This model is trained in advance using historical data. The training process involves multiple rounds of training, using historical data as input, historical carbon emissions values ​​as label data, and predicted total carbon emissions as output. This historical data includes external environmental factors, construction site characteristics, and dynamic change factors. Then, during the actual construction process, the actual values ​​of this data are input into the multivariate regression model in real time to predict the total carbon emissions.

[0003] However, this prediction method is based on historical data from past construction processes. Incidental circumstances during actual construction, as well as special circumstances not encountered in historical data, are not factored into the carbon emissions forecast. This results in a discrepancy between the carbon emissions predicted by the multivariate regression model based on construction data and the actual carbon emissions. When this discrepancy is too large, the carbon emissions forecast loses its guiding significance for subsequent adjustments to the construction plan.

[0004] Therefore, it is necessary to monitor and calibrate the accuracy of the multivariate regression model's prediction of carbon emissions during the construction process according to the actual situations encountered during the construction process. Summary of the Invention

[0005] In view of the above-mentioned defects or deficiencies in the prior art, it is desired to provide a method for predicting and correcting carbon emissions during construction.

[0006] The present invention provides a method for predicting and correcting carbon emissions during a construction process, comprising:

[0007] Obtaining a first carbon emission prediction value and an actual carbon emission value during a period from a first moment to a second moment in a construction process, and calculating a first total error between the first carbon emission prediction value and the actual carbon emission value; the first carbon emission prediction value is predicted by a first carbon emission prediction model;

[0008] If the first total error value is greater than a first set value, constructing a weighted relationship between the first total error value and a plurality of influencing factors;

[0009] Substituting the first total error value into the weighted relationship, and sequentially performing nonlinear correction and local optimization on the weighted relationship to obtain an optimized weighted relationship; the nonlinear correction process introduces correction parameters;

[0010] According to the optimized weighting relationship, the first model parameter of the first carbon emission prediction model and the correction parameter are optimized using a genetic algorithm to obtain an optimized model parameter and an optimized correction parameter;

[0011] An error term is introduced for the first carbon emission prediction model, the optimized model parameter is substituted into the first carbon emission prediction model, the optimized weighting relationship with the optimized correction parameter is substituted into the error term, and a second carbon emission prediction model is obtained.

[0012] According to the technical scheme provided by the application, after obtaining the second carbon emission prediction model, the following steps are further included:

[0013] A second carbon emission prediction value during a first time to a second time in the construction process is obtained, and a second total error value between the second carbon emission prediction value and the actual carbon emission value is calculated; the second carbon emission prediction value is predicted by the second carbon emission prediction model;

[0014] If the second total error value is greater than the first set value, the second carbon emission prediction model is cyclically optimized according to the second total error value until the second total error value between the second carbon emission prediction value predicted by the second carbon emission prediction model and the actual carbon emission value is less than or equal to the first set value.

[0015] According to the technical scheme provided by the application, the first carbon emission prediction model is obtained in the following manner:

[0016] Obtain historical data; the historical data includes a plurality of influence factors in the past construction process;

[0017] Obtain the carbon emission value in the past construction process to obtain a historical carbon emission value;

[0018] Obtain a multiple regression model;

[0019] The plurality of influence factors in the past construction process are used as the input of the multiple regression model, the historical carbon emission value is used as the label data, and the first carbon emission prediction value is used as the output. The multiple regression model is trained to obtain the first carbon emission prediction model.

[0020] According to the technical scheme provided by the application, the first set value satisfies formula one:

[0021] Formula one;

[0022] Wherein, The first set value is represented by T 总 The total construction time is represented by C 0 represents the carbon emission standard value of the entire construction process,t’ Indicates the monitoring duration from the first moment to the second moment, z Indicates the error tolerance threshold.

[0023] According to the technical solution provided by the present invention, the weighted relationship is:

[0024] Formula 2;

[0025] in, Represents a weighted relationship, whose value is equal to the first total error value, Indicates the i The weight of the error caused by the influencing factors, Indicates the i The error caused by the influencing factors, Indicates the i The nonlinear effect of the error caused by the influencing factors, n Indicates the total number of impact factors, i =1, 2, 3... n .

[0026] According to the technical solution provided by the present invention, the weighted relationship is nonlinearly corrected using Formula 3 to obtain an expression for the total value of the corrected error;

[0027] Formula 3;

[0028] in, represents the predicted carbon emission value after nonlinear correction, Indicates the first carbon emission prediction value, Indicates the correction factor, represents the nonlinear correction factor, Represents the weighted relationship, Represents the total value expression of the corrected error;

[0029] The correction parameters include the correction coefficient and the nonlinear correction factor.

[0030] According to the technical solution provided by the present invention, the expression of the total value of the corrected error is locally optimized to obtain the optimized weighted relationship;

[0031] Local optimization includes:

[0032] Obtain the actual carbon emission sub-value generated by each influencing factor from the first moment to the second moment;

[0033] Obtain the error sub-value caused by each influencing factor;

[0034] Calculating ratios of the error sub-values ​​of the plurality of influencing factors to the actual carbon emission sub-values ​​to obtain a plurality of first ratios;

[0035] Re-assigning weights to multiple influencing factors according to the multiple first ratios to obtain optimized weights;

[0036] Calculating an average of a plurality of the first ratios to obtain a first average;

[0037] Taking the impact factors corresponding to the multiple first ratios greater than the first mean as local optimization targets;

[0038] Using genetic algorithms, the correction coefficient and nonlinear correction factor of the local optimization target are optimized to obtain the optimized correction coefficient and the optimized nonlinear correction factor;

[0039] Substitute the optimized correction coefficient, the optimized nonlinear correction factor, and the optimized weight into Formula 2 to obtain the optimized weighted relationship.

[0040] According to the technical solution provided by the present invention, according to the optimized weighted relationship, a genetic algorithm is used to optimize the first model parameters and the correction parameters of the first carbon emission prediction model to obtain optimized model parameters and optimized correction parameters, including:

[0041] Setting algorithm parameters, fitness function and end conditions; the algorithm parameters include crossover probability and mutation probability;

[0042] Before the genetic algorithm optimization iteration, the values ​​of the crossover probability and the mutation probability are updated according to the first total error value;

[0043] During the iterative optimization process of the genetic algorithm, the crossover probability and the mutation probability are reduced as the number of iterations increases;

[0044] After the genetic algorithm reaches the end condition, the historical optimal model parameters and historical optimal correction parameters are obtained through optimization;

[0045] Under the condition that the loss function satisfies the gradient descent, adjusting the historical optimal model parameters to obtain the optimized model parameters;

[0046] The historical optimal correction parameter is used as the optimized correction parameter.

[0047] According to the technical solution provided by the present invention, updating the values ​​of the crossover probability and the mutation probability according to the first total error value includes:

[0048] Obtaining initial values ​​of the crossover probability and the mutation probability to obtain the initial crossover probability and the initial mutation probability;

[0049] Calculating a ratio of the first total error value to the actual carbon emission value to obtain a second ratio;

[0050] Making the value of the crossover probability equal to the initial crossover probability plus the product of the initial crossover probability and the second ratio;

[0051] The value of the mutation probability is made equal to the initial mutation probability plus the product of the initial mutation probability and the second ratio.

[0052] According to the technical solution provided by the present invention, the genetic algorithm also has an evolution equation with a nonlinear penalty function, as shown in Formula 4:

[0053] Formula 4;

[0054] in, Indicates that the genetic algorithm t +1 generation later, Indicates the current t Generation of individuals, t Indicates the number of current iterations, and V are all regulatory factors that regulate the evolutionary step length, represents the adaptive fitness function, represents the gradient of the adaptive fitness function, represents the random perturbation factor, represents the random disturbance term, G Represents an individual of the genetic algorithm.

[0055] The beneficial effects of the present invention are:

[0056] During the construction process, the first total error value between the first carbon emission prediction value and the actual carbon emission value of the first carbon emission prediction model is monitored in real time. When the first total error value is greater than the first set value, calibration is performed, including: weighted modeling of the first total error value in sequence (i.e., constructing a weighted relationship between the first total error value and multiple influencing factors), nonlinear correction, and local optimization. A genetic algorithm is used to optimize the first model parameters of the first carbon emission prediction model and the correction parameters introduced by the nonlinear correction, and the optimized model parameters and correction parameters are substituted into the first carbon emission prediction model. The prediction error of the optimized first carbon emission prediction model is compared with the first set value again. If it is still greater than the first set value, the calibration process is repeated until the prediction error is less than the first set value. Based on the above scheme, the first carbon emission prediction model can be adjusted in real time according to the actual situation encountered during the actual construction process, thereby improving the accuracy of the carbon emission prediction results in the face of incidental situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0058] Figure 1Schematic diagram of a process for predicting and correcting carbon emissions during construction. DETAILED DESCRIPTION

[0059] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the invention are shown in the accompanying drawings.

[0060] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0061] Please refer to Figure 1 The present invention provides a method for predicting and correcting carbon emissions during a construction process, comprising:

[0062] S1: Obtain the first carbon emission prediction model, including:

[0063] S1-1: Acquire historical data; the historical data includes multiple influencing factors in previous construction processes;

[0064] S1-2: Obtain carbon emission values ​​during previous construction processes to obtain historical carbon emission values;

[0065] S1-3: Obtaining a multiple regression model;

[0066] S1-4: Using multiple influencing factors in previous construction processes as inputs of the multivariate regression model, the historical carbon emission values ​​as label data, and the first carbon emission prediction value as output, the multivariate regression model is trained to obtain the first carbon emission prediction model.

[0067] Specifically, the first carbon emission prediction model is used to input multiple influencing factors at a first moment and predict the carbon emission value of the construction process from the first moment to the second moment; the first carbon emission prediction model has a first model parameter; the multiple influencing factors include external environmental factors, construction site characteristic factors, and dynamic change factors.

[0068] Among them, external environmental factors include: temperature and humidity at the construction site;

[0069] Construction site characteristic factors: type of construction equipment, amount of materials used;

[0070] Dynamically changing factors: construction efficiency and working hours.

[0071] The multiple regression model (the first carbon emission prediction model) is shown in Formula 8:

[0072] Formula eight.

[0073] Thus, the relationship between the first carbon emission prediction value and the carbon emission actual value can be represented by formula nine:

[0074] Formula nine;

[0075] Wherein:

[0076] represents the carbon emission actual value,

[0077] represents the first carbon emission prediction value,

[0078] X is an external environmental factor; such as temperature, humidity, etc.

[0079] W is a construction site characteristic factor; such as equipment type, material use, etc. In specific calculation, the specific value of equipment type is the carbon emission value generated by the normal operation of the equipment per unit time; material use represents the carbon emission value generated by consuming unit mass of the material;

[0080] Z is a dynamic change factor in the construction process; such as the efficiency of the workers, the working time, etc. The efficiency is the ratio of the actual amount of work completed by the workers in a unit time to the expected amount of work completed;

[0081] β 0 is a constant term,

[0082] β 1, β 2, β 3 is the regression coefficient of each factor, that is, the model parameter; for example, in the first carbon emission prediction model, it is the first model parameter;

[0083] is a weighted relationship.

[0084] S2: Obtain the first carbon emission prediction value and the carbon emission actual value during the first time to the second time in the construction process; the first carbon emission prediction value is predicted by the first carbon emission prediction model;

[0085] Calculate the first error total value between the first carbon emission prediction value predicted by the first carbon emission prediction model and the carbon emission actual value generated by the construction from the first time to the second time during the first time to the second time in the construction process, including:

[0086] S2-1: Obtain a plurality of influence factors at the first time;

[0087] S2-2: Inputting the multiple influencing factors at the first moment into the first carbon emission prediction model to predict and obtain the first carbon emission prediction value;

[0088] S2-3: Obtain the actual carbon emissions generated from the first moment to the second moment of construction;

[0089] Specifically, different data are measured by various types of sensors at the actual construction site, and the actual value of each factor is obtained by calculation; this calculation process belongs to the existing technology.

[0090] S2-4: Calculating the absolute value of the difference between the first carbon emission prediction value and the actual carbon emission value;

[0091] S2-5: Taking the absolute value of the difference as the first total error value.

[0092] Specifically, the first moment and the second moment of the present invention are set to be 5-10 minutes apart; thus, when an accidental situation occurs, the impact of the accidental situation on the carbon emission forecast can be taken into account in a shorter period of time; and the amount of data calculation will not be increased due to too frequent detection and correction.

[0093] S3: Determine whether the first total error value is greater than a first set value;

[0094] If the first total error value is greater than the first set value, proceed to steps S5 to S10; otherwise, proceed to step S4;

[0095] Furthermore, the first setting value satisfies Formula 1:

[0096] Formula 1;

[0097] in, Indicates the first set value, T 总 Indicates the total construction time, C 0 represents the carbon emission standard value of the entire construction process, t’ Indicates the monitoring duration from the first moment to the second moment, z Indicates the error tolerance threshold.

[0098] The carbon emission standard value is manually set according to the specified emission standards. The error tolerance threshold is also manually set. A larger value allows for a larger error between the predicted value and the actual value of the first carbon emission prediction model; conversely, a smaller error is allowed, which can make the prediction of the calibrated first carbon emission prediction model more accurate. The present invention sets the error tolerance threshold to 1; thus, the prediction error of the calibrated first carbon emission prediction model is limited to less than 1%.

[0099] S4: Update the first moment and the second moment, and repeat steps S2 to S3;

[0100] Specifically, the method of updating the first moment and the second moment is:

[0101] After the monitoring time has passed, the current moment is taken as the first moment, and the moment after the monitoring time has passed is taken as the second moment.

[0102] Assuming that the monitoring duration is set to 1 minute and the first monitoring time is 8:20 am, the first moment is 8:20 am and the second moment is 8:21 am.

[0103] After updating the first time and the second time, the second monitoring time is 8:21 am, so the first time is 8:21 am, and the second time is 8:22 am.

[0104] S5: constructing a weighted relationship between the first total error value and multiple influencing factors (i.e., performing weighted modeling);

[0105] Substituting the first total error value into the weighted relationship, and sequentially performing nonlinear correction and local optimization on the weighted relationship to obtain an optimized weighted relationship; the nonlinear correction process introduces correction parameters into the weighted relationship;

[0106] At the same time, an error term should be introduced into the first carbon emission prediction model;

[0107] Furthermore, in order to more accurately capture the source of error, a weighted error modeling method is introduced to divide the total error value into the sum of the errors caused by multiple influencing factors, so that the error can more realistically reflect the current construction status.

[0108] Using formula 2, weighted modeling is performed on the first total error value to obtain a weighted relationship;

[0109] Formula 2;

[0110] in, Represents a weighted relationship, whose value is equal to the total value of the first error, Indicates the i The weight of the error caused by the influencing factors, Indicates the i The error caused by the influencing factors, Indicates the i The nonlinear effect of the error caused by the influencing factors, n Indicates the total number of impact factors, i =1, 2, 3... n .

[0111] Furthermore, in order to deal with unforeseen nonlinear errors during the construction process, nonlinear correction and local optimization mechanisms are introduced.

[0112] Using formula 3, the weighted relationship is corrected nonlinearly to obtain the total value expression of the corrected error;

[0113] Formula 3;

[0114] in, represents the predicted carbon emission value after nonlinear correction, Indicates the first carbon emission prediction value, Indicates the correction factor, represents the nonlinear correction factor, Represents the weighted relationship, It represents the total value expression of the corrected error. The nonlinear correction factor is usually less than 1, which can control the intensity of the error correction, making the correction more suitable for the actual situation and avoiding excessive adjustment of the error.

[0115] The correction parameters include the correction coefficient and the nonlinear correction factor.

[0116] Specifically, by introducing dynamic weighting and nonlinear influencing factors, the model can adjust the contribution of each error to the prediction error, thereby improving the model's fitting performance. This error adjustment method enables the regression model to more flexibly adapt to the complex and changing environmental conditions during the construction process, improving the accuracy and reliability of predictions because it accurately reflects the specific impact of different errors on the final prediction results.

[0117] Furthermore, the total value expression of the corrected error is locally optimized to obtain the optimized weighted relationship;

[0118] Local optimization includes:

[0119] Obtaining the actual carbon emission sub-value generated by each influencing factor between the first moment and the second moment; the specific acquisition method is to detect the specific value of each influencing factor and substitute it into the corresponding formula to calculate the carbon emission value corresponding to the influencing factor, and the formula calculation method belongs to the existing technology;

[0120] Obtain the error sub-value caused by each influencing factor;

[0121] Calculating ratios of the error sub-values ​​of the plurality of influencing factors to the actual carbon emission sub-values ​​to obtain a plurality of first ratios;

[0122] Re-assigning weights to multiple influencing factors according to the multiple first ratios to obtain optimized weights;

[0123] Specifically, the weights are redistributed as follows:

[0124] The weights are distributed according to the size of the corresponding first ratios. The larger the first ratio, the greater the redistributed weight; at the same time, the sum of all weights is ensured to be equal to 1.

[0125] For example, the sum of all first ratios is calculated first; then the optimization weight corresponding to each of the multiple influencing factors is set equal to the ratio of the first ratio corresponding to each influencing factor to the sum of all first ratios.

[0126] Local optimization also includes:

[0127] Calculating an average of a plurality of the first ratios to obtain a first average;

[0128] The influencing factors corresponding to the multiple first ratios that are greater than the first mean are used as local optimization targets; the purpose here is to screen out the influencing factors that have a greater impact on the error.

[0129] Using genetic algorithms, the correction coefficient and nonlinear correction factor of the local optimization target are optimized to obtain the optimized correction coefficient and the optimized nonlinear correction factor;

[0130] Substitute the optimized correction coefficient, the optimized nonlinear correction factor, and the optimized weight into Formula 2 to obtain the optimized weighted relationship.

[0131] Specifically, the genetic algorithm applied in the local optimization process has variable ranges set to the ranges of the correction coefficient and the nonlinear correction factor;

[0132] The loss function is set as:

[0133] Formula 5;

[0134] in, i represents variables, which in this embodiment are correction coefficients and nonlinear correction factors, represents the loss function, Indicates impact factor i In the sensor j The weight of Indicates the j The corresponding factors detected by each sensor are used to calculate the actual carbon emission value, that is, the real carbon emission data obtained on site. Represents a variable based i The calculated j The predicted carbon emission value of observation samples, i =1, 2, 3... n 、 n represents the total number of impact factors, j =1, 2, 3...m , m Indicates the total number of sensors.

[0135] The remaining implementation is consistent with that in step S6.

[0136] Specifically, the main optimization goal of the genetic algorithm is to correct the coefficient c and nonlinear correction factor α These parameters directly affect the strength and method of error correction. By simulating the process of natural selection, the genetic algorithm gradually and iteratively optimizes these parameters, finding the parameter combination that best suits the current error pattern to improve the accuracy and stability of the prediction model.

[0137] The criterion for finding the optimal solution is to minimize the error between the revised carbon emission forecast and the actual carbon emission value. Specifically, the optimization process measures the quality of the solution by minimizing a loss function (usually the sum of squared prediction errors). Ultimately, the optimal solution is the parameter combination that minimizes the error and maximizes model stability within certain constraints.

[0138] During the construction process, certain factors or errors (such as equipment errors and external environmental factors) significantly impact predicted carbon emissions, resulting in particularly pronounced errors in these localized areas. The system analyzes the distribution of current errors, identifies localized areas or factors with significant error variation, and focuses on optimizing parameters in these areas. By dynamically monitoring changes at the construction site, the algorithm identifies key areas requiring local optimization, enabling more precise error correction.

[0139] S6: Optimizing the first model parameters and the correction parameters of the first carbon emission prediction model using a genetic algorithm according to the optimized weighted relationship to obtain optimized model parameters and optimized correction parameters;

[0140] Among them, the first model parameter is the one mentioned above β 0. β 1. β 2. β 3; The correction parameters include the ones mentioned above and ;

[0141] Specifically include:

[0142] Setting algorithm parameters, fitness function and end conditions; the algorithm parameters include crossover probability and mutation probability;

[0143] The algorithm parameters include:

[0144] Population size, chromosome length, crossover probability, mutation probability, selection operator, variable range (in this step, the range of model parameters and correction parameters), and maximum number of iterations;

[0145] In this embodiment, the population size is set to: 100;

[0146] Chromosome length: 30~50;

[0147] The initial value of the crossover probability is: 0.8;

[0148] The initial value of the mutation probability is: 0.05;

[0149] The maximum number of iterations is any value between 100 and 1000; in this embodiment, it is set to 200.

[0150] Fitness function:

[0151]

[0152] Formula 6;

[0153] in, represents the fitness function, and Both represent the weight factors in the fitness function, represents the error change weight factor, Indicates the change in error compared to the previous iteration, Indicates monitoring time, Indicates the actual value of carbon emissions, represents the first carbon emission prediction value, m is the number of sensors.

[0154] End condition: The minimum value of the fitness function does not decrease after multiple consecutive iterations and reaches the maximum number of iterations.

[0155] Furthermore, before the genetic algorithm optimization iteration, updating the values ​​of the crossover probability and the mutation probability according to the first total error value includes:

[0156] Obtaining initial values ​​of the crossover probability and the mutation probability to obtain the initial crossover probability and the initial mutation probability;

[0157] Calculating a ratio of the first total error value to the actual carbon emission value to obtain a second ratio;

[0158] Making the value of the crossover probability equal to the initial crossover probability plus the product of the initial crossover probability and the second ratio;

[0159] The value of the mutation probability is made equal to the initial mutation probability plus the product of the initial mutation probability and the second ratio.

[0160] Specifically, the error generally accounts for a small proportion, and the crossover probability and mutation probability will not be greater than 1 after improvement; in some special cases, when the crossover probability and mutation probability are greater than 1 after improvement, the crossover probability and mutation probability are set to 1.

[0161] With each iteration of the genetic algorithm, the system calculates the change in prediction error as the external environment changes. If the error increases (for example, due to abnormal errors caused by equipment failure or material changes), it indicates that the system's current search strategy may not fully explore the problem space. Therefore, the crossover and mutation probabilities can be appropriately increased. Crossover helps explore new solution spaces, while mutation increases the diversity of solutions, thereby preventing the algorithm from becoming trapped in local optima.

[0162] Furthermore, during the iterative optimization process of the genetic algorithm, the crossover probability and the mutation probability are reduced as the number of iterations increases;

[0163] The specific reduction method is to reduce the fixed value each generation. The fixed value of the crossover probability is equal to the ratio of the initial crossover probability to the maximum number of iterations. The fixed value of the mutation probability is equal to the ratio of the initial mutation probability to the maximum number of iterations.

[0164] If the external environment undergoes drastic changes (for example, a sudden equipment failure or material change at a construction site), the carbon emission forecast error may increase significantly. Based on the above settings, in this case, the system automatically increases the intensity of the mutation operation. By increasing the mutation probability, the algorithm can quickly escape the local area of ​​the current solution and search for a more suitable solution to adapt to the new environment. The intensity of the mutation operation can be adjusted based on the magnitude of the error change: when the error changes significantly, the mutation rate is increased to enhance the search capability; when the error changes steadily, the mutation rate is reduced to reduce the risk of over-searching.

[0165] Furthermore, after the genetic algorithm reaches the end condition, the historical optimal model parameters and historical optimal correction parameters are obtained through optimization;

[0166] Under the condition that the loss function satisfies the gradient descent, adjusting the historical optimal model parameters to obtain the optimized model parameters;

[0167] The gradient descent formula is shown in Formula 7:

[0168] Formula 7;

[0169] in, i t Indicates the current variable. i t+1 Represents the updated variable, represents the learning rate, Represents the gradient of the loss function with respect to the variable.

[0170] Finally, the historical optimal correction parameter is used as the optimized correction parameter.

[0171] Specifically, over multiple iterations, all weights are gradually adjusted based on the current error. With each update, all model parameters of the regression model are adjusted based on the calculated gradient to minimize the loss function and gradually bring the predicted results closer to the actual values. This process is gradual, and after multiple updates, the model weights converge to the optimal solution, thereby improving prediction accuracy.

[0172] Furthermore, the genetic algorithm also has an evolution equation with a nonlinear penalty function, as shown in Formula 4:

[0173] Formula 4;

[0174] in, Indicates that the genetic algorithm t +1 generation of individuals, that is, the new generation of individuals or optimized parameters after this round of evolution and optimization, Represents the individual of the current t generation, that is, the solution that the current genetic algorithm is optimizing. t Indicates the number of current iterations, and V are all regulatory factors that regulate the evolutionary step length, represents the adaptive fitness function, represents the gradient of the adaptive fitness function, Represents a random perturbation factor, used to avoid premature convergence, represents the random disturbance term, G Represents the individual of the genetic algorithm, that is, the solution vector involved in the optimization process of the genetic algorithm.

[0175] Specifically, applying the evolutionary equation of the nonlinear penalty function can further increase the complexity of the algorithm, making it closer to the actual construction situation, while also improving the optimization ability.

[0176] S7: Substitute the optimized model parameters into the first carbon emission prediction model, and substitute the optimized weighted relationship with the optimized correction parameters into the error term to obtain a second carbon emission prediction model, referring to Formula 10:

[0177] Formula 10;

[0178] in, represents the second carbon emission prediction model, β 0. β 1. β 2. β 3 all indicate optimized model parameters; represents the optimization weight;

[0179] Optimization correction parameters include: optimization correction coefficient and optimize the nonlinear correction factor .

[0180] The second carbon emission prediction model can be obtained through the above steps. Since the model parameters of this model are obtained through algorithm optimization, there is still a low probability that the second carbon emission model still cannot accurately predict the carbon emission value. In order to eliminate this low probability event, after obtaining the second carbon emission prediction model, the following steps are also included:

[0181] S8: Obtaining a second carbon emission prediction value from a first moment to a second moment during the construction process; the second carbon emission prediction value is predicted by the second carbon emission prediction model;

[0182] Calculating a second total error between a second carbon emission prediction value obtained by using the second carbon emission prediction model and the actual carbon emission value during a period from a first moment to a second moment of the construction process;

[0183] S9: Determine whether the second total error value is greater than the first set value;

[0184] If the second total error value is greater than the first set value, the second carbon emission prediction model is cyclically optimized according to the second total error value until the second total error value between the second carbon emission prediction value predicted by the second carbon emission prediction model and the actual carbon emission value is less than or equal to the first set value; then step S10 is performed;

[0185] The loop optimization specifically includes: using the second total error value as the first total error value; using the current model parameters as the first model parameters, and then repeating steps S5 and S9;

[0186] If the second total error value is less than or equal to the first set value, directly proceed to step S10.

[0187] S10: Use the second carbon emission prediction model finally obtained by the cyclic optimization to predict the carbon emission value during the construction process; and return to step S4.

[0188] Specifically, after the carbon emission prediction is completed using the second carbon emission prediction model, the process returns to step S4 to repeatedly monitor the current carbon emission prediction model throughout the construction process to see if it can still accurately predict carbon emission values. If the actual conditions change, resulting in a significant error between the actual carbon emission values ​​and the predicted carbon emission values, the second carbon emission prediction model, which has been repeatedly optimized, will need to be optimized again to adapt it to the new actual conditions and ensure prediction accuracy.

[0189] Specifically, after steps S7 to S10, in order to avoid the situation where the error cannot be less than the first set value after the optimized correction parameters and optimized model parameters obtained by the genetic algorithm are substituted into the first carbon emission prediction model, the present invention also sets up a cyclic optimization mechanism until the optimized correction parameters and optimized model parameters are optimized and substituted into the first carbon emission prediction model, and the predicted error is less than the first set value.

[0190] Based on this mechanism, it can be guaranteed that the final optimization results meet the requirements.

[0191] The above description is merely a preferred embodiment of the present invention and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the inventive concept. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in the present invention.

Claims

1. A method for predicting and correcting carbon emissions during construction, characterized in that: include: Obtaining a first carbon emission prediction value and an actual carbon emission value during a period from a first moment to a second moment in a construction process, and calculating a first total error between the first carbon emission prediction value and the actual carbon emission value; The first carbon emission prediction value is obtained by prediction by a first carbon emission prediction model; If the first total error value is greater than a first set value, constructing a weighted relationship between the first total error value and a plurality of influencing factors; Wherein, the weighted relationship is: Formula 2; in, Represents a weighted relationship, whose value is equal to the first total error value, Indicates the i The weight of the error caused by the influencing factors, Indicates the i The error caused by the influencing factors, Indicates the i The nonlinear effect of the error caused by the influencing factors, n Indicates the total number of impact factors, i =1, 2, 3... n ; Substituting the first total error value into the weighted relationship, and sequentially performing nonlinear correction and local optimization on the weighted relationship to obtain an optimized weighted relationship; the nonlinear correction process introduces correction parameters; According to the optimized weighted relationship, a genetic algorithm is used to optimize the first model parameters and the correction parameters of the first carbon emission prediction model to obtain optimized model parameters and optimized correction parameters; Introducing an error term into the first carbon emission prediction model, substituting the optimized model parameters into the first carbon emission prediction model, and substituting the optimized weighted relationship with the optimized correction parameters into the error term to obtain a second carbon emission prediction model; Among them, local optimization includes: Obtain the actual carbon emission sub-value generated by each influencing factor from the first moment to the second moment; Obtain the error sub-value caused by each influencing factor; Calculating ratios of the error sub-values ​​of the plurality of influencing factors to the actual carbon emission sub-values ​​to obtain a plurality of first ratios; Re-assigning weights to multiple influencing factors according to the multiple first ratios to obtain optimized weights; Calculating an average of a plurality of the first ratios to obtain a first average; Taking the impact factors corresponding to the multiple first ratios greater than the first mean as local optimization targets; Using genetic algorithms, the correction coefficient and nonlinear correction factor of the local optimization target are optimized to obtain the optimized correction coefficient and the optimized nonlinear correction factor; Substitute the optimized correction coefficient, the optimized nonlinear correction factor, and the optimized weight into Formula 2 to obtain the optimized weighted relationship.

2. A method for predicting and correcting carbon emissions during construction according to claim 1, characterized in that: After obtaining the second carbon emission prediction model, it also includes: Obtaining a second carbon emission prediction value from a first moment to a second moment during the construction process, and calculating a second total error between the second carbon emission prediction value and the actual carbon emission value; the second carbon emission prediction value is predicted by the second carbon emission prediction model; If the second total error value is greater than the first set value, the second carbon emission prediction model is cyclically optimized according to the second total error value until the second carbon emission prediction value predicted using the second carbon emission prediction model and the second total error value between the actual carbon emission value are less than or equal to the first set value.

3. A method for predicting and correcting carbon emissions during construction according to claim 1, characterized in that: The first carbon emission prediction model is obtained in the following way: Acquiring historical data; the historical data includes multiple influencing factors in previous construction processes; Obtain carbon emission values ​​during previous construction processes to obtain historical carbon emission values; Obtain a multiple regression model; The multiple influencing factors in the previous construction process are used as the input of the multivariate regression model, the historical carbon emission values ​​are used as label data, and the first carbon emission prediction value is used as the output. The multivariate regression model is trained to obtain the first carbon emission prediction model.

4. A construction process carbon emission prediction and correction method according to claim 1, characterized in that: The first setting value satisfies Formula 1: Formula 1; in, Indicates the first set value, T 总 Indicates the total construction time, C 0 represents the carbon emission standard value of the entire construction process, t’ Indicates the monitoring duration from the first moment to the second moment, z Indicates the error tolerance threshold.

5. A construction process carbon emission prediction and correction method according to claim 1, characterized in that: Using formula 3, the weighted relationship is nonlinearly corrected to obtain the total value expression of the corrected error; Formula 3; in, represents the predicted carbon emission value after nonlinear correction, Indicates the first carbon emission prediction value, Indicates the correction factor, represents the nonlinear correction factor, Represents the weighted relationship, Represents the total value expression of the corrected error; The correction parameters include the correction coefficient and the nonlinear correction factor; the expression of the total value of the correction error is locally optimized to obtain the optimized weighted relationship.

6. A method for predicting and correcting carbon emissions during construction according to claim 1, characterized in that: According to the optimized weighted relationship, a genetic algorithm is used to optimize the first model parameters and the correction parameters of the first carbon emission prediction model to obtain optimized model parameters and optimized correction parameters, including: Setting algorithm parameters, fitness function and end conditions; the algorithm parameters include crossover probability and mutation probability; Before the genetic algorithm optimization iteration, the values ​​of the crossover probability and the mutation probability are updated according to the first total error value; During the iterative optimization process of the genetic algorithm, the crossover probability and the mutation probability are reduced as the number of iterations increases; After the genetic algorithm reaches the end condition, the historical optimal model parameters and historical optimal correction parameters are obtained through optimization; Under the condition that the loss function satisfies the gradient descent, adjusting the historical optimal model parameters to obtain the optimized model parameters; The historical optimal correction parameter is used as the optimized correction parameter.

7. A method for predicting and correcting carbon emissions during construction according to claim 6, characterized in that: Updating the values ​​of the crossover probability and the mutation probability according to the first total error value includes: Obtaining initial values ​​of the crossover probability and the mutation probability to obtain the initial crossover probability and the initial mutation probability; Calculating a ratio of the first total error value to the actual carbon emission value to obtain a second ratio; Making the value of the crossover probability equal to the initial crossover probability plus the product of the initial crossover probability and the second ratio; The value of the mutation probability is made equal to the initial mutation probability plus the product of the initial mutation probability and the second ratio.

8. A method for predicting and correcting carbon emissions during construction according to claim 6, characterized in that: The genetic algorithm also has an evolution equation with a nonlinear penalty function, as shown in Formula 4: Formula 4; in, Indicates that the genetic algorithm t +1 generation later, Indicates the current t Generation of individuals, t Indicates the number of current iterations, and V are all regulatory factors that regulate the evolutionary step length, represents the adaptive fitness function, represents the gradient of the adaptive fitness function, represents the random perturbation factor, represents the random disturbance term, G Represents an individual of the genetic algorithm.

Images