Commuting carbon efficiency causal inference method and device for multiple bus priority strategies, and medium

Abstract

The invention discloses a commuting carbon efficiency causal inference method and device for multiple bus priority strategies, and a medium. The commuting carbon efficiency causal inference method comprises the following steps: constructing a variable set containing control variables (bus lanes and signal priority strategies), covariables (carbon efficiency influence factors) and carbon efficiency true values; and splicing the control variable and the covariable into a joint feature vector, and inputting the joint feature vector into the trained multi-strategy causal neural network. The network outputs carbon efficiency estimation values under four situations through completely decoupled branches: no strategy, only special channel, only signal priority and joint strategy. And on the basis of no strategy, the carbon efficiency improvement rates under the implementation of the three strategies are calculated respectively, that is, the ratio of the difference value to the basic value is estimated, so that the emission reduction effect of each strategy is quantified.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban transportation carbon efficiency, and in particular relates to a commuting carbon efficiency causal inference method, device, and medium for a multi-public transport priority strategy. Background Art

[0002] With the continuous increase in urban traffic loads, bus priority strategies have become an important means of alleviating traffic congestion, improving travel efficiency, and promoting green travel. Bus priority strategies (such as dedicated bus lanes and bus signal priority) can improve bus speed and punctuality, enhance the passenger experience, and shift private car travel demand toward the bus system, thereby optimizing the transportation structure and improving the overall energy efficiency of the transportation system. Compared to traditional indicators that only examine the total carbon emissions of the bus system, the bus carbon efficiency indicator focuses on energy efficiency per unit travel distance or per unit passenger flow, providing a more scientific and quantitative basis for the formulation and optimization of transportation policies.

[0003] However, existing research has significant shortcomings in its causal inference of commuting carbon efficiency. First, mainstream studies have mostly focused solely on assessing the carbon efficiency of public transportation systems, overlooking the combined carbon effects of public transportation and private car ownership. In reality, the ultimate goal of optimizing and prioritizing public transportation systems is to encourage private car users to migrate to public transportation and, from a global perspective, reduce the overall carbon emissions of urban commuting. Therefore, carbon efficiency analysis of a single public transportation system fails to accurately reflect the true emission reduction effects and changes in transportation structure.

[0004] Second, current evaluations of bus priority strategies are often limited to single measures (e.g., examining only bus lanes or signal priority), failing to systematically capture the synergistic effects and complex interactions of multiple bus priority strategies. In reality, measures like bus lanes and signal priority are often implemented in parallel. While their synergistic effects can lead to greater emissions reduction potential, they can also lead to suppression or even diminishing marginal effects between strategies. Policy evaluations based on a single strategy perspective struggle to reveal the actual impacts and mechanisms of a multi-strategy combination. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a commuting carbon efficiency causal inference method, equipment and medium for multiple public transport priority strategies.

[0006] In a first aspect, an embodiment of the present invention provides a causal inference method for commuting carbon efficiency with multiple public transport priority strategies, the method comprising:

[0007] Construct a bus carbon efficiency variable set, which includes control variables, covariates, and the true value of commuting carbon efficiency; the control variables are bus priority strategies, including bus lane setting strategies and bus signal priority strategies; the covariates are factors affecting bus carbon efficiency;

[0008] The control variables and covariates are concatenated to obtain a joint feature vector; the joint feature vector is input into the trained multi-strategy causal neural network model, and the fully decoupled branch network outputs the first commuting carbon efficiency estimate when the bus priority strategy is not implemented under the covariate condition, the second commuting carbon efficiency estimate when only the bus lane setting strategy is implemented under the covariate condition, the third commuting carbon efficiency estimate when only the bus signal priority strategy is implemented under the covariate condition, and the fourth commuting carbon efficiency estimate when the bus lane setting strategy and the bus signal priority strategy are jointly implemented under the covariate condition;

[0009] The ratio of the difference between the second commuting carbon efficiency estimate and the first commuting carbon efficiency estimate in the first commuting carbon efficiency estimate is used as the commuting carbon efficiency improvement rate when only the bus lane setting strategy is implemented under the covariate condition; the ratio of the difference between the third commuting carbon efficiency expectation and the first commuting carbon efficiency estimate in the first commuting carbon efficiency estimate is used as the commuting carbon efficiency improvement rate when only the bus signal priority strategy is implemented under the covariate condition; the ratio of the difference between the fourth commuting carbon efficiency estimate and the first commuting carbon efficiency estimate in the first commuting carbon efficiency estimate is used as the commuting carbon efficiency improvement rate when the bus lane setting strategy and the bus signal priority strategy are jointly implemented under the covariate condition.

[0010] In a second aspect, an embodiment of the present invention provides an electronic device, including:

[0011] at least one processor; and

[0012] a memory communicatively connected to the at least one processor; wherein,

[0013] The memory stores one or more computer programs that can be executed by the at least one processor, and the one or more computer programs are executed by the at least one processor to enable the at least one processor to perform the commuting carbon efficiency causal inference method of multiple bus priority strategies as described above.

[0014] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned commuting carbon efficiency causal inference method for multiple public transport priority strategies.

[0015] In a fourth aspect, an embodiment of the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the above-mentioned commuting carbon efficiency causal inference method of multiple public transport priority strategies.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] The application provides a commuting carbon efficiency causal inference method for multiple bus priority strategies, which measures the commuting carbon efficiency improvement rate under single and joint bus priority strategies, quantifies the promotion, synergy and inhibition effects of various bus priority strategies (including single and joint measures) on carbon efficiency, and provides strong technical support for urban traffic green transformation and scientific decision-making. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0019] Figure 1 is a flowchart of the commuting carbon efficiency causal inference method for multiple bus priority strategies provided by the embodiments of the present application;

[0020] Figure 2 is a technical roadmap of the commuting carbon efficiency causal inference method for multiple bus priority strategies provided by the embodiments of the present application;

[0021] Figure 3 is a framework diagram of the multi-strategy causal neural network model provided by the embodiments of the present application;

[0022] Figure 4 is a schematic diagram of an electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0024] It should be noted that the features in the following embodiments and implementation manners can be combined with each other without conflict.

[0025] Existing researches mostly focus on the causal inference of single bus priority strategy on commuting carbon efficiency, which is difficult to comprehensively reveal the emission reduction effect and mechanism under the coordination or interaction of multiple bus priority strategies (such as bus lane and bus signal priority). In reality, different bus priority measures may have complex interaction relationships such as mutual promotion or mutual inhibition, and existing methods are difficult to accurately identify and quantify the synergistic effect and mechanism of multiple strategies. In view of the importance of bus priority strategy to improve bus efficiency and promote green travel, embodiments of the present invention provide a commuting carbon efficiency causal inference method for multiple bus priority strategies, as shown in Figure 1 and Figure 2 The method comprises the following steps:

[0026] Step S1, a bus carbon efficiency variable set is constructed, the bus carbon efficiency variable set comprising control variables, covariates, and true values of commuting carbon efficiency; the control variables are bus priority strategies, including: bus lane setting strategy, bus signal priority strategy; the covariates are bus carbon efficiency influencing factors.

[0027] Further, the bus multi-source heterogeneous data includes bus smart cards, mobile phone signaling, road networks, points of interest (POI), etc., and a multi-dimensional variable set is established based on the bus multi-source heterogeneous data , wherein: is the variable set, representing the implementation state (0 / 1, not implemented / implemented) of the bus lane ( ) and bus signal priority ( ) strategy, respectively. is the covariate set, the covariates being bus carbon efficiency influencing factors, including: private car density, vehicle energy type, road type, station density, GDP density, land mix degree, road congestion, etc.

[0028] Further, in the bus lane setting strategy, the mileage of the bus line running on the bus lane and the proportion of the bus lane mileage are estimated, and clustering is performed, so as to divide the bus line into a first bus lane proportion line mainly running on the bus lane, a second bus lane proportion line mainly running on the ordinary lane; in the bus signal priority strategy, the bus line is clustered into a first bus signal priority proportion line (high bus signal priority proportion line with higher bus signal priority rate), and a second bus signal priority proportion line (low bus signal priority proportion line with lower bus signal priority rate).

[0029] Furthermore, in this example, the process of estimating the mileage of the bus route traveling on the bus lane and the mileage of the bus route traveling on the ordinary lane includes:

[0030] Get bus GIS route data;

[0031] Obtain all bus routes and bus lanes based on bus GIS route data;

[0032] Divide each bus route into several equally spaced sub-route segments;

[0033] For a sub-route segment, obtain the bus-only road segment closest to the midpoint of the sub-route segment; calculate the vertical projection distance from the midpoint to the bus-only road segment; if the vertical projection distance is less than the total lane width of the bus-only road segment, determine that the bus is traveling on the bus lane on the sub-route segment; otherwise, determine that the bus is traveling on the ordinary lane on the sub-route segment;

[0034] Traverse each sub-route segment and determine whether the bus is traveling on the bus lane or the ordinary lane in each sub-route segment; accumulate the length of the sub-route segments traveling on the bus lane to obtain the bus lane mileage of the current bus route;

[0035] Traverse all bus routes and obtain the bus lane mileage corresponding to each bus route.

[0036] Furthermore, in this example, the calculation process of the bus signal priority rate includes:

[0037] For each intersection signal timing data passed by each bus line, the Transit Signal Priority (TSP) field, signal phase information and geographic location information of each intersection are extracted; thus obtaining the intersection set passed by a bus line. and signalized priority intersection collection ; It should be noted that if the intersection Equipped with bus signal priority function, and its priority phase is the same as the line If the running direction is consistent, the intersection is classified as the bus priority intersection of the line, thus forming a line Signal priority intersection set ;

[0038] The ratio of the total length of all consecutive sections with bus signal priority on the bus route to the total length of the bus route is used as the route spatial continuity factor; the expression is as follows:

[0039]

[0040] Where, represents the spatial continuity factor of the line; Indicates line The total length of all continuous sections with bus signal priority function. If two adjacent intersections are equipped with bus signal priority lights and the priority phase is consistent with the line running direction, the middle section is regarded as a continuous priority section; Indicates line Total length;

[0041] Calculate the ratio of the sum of peak-hour passenger flows at all signalized priority intersections to the sum of peak-hour passenger flows at all intersections, and then multiply it by the spatial continuity factor to obtain the bus signal priority rate. The expression is as follows:

[0042]

[0043] Where, Indicates the bus signal priority rate, represents the passenger flow weighting factor;

[0044] It should be noted that in this example, the continuity factor It can effectively distinguish between lines with scattered priority signal distribution (low continuity) and lines with high concentration (high continuity), which is of great significance to reflecting the efficiency of bus operation. In addition, considering the actual impact of key nodes such as transportation hubs and main road intersections on bus travel, this example further introduces the passenger flow of each intersection during peak hours. As a weighting factor, it effectively quantifies the contribution of different intersection signal priority configurations to the overall operation of the line. The bus signal priority rate indicator not only reflects the continuity of the spatial distribution of bus priority signals, but also takes into account the weighted influence of key nodes, more comprehensively depicting the actual implementation of bus line signal priority strategies.

[0045] Step S2, concatenate the control variables and the covariates to obtain a joint feature vector; input the joint feature vector into the trained multi-strategy causal neural network model, and output the first commuting carbon efficiency estimate when the bus priority strategy is not implemented under the covariate condition, the second commuting carbon efficiency estimate when only the bus lane setting strategy is implemented under the covariate condition, the third commuting carbon efficiency estimate when only the bus signal priority strategy is implemented under the covariate condition, and the fourth commuting carbon efficiency estimate when the bus lane setting strategy and the bus signal priority strategy are jointly implemented under the covariate condition.

[0046] It should be noted that for any given sample, it is impossible to observe all possible counterfactual situations at the same time, that is, Estimated carbon efficiency of the first commute without implementing the bus priority strategy , in the covariate Estimated carbon efficiency of the second commuting scenario with only the bus lane strategy implemented under the same conditions , the estimated carbon efficiency of the third commute when only the bus signal priority strategy is implemented under the covariate condition and in the covariate Estimated carbon efficiency of the fourth commuting trip under the combined implementation of bus lane setting strategy and bus signal priority strategy Only one of them can be actually observed, and the rest are counterfactual results. Therefore, traditional causal inference methods are difficult to accurately estimate the joint effect of multiple strategies. To this end, this example proposes a multi-policy causal neural network model (Multi-PolicyCausal Neural Network, MCN), whose model structure is as follows Figure 3 As shown, the multi-strategy causal neural network model can estimate the individualized causal effects of commuting carbon efficiency for individual samples under various strategy combinations. Based on an end-to-end deep learning architecture, this multi-strategy causal neural network model jointly models the impact of processing variables and covariates on outcome indicators, improving the reliability and generalization of multi-strategy causal inference.

[0047] Specifically, one-hot encoding is used to process variables Vectorize and concatenate with normalized covariates to form a joint feature vector , as the input of the multi-strategy causal neural network model. Through the backbone shared network in the multi-strategy causal neural network model , the joint feature vector of the input Extracting the high-level shared representation, the expression is as follows:

[0048]

[0049] And set up four fully parameter decoupled branch networks (Right now 、 、 and ), outputs the estimated carbon efficiency of commuting, expressed as follows:

[0050]

[0051] In step S3, the ratio of the difference between the second commuting carbon efficiency estimate and the first commuting carbon efficiency estimate in the first commuting carbon efficiency estimate is used as the commuting carbon efficiency improvement rate when only the bus lane setting strategy is implemented under the covariate condition; the ratio of the difference between the third commuting carbon efficiency expectation and the first commuting carbon efficiency estimate in the first commuting carbon efficiency estimate is used as the commuting carbon efficiency improvement rate when only the bus signal priority strategy is implemented under the covariate condition; the ratio of the difference between the fourth commuting carbon efficiency estimate and the first commuting carbon efficiency estimate in the first commuting carbon efficiency estimate is used as the commuting carbon efficiency improvement rate when the bus lane setting strategy and the bus signal priority strategy are jointly implemented under the covariate condition.

[0052] Furthermore, to quantify the causal effects of different bus priority strategies on improving commuting carbon efficiency, we define the commuting carbon efficiency improvement rate (CER) under single and combined strategies, as follows:

[0053]

[0054]

[0055]

[0056] Where, and The covariates are The improvement rate of commuting carbon efficiency when implementing bus lane strategy and bus signal priority strategy separately under the same conditions is: Refers to the covariate The improvement rate of commuting carbon efficiency by jointly implementing bus lane strategy and bus signal priority strategy under the same conditions is: Refers to the covariate The estimated carbon efficiency of the first commute without implementing the bus priority strategy under the condition of Refers to the covariate The estimated carbon efficiency of the second commuting scenario with only the bus lane strategy implemented under the condition of Refers to the estimated carbon efficiency of the third commute when only the bus signal priority strategy is implemented under the covariate condition. Refers to the covariate The estimated carbon efficiency of the fourth commuter under the combined implementation of the bus lane setting strategy and bus signal priority strategy is: is the sample number.

[0057] Furthermore, to describe the interaction between multiple bus priority strategies, that is, whether there is mutual synergy or mutual inhibition, the specific calculation formula is as follows:

[0058]

[0059] Where, is the covariate The synergy index of various bus priority strategies under the condition of , indicating that the strategies have positive synergy (superadditivity); if , indicating that there may be an inhibitory effect (subadditivity).

[0060] Furthermore, the process of training the multi-strategy causal neural network model includes:

[0061] Step S100: constructing a joint loss function, where the joint loss function is a weighted sum of sample balance loss, causal embedding alignment loss, multi-strategy pairing loss, and regularization loss;

[0062] Among them, the sample balance loss uses the mean square error between the true value of commuting carbon efficiency and the estimated value of commuting carbon efficiency to be calculated based on the inverse propensity weight; the causal embedding alignment loss measures the distribution deviation of the embedding vectors corresponding to each two strategies based on the distribution distance; the multi-strategy pairing loss is based on the adaptive pairing weight to constrain the consistency of treatment effects between sample pairs with similar covariate characteristics in the embedding space.

[0063] Among them, the joint loss function The expression is as follows:

[0064]

[0065] in, represents the sample balance loss, represents the causal embedding alignment loss, represents the multi-strategy pairing loss, represents the regularization loss, represents the weight corresponding to the causal embedding alignment loss, represents the weight corresponding to the multi-strategy pairing loss, Represents the weight corresponding to the regularization loss.

[0066] Among them, the sample balance loss adopts the weighted mean square error loss to ensure the global balance of samples in different strategy groups during the optimization process of the multi-strategy causal neural network model. The expression is as follows:

[0067]

[0068] Where, represents the number of samples, i represents the i-th sample, represents the true value of commuting carbon efficiency corresponding to the i-th sample, represents the estimated carbon efficiency of commuting corresponding to the i-th sample, represents the inverse propensity weight.

[0069] Among them, the inverse tendency weight The setup process includes:

[0070] To address the significant heterogeneity in the distribution of covariates across different strategy groups, traditional methods typically rely on simplified models such as linear regression to estimate propensity scores. These methods are only applicable to scenarios with low covariate dimensionality or a limited number of strategy groups, and are unable to effectively address the complex distribution biases found in high-dimensional, multi-strategy combinations. Therefore, this example proposes a Deep Auxiliary Propensity (DAP) module, which uses a neural network to perform high-dimensional nonlinear modeling of the probability that a sample belongs to each strategy group, thereby accurately estimating the probability distribution of multiple strategy groups. The specific steps are as follows:

[0071] First, DAP is used to analyze the covariate characteristics. Encode and output the probability distribution of each strategy group. The calculation formula is:

[0072]

[0073] Where, Representation sample belong The probability of the strategy group. In order to correct the problem of uneven sample distribution between strategy groups, the inverse tendency weight is further introduced. Perform sample reweighting, and the calculation formula is as follows:

[0074]

[0075] Where, To prevent small positive numbers with a denominator of zero, effectively improve numerical stability.

[0076] The calculation process of the multi-strategy alignment loss includes the following: To eliminate the distribution differences in the covariate embedding space between different strategy groups, this paper proposes a Causal Embedding Alignment (CEA) module. This module uses a distribution distance metric (such as Maximum Mean Discrepancy (MMD) or Wasserstein distance) to quantitatively evaluate the distribution deviation of the embedding space of each strategy group. The expression is as follows:

[0077]

[0078] in, and Traversing all different strategy group combinations. The multi-strategy alignment loss improves the distribution consistency of the embedding space under multiple strategy dimensions, laying a solid foundation for subsequent individualized causal effect estimation.

[0079] The calculation process of multi-strategy pairing loss includes: On the basis of achieving global distribution consistency, this example further proposes a multi-strategy group pairing weight correction mechanism for the individual level to enhance the local consistency and causal explanation ability of the model. In specific implementation, for each target sample , dynamically retrieve the closest covariate features in the high-dimensional embedding space Different strategy group samples form a paired sample set Based on the Euclidean distance in the embedding space, adaptive pairing weights are defined , the expression is as follows:

[0080]

[0081] Where, is the bandwidth parameter, which is used to adjust the sensitivity of embedding distance to weight distribution; The larger the value, the closer the features are. This weighting mechanism can flexibly adapt to the complexity of high-dimensional covariate distribution. Furthermore, we define the multi-strategy pairing loss , which is used to constrain the consistency of treatment effects between sample pairs with similar covariate characteristics in the embedding space. The expression is as follows:

[0082]

[0083] This multi-strategy pairing loss aims at the local consistency of "individual treatment effects under the same strategy switching" in the covariate space, expands the traditional "double difference" idea, and innovatively introduces the "four-difference" structure, which significantly improves the smoothness and matching fairness of the local treatment effect and reduces heterogeneity bias.

[0084] Step S200: inputting the bus carbon efficiency variable set into the multi-strategy causal neural network model, and training the multi-strategy causal neural network model based on the joint loss function.

[0085] Furthermore, the backbone network The parameters of the branch network are optimized by back propagating the total loss of all samples. Only observations from the corresponding strategy group are used for parameter updates. The multi-strategy causal neural network model achieves an organic fusion of global distribution alignment, local pairwise consistency, and regularization constraints, significantly improving the consistency of the multi-strategy causal inference model in evaluation and mechanism attribution.

[0086] Furthermore, the calculation process of the true value of commuting carbon efficiency includes:

[0087]

[0088] Where, Indicates line The true value of commuting carbon efficiency, represents the city-level commuting carbon emissions per passenger mile, Indicates line of commuting carbon emissions per passenger mile. , it means that the carbon efficiency of the line is higher than the average level of the city (that is, the unit carbon emissions are lower than the average); otherwise, it is lower than the average level.

[0089] Among them, the line The calculation process of commuting carbon emissions per passenger mile includes:

[0090]

[0091] Where, is the expansion coefficient of public transportation travel demand; The carbon emission factor of the entire bus life cycle covers carbon emissions from bus production, operation, scrapping, and infrastructure (charging stations, stops, dedicated lanes, parking lots, etc.); It is a bus network link length; It is bus trip link passenger capacity at the time; is the expansion coefficient of private car travel demand; It is the full lifecycle carbon emission factor of a private car, covering the full lifecycle emissions of vehicle production, operation, scrapping and related infrastructure; When private car users switch to public transportation, they choose public transportation routes. probability; It's a bus route Middle Line The mileage ratio reflects the contribution of different routes to the overall path; It is The travel distance of each private car trip; It is The passenger capacity of each private car trip can be taken as the average passenger capacity of private cars in the selected city’s traffic annual report or traffic survey data (usually 1.2-1.5).

[0092] The calculation process of city-level commuting carbon emissions per passenger mile includes:

[0093]

[0094] Among them, private car users choose bus routes when switching to public transportation Probability The calculation process includes:

[0095] For each candidate bus route ,construct a multi-dimensional feature input vector, including: total travel time, fare, comfort index, number of transfers, weather, real-time congestion, user historical preferences, etc.

[0096] Among them, candidate bus routes The calculation process of the corresponding comfort index includes:

[0097]

[0098]

[0099] Where, For the line The number of seats in the vehicle, For Link Passenger capacity during the period, is the standing comfort coefficient.

[0100] Input the multidimensional feature input vector into the neural logit model to obtain the bus route The observable utility value is expressed as follows:

[0101]

[0102]

[0103]

[0104]

[0105] Where, It's a bus route The utility value of It's a bus route The observable utility value of It's a bus route The random disturbance term of , which represents the unobserved preference or error term, It's a bus route The multi-dimensional feature input vector, and is the hidden layer output, and is the trainable hidden layer weight matrix, and is the trainable hidden layer bias vector, is the activation layer.

[0106] By performing Softmax normalization on the observable utility value, the selection probability of each candidate bus route is Calculated as follows:

[0107]

[0108] Furthermore, this paper proposes a SHAP value calculation method for multiple bus priority strategies (hereinafter referred to as "strategy SHAP"), which can more accurately quantify the individual and global contributions of each feature to the improvement rate of commuting carbon efficiency in a multi-strategy scenario. While retaining the advantages of traditional SHAP feature attribution and model interpretability, this method decomposes the characteristic effects of different bus priority strategies (such as bus lanes, bus signal priority, and their combined implementation), thereby systematically revealing the impact changes and causal contribution mechanisms of each characteristic variable under different policy settings. Specifically, the calculation formula of the strategy SHAP value is as follows:

[0109]

[0110]

[0111]

[0112] in, To implement only bus lane strategy ( ) conditions, the characteristics Individual marginal contribution to the improvement in commuting carbon efficiency; To implement only bus signal priority strategy ( ) individual contribution under the conditions; It is implemented in conjunction with bus lanes and signal priority ( ) individual contribution under certain conditions. Represents the full set of features Medium Any subset outside and are the number of elements in the full set and subset of features, respectively.

[0113] For each feature , for all samples respectively 、 、 The mean of the absolute values ​​is calculated to obtain the global strategy SHAP value under the bus lane, bus signal priority, and the combined strategy. The calculation formula is as follows:

[0114]

[0115]

[0116]

[0117] By comparing different features in 、 and The high and low values ​​under can be used to identify the key factors that have the most significant impact on the improvement rate of commuting carbon efficiency under the two scenarios of implementing the bus lane strategy alone, the bus signal priority strategy alone, and the combined implementation.

[0118] This method can comprehensively reveal the marginal effects, heterogeneity and causal mechanisms of characteristic variables under different bus priority strategy combinations, provide quantitative and explainable technical support for the explanation of emission reduction mechanisms, attribution of influencing factors and policy optimization under the coordinated implementation of multiple strategies, and further provide strong guarantees for scientific decision-making and precise governance in different cities, routes and multi-strategy coordinated application scenarios.

[0119] In summary, the present invention provides a causal inference method for commuting carbon efficiency under multiple bus priority strategies. By measuring the improvement rate of commuting carbon efficiency under single and combined bus priority strategies, the improvement, synergy, and inhibition effects of multiple bus priority strategies (including single and combined measures) on carbon efficiency are quantified, providing strong technical support for the green transformation of urban transportation and scientific decision-making.

[0120] Accordingly, the present application also provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the commuting carbon efficiency causal inference method of the multi-public transport priority strategy as described above. Figure 4 As shown, the hardware structure diagram of any device with data processing capability is provided for the causal inference method of commuting carbon efficiency of multiple bus priority strategies provided by the embodiment of the present invention. Figure 4 In addition to the processor, memory, and network interface shown, any device with data processing capabilities in which the apparatus in the embodiment is located may also include other hardware, generally based on the actual functions of the device with data processing capabilities, which will not be described in detail.

[0121] Accordingly, the present application also provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the commuting carbon efficiency causal inference method for the multi-public transport priority strategy described above. The computer-readable storage medium can be the internal storage unit of any device with data processing capabilities described in any of the aforementioned embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc. equipped on the device. Furthermore, the computer-readable storage medium can also include both the internal storage unit and the external storage device of any device with data processing capabilities. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and can also be used to temporarily store data that has been output or is to be output.

[0122] The above embodiments are intended only to illustrate the design concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. The scope of protection of the present invention is not limited to the above embodiments. Therefore, any equivalent changes or modifications made based on the principles and design concepts disclosed in the present invention are within the scope of protection of the present invention.

Claims

1. A causal inference method for commuting carbon efficiency with multiple bus priority strategies, characterized by: The method comprises: Construct a bus carbon efficiency variable set, which includes control variables, covariates, and the true value of commuting carbon efficiency; the control variables are bus priority strategies, including bus lane setting strategies and bus signal priority strategies; the covariates are factors affecting bus carbon efficiency; The control variables and covariates are concatenated to obtain a joint feature vector; the joint feature vector is input into the trained multi-strategy causal neural network model, and the fully decoupled branch network outputs the first commuting carbon efficiency estimate when the bus priority strategy is not implemented under the covariate condition, the second commuting carbon efficiency estimate when only the bus lane setting strategy is implemented under the covariate condition, the third commuting carbon efficiency estimate when only the bus signal priority strategy is implemented under the covariate condition, and the fourth commuting carbon efficiency estimate when the bus lane setting strategy and the bus signal priority strategy are jointly implemented under the covariate condition; The ratio of the difference between the second commuting carbon efficiency estimate and the first commuting carbon efficiency estimate in the first commuting carbon efficiency estimate is used as the commuting carbon efficiency improvement rate when only the bus lane setting strategy is implemented under the covariate condition; the ratio of the difference between the third commuting carbon efficiency expectation and the first commuting carbon efficiency estimate in the first commuting carbon efficiency estimate is used as the commuting carbon efficiency improvement rate when only the bus signal priority strategy is implemented under the covariate condition; the ratio of the difference between the fourth commuting carbon efficiency estimate and the first commuting carbon efficiency estimate in the first commuting carbon efficiency estimate is used as the commuting carbon efficiency improvement rate when the bus lane setting strategy and the bus signal priority strategy are jointly implemented under the covariate condition.

2. The causal inference method for commuting carbon efficiency with multiple public transport priority strategies according to claim 1 is characterized in that: In the bus lane setting strategy, the mileage of bus routes on bus lanes and the proportion of bus lane mileage are estimated and clustered, thereby dividing bus routes into routes with the first bus lane proportion and routes with the second bus lane proportion; In the bus signal priority strategy, bus routes are clustered into first bus signal priority ratio routes and second bus signal priority ratio routes according to the bus signal priority ratio.

3. The causal inference method for commuting carbon efficiency with multiple public transport priority strategies according to claim 2 is characterized in that: The calculation process of the bus signal priority rate includes: Obtain a set of intersections that a bus route passes through and a set of signal priority intersections; The ratio of the total length of all consecutive sections with bus signal priority on the bus route to the total length of the bus route is used as the route spatial continuity factor; Calculate the proportion of the sum of the peak-hour passenger flows corresponding to all signal priority intersections in the sum of the peak-hour passenger flows corresponding to all intersections, and then multiply it by the line space continuity factor to obtain the bus signal priority rate.

4. The causal inference method for commuting carbon efficiency with multiple public transport priority strategies according to claim 1 is characterized in that: The calculation process of the true value of commuting carbon efficiency includes: The ratio of commuting carbon emissions per passenger mile to the city-level commuting carbon emissions per passenger mile is taken as the true value of commuting carbon efficiency; The calculation process for commuting carbon emissions per passenger mile includes: calculating the sum of carbon emissions from buses and private cars, and the sum of travel distances of bus users and private car users, and taking the ratio of the two as the commuting carbon emissions per passenger mile. The carbon emissions and travel distance of private car users are obtained based on the probability of selecting each candidate bus route when switching from private cars to buses, and the mileage ratio of the bus route. The calculation process of commuting carbon emissions per passenger mile at the city level includes: calculating the sum of carbon emissions from buses and private cars, the sum of travel distances of bus users and private car users, and taking the ratio of the two as the city-level commuting carbon emissions per passenger mile.

5. The causal inference method for commuting carbon efficiency with multiple public transport priority strategies according to claim 4 is characterized in that: The calculation process of the selection probability of each candidate bus route when switching from private car to bus travel includes: For each candidate bus route, the total travel time, fare, comfort index, number of rides, weather characteristics, and road condition characteristics are obtained and input into the neural logit model to obtain the observable utility value of each candidate bus route; The observable utility value corresponding to each candidate bus route is Softmax normalized to obtain the selection probability of each candidate bus route.

6. The causal inference method for commuting carbon efficiency with multiple public transport priority strategies according to claim 1 is characterized in that: The process of training a multi-strategy causal neural network model includes: Constructing a joint loss function, which is a weighted sum of sample balance loss, causal embedding alignment loss, multi-strategy pairing loss, and regularization loss; The sample balance loss uses the mean squared error between the true value of commuting carbon efficiency and the estimated value of commuting carbon efficiency to be calculated based on the inverse propensity weight; the causal embedding alignment loss measures the distribution deviation of the embedding vectors corresponding to each two strategies based on the distribution distance; and the multi-strategy pairing loss uses adaptive pairing weights to constrain the consistency of treatment effects between sample pairs with similar covariate characteristics in the embedding space. The bus carbon efficiency variable set is input into the multi-strategy causal neural network model, and the multi-strategy causal neural network model is trained based on the joint loss function.

7. The causal inference method for commuting carbon efficiency with multiple public transport priority strategies according to claim 6 is characterized in that: The process of setting the inverse tendency weight includes: inputting the covariate into the deep auxiliary tendency network to obtain the probability distribution of each strategy, and normalizing each probability distribution to obtain the inverse tendency weight; The process of setting the adaptive pairing weights includes defining the similarity between sample pairs based on the Euclidean distance in the embedding space.

8. The causal inference method for commuting carbon efficiency with multiple public transport priority strategies according to claim 1 is characterized in that: The method further comprises: Obtain the SHAP value of each covariate characteristic on the improvement rate of commuting carbon efficiency under the condition of implementing only the bus lane strategy, and take the average to obtain the global strategy SHAP value under the condition of implementing only the bus lane strategy; Obtain the SHAP value of each covariate characteristic on the improvement rate of commuting carbon efficiency under the condition of implementing only the bus signal priority strategy, and take the average to obtain the global strategy SHAP value under the condition of implementing only the bus signal priority strategy; Obtain the SHAP value of each covariate characteristic on the improvement rate of commuting carbon efficiency under the conditions of jointly implementing the bus lane strategy and the bus signal priority strategy, and take the average to obtain the global strategy SHAP value under the conditions of jointly implementing the bus lane strategy and the bus signal priority strategy; The SHAP values ​​of the corresponding global strategies under the conditions of implementing only the bus lane strategy, only the bus signal priority strategy, and the combined implementation of the bus lane strategy and the bus signal priority strategy were compared to obtain the key factors that have the most significant impact on the improvement rate of commuting carbon efficiency.

9. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores one or more computer programs that can be executed by the at least one processor, and the one or more computer programs are executed by the at least one processor to enable the at least one processor to perform the commuting carbon efficiency causal inference method of multiple bus priority strategies according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program implements the commuting carbon efficiency causal inference method of multiple public transport priority strategies according to any one of claims 1 to 8.

Images