Intelligent carbon emission prediction and optimization method in tunnel construction stage

Abstract

The invention provides an intelligent carbon emission prediction and optimization method in a tunnel construction stage. The method comprises the following steps: collecting tunnel construction historical data and carbon emission data thereof, carrying out primary processing, taking the data as a machine learning data set, and carrying out training; importing an engineering list, and predicting oil consumption, power consumption, the number of personnel and the like required by construction so as to obtain a carbon emission prediction value; analyzing the most key factors influencing the carbon emission by using an SHAP method, and giving construction adjustment suggestions; a CO2 sensor is used for transmitting data, recording actually-measured CO2 concentration information, recording actual construction electricity consumption, oil consumption and the number of personnel, calculating actually-measured carbon emission, comparing the actually-measured carbon emission with prediction data, judging prediction accuracy and carrying out continuous machine learning. According to the method, the carbon emission can be quickly and accurately predicted, continuous updating can be kept, the prediction performance is continuously improved, and a basis is provided for energy conservation and emission reduction measurement of a specified tunnel.

Description

Technical Field

[0001] The present invention relates to the field of tunnels, and in particular to an intelligent carbon emission prediction and optimization method during a tunnel construction phase. Background Art

[0002] Climate change has become a hot issue related to the sustainable development of the earth's ecological environment. Its impacts include rising temperatures and sea levels, droughts, floods, heat waves and frequent typhoons. Numerous previous studies have confirmed that global climate change is directly related to greenhouse gas emissions. Various human economic activities (electricity, transportation and agriculture) will emit a large amount of greenhouse gases, mainly composed of carbon-containing gases such as carbon dioxide and methane.

[0003] With the continuous development of infrastructure construction, tunnel projects are becoming more and more important as an important part of the transportation network. However, the problem of carbon emissions during tunnel construction has gradually attracted people's attention. Carbon emissions mainly come from the use of construction machinery and equipment, blasting operations, concrete production and transportation. For example, large-scale machinery and equipment such as drills, excavators and loaders are essential tools for tunnel construction. These equipment are usually driven by diesel engines and will produce a large amount of carbon dioxide when running for a long time. Due to the complex construction environment, frequent use of machinery and equipment, high demand for blasting operations, and carbon emissions generated during concrete production and transportation, tunnel construction has become one of the significant sources of carbon emissions.

[0004] Due to the improvement of environmental awareness and technological progress, predicting and reducing carbon emissions during construction is an important content. The traditional method of calculating carbon emissions is generally the carbon emission factor method. This traditional method requires the use of statistical results of on-site data. This method is time-consuming and labor-intensive. It can only perform data statistics, and it is difficult to provide real-time adjustment opinions. In addition, the statistical process may even cause secondary carbon emissions, which is relatively inefficient. Therefore, it is important and necessary to use a fast and simple method to make a preliminary and accurate prediction of carbon emissions in the early stage of construction. Summary of the invention

[0005] The present invention focuses on the difficulties of carbon emissions during tunnel construction, provides a method for predicting tunnel carbon emissions, solves the problem that the traditional carbon emission factor method requires manpower and material resources for data statistics, uses machine learning methods to improve prediction speed and accuracy, so that a preliminary understanding of the carbon footprint can be obtained before construction, and uses the SHAP method to analyze the most critical factors affecting carbon emissions to provide data support for adjusting the construction process. Finally, in order to ensure the advancement and accuracy of this model, a CO2 sensor is used to obtain the actual concentration, and the power consumption is calculated according to the electric meter, the measured concentration is compared with the predicted concentration, and the training model is corrected to ensure that the model can self-learn and update. The more this model is used, the more accurate the prediction is, which improves the prediction efficiency and accuracy, and provides technical support and ideas for reducing carbon emissions in actual tunnel construction.

[0006] In order to solve the problem that the carbon emissions from tunnel construction are huge and difficult to detect, this application invents an intelligent carbon emission prediction and analysis method for the tunnel construction stage. The present invention first extracts key construction data, such as power consumption data, fuel consumption data, construction equipment usage, engineering quantity data, etc., based on previous tunnel construction data, and uses advanced machine learning methods to train the processed data in the model, and uses the random forest method to predict power consumption, fuel consumption and manpower, and then calculates carbon emissions according to the emission factor method. In order to analyze the impact of all influencing factors on carbon emissions, the method of calculating SHAP values ​​is used to quantify the specific impact of each factor. Finally, in order to make the prediction of machine learning more accurate and keep the database continuously updated, the CO2 sensor is combined to measure the CO2 concentration value, calculate the actual emissions, and compare it with the predicted value to judge the accuracy of the prediction and conduct continuous learning to improve the accuracy of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is an overall flow chart of the implementation process of the present invention;

[0008] Figure 2 A flowchart is shown for collecting tunnel construction data and its carbon emission data, performing preliminary processing, and using them as machine learning data sets for training;

[0009] Figure 3 A flowchart is presented for importing a project list, predicting the fuel consumption, electricity consumption, and number of personnel required for construction, and thus obtaining a carbon emission forecast value;

[0010] Figure 4 It is a flowchart that uses random forest method for machine learning. Random forest method combines multiple decision tree models to improve the accuracy and stability of overall prediction;

[0011] Figure 5Describes the use of the SHAP method to analyze the most critical factors affecting carbon emissions and provides a flowchart of construction adjustment suggestions;

[0012] Figure 6 The flow chart shows the use of CO2 sensors to transmit data, record the measured CO2 concentration information, record the electricity consumption, calculate the measured carbon emissions, compare them with the predicted data, judge the prediction accuracy and conduct continuous learning. DETAILED DESCRIPTION

[0013] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] like Figure 1 As shown, the embodiment of the present application provides an intelligent carbon emission prediction and optimization method in the tunnel construction stage, and the overall implementation steps are as follows:

[0015] Step 1: Collect tunnel construction data and its carbon emission data, perform preliminary processing, and use them as machine learning data sets for machine learning training, such as Figure 2 As shown in the flow chart.

[0016] Firstly, based on the existing tunnel construction information, the key information of tunnel construction is extracted, including power consumption data, oil consumption data, construction equipment usage, engineering quantity data, geological condition information, construction environment information, construction personnel and material information.

[0017] Furthermore, the collected data are classified into one category as input data and the other as output data. Since there are few statistical data on carbon emissions in previous cases, power consumption, oil consumption and the number of construction workers are output data. The main input data should include: construction methods, equipment information, construction time, transportation distance, tunnel length, tunnel type, geological conditions, construction environment and construction materials, etc.

[0018] Furthermore, before model training, the data required for input is pre-processed. The embodiment of the present application adopts the random forest method for prediction, which mainly processes the missing values ​​and outliers of the data and makes up for the missing values. Since the data required for input are generally those that must be considered during construction, there are generally few missing values. In case of missing values, the median can be used instead. For obvious outliers, they should be eliminated or processed.

[0019] Furthermore, the random forest method is used for machine learning. The random forest method combines multiple decision tree models to improve the accuracy and stability of the overall prediction. At each node split, the decision tree selects a part of the features from all the input features instead of all the features. The final prediction result is obtained by solving the results of each decision tree and averaging them. The random forest model machine learning creation process is as follows: Figure 4 shown.

[0020] Preferably, the content of model training using the random forest method includes:

[0021] The model should include an input module, a data processing module, a training module, a testing module, an output module and a carbon emission calculation module.

[0022] The input module should include a data import part, which can directly import the collected and processed data;

[0023] The data processing module should be connected to the input module, and be able to classify the imported data and determine each type of data for subsequent model training;

[0024] The training module uses an automatic sampling method to select 80% of the data from the original training set for learning, finds the rules in the data, and establishes a training model;

[0025] The test module uses 20% of the remaining data set to verify and evaluate the model to ensure the accuracy of the model;

[0026] The output module should be able to output the predicted fuel consumption, power consumption and manpower to calculate carbon emissions;

[0027] The calculation module should be connected to the output module to calculate the carbon emissions using the output energy consumption.

[0028] Step 2: Import the project list to predict the fuel consumption, electricity consumption and number of personnel required for construction, so as to obtain the carbon emission prediction value. The specific implementation method is as follows:

[0029] Combination process Figure 3 ,First, process the engineering information, collect the above key information required, and process the ,information for missing values ​​and outliers, and make the processed data into an ,Excel table to facilitate data import.

[0030] Furthermore, by importing the data, the model can automatically predict all the electricity, oil and number of construction workers consumed during the construction of this tunnel project.

[0031] Further, the carbon emissions are calculated, and then the predicted carbon emissions are calculated according to the carbon emission factor method. This calculation process can also be solved automatically. The calculation formula is as follows:

[0032] ECF=ECF1×N1+ECF2×N2+ECF3×N3×t

[0033] Among them, ECF refers to total carbon emissions, ECF1 refers to the carbon emission factor of local electricity consumption, ECF2 refers to the carbon emission factor of fuel consumption, ECF3 refers to the carbon emission factor of human breathing, N1 refers to electricity consumption, N2 refers to fuel consumption, N3 refers to the number of labor; t refers to working hours.

[0034] Step 3, Combine Figure 5 Process, using SHAP method to analyze the most critical factors affecting carbon emissions, and give specific implementation methods for construction adjustment suggestions;

[0035] First, based on the trained model above, a baseline value is obtained.

[0036] Preferably, the baseline value in this training model refers to the average carbon emission value obtained based on all data training sets. This value is used as the baseline value, that is, the most reasonable carbon emission prediction value that the model will give when nothing is known about the new sample.

[0037] Furthermore, all possible feature combinations are then calculated to evaluate the impact of adding or removing a feature on the prediction results and calculate its marginal contribution.

[0038] Preferably, when there are many input parameters, it is unrealistic to combine all the features. Therefore, we can use the Tree SHAP method to simplify the calculation and efficiently calculate the SHAP value. In Python, we can directly use this method for efficient approximation.

[0039] Furthermore, the boundary contributions of all feature combinations are averaged to obtain the SHAP value of the influencing factor. The higher the SHAP value, the greater its contribution, and vice versa.

[0040] Preferably, the calculation formula of the SHAP value is as follows:

[0041]

[0042] In the formula, φ i Refers to the SHAP value of feature i;

[0043] The set of all features of N value

[0044] All feature subsets whose values ​​do not contain feature i;

[0045] S refers to the size of subset S (i.e. the number of features)

[0046] f(S) refers to the model output using only the feature subset S for prediction;

[0047] f(S∪{i}) refers to the model output using the feature subset plus feature i for prediction;

[0048] Refers to the weighting coefficient of the feature subset S.

[0049] For the calculated SHAP value, if it is positive, it means that this item will make a positive contribution to carbon emissions, that is, it will cause carbon emissions. If it is negative, it means that this item will reduce carbon emissions.

[0050] Furthermore, based on the SHAP value, the most obvious factors affecting carbon emissions can be analyzed, and construction suggestions can be provided for similar projects in the future to reduce carbon emissions.

[0051] Step 4: Use the CO2 sensor to transmit data, record the measured CO2 concentration information, record the power consumption, calculate the measured carbon emissions, compare with the predicted data, and combine Figure 6 Process diagram The specific implementation method for judging prediction accuracy and conducting continuous learning is:

[0052] First, CO2 concentration detectors are arranged in real time according to the length of tunnel construction, and real-time CO2 concentration is detected around the clock. The concentration data is transmitted in real time through 4G / 5G network or Bluetooth, and a concentration change curve over time is drawn every day.

[0053] Further, the method of using the CO2 sensor is as follows:

[0054] In order to obtain accurate CO2 concentration values, sensors should be arranged at a certain density according to the real-time construction length. 2-3 sensors should be arranged near the construction face, and one sensor can be arranged every 50-100m at other constructed locations.

[0055] Preferably, the CO2 concentration detection aspect includes a CO2 concentration sensor, a signal processor and a data storage system. First, the concentration change information received by the CO2 concentration sensor is transmitted to the signal processor, the CO2 concentration information is processed, stored in the form of data and a concentration change curve over time is automatically drawn.

[0056] Furthermore, the actual CO2 emissions are calculated based on the detected concentration.

[0057] Preferably, the calculation method is to first determine the carbon emissions in a day when no construction is carried out. This value may be the lowest value of the concentration curve, or it may be appropriately determined based on sensors arranged in real time. After finding the normal carbon emission value, the measured value is subtracted from the normal value, which is the additional carbon emissions caused by electricity, oil and breathing. The specific value can be obtained by calculation.

[0058] Preferably, the calculation formula is:

[0059] ECF = ∑(c×A×D)

[0060] In the formula, ECF refers to the CO2 emission value (kg CO2) calculated based on actual measurements;

[0061] C refers to the measured CO2 concentration value (kgCO2 / m 3 );

[0062] A refers to the cross-sectional area of ​​the tunnel;

[0063] D refers to the CO2 sensor spacing, that is, the sensor coverage distance.

[0064] Furthermore, the amount of electricity can be directly obtained based on the electricity consumption statistics during the construction period, and the carbon emissions caused by electricity consumption can be calculated in real time.

[0065] Furthermore, during real-time detection, the data can be compared with the predicted value to determine the accuracy of the prediction. If the calculated error is within 10%, the prediction can be considered good. If the error is large, it means that the prediction accuracy of this solution is poor. This data can be used for learning to improve the accuracy, and then the accuracy can be continuously determined. This process is repeated until the accuracy meets the requirements.

[0066] The above is a detailed introduction to the intelligent carbon emission prediction and optimization method for the tunnel construction stage provided by the embodiment of the present application. The description of the above embodiment is only used to help understand the method and its core idea of ​​the present application; at the same time, for those of ordinary skill in the art, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. An intelligent carbon emission prediction and optimization method for tunnel construction stage, It is characterized in that The following steps are involved: Step 1: Collect tunnel construction data for preliminary processing and use it as a data set for machine learning for training; Step 2: Import the project list to predict the fuel consumption, electricity consumption and number of personnel required for construction, so as to obtain the carbon emission prediction value; Step 3: Use the SHAP method to analyze the most critical factors affecting carbon emissions and provide construction adjustment and optimization suggestions; Step 4: Record the measured CO2 concentration information, as well as the electricity consumption, fuel consumption and number of personnel, calculate the measured carbon emissions, compare them with the predicted data, determine the prediction accuracy and conduct continuous learning.

2. According to claim 1, a method for intelligent carbon emission prediction and optimization in the tunnel construction stage, Features: The step 1 of collecting tunnel construction data and performing preliminary processing is specifically as follows: Extract key information of tunnel construction based on the existing tunnel construction information; Classify the extracted key information to determine the input data and output data of machine learning; Random forest method is used for machine learning.

3. According to claim 2, a method for intelligent carbon emission prediction and optimization in the tunnel construction stage, Features: The key information includes power consumption data, oil consumption data, construction equipment usage, engineering quantity data, geological condition information, construction environment information, construction personnel and material information.

4. According to claim 2, a method for intelligent carbon emission prediction and optimization in the tunnel construction stage, Features: When using the random forest method for machine learning, it also includes the processing of missing values ​​and outliers in the data.

5. According to claim 1, a method for intelligent carbon emission prediction and optimization in the tunnel construction stage, Features: The carbon emission prediction value of step 2 is calculated as follows: ； in ECF Refers to total carbon emissions, ECF 1 Refers to the carbon emission factor of local electricity consumption, ECF 2 refers to the carbon emission factor of fuel consumption, ECF 3 refers to the carbon emission factor of human breathing, N 1 Refers to power consumption, N 2 Refers to fuel consumption, N 3 Refers to the number of workers; t Refers to working hours.

6. According to claim 1, a method for intelligent carbon emission prediction and optimization in the tunnel construction stage, Features: The specific implementation process of the SHAP method in step 3 is: According to the above trained model, the baseline value is obtained; Calculate all possible feature combinations, evaluate the impact of adding or removing a feature on the prediction results, and calculate its marginal contribution; The boundary contributions of all feature combinations are averaged to obtain the SHAP value of the influencing factor.

7. According to claim 6, a method for intelligent carbon emission prediction and optimization in the tunnel construction stage, Features: The baseline value is the average carbon emission value obtained from all data training sets.

8. According to claim 1, a method for intelligent carbon emission prediction and optimization in the tunnel construction stage, Features: The step 4 is specifically as follows: Arrange CO2 concentration detectors in real time according to the length of tunnel construction, detect real-time CO2 concentration around the clock, and draw a curve of concentration change over time; Calculate the actual CO2 emissions based on the detected concentration; While detecting CO2 in real time, it is compared with the predicted value to determine the accuracy of the prediction. If the calculated error is within 10%, it is considered to be predictive; if the error exceeds 10%, it means that the prediction accuracy is poor.

9. According to claim 8, a method for intelligent carbon emission prediction and analysis in the tunnel construction stage, It is characterized in that In order to obtain accurate CO2 concentration values, sensors are arranged at a certain density according to the real-time construction length. 2-3 CO2 sensors should be arranged near the construction face, and a CO2 sensor should be arranged every 50-100 m at other constructed locations.

Images