An ecological environment governance technology applicability evaluation method and system

Abstract

This invention discloses a method and system for assessing the applicability of ecological environment governance technologies. First, it acquires a set of historical governance case data, target ecological environment scenario parameters, and a set of candidate governance technologies. Based on the historical governance case data, it constructs a causal forest scenario adaptability prediction model to predict the governance effects of each candidate technology in the target scenario using counterfactual methods, generating a target scenario evaluation matrix. The matrix is ​​standardized, and the weights of the indicators are determined according to the distribution of the indicator data to form a weighted evaluation matrix. Positive and negative ideal solutions are determined, and the distance evaluation parameters and correlation evaluation parameters of each candidate governance technology are calculated. These are then fused according to preset preference coefficients to obtain a comprehensive applicability evaluation value, which is then sorted to generate the governance technology applicability assessment result corresponding to the target scenario.

Description

Technical Field

[0001] This invention relates to the field of ecological environment governance technology, and more specifically, to a method and system for assessing the applicability of ecological environment governance technologies. Background Technology

[0002] With increasingly stringent requirements for ecological and environmental governance, various governance technologies have emerged in the fields of pollution prevention and control, energy conservation and carbon reduction, and resource recycling. Different governance technologies have varying degrees of adaptability to different scenario conditions, such as temperature, pollutant inlet concentration, environmental humidity, treatment scale, and investment budget. Therefore, it is necessary to comprehensively consider multiple indicators, including governance effectiveness, economic cost, technological maturity, operational stability, and promotional value, to conduct an applicability assessment of candidate governance technologies.

[0003] Existing governance technology assessment methods primarily rely on expert experience, fixed weights, or static evaluation matrices for comprehensive analysis. Subjective factors significantly influence the assessment results, making it difficult to dynamically adjust indicator weights based on the distribution of indicator data in different scenarios. Furthermore, static evaluation matrices typically only record the governance effects of technologies in existing scenarios, failing to leverage historical governance cases to extrapolate the potential performance of each technology in target ecological and environmental scenarios. This results in a lack of comparability among different governance technologies under unified scenario conditions. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for assessing the applicability of ecological environment governance technologies.

[0005] In a first aspect, embodiments of the present invention provide a method for assessing the applicability of ecological environment governance technologies, the method comprising:

[0006] Acquire a set of historical governance case data, target ecological environment scenario parameters, and a set of candidate governance technologies. The set of historical governance case data includes historical governance scenario information, governance technology information, and governance effect indicator information.

[0007] Based on the historical governance case data set, a scenario-adaptive prediction model based on causal forest is constructed. The scenario-adaptive prediction model is used to predict the governance effect of each candidate governance technology in the candidate governance technology set under the target ecological environment scenario parameters, and generate a target scenario evaluation matrix.

[0008] Based on the indicator attributes corresponding to each governance effect indicator in the target scenario evaluation matrix, the target scenario evaluation matrix is ​​standardized to obtain a standardized evaluation matrix.

[0009] The weights of each governance effectiveness indicator are determined based on the data distribution of each governance effectiveness indicator in the standardized evaluation matrix, and the standardized evaluation matrix is ​​weighted according to the indicator weights to obtain a weighted evaluation matrix.

[0010] The positive and negative ideal solutions are determined based on the weighted evaluation matrix, and the distance evaluation parameters and correlation evaluation parameters of each candidate governance technology relative to the positive and negative ideal solutions are calculated.

[0011] The distance evaluation parameter and the correlation evaluation parameter are fused according to the preset preference coefficient to obtain the comprehensive applicability evaluation value of each candidate governance technology;

[0012] Based on the comprehensive applicability evaluation value, each candidate governance technology is ranked, and the applicability assessment results of the governance technologies corresponding to the target ecological environment scenario parameters are generated.

[0013] Secondly, embodiments of the present invention provide an ecological environment governance technology applicability assessment system, including at least one service node;

[0014] The service node includes a storage unit and a computing unit; the storage unit is used to store program code; the computing unit is used to run the program code to execute the method described in the first aspect.

[0015] Compared with existing technologies, the beneficial effects provided by this invention include: compared with existing technologies, this invention can use historical governance cases to make counterfactual predictions of the governance effects of candidate governance technologies in target ecological environment scenarios, thereby improving the comparability of different governance technologies in different scenarios; objectively determine the weight of indicators based on the distribution of indicator data, thereby reducing the influence of subjective weighting; integrate distance evaluation and correlation evaluation, taking into account the overall proximity and the correlation characteristics of indicator changes, thereby improving the accuracy, stability and interpretability of the evaluation results, and realizing the automated screening and ranking of governance technologies. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating the steps of the ecological environment governance technology applicability assessment method provided in an embodiment of the present invention;

[0018] Figure 2 This is a schematic diagram of the structure of the target scene evaluation matrix generated based on causal forest provided in an embodiment of the present invention;

[0019] Figure 3 This is a schematic diagram of the Euclidean distance and grey relational fusion evaluation structure provided in an embodiment of the present invention;

[0020] Figure 4 This is a schematic diagram of the counterfactual prediction confidence correction structure provided in an embodiment of the present invention;

[0021] Figure 5 This is a schematic diagram of the interactive correction structure of the adjacent governance technology provided in an embodiment of the present invention;

[0022] Figure 6 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0024] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0025] In order to solve the technical problems mentioned in the background art Figure 1 This is a flowchart illustrating the applicability assessment method for ecological environment governance technologies provided in this embodiment. The following is a detailed description of this ecological environment governance technology applicability assessment method.

[0026] Step S201: Obtain a set of historical governance case data, target ecological environment scenario parameters, and a set of candidate governance technologies. The set of historical governance case data includes historical governance scenario information, governance technology information, and governance effect indicator information.

[0027] Step S202: Construct a scenario-adaptive prediction model based on causal forest based on the historical governance case data set. Use the scenario-adaptive prediction model to perform counterfactual prediction on the governance effect of each candidate governance technology in the candidate governance technology set under the target ecological environment scenario parameters, and generate a target scenario evaluation matrix.

[0028] Step S203: Based on the indicator attributes corresponding to each governance effect indicator in the target scenario evaluation matrix, the target scenario evaluation matrix is ​​standardized to obtain a standardized evaluation matrix.

[0029] Step S204: Determine the index weight of each governance effect index according to the data distribution of each governance effect index in the standardized evaluation matrix, and perform weighted processing on the standardized evaluation matrix according to the index weight to obtain a weighted evaluation matrix;

[0030] Step S205: Determine the positive ideal solution and the negative ideal solution based on the weighted evaluation matrix, and calculate the distance evaluation parameters and correlation evaluation parameters of each candidate governance technology relative to the positive ideal solution and the negative ideal solution;

[0031] Step S206: The distance evaluation parameter and the correlation evaluation parameter are fused according to the preset preference coefficient to obtain the comprehensive applicability evaluation value of each candidate governance technology;

[0032] Step S207: Sort each candidate governance technology according to the comprehensive applicability evaluation value, and generate the governance technology applicability assessment result corresponding to the target ecological environment scenario parameters.

[0033] This embodiment takes the selection of carbon pollution co-treatment technology for steel enterprises as an example, and the server executes the ecological environment governance technology applicability assessment method. The server first reads a set of historical governance case data from the historical governance case database. Each historical governance case data is linked according to historical governance scenario information, governance technology information, and governance effect indicator information. The historical governance scenario information includes flue gas temperature, pollutant inlet concentration, ambient humidity, treatment scale, and investment budget. The governance technology information includes a governance technology identifier. The governance effect indicator information includes sulfur dioxide emission reduction, nitrogen oxide emission reduction, particulate matter emission reduction, carbon dioxide emission reduction, technology adoption rate, energy savings, unit investment cost, technology maturity, technology stability, and promotion value. The server also receives target ecological environment scenario parameters submitted by the target steel enterprise. These parameters record a flue gas temperature of 28 degrees Celsius, a pollutant inlet concentration of 115 mg / m³, an ambient humidity of 62%, a treatment scale of large-scale, and an investment budget of 5.2 million yuan. The server determines a set of candidate governance technologies consisting of T1 to T21 based on the technical foundation database, and cleans up missing, abnormal, and duplicate data in historical governance cases to form a one-to-one correspondence of scenario parameters, governance technology identifiers, and governance effect indicators for each historical governance case.

[0034] The server encodes historical governance scenario information, such as processing scale, and the corresponding governance technology identifiers for T1 to T21. The encoded flue gas temperature, pollutant inlet concentration, ambient humidity, treatment scale, and investment budget are used as covariates, and the encoded governance technology identifiers are used as intervention variables. Ten governance effectiveness indicators are then used as outcome variables, thus constructing ten sets of causal training samples. The server uses these ten sets of training samples to train ten causal forest regressors, enabling each regressor to learn the response relationship of specific governance effectiveness indicators to changes in scenario parameters and governance technologies. These ten regressors are then combined into a scenario-adaptive prediction model. After model training, the server encodes the target ecological environment scenario parameters using the same encoding process as the historical governance scenario information, forming target scenario covariates. While keeping the target scenario covariates unchanged, the server sequentially sets the governance technology identifiers for T1 to T21 as target intervention variables and inputs them into the ten causal forest regressors, obtaining predicted values ​​for ten counterfactual governance effectiveness indicators for each candidate governance technology under the same target scenario. The server arranges the ten predicted values ​​corresponding to each candidate governance technology into a counterfactual governance effect indicator prediction vector, and then forms a target scenario evaluation matrix of 21 rows and 10 columns according to the structure of candidate governance technologies as rows and governance effect indicators as columns, so that different candidate governance technologies can be compared under the same target scenario conditions.

[0035] The server reads the attribute values ​​of each governance effectiveness indicator in the target scenario evaluation matrix. It identifies sulfur dioxide emission reduction, nitrogen oxide emission reduction, particulate matter emission reduction, carbon dioxide emission reduction, technology adoption rate, energy saving, technology maturity, technology stability, and promotion value as benefit-type indicators, and unit investment cost as a cost-type indicator. For each column of governance effectiveness indicators, the server calculates the maximum and minimum values ​​among 21 candidate governance technologies. The server performs linear standardization on benefit-type indicators based on the difference between the corresponding indicator value and the minimum value, and performs linear standardization on cost-type indicators based on the difference between the maximum and corresponding indicator values, ensuring that larger standardized values ​​indicate better governance effectiveness. When the maximum and minimum values ​​of the same governance effectiveness indicator are equal, the server sets all standardized indicator values ​​corresponding to that indicator to 1 to avoid a denominator of zero, thus obtaining a standardized evaluation matrix.

[0036] Subsequently, the server calculates the proportion of each standardized indicator value to the sum of the standardized indicator values ​​in its corresponding column, forming a probability matrix. Zero values ​​in the probability matrix are replaced with preset small positive numbers to avoid invalid values ​​when calculating the logarithm. Based on the smoothed probability matrix, the server calculates the entropy values ​​of the ten governance effectiveness indicators. The difference between 1 and the corresponding indicator entropy value is used as the difference coefficient. The proportion of each difference coefficient to the sum of all difference coefficients is then calculated to obtain the indicator weights of the ten governance effectiveness indicators. The more significant the difference in standardized indicator values ​​between different candidate governance technologies, the larger the corresponding difference coefficient and indicator weight. The server multiplies each standardized indicator value in each column of the standardized evaluation matrix by its corresponding indicator weight to form a weighted evaluation matrix.

[0037] The server selects the maximum weighted index value from each column of the weighted evaluation matrix to form the positive ideal solution and the minimum weighted index value to form the negative ideal solution. Since cost-type indicators have been converted into unidirectional indicators where larger values ​​are better during the standardization phase, the positive ideal index value corresponding to unit investment cost is also taken as the maximum value of that column. The server calculates the first Euclidean distance between the weighted index vector of each candidate governance technology and the positive ideal solution, and the second Euclidean distance between it and the negative ideal solution. Simultaneously, using 0.5 as the resolution coefficient, the server calculates the first grey relational coefficient between each candidate governance technology and the positive ideal solution, and the second grey relational coefficient between it and the negative ideal solution under the ten governance effect indicators. The server then averages the ten first grey relational coefficients and the ten second grey relational coefficients to obtain the grey relational degree of the positive ideal solution and the grey relational degree of the negative ideal solution.

[0038] The server performs dimensionless processing on the first Euclidean distance, the second Euclidean distance, the positive ideal solution grey relational degree, and the negative ideal solution grey relational degree, and sets a preset preference coefficient to 0.5. According to the preset preference coefficient, the server performs a weighted fusion of the dimensionless second Euclidean distance and the positive ideal solution grey relational degree to obtain the positive proximity degree; and performs a weighted fusion of the dimensionless first Euclidean distance and the negative ideal solution grey relational degree to obtain the negative proximity degree. The server calculates the proportion of the positive proximity degree in the sum of the positive and negative proximity degrees, and determines the obtained proportion as the comprehensive applicability evaluation value of the corresponding candidate governance technology. The larger the comprehensive applicability evaluation value, the closer the candidate governance technology is to the ideal governance state in the target scenario.

[0039] Finally, the server sorts T1 to T21 according to their comprehensive applicability evaluation scores from highest to lowest, and outputs the technical identifier, predicted values ​​of ten counterfactual governance effect indicators, indicator weights, distance evaluation parameters, correlation evaluation parameters, comprehensive applicability evaluation scores, and ranking for each candidate governance technology, generating the governance technology applicability assessment results for the target steel enterprise. The server also saves the target scenario evaluation matrix, standardized evaluation matrix, and weighted evaluation matrix, enabling the assessment results to be further used for ranking stability analysis, counterfactual prediction confidence correction, and governance technology combination chain evaluation under different preset preference coefficients.

[0040] In this embodiment of the invention, the historical treatment scenario information includes flue gas temperature, pollutant inlet concentration, ambient humidity, treatment scale, and investment budget; the treatment technology information includes treatment technology identifiers; the treatment effect index information includes sulfur dioxide emission reduction, nitrogen oxide emission reduction, particulate matter emission reduction, carbon dioxide emission reduction, technology adoption rate, energy saving, unit investment cost, technology maturity, technology stability, and promotion value; the index attribute corresponding to the unit investment cost is a cost-type attribute, and the index attributes corresponding to the treatment effect indicators other than the unit investment cost are benefit-type attributes.

[0041] In this embodiment of the invention, exemplarily, the application scenario is the applicability assessment of carbon pollution co-treatment technology for steel enterprises. The server collects and structures historical treatment case data. The server reads historical treatment scenario information one by one from the historical project database. Here, flue gas temperature represents the inlet flue gas temperature of the treatment facility, pollutant inlet concentration represents the concentration of the target pollutant in the flue gas before treatment, ambient humidity represents the relative humidity of the operating environment of the treatment facility, treatment scale represents the amount of flue gas to be treated per unit time, and investment budget represents the construction funds that can be invested in the corresponding treatment project. The server uniformly converts the flue gas temperature to degrees Celsius, the pollutant inlet concentration to milligrams per cubic meter, the ambient humidity to a percentage, the treatment scale to a uniform flue gas treatment volume unit, and the investment budget to ten thousand yuan, thereby eliminating the differences in data units between different historical projects. A historical treatment case records a flue gas temperature of 25 degrees Celsius, a pollutant inlet concentration of 100 milligrams per cubic meter, an ambient humidity of 60%, a treatment scale of large-scale, an investment budget of 5 million yuan, and a treatment technology identifier of T1. The historical governance cases correspond to sulfur dioxide emission reductions of 0.13, nitrogen oxide emission reductions of 0.95, particulate matter emission reductions of 0.31, and carbon dioxide emission reductions of 1230.91. The technology adoption rate is 9, energy savings are 6.07, unit investment cost is 23.10, technology maturity is 4.00, technology stability is 4.00, and promotion value is 4.33. The server associates the historical governance scenario information, governance technology identifiers, and ten governance effect indicator data into a single historical governance case record and writes this record into the historical governance case dataset.

[0042] After constructing the historical governance case dataset, the server separately receives the target ecological environment scenario parameters. These parameters include a flue gas temperature of 28 degrees Celsius, a pollutant inlet concentration of 115 milligrams per cubic meter, an ambient humidity of 62%, a large-scale treatment scale, and an investment budget of 5.2 million yuan. These target ecological environment scenario parameters are used only as query conditions for the scenario adaptive prediction model to perform counterfactual predictions; they are not included as historical governance cases in the causal training samples used to train the model. This distinguishes the historical governance cases used in the model training phase from the target ecological environment scenario used in the model prediction phase.

[0043] The server reads the actual governance technology used in each historical governance case and assigns a unique governance technology identifier to that technology. The server associates source control technology, process emission reduction technology, end-of-pipe treatment technology, and waste heat recovery technology with the corresponding governance technology identifiers in T1 to T21, ensuring that the same governance technology uses the same identifier in different historical governance cases. The server thus establishes a correspondence between historical governance scenario information and governance technology identifiers, representing the actual application of a specific governance technology under specific flue gas temperature, pollutant inlet concentration, ambient humidity, treatment scale, and investment budget conditions.

[0044] The server continues to read the governance effectiveness index information from the acceptance reports and operational statistics records of each historical governance case. This governance effectiveness index information includes sulfur dioxide emission reduction, nitrogen oxide emission reduction, particulate matter emission reduction, carbon dioxide emission reduction, technology adoption rate, energy savings, unit investment cost, technology maturity, technology stability, and promotion value. Specifically, the server determines the emission reduction of each pollutant based on pollutant emission data before and after governance, determines the technology adoption rate based on the number of technology application projects, determines energy savings based on changes in energy consumption, and determines the unit investment cost based on project investment and governance capacity. It also records the evaluation data for technology maturity, technology stability, and promotion value. The server configures the unit investment cost as a cost-type attribute; a lower unit investment cost indicates better economic applicability of the candidate governance technology. The server configures the remaining nine governance effectiveness indicators as benefit-type attributes; higher values ​​for these nine indicators indicate better emission reduction, energy-saving, engineering reliability, or promotion applicability of the candidate governance technology. Finally, the server writes the historical governance scenario information, governance technology identifiers, and ten governance effectiveness index information into the same historical governance case record, providing structurally consistent training data for the subsequent construction of a scenario-adaptive prediction model based on causal forests.

[0045] In this embodiment of the invention, the construction of a scenario-adaptive prediction model based on causal forest based on the historical governance case data set and the generation of the target scenario evaluation matrix can be implemented through the following example.

[0046] The categorical historical governance scenario information and governance technology identifiers in the historical governance case dataset are encoded.

[0047] By using encoded historical governance scenario information as covariates, encoded governance technology identifiers as intervention variables, and governance effect indicator data corresponding to different governance effect indicators as outcome variables, causal training samples corresponding to different governance effect indicators are constructed.

[0048] Based on the causal training samples, causal forest regressors corresponding to different governance effect indicators are trained respectively, and multiple causal forest regressors are combined to form the scenario adaptive prediction model;

[0049] The target ecological environment scenario parameters are encoded to obtain target scenario covariates. The target scenario covariates are kept unchanged, and the governance technology identifiers corresponding to each candidate governance technology in the historical governance case data set are set as target intervention variables in turn.

[0050] The target scenario covariate and each target intervention variable are respectively input into the multiple causal forest regressors to obtain the counterfactual governance effect index prediction vector corresponding to each candidate governance technology;

[0051] Arrange multiple counterfactual governance effect indicator prediction vectors according to the correspondence between each candidate governance technology and each governance effect indicator, and generate the target scenario evaluation matrix with candidate governance technologies as rows and governance effect indicators as columns.

[0052] In this embodiment, the server reads data on flue gas temperature, pollutant inlet concentration, ambient humidity, treatment scale, investment budget, treatment technology identifiers, and ten treatment effect indicators from a historical treatment case dataset. The server retains continuous values ​​corresponding to flue gas temperature, pollutant inlet concentration, ambient humidity, and investment budget, and encodes categorical information such as treatment scale, converting small, medium, and large into corresponding category codes. Following a pre-established technology coding table, the server converts the treatment technology identifiers corresponding to T1 to T21 into mutually distinguishable intervention codes and saves the mapping relationship between the coded values ​​and the treatment technology identifiers.

[0053] The server combines the encoded flue gas temperature, pollutant inlet concentration, ambient humidity, treatment scale, and investment budget as scenario covariates, and uses the encoded treatment technology identifiers as technology intervention variables. Instead of using the numerical values ​​of different treatment technology identifiers as the technical order or distance, the server employs multi-category intervention coding to represent the different treatment technologies corresponding to T1 to T21, allowing each treatment technology to participate in causal forest training as a mutually distinguishable intervention state.

[0054] The server denotes the number of historical governance cases in the historical governance case dataset as N, the number of candidate governance technologies as K, and the number of governance effectiveness indicators as J. For the j-th governance effectiveness indicator, the server constructs the j-th set of causal training samples according to the correspondence between the scenario covariate, the technology intervention variable, and the governance effectiveness indicator data in the same historical governance case. In the formula, This represents the causal training sample corresponding to the j-th governance effectiveness indicator; This represents the scenario covariate corresponding to the r-th historical governance case; This represents the technical intervention variable corresponding to the r-th historical governance case; This represents the governance effectiveness index data of the r-th historical governance case under the j-th governance effectiveness index; N represents the number of historical governance cases.

[0055] The server trains a multi-valued intervention causal forest regressor for each governance effectiveness indicator. The server randomly selects training samples for each causal tree and divides these samples into a node splitting sample subset and a leaf node estimation sample subset. The node splitting sample subset is used to determine node splitting conditions based on the differences in conditional treatment effects between different governance techniques. The leaf node estimation sample subset is used to calculate the potential governance effects of each governance technique within the leaf node after the split, thus avoiding the use of the same historical governance case to simultaneously determine the node structure and estimate the governance effects of leaf nodes.

[0056] The server estimates the conditional probability of the k-th candidate governance technology being adopted under scenario covariates based on historical governance scenario information. In the formula, Indicates the scene covariate The conditional probability of adopting the k-th candidate governance technology; the value of k ranges from 1 to K.

[0057] The server uses the conditional probability to perform sample balancing on historical governance cases corresponding to different governance technologies. The sample weight of the r-th historical governance case relative to the k-th candidate governance technology is determined according to the following formula. In the formula, This represents the sample weight of the r-th historical governance case relative to the k-th candidate governance technology; This indicates an indicator function that takes a value of 1 when the r-th historical governance case adopts the k-th candidate governance technology, and a value of 0 otherwise. This represents a preset small positive number to prevent the denominator from being zero.

[0058] Target scenario covariates After falling into the target leaf node of the b-th causal tree, the server calculates the tree-level potential governance effect prediction value corresponding to the j-th governance effect index based on the historical governance cases that match the k-th candidate governance technology within the target leaf node. In the formula, This represents the predicted potential governance effect of the k-th candidate governance technology output by the b-th causal tree under the j-th governance effect index. This represents the target leaf node corresponding to the target scenario covariate in the b-th causal tree.

[0059] The server averages the tree-level potential governance effect predictions output by B causal trees to obtain the counterfactual governance effect index predictions output by the causal forest regressor. In the formula, B represents the number of causal trees contained in the causal forest regressor; This represents the predicted value of the j-th counterfactual governance effect index corresponding to the k-th candidate governance technology under the target scenario covariates.

[0060] The conditional treatment effect between any two candidate governance technologies is determined by the following formula: In the formula, τ This represents the conditional treatment effect of the k-th candidate governance technique relative to the l-th candidate governance technique on the j-th governance effectiveness index, under the same target scenario covariates. The server uses this conditional treatment effect to determine the difference in governance effectiveness when changing the governance technique intervention while keeping the target scenario covariates constant.

[0061] The server arranges the J counterfactual governance effect index prediction values ​​corresponding to the same candidate governance technology in order of governance effect index, obtains the counterfactual governance effect index prediction vector of the corresponding candidate governance technology, and generates the target scenario evaluation matrix with the candidate governance technology as the row and the governance effect index as the column.

[0062] In the formula, K represents the target scenario evaluation matrix; J represents the number of candidate governance technologies; and J represents the number of governance effectiveness indicators.

[0063] Please refer to the following: Figure 2 , Figure 2 This is a schematic diagram of the structure of the target scenario evaluation matrix generated based on causal forest provided in an embodiment of the present invention. To ensure that the counterfactual governance effects under different governance technologies are supported by historical data, the server performs a sample coverage check on the historical governance case dataset. In the historical governance case dataset, the same governance technology corresponds to multiple historical governance cases under different ecological and environmental scenarios, and the same or similar ecological and environmental scenarios correspond to multiple historical governance cases with different governance technologies. When the server does not find a historical governance case matching the target intervention variable in the target leaf node of the causal tree, it does not directly output the corresponding tree-level prediction value, and the counterfactual prediction confidence coefficient of the corresponding candidate governance technology is reduced by the subsequent counterfactual prediction confidence correction step.

[0064] The server receives target ecological environment scenario parameters: flue gas temperature of 28 degrees Celsius, pollutant inlet concentration of 115 mg / m³, ambient humidity of 62%, treatment scale of large-scale, and investment budget of 5.2 million yuan. It processes these parameters according to the coding rules used for historical treatment scenario information to obtain target scenario covariates. Throughout the prediction process, the server keeps these target scenario covariates constant. It first sets the intervention code corresponding to T1 as the target intervention variable, then sequentially replaces it with the intervention codes corresponding to T2 through T21, thus forming 21 sets of model input data that differ only in the treatment technology.

[0065] The server inputs the data from each model into ten causal forest regressors to obtain counterfactual predictions of the corresponding candidate governance technologies for sulfur dioxide emission reduction, nitrogen oxide emission reduction, particulate matter emission reduction, carbon dioxide emission reduction, technology adoption rate, energy saving, unit investment cost, technology maturity, technology stability, and promotion value under the same target scenario. These ten counterfactual predictions are then arranged into a counterfactual governance effect indicator prediction vector for the corresponding candidate governance technologies. The server arranges 21 prediction vectors in the order of governance technologies T1 to T21, treating each candidate governance technology as a row and the ten governance effect indicators as ten columns, generating a 21-row, 10-column target scenario evaluation matrix. This target scenario evaluation matrix, composed of counterfactual prediction results under the same target scenario, can be directly used for subsequent standardization, indicator weighting, and comprehensive applicability evaluation.

[0066] In one data fault-tolerant implementation, the server simultaneously stores a static evaluation matrix and a target scenario evaluation matrix generated by a scenario-adaptive prediction model. The static evaluation matrix, with candidate governance technologies as rows and governance effectiveness indicators as columns, records the baseline governance effectiveness indicator data corresponding to each candidate governance technology. The target scenario evaluation matrix records the predicted counterfactual governance effectiveness indicators for each candidate governance technology under the target ecological environment scenario parameters.

[0067] The server monitors the model status of the scenario-adaptive prediction model, the validity of the historical governance case dataset, and the generation status of the target scenario evaluation matrix. When the historical governance case dataset meets the model training requirements, the scenario-adaptive prediction model has completed training, and the target scenario evaluation matrix is ​​successfully generated, the server determines the target scenario evaluation matrix as the input evaluation matrix for subsequent standardization processing. When the historical governance case dataset is missing, the number of historical governance cases is lower than the preset training quantity, the scenario-adaptive prediction model training fails, or the user selects to perform static evaluation, the server determines the static evaluation matrix as the input evaluation matrix for subsequent standardization processing. In the formula, This represents the input evaluation matrix used in subsequent standardization processing; V represents the target scene evaluation matrix generated by the scene adaptive prediction model; V represents the static evaluation matrix; q represents the matrix selection identifier. When the scene adaptive prediction model is effective and the target scene evaluation matrix is ​​successfully generated, q is 1; when the static evaluation fallback mode is used, q is 0.

[0068] The server records the data source identifier of the input evaluation matrix in the assessment results of the applicability of governance technologies. When the input evaluation matrix is ​​the target scenario evaluation matrix, the server marks the evaluation result as a counterfactual prediction evaluation result; when the input evaluation matrix is ​​a static evaluation matrix, the server marks the evaluation result as a static benchmark evaluation result, thereby avoiding mistaking the evaluation result formed by the static benchmark value as a counterfactual prediction result under the target ecological environment scenario.

[0069] In this embodiment of the invention, the standardization of the target scenario evaluation matrix and the determination of the indicator weights of each governance effect indicator can be implemented through the following example.

[0070] Obtain the maximum and minimum values ​​of each governance effectiveness indicator in the candidate governance technology set;

[0071] For governance effectiveness indicators with benefit-type attributes, linear standardization is performed based on the difference between the corresponding governance effectiveness indicator data and the minimum value of the indicator. For governance effectiveness indicators with cost-type attributes, linear standardization is performed based on the difference between the maximum value of the indicator and the corresponding governance effectiveness indicator data.

[0072] When the maximum value of any governance effectiveness indicator is equal to the minimum value of the indicator, the standardized indicator value corresponding to any governance effectiveness indicator is set to 1;

[0073] A probability matrix is ​​constructed based on the proportion of multiple standardized indicator values ​​corresponding to the same governance effectiveness indicator in the sum of the multiple standardized indicator values, and the zero values ​​in the probability matrix are smoothed.

[0074] The entropy value of each governance effectiveness indicator is calculated based on the smoothed probability matrix. The difference coefficient of each governance effectiveness indicator is determined based on the entropy value. The weight of each indicator is determined based on the proportion of each difference coefficient in the sum of all difference coefficients.

[0075] Based on the correspondence between the various governance effectiveness indicators, the standardized indicator values ​​in the standardized evaluation matrix are multiplied by their corresponding indicator weights to obtain the weighted evaluation matrix.

[0076] In this embodiment, the server reads the 21-row, 10-column target scenario evaluation matrix and iterates through the counterfactual predicted values ​​corresponding to the 21 candidate governance technologies column by column according to the governance effectiveness indicators. The server determines the maximum and minimum values ​​of each governance effectiveness indicator. In the column for sulfur dioxide emission reduction, the server determines the maximum value to be 0.56 and the minimum value to be 0.01; in the column for unit investment cost, the server determines the maximum value to be 66.26 and the minimum value to be 0.91, and establishes a correlation between the maximum and minimum values ​​of each column and the corresponding governance effectiveness indicators.

[0077] The server performs forward linear standardization on nine benefit-related indicators, including sulfur dioxide emission reduction. This involves subtracting the minimum value from the corresponding indicator data and then dividing by the difference between the maximum and minimum values. When the sulfur dioxide emission reduction for T1 is 0.13, the server calculates a standardized indicator value of 0.2182. The server also performs backward linear standardization on unit investment cost. This involves subtracting the corresponding unit investment cost from the maximum value and then dividing by the difference between the maximum and minimum values. When the unit investment cost for T1 is 23.10, the server calculates a standardized indicator value of 0.6604. After these processes, all ten governance effectiveness indicators are converted to unidirectional indicators between 0 and 1, with higher values ​​indicating better governance effectiveness. The server checks the maximum and minimum values ​​of each governance effectiveness indicator. When the predicted values ​​of governance effectiveness indicators such as technical stability are exactly the same across the 21 candidate governance technologies, the server sets all 21 standardized indicator values ​​in that column to 1 to avoid division by zero errors during linear standardization.

[0078] The server accumulates standardized index values ​​column by column according to the governance effectiveness indicators, and calculates the proportion of each standardized index value in the sum of the standardized index values ​​in the corresponding column, thus forming a 21-row, 10-column probability matrix. For a probability zero value formed by a standardized index value of 0, the server replaces the probability zero value with a preset small positive number to ensure that subsequent logarithmic operations can be performed normally. The server calculates the index entropy value of each governance effectiveness indicator based on the smoothed probability matrix, and subtracts the corresponding index entropy value from 1 to obtain the difference coefficient. Governance effectiveness indicators with more significant differences between candidate governance technologies have smaller index entropy values ​​and larger difference coefficients; governance effectiveness indicators with the same values ​​for all candidate governance technologies have an index entropy value of 1 and a difference coefficient of 0.

[0079] The server calculates the proportion of each difference coefficient in the sum of the ten difference coefficients, and determines the resulting proportion as the indicator weight of the corresponding governance effect indicator, so that the sum of all indicator weights is 1. Finally, according to the column correspondence, the server multiplies the values ​​of the ten standardized indicators, such as sulfur dioxide emission reduction, in the standardized evaluation matrix by their corresponding indicator weights to obtain the weighted indicator values ​​of each candidate governance technology under each governance effect indicator, and arranges the weighted indicator values ​​according to the original 21-row, 10-column structure to generate a weighted evaluation matrix.

[0080] In this embodiment of the invention, the step of determining the positive ideal solution and the negative ideal solution based on the weighted evaluation matrix, and calculating the distance evaluation parameter and the correlation evaluation parameter, can be implemented through the following example.

[0081] The weighted index values ​​of each candidate governance technology under each governance effect index are obtained from the weighted evaluation matrix. The maximum value among multiple weighted index values ​​corresponding to the same governance effect index is determined as the positive ideal index value, and the minimum value among multiple weighted index values ​​corresponding to the same governance effect index is determined as the negative ideal index value, thus obtaining the positive ideal solution and the negative ideal solution.

[0082] Calculate the first Euclidean distance between the weighted index value corresponding to each candidate governance technology and the positive ideal solution, and the second Euclidean distance between the weighted index value and the negative ideal solution, and use the first Euclidean distance and the second Euclidean distance as the distance evaluation parameters;

[0083] Based on the preset resolution coefficient, calculate the first grey relational coefficient between each candidate governance technology and the positive ideal solution and the second grey relational coefficient between each candidate governance technology and the negative ideal solution under each governance effect index;

[0084] The average values ​​of multiple first grey relation coefficients and multiple second grey relation coefficients corresponding to the same candidate governance technology are respectively processed to obtain the positive ideal solution grey relation degree and the negative ideal solution grey relation degree, and the positive ideal solution grey relation degree and the negative ideal solution grey relation degree are used as the relation evaluation parameters.

[0085] In this embodiment, the server reads a 21-row, 10-column weighted evaluation matrix and obtains the weighted index values ​​corresponding to T1 to T21 column by column according to the governance effect indicators. For sulfur dioxide emission reduction, the server determines the maximum value among the 21 weighted index values ​​as the positive ideal index value for sulfur dioxide emission reduction and the minimum value as the negative ideal index value for sulfur dioxide emission reduction. The server processes nitrogen oxide emission reduction, particulate matter emission reduction, carbon dioxide emission reduction, technology adoption rate, energy saving, unit investment cost, technology maturity, technology stability, and promotion value in the same manner. Since unit investment cost has been converted into a positive indicator that the larger the value, the better the cost performance during the standardization stage, the server also uses the maximum weighted index value corresponding to unit investment cost as the positive ideal index value. The server arranges the ten positive ideal index values ​​according to the column order of the ten governance effect indicators to form a positive ideal solution, and arranges the ten negative ideal index values ​​according to the same column order to form a negative ideal solution.

[0086] The server reads the ten weighted index values ​​corresponding to each candidate governance technology line by line, calculates the difference between each of the ten weighted index values ​​and the ten positive ideal index values, squares each difference, sums them, and then takes the square root of the sum to obtain the first Euclidean distance between the corresponding candidate governance technology and the positive ideal solution. The server calculates the second Euclidean distance between the corresponding candidate governance technology and the negative ideal solution in the same way. The first Euclidean distance of T9 is 0.083 and the second Euclidean distance is 0.214, indicating that the overall index state of T9 is closer to the positive ideal solution. The server records the first and second Euclidean distances corresponding to each candidate governance technology together as distance evaluation parameters.

[0087] The server further calculates the first absolute difference between each weighted index value and its corresponding positive ideal index value, and the second absolute difference between each weighted index value and its corresponding negative ideal index value in the weighted evaluation matrix. It then calculates the maximum and minimum differences among all first and second absolute differences. The server sets a preset resolution coefficient to 0.5 and calculates the first and second grey relational coefficients for each candidate governance technology under the ten governance effectiveness indicators based on the minimum and maximum differences, the preset resolution coefficient, and the current absolute difference. When a weighted index value equals its corresponding ideal index value, the server sets the corresponding grey relational coefficient to 1.

[0088] Finally, the server calculates the arithmetic mean of the ten first grey relational coefficients corresponding to each candidate governance technology from T1 to T21 to obtain the grey relational degree of the positive ideal solution, and calculates the arithmetic mean of the ten second grey relational coefficients to obtain the grey relational degree of the negative ideal solution. The grey relational degree of the positive ideal solution for T9 is 0.82, and the grey relational degree of the negative ideal solution is 0.47, indicating that the change patterns of the various governance effect indicators of T9 are highly consistent with the positive ideal solution. The server uses the grey relational degrees of the positive and negative ideal solutions as correlation evaluation parameters, which, together with the distance evaluation parameters, are used for subsequent comprehensive applicability evaluation.

[0089] In this embodiment of the invention, the fusion processing of the distance evaluation parameter and the correlation evaluation parameter according to the preset preference coefficient can be performed through the following example.

[0090] The first Euclidean distance, the second Euclidean distance, the positive ideal solution grey relational degree, and the negative ideal solution grey relational degree are respectively subjected to dimensionless processing;

[0091] Set the value that is greater than or equal to 0 and less than or equal to 1 as the preset preference coefficient;

[0092] Based on the preset preference coefficient and the difference between 1 and the preset preference coefficient, the dimensionless second Euclidean distance and the dimensionless positive ideal solution grey relational degree are weighted and fused to obtain the positive proximity degree, and the dimensionless first Euclidean distance and the dimensionless negative ideal solution grey relational degree are weighted and fused to obtain the negative proximity degree.

[0093] The comprehensive applicability evaluation value is determined based on the proportion of the positive proximity degree in the sum of the positive proximity degree and the negative proximity degree.

[0094] The comprehensive applicability evaluation value of each candidate governance technology is calculated using multiple different preset preference coefficients. The ranking change results of each candidate governance technology under different preset preference coefficients are determined based on the multiple comprehensive applicability evaluation values. A comparison chart of comprehensive applicability evaluation values ​​and a heat map of the ranking change of candidate governance technologies are generated based on the ranking change results.

[0095] In this embodiment, the server obtains the first Euclidean distance, the second Euclidean distance, the positive ideal solution grey relational degree, and the negative ideal solution grey relational degree corresponding to T1 to T21, respectively, and performs dimensionless processing on the four sets of evaluation data. The server obtains the maximum value in each set of evaluation data, and then divides each evaluation data in the same set by the corresponding maximum value, so that all four sets of evaluation data are converted to a numerical range of 0 to 1. When the first Euclidean distance of T9 is 0.083, the second Euclidean distance is 0.214, the positive ideal solution grey relational degree is 0.82, and the negative ideal solution grey relational degree is 0.47, and the maximum values ​​of the four sets of evaluation data are 0.30, 0.25, 0.90, and 0.88 respectively, the server obtains the dimensionless values ​​corresponding to T9 as 0.2767, 0.8560, 0.9111, and 0.5341 respectively.

[0096] The server reads the preset preference coefficient from the evaluation parameter configuration table and verifies that the preset preference coefficient is within the closed interval of 0 to 1. The preset preference coefficient used in this evaluation is 0.5, and the difference between 1 and the preset preference coefficient is also 0.5. The server multiplies the dimensionless second Euclidean distance by 0.5, multiplies the dimensionless gray relational degree of the positive ideal solution by 0.5, and then sums the two products to obtain the positive proximity of T9, 0.8836. The server multiplies the dimensionless first Euclidean distance by 0.5, multiplies the dimensionless gray relational degree of the negative ideal solution by 0.5, and then sums the two products to obtain the negative proximity of T9, 0.4054. A larger positive proximity indicates that the candidate governance technology is closer to the positive ideal solution; a larger negative proximity indicates that the candidate governance technology is closer to the negative ideal solution.

[0097] The server divides the positive proximity score of T9 by the sum of the positive and negative proximity scores to obtain a comprehensive applicability evaluation value of 0.6855 for T9. The server calculates the comprehensive applicability evaluation values ​​of T1 to T21 in the same way, and ranks the candidate governance technologies according to their comprehensive applicability evaluation values ​​from largest to smallest.

[0098] To analyze the sensitivity of the fusion results to the preset preference coefficient, the server repeatedly performed the fusion calculation with the preset preference coefficient set to 0, 0.5, and 1 respectively. When the preset preference coefficient is 0, the evaluation result fully reflects the grey relational characteristics; when the preset preference coefficient is 1, the evaluation result fully reflects the Euclidean distance characteristics; and when the preset preference coefficient is 0.5, the evaluation result considers both distance and relational characteristics. The server records the comprehensive applicability evaluation values ​​and rankings of T1 to T21 under the three preset preference coefficients. A comprehensive applicability evaluation value comparison chart is generated with the candidate governance technology identifier as the x-axis and the comprehensive applicability evaluation value as the y-axis. A heatmap of candidate governance technology ranking changes is generated with the candidate governance technology identifier as the row, the preset preference coefficient as the column, and the ranking ranking as the cell data, thus visually displaying the ranking stability of each candidate governance technology.

[0099] Please refer to the following: Figure 3 , Figure 3 This is a schematic diagram of the Euclidean distance and grey relational analysis fusion evaluation structure provided in an embodiment of the present invention. In this embodiment, the method further includes:

[0100] According to the preset technology types, each candidate governance technology in the candidate governance technology set is assigned a technology type label. The preset technology types include source control type, process emission reduction type, end-of-pipe treatment type and waste heat recovery type arranged in the order of governance.

[0101] From different preset technology types, select no more than one candidate governance technology, and arrange the selected candidate governance technologies according to the governance order to generate a candidate technology combination chain containing at least two and no more than four candidate governance technologies.

[0102] The governance effect index data corresponding to the candidate governance technologies contained in each candidate technology combination chain are obtained from the target scenario evaluation matrix. The governance effect index data is then aggregated according to the preset index aggregation rules corresponding to each governance effect index to obtain combined governance effect index data. The preset index aggregation rules include summation rules, weighted average rules, or nonlinear combination rules executed according to the governance order.

[0103] Construct a combined chain evaluation matrix based on the combined governance effect index data corresponding to each candidate technology combined chain;

[0104] The combined chain evaluation matrix is ​​subjected to standardization, index weight determination, weighting, positive and negative ideal solution determination, distance evaluation parameter calculation, correlation evaluation parameter calculation and fusion processing in sequence to obtain the combined chain comprehensive applicability evaluation value of each candidate technology combined chain;

[0105] Based on the comprehensive applicability evaluation value of the combined chain, each candidate technology combined chain is ranked, and the applicability evaluation result of the governance technology combined chain corresponding to the target ecological environment scenario parameters is generated.

[0106] In this embodiment, the server reads the governance technology identifiers and corresponding technical function configuration data of T1 to T21, and configures the technology type labels according to the preset technology types. The server marks T1 to T4 as source control type, T5 to T19 as process emission reduction type, T20 as end-of-pipe treatment type, and T21 as waste heat recovery type, while saving the governance order of "source control, process emission reduction, end-of-pipe treatment, and waste heat recovery". The governance order is consistent with the actual governance process of steel enterprises from pollution generation control, production process emission reduction, end-of-pipe emission treatment to waste heat recovery.

[0107] The server groups technologies T1 to T21 according to their technology types and selects no more than one candidate governance technology from each technology type. The server then arranges the selected candidate governance technologies according to a preset governance order, generating candidate technology combination chains. These chains include formats such as "T2→T9", "T3→T12→T20", and "T2→T9→T20→T21". Each chain contains two to four candidate governance technologies, and the same technology type appears only once in the same chain. The server deletes combinations with fewer than two technologies, more than four technologies, duplicate technologies of the same type, or combinations that do not conform to the governance order, resulting in a set of candidate technology combination chains that satisfy the combination constraints.

[0108] For the candidate technology combination chain “T2→T9→T20→T21”, the server reads the ten counterfactual governance effect indicators corresponding to T2, T9, T20 and T21 from the target scenario evaluation matrix. The server processes the data according to the preset index aggregation rules corresponding to each governance effect indicator: for emission reductions and energy savings that can be directly accumulated, an additive rule is used; for unit investment cost, technology maturity, technology stability and promotion value, a weighted average rule corresponding to the configuration weight is used; for governance effect indicators that have a relationship of influence between previous and subsequent governance, a non-linear combination rule is executed according to the governance order of “T2, T9, T20, T21”, thereby avoiding the inclusion of pollutants already reduced in the previous governance stage in the subsequent governance stage. The server arranges the aggregated sulfur dioxide emission reduction, nitrogen oxide emission reduction, particulate matter emission reduction, carbon dioxide emission reduction, technology adoption rate, energy savings, unit investment cost, technology maturity, technology stability and promotion value as the combined governance effect indicator data of the candidate technology combination chain.

[0109] The server calculates the combined governance effect index data of all candidate technology combination chains using the same method, and constructs a combination chain evaluation matrix with candidate technology combination chains as rows and ten governance effect indicators as columns. The server performs standardization processing on the combination chain evaluation matrix using benefit-type and cost-type indicators, calculates indicator weights based on the distribution of standardized indicator data, and forms a weighted combination chain evaluation matrix. Subsequently, it determines the positive and negative ideal solutions, calculates the first Euclidean distance, second Euclidean distance, gray relational degree of the positive ideal solution, and gray relational degree of the negative ideal solution for each candidate technology combination chain, and fuses them using a preset preference coefficient to obtain the comprehensive applicability evaluation value of the combination chain. Finally, the server ranks each candidate technology combination chain from largest to smallest according to the comprehensive applicability evaluation value, outputs the combination chain identifier, technology ranking order, combined governance effect index data, comprehensive applicability evaluation value, and ranking, generating the governance technology combination chain applicability assessment result corresponding to the current target ecological environment scenario.

[0110] In this embodiment of the invention, the method further includes:

[0111] Obtain the predicted values ​​of governance effect indicators output by multiple causal trees in each causal forest regressor for the target scenario covariate and each target intervention variable, and construct the inter-tree prediction value sequence corresponding to each candidate governance technology;

[0112] Based on the degree of dispersion among the predicted values ​​of multiple governance effectiveness indicators in the same inter-tree prediction value sequence, the prediction dispersion parameters of each candidate governance technology under each governance effectiveness indicator are determined.

[0113] Obtain the target leaf nodes corresponding to the target scenario covariates in each causal tree, and count the number of historical governance cases that match the corresponding target intervention variables within the target leaf nodes. Determine the sample support parameters of each candidate governance technology under each governance effect indicator based on the number of historical governance cases.

[0114] The counterfactual prediction confidence coefficients for each candidate governance technology are determined based on the predicted discrete parameters and the sample support parameters. The counterfactual prediction confidence coefficients decrease as the predicted discrete parameters increase and increase as the sample support parameters increase.

[0115] The comprehensive applicability evaluation value corresponding to each candidate governance technology is multiplied by the corresponding counterfactual prediction confidence coefficient to obtain the confidence-corrected applicability evaluation value, and the governance technology applicability assessment result is updated based on the confidence-corrected applicability evaluation value.

[0116] In this embodiment of the invention, for example, the scenario-adaptive prediction model includes causal forest regressors corresponding to ten governance effect indicators, each containing 100 causal trees. The server inputs the target scenario covariates (flue gas temperature 28 degrees Celsius, pollutant inlet concentration 115 mg / m³, ambient humidity 62%, large-scale treatment, and investment budget of 5.2 million yuan) and the target intervention variable corresponding to T1 into each causal tree. It then reads the predicted governance effect indicators output by each of the 100 causal trees and sequentially replaces the target intervention variables with the intervention codes corresponding to T2 to T21. The server arranges the causal tree outputs according to the candidate governance technologies and governance effect indicators, forming an inter-tree prediction value sequence containing 100 predicted values ​​for each candidate governance technology under each governance effect indicator.

[0117] The server calculates the mean and standard deviation of each tree-based predicted value sequence, and determines the prediction dispersion parameter based on the ratio of the standard deviation to the mean. The inter-tree predicted values ​​for sulfur dioxide emission reduction corresponding to T9 are concentrated between 0.44 and 0.50, with a mean of 0.47 and a standard deviation of 0.018, indicating a small prediction dispersion parameter. The inter-tree predicted values ​​for sulfur dioxide emission reduction corresponding to T18 have a wider distribution range and a standard deviation of 0.095, indicating a larger prediction dispersion parameter. A larger prediction dispersion parameter indicates lower consistency in predictions of the same candidate treatment technology among different causal trees.

[0118] The server further determines the target leaf nodes in each causal tree where the target scenario covariates fall, and retrieves historical governance cases stored within those target leaf nodes. The server matches the target intervention variables, counts the number of historical governance cases using the same candidate governance technology within each target leaf node, and then summarizes the number of historical governance cases corresponding to the 100 causal trees to obtain sample support parameters. T9 has a large number of historical governance cases with similar scenarios and consistent technology identifiers within multiple target leaf nodes, with an average matching number of 24; T18 has an average matching number of 4, therefore T9 obtains a higher level of sample support than T18.

[0119] The server records the predicted value of the governance effect index of the i-th candidate governance technology under the j-th governance effect index, output by the b-th causal tree. The server first calculates the mean of the predicted values ​​for the same tree sequence. In the formula, B represents the inter-tree prediction mean of the i-th candidate governance technology under the j-th governance effectiveness index; B represents the number of causal trees.

[0120] The server calculates the inter-tree prediction standard deviation based on the degree of fluctuation of the inter-tree predictions relative to the prediction mean. The server determines the ratio between the standard deviation of the inter-tree predictions and the absolute value of the mean of the inter-tree predictions as the prediction discrete parameter. In the formula, Indicates the prediction of discrete parameters; This represents a small positive number to prevent division by zero errors when the predicted mean is zero. The larger the prediction discrete parameter, the lower the consistency between the predicted governance effectiveness indicators output by different causal trees.

[0121] The server records the number of historical governance cases in the target leaf node corresponding to the i-th candidate governance technology that match the target scenario covariate in the b-th causal tree. The average number of matching historical governance cases within multiple target leaf nodes is calculated to obtain the sample support parameters. In the formula, This represents the sample support parameter for the i-th candidate governance technology under the j-th governance effectiveness indicator. The larger the sample support parameter, the more historical governance cases using the same governance technology exist near the target ecological environment scenario.

[0122] The server converts the predicted discrete parameters into discrete confidence components. In the formula, This represents the discrete confidence component. As the prediction discrete parameter increases, the discrete confidence component decreases monotonically.

[0123] The server converts the sample support parameters into sample support confidence components. In the formula, This indicates that the sample supports the confidence component; This indicates that the preset samples support the scale parameter, and As the sample support parameter increases, the sample support confidence component increases monotonically and gradually approaches 1.

[0124] The server performs weighted fusion of discrete confidence components and sample support confidence components based on pre-set confidence preference parameters to obtain the index confidence value. In the formula, This represents the confidence value of the i-th candidate governance technology under the j-th governance effectiveness indicator; This indicates the preset credit preference parameters, and This embodiment will Setting it to 0.5 gives the same fusion weight to the degree of dispersion in prediction and the degree of support from historical samples.

[0125] The server performs a weighted summation of the confidence values ​​of multiple indicators corresponding to the same candidate governance technology according to the indicator weights of each governance effectiveness indicator, and obtains the counterfactual prediction confidence coefficient. In the formula, This represents the counterfactual prediction confidence coefficient corresponding to the i-th candidate governance technology; represents the weight of the j-th governance effectiveness indicator; J represents the number of governance effectiveness indicators.

[0126] Please refer to the following: Figure 4 , Figure 4 This is a schematic diagram of the counterfactual prediction confidence correction structure provided in an embodiment of the present invention. The server multiplies the comprehensive applicability evaluation value with the counterfactual prediction confidence coefficient to obtain the confidence correction applicability evaluation value. In the formula, This represents the comprehensive applicability evaluation value of the i-th candidate governance technology; This represents the confidence correction applicability evaluation value of the i-th candidate governance technology. The server rearranges the candidate governance technologies in descending order of their confidence correction applicability evaluation values ​​and updates the governance technology applicability assessment results.

[0127] Finally, the server multiplies the overall applicability evaluation value of each candidate governance technology by its corresponding counterfactual prediction confidence coefficient. The overall applicability evaluation value of T9 is 0.6855, which, multiplied by 0.91, yields a confidence-corrected applicability evaluation value of 0.6238. The server then rearranges T1 to T21 according to their confidence-corrected applicability evaluation values ​​from largest to smallest, and simultaneously records the original overall applicability evaluation value, counterfactual prediction confidence coefficient, confidence-corrected applicability evaluation value, and updated ranking in the evaluation results. This reduces the ranking priority of candidate governance technologies with insufficient historical samples or large fluctuations in inter-tree predictions in the final evaluation results.

[0128] In this embodiment of the invention, the method further includes:

[0129] Obtain a historical governance technology combination case data set, which includes combined governance scenario information, multiple governance technology identifiers arranged in the actual governance order, and combined governance effect index data;

[0130] Using the combined governance scenario information as a combined covariate, the adjacent governance technology identifiers arranged in the actual governance order as a combined intervention variable, and the combined governance effect index data as a combined result variable, a governance technology interaction training sample is constructed, and a technology interaction causal forest model is trained based on the governance technology interaction training sample.

[0131] For two adjacent candidate governance technologies in each candidate technology combination chain according to the governance order, the combination covariates corresponding to the target ecological environment scenario parameters are kept unchanged. The governance technology identifiers corresponding to the two adjacent candidate governance technologies are set as target combination intervention variables. The joint counterfactual governance effect index prediction vectors corresponding to the adjacent governance technologies are generated through the technology interaction causal forest model.

[0132] According to the preset index aggregation rules, the counterfactual governance effect index prediction vectors corresponding to two adjacent candidate governance technologies are aggregated to obtain an initial binary aggregation vector. The difference between the joint counterfactual governance effect index prediction vector and the initial binary aggregation vector is calculated to obtain the interaction correction vector corresponding to the adjacent governance technologies.

[0133] According to the governance order, the counterfactual governance effect index prediction vectors corresponding to the candidate governance technologies contained in each candidate technology combination chain and the interaction correction vectors corresponding to adjacent governance technologies are aggregated to obtain the combined governance effect index data after interaction correction.

[0134] The combined governance effect index data after the interaction correction is updated to update the combined chain evaluation matrix, and the applicability assessment result of the combined governance technology chain is updated according to the updated combined chain evaluation matrix.

[0135] In this embodiment, the server retrieves a collection of historical combined treatment technology case studies from a historical combined treatment project database. Each historical combined treatment technology case study includes combined treatment scenario information consisting of flue gas temperature, pollutant inlet concentration, ambient humidity, treatment scale, and investment budget; treatment technology identifiers arranged in the actual treatment sequence; and ten combined treatment effect indicators obtained after combined implementation. The treatment technology sequence recorded in a historical case study is "T2→T9→T20→T21", indicating that the corresponding project implements source control, process emission reduction, end-of-pipe treatment, and waste heat recovery in sequence.

[0136] The server performs a validity check on the historical governance technology combination case dataset. For historical governance technology combination cases containing only two governance technologies, the server directly determines the combined governance effect index data obtained after the joint implementation of the two governance technologies as the combined result variable of the corresponding adjacent governance technology pair. For historical governance technology combination cases containing three or more governance technologies, the server reads the stage combined governance effect index data formed within the governance stages corresponding to two adjacent governance technologies, and determines the stage combined governance effect index data as the combined result variable of the corresponding adjacent governance technology pair.

[0137] The server does not directly copy the overall governance effect index data of the same multi-technology combination chain into the combination result variable of each adjacent governance technology pair in the multi-technology combination chain. Multi-technology combination cases that do not record the stage-specific governance effect index data do not participate in the construction of adjacent governance technology interaction training samples, but continue to be used for the overall governance effect verification of the combination chain, thereby avoiding the incorrect attribution of the governance effects generated by other governance technologies in the combination chain to the current adjacent governance technology pair.

[0138] The server encodes two governance technology identifiers arranged in the actual governance order into an ordered technology pair, so that the governance order between the first and second governance technologies participates in the determination of the combined intervention variable. The combined intervention variable formed when governance technology identifiers T2 and T9 are arranged in the order of "T2→T9" is not the same as the combined intervention variable formed in the order of "T9→T2". In the formula, This represents the combined intervention variable corresponding to the k-th adjacent governance technology pair in the r-th historical governance technology combination case; and These represent two adjacent governance technology identifiers arranged in the actual governance order.

[0139] The interaction training samples for the governance technologies corresponding to the j-th governance effectiveness indicator are constructed according to the following formula: In the formula, This represents the interactive training sample of the governance technology corresponding to the j-th governance effectiveness indicator; This represents the combined covariate corresponding to the r-th historical governance technology combination case; This represents the combined intervention variables corresponding to ordered adjacency governance techniques; This represents the data of the combined governance effect index of the j-th stage corresponding to the joint implementation of the ordered adjacent governance technology.

[0140] The server uses the same causal forest training method as the scenario-adaptive prediction model, trains technology interaction causal forest regressors for different governance effect indicators, and combines multiple technology interaction causal forest regressors into a technology interaction causal forest model.

[0141] The server encodes the combined governance scenario information and governance technology identifiers, using the encoded combined governance scenario information as a combined covariate. It then extracts the identifiers of adjacent governance technologies according to the actual governance sequence, forming three combined intervention variables: "T2 and T9", "T9 and T20", and "T20 and T21". The server uses sulfur dioxide emission reduction, nitrogen oxide emission reduction, particulate matter emission reduction, carbon dioxide emission reduction, technology adoption rate, energy saving, unit investment cost, technology maturity, technology stability, and promotion value as combined outcome variables to construct a governance technology interaction training sample containing combined covariates, combined intervention variables, and combined outcome variables. This training sample is then used to train a technology interaction causal forest model, enabling the model to learn the synergistic enhancement or mutual inhibition relationships that arise when adjacent governance technologies are implemented together.

[0142] The server arranges the candidate governance technology a and the candidate governance technology b into ordered adjacent governance technology pairs according to the governance order, and predicts the joint counterfactual governance effect index prediction vector when the ordered adjacent governance technology pairs are jointly implemented under the target ecological environment scenario parameters through a technology interaction causal forest model. The server performs initial binary aggregation of the counterfactual governance effect index prediction values ​​corresponding to the candidate governance technology a and the candidate governance technology b, respectively, according to the preset index aggregation rules corresponding to the j-th governance effect index. In the formula, This represents the preset index aggregation function corresponding to the j-th governance effectiveness index; and denoted as the counterfactual governance effect index prediction values ​​of candidate governance technologies a and b under the governance effect index j, respectively.

[0143] The server calculates the difference between the predicted value of the joint counterfactual governance effectiveness index and the initial binary aggregate value, and obtains the interactive correction value of the corresponding governance effectiveness index. The server arranges multiple interactive correction values ​​in order of governance effectiveness indicators to obtain an interactive correction vector. When the interaction correction value is greater than zero, it indicates that the actual combined effect of the joint implementation of adjacent governance technologies is higher than the effect obtained by the initial aggregation based on the prediction results of individual governance technologies; when the interaction correction value is less than zero, it indicates that there is effect inhibition between adjacent governance technologies; when the interaction correction value is equal to zero, it indicates that the combined effect of adjacent governance technologies is consistent with the initial aggregation effect.

[0144] When the server processes the candidate technology combination chain "T2→T9→T20→T21", it keeps the combination covariates corresponding to the target ecological environment scenario unchanged, and sequentially sets "T2 and T9", "T9 and T20", and "T20 and T21" as the target combination intervention variables. The technology interaction causal forest model outputs the joint counterfactual governance effect index prediction vectors for the three adjacent technology pairs.

[0145] Please refer to the following: Figure 5 , Figure 5 This is a schematic diagram of the adjacent governance technology interaction correction structure provided in an embodiment of the present invention. For T2 and T9, the server aggregates the predicted vectors of the counterfactual governance effect indicators for T2 and T9 according to a preset indicator aggregation rule. The predicted sulfur dioxide emission reduction values ​​for T2 and T9 are 0.10 and 0.45, respectively, and the initial binary aggregation value of 0.55 is obtained by using the summation rule; the joint counterfactual sulfur dioxide emission reduction output by the technology interaction causal forest model is 0.62, and the server calculates an interaction correction value of 0.07 for sulfur dioxide emission reduction. The server performs the same vector difference calculation on the remaining nine governance effect indicators to form the interaction correction vectors corresponding to T2 and T9.

[0146] The server represents the c-th candidate technology combination chain as a sequence of governance technologies arranged in governance order. In the formula, This represents the c-th candidate technology combination chain; L represents the k-th candidate governance technology in the candidate technology combination chain; L represents the number of candidate governance technologies included in the candidate technology combination chain.

[0147] For governance effectiveness metrics that use the summation rule, the server sums the counterfactual governance effectiveness metric predictions for each candidate governance technology and adds the corresponding interaction correction value for each ordered adjacent governance technology to the summation result. In the formula, This represents the combined governance effect index data of the c-th candidate technology combination chain after interactive correction under the j-th governance effect index.

[0148] For governance effectiveness metrics that use a weighted average rule, the server calculates an initial weighted average based on the technical weights of each candidate governance technology, and then weights the interaction correction values ​​based on the interaction weights of adjacent governance technologies. The interaction weights for adjacent governance technologies are determined according to the following formula: In the formula, This represents the technical weight of the k-th candidate governance technology relative to the j-th governance effectiveness indicator. This represents the interaction weight of the k-th ordered neighbor governance technology pair relative to the j-th governance effect index; the sum of all interaction weights is 1. For governance effect indicators that adopt nonlinear combination rules executed in order of governance, the server first calculates the initial combination indicator value through the corresponding nonlinear aggregation function, and then adds the corresponding interactive correction value of each ordered adjacent governance technology through the interactive aggregation function. In the formula, This represents the nonlinear combination function corresponding to the j-th governance effectiveness indicator; This represents the interactive correction aggregation function corresponding to the j-th governance effectiveness indicator.

[0149] For governance effectiveness indicators with physical upper and lower limits, the server will restrict the combined governance effectiveness indicator data after interactive correction to the physical value range of the corresponding indicator. In the formula, and These represent the lower limit and upper limit of the j-th governance effectiveness indicator, respectively; This represents the combined governance effect index data after the value range correction has been completed.

[0150] The server adds an interaction correction vector only once to each ordered pair of adjacent governance technologies, and processes adjacent governance technology pairs in the candidate technology combination chain in the order of governance, thereby avoiding the interaction effect of the same adjacent governance technology being repeatedly included in the combined governance effect index data. The server arranges the interaction-corrected combined governance effect index data with the candidate technology combination chain as rows and the governance effect index as columns, to obtain the updated combination chain evaluation matrix. In the formula, denoted as the evaluation matrix of the combined chains after interactive correction; C represents the number of candidate technology combined chains; J represents the number of governance effectiveness indicators.

[0151] The server aggregates the counterfactual governance effect prediction vectors corresponding to T2, T9, T20, and T21 according to the governance order, and sequentially adds the interaction correction vectors corresponding to T2 and T9, T9 and T20, and T20 and T21 to obtain combined governance effect indicator data that simultaneously reflects the effects of individual technologies and the interaction effects of adjacent technologies. The server uses the combined governance effect indicator data to replace the original row data corresponding to "T2→T9→T20→T21" in the combined chain evaluation matrix, and updates other candidate technology combined chains in the same way. Finally, the server re-performs standardization, indicator weighting, weighting, ideal solution determination, distance calculation, correlation calculation, and fusion evaluation on the updated combined chain evaluation matrix to obtain the updated comprehensive applicability evaluation value of the combined chain, and regenerates the governance technology combined chain applicability assessment result.

[0152] This invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 performs the aforementioned method for assessing the applicability of ecological environment governance technologies. Figure 6 As shown, Figure 6 This is a structural block diagram of a computer device 100 provided in an embodiment of the present invention. The computer device 100 includes a memory 111, a processor 112, and a communication unit 113. To enable data transmission or interaction, the memory 111, processor 112, and communication unit 113 are electrically connected to each other directly or indirectly. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0153] For illustrative purposes, the foregoing description has been made with reference to specific embodiments. However, the foregoing illustrative discussions are not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed. Numerous modifications and variations are possible in accordance with the foregoing teachings. These embodiments were chosen and described in order to best illustrate the principles of the present disclosure and its practical application, thereby enabling those skilled in the art to best utilize the disclosure and to employ various embodiments with different modifications to suit a particular intended application.

Claims

1. A method for assessing the applicability of ecological environment governance technologies, characterized in that, The method includes: Acquire a set of historical governance case data, target ecological environment scenario parameters, and a set of candidate governance technologies. The set of historical governance case data includes historical governance scenario information, governance technology information, and governance effect indicator information. Based on the historical governance case data set, a scenario-adaptive prediction model based on causal forest is constructed. The scenario-adaptive prediction model is used to predict the governance effect of each candidate governance technology in the candidate governance technology set under the target ecological environment scenario parameters, and generate a target scenario evaluation matrix. Based on the indicator attributes corresponding to each governance effect indicator in the target scenario evaluation matrix, the target scenario evaluation matrix is ​​standardized to obtain a standardized evaluation matrix. The weights of each governance effectiveness indicator are determined based on the data distribution of each governance effectiveness indicator in the standardized evaluation matrix, and the standardized evaluation matrix is ​​weighted according to the indicator weights to obtain a weighted evaluation matrix. The positive and negative ideal solutions are determined based on the weighted evaluation matrix, and the distance evaluation parameters and correlation evaluation parameters of each candidate governance technology relative to the positive and negative ideal solutions are calculated. The distance evaluation parameter and the correlation evaluation parameter are fused according to the preset preference coefficient to obtain the comprehensive applicability evaluation value of each candidate governance technology; Based on the comprehensive applicability evaluation value, each candidate governance technology is ranked, and the applicability assessment results of the governance technologies corresponding to the target ecological environment scenario parameters are generated.

2. The method according to claim 1, characterized in that, The historical governance scenario information includes flue gas temperature, pollutant inlet concentration, ambient humidity, treatment scale, and investment budget; the governance technology information includes governance technology identifiers; the governance effect index information includes sulfur dioxide emission reduction, nitrogen oxide emission reduction, particulate matter emission reduction, carbon dioxide emission reduction, technology adoption rate, energy saving, unit investment cost, technology maturity, technology stability, and promotion value; the index attribute corresponding to the unit investment cost is a cost-type attribute, and the index attributes corresponding to the governance effect indicators other than the unit investment cost are benefit-type attributes.

3. The method according to claim 1, characterized in that, The process of constructing a scenario-adaptive prediction model based on causal forests using the historical governance case dataset and generating the target scenario evaluation matrix includes: The categorical historical governance scenario information and governance technology identifiers in the historical governance case dataset are encoded. By using encoded historical governance scenario information as covariates, encoded governance technology identifiers as intervention variables, and governance effect indicator data corresponding to different governance effect indicators as outcome variables, causal training samples corresponding to different governance effect indicators are constructed. Based on the causal training samples, causal forest regressors corresponding to different governance effect indicators are trained respectively, and multiple causal forest regressors are combined to form the scenario adaptive prediction model; The target ecological environment scenario parameters are encoded to obtain target scenario covariates. The target scenario covariates are kept unchanged, and the governance technology identifiers corresponding to each candidate governance technology in the historical governance case data set are set as target intervention variables in turn. The target scenario covariate and each target intervention variable are respectively input into the multiple causal forest regressors to obtain the counterfactual governance effect index prediction vector corresponding to each candidate governance technology; Arrange multiple counterfactual governance effect indicator prediction vectors according to the correspondence between each candidate governance technology and each governance effect indicator, and generate the target scenario evaluation matrix with candidate governance technologies as rows and governance effect indicators as columns.

4. The method according to claim 1, characterized in that, The standardization process for the target scenario evaluation matrix and the determination of the weights of each governance effectiveness indicator include: Obtain the maximum and minimum values ​​of each governance effectiveness indicator in the candidate governance technology set; For governance effectiveness indicators with benefit-type attributes, linear standardization is performed based on the difference between the corresponding governance effectiveness indicator data and the minimum value of the indicator. For governance effectiveness indicators with cost-type attributes, linear standardization is performed based on the difference between the maximum value of the indicator and the corresponding governance effectiveness indicator data. When the maximum value of any governance effectiveness indicator is equal to the minimum value of the indicator, the standardized indicator value corresponding to any governance effectiveness indicator is set to 1; A probability matrix is ​​constructed based on the proportion of multiple standardized indicator values ​​corresponding to the same governance effectiveness indicator in the sum of the multiple standardized indicator values, and the zero values ​​in the probability matrix are smoothed. The entropy value of each governance effectiveness indicator is calculated based on the smoothed probability matrix. The difference coefficient of each governance effectiveness indicator is determined based on the entropy value. The weight of each indicator is determined based on the proportion of each difference coefficient in the sum of all difference coefficients. Based on the correspondence between the various governance effectiveness indicators, the standardized indicator values ​​in the standardized evaluation matrix are multiplied by their corresponding indicator weights to obtain the weighted evaluation matrix.

5. The method according to claim 1, characterized in that, The step of determining the positive and negative ideal solutions based on the weighted evaluation matrix, and calculating the distance evaluation parameters and the correlation evaluation parameters, includes: The weighted index values ​​of each candidate governance technology under each governance effect index are obtained from the weighted evaluation matrix. The maximum value among multiple weighted index values ​​corresponding to the same governance effect index is determined as the positive ideal index value, and the minimum value among multiple weighted index values ​​corresponding to the same governance effect index is determined as the negative ideal index value, thus obtaining the positive ideal solution and the negative ideal solution. Calculate the first Euclidean distance between the weighted index value corresponding to each candidate governance technology and the positive ideal solution, and the second Euclidean distance between the weighted index value and the negative ideal solution, and use the first Euclidean distance and the second Euclidean distance as the distance evaluation parameters; Based on the preset resolution coefficient, calculate the first grey relational coefficient between each candidate governance technology and the positive ideal solution and the second grey relational coefficient between each candidate governance technology and the negative ideal solution under each governance effect index; The average values ​​of multiple first grey relation coefficients and multiple second grey relation coefficients corresponding to the same candidate governance technology are respectively processed to obtain the positive ideal solution grey relation degree and the negative ideal solution grey relation degree, and the positive ideal solution grey relation degree and the negative ideal solution grey relation degree are used as the relation evaluation parameters.

6. The method according to claim 5, characterized in that, The step of fusing the distance evaluation parameter and the correlation evaluation parameter according to the preset preference coefficient includes: The first Euclidean distance, the second Euclidean distance, the positive ideal solution grey relational degree, and the negative ideal solution grey relational degree are respectively subjected to dimensionless processing; Set the value that is greater than or equal to 0 and less than or equal to 1 as the preset preference coefficient; Based on the preset preference coefficient and the difference between 1 and the preset preference coefficient, the dimensionless second Euclidean distance and the dimensionless positive ideal solution grey relational degree are weighted and fused to obtain the positive proximity degree, and the dimensionless first Euclidean distance and the dimensionless negative ideal solution grey relational degree are weighted and fused to obtain the negative proximity degree. The comprehensive applicability evaluation value is determined based on the proportion of the positive proximity degree in the sum of the positive proximity degree and the negative proximity degree. The comprehensive applicability evaluation value of each candidate governance technology is calculated using multiple different preset preference coefficients. The ranking change results of each candidate governance technology under different preset preference coefficients are determined based on the multiple comprehensive applicability evaluation values. A comparison chart of comprehensive applicability evaluation values ​​and a heat map of the ranking change of candidate governance technologies are generated based on the ranking change results.

7. The method according to claim 1, characterized in that, The method further includes: According to the preset technology types, each candidate governance technology in the candidate governance technology set is assigned a technology type label. The preset technology types include source control type, process emission reduction type, end-of-pipe treatment type and waste heat recovery type arranged in the order of governance. From different preset technology types, select no more than one candidate governance technology, and arrange the selected candidate governance technologies according to the governance order to generate a candidate technology combination chain containing at least two and no more than four candidate governance technologies. The governance effect index data corresponding to the candidate governance technologies contained in each candidate technology combination chain are obtained from the target scenario evaluation matrix. The governance effect index data is then aggregated according to the preset index aggregation rules corresponding to each governance effect index to obtain combined governance effect index data. The preset index aggregation rules include summation rules, weighted average rules, or nonlinear combination rules executed according to the governance order. Construct a combined chain evaluation matrix based on the combined governance effect index data corresponding to each candidate technology combined chain; The combined chain evaluation matrix is ​​subjected to standardization, index weight determination, weighting, positive and negative ideal solution determination, distance evaluation parameter calculation, correlation evaluation parameter calculation and fusion processing in sequence to obtain the combined chain comprehensive applicability evaluation value of each candidate technology combined chain; Based on the comprehensive applicability evaluation value of the combined chain, each candidate technology combined chain is ranked, and the applicability evaluation result of the governance technology combined chain corresponding to the target ecological environment scenario parameters is generated.

8. The method according to claim 3, characterized in that, The method further includes: Obtain the predicted values ​​of governance effect indicators output by multiple causal trees in each causal forest regressor for the target scenario covariate and each target intervention variable, and construct the inter-tree prediction value sequence corresponding to each candidate governance technology; Based on the degree of dispersion among the predicted values ​​of multiple governance effectiveness indicators in the same inter-tree prediction value sequence, the prediction dispersion parameters of each candidate governance technology under each governance effectiveness indicator are determined. Obtain the target leaf nodes corresponding to the target scenario covariates in each causal tree, and count the number of historical governance cases that match the corresponding target intervention variables within the target leaf nodes. Determine the sample support parameters of each candidate governance technology under each governance effect indicator based on the number of historical governance cases. The counterfactual prediction confidence coefficients for each candidate governance technology are determined based on the predicted discrete parameters and the sample support parameters. The counterfactual prediction confidence coefficients decrease as the predicted discrete parameters increase and increase as the sample support parameters increase. The comprehensive applicability evaluation value corresponding to each candidate governance technology is multiplied by the corresponding counterfactual prediction confidence coefficient to obtain the confidence-corrected applicability evaluation value, and the governance technology applicability assessment result is updated based on the confidence-corrected applicability evaluation value.

9. The method according to claim 7, characterized in that, The method further includes: Obtain a historical governance technology combination case data set, which includes combined governance scenario information, multiple governance technology identifiers arranged in the actual governance order, and combined governance effect index data; Using the combined governance scenario information as a combined covariate, the adjacent governance technology identifiers arranged in the actual governance order as a combined intervention variable, and the combined governance effect index data as a combined result variable, a governance technology interaction training sample is constructed, and a technology interaction causal forest model is trained based on the governance technology interaction training sample. For two adjacent candidate governance technologies in each candidate technology combination chain according to the governance order, the combination covariates corresponding to the target ecological environment scenario parameters are kept unchanged. The governance technology identifiers corresponding to the two adjacent candidate governance technologies are set as target combination intervention variables. The joint counterfactual governance effect index prediction vectors corresponding to the adjacent governance technologies are generated through the technology interaction causal forest model. According to the preset index aggregation rules, the counterfactual governance effect index prediction vectors corresponding to two adjacent candidate governance technologies are aggregated to obtain an initial binary aggregation vector. The difference between the joint counterfactual governance effect index prediction vector and the initial binary aggregation vector is calculated to obtain the interaction correction vector corresponding to the adjacent governance technologies. According to the governance order, the counterfactual governance effect index prediction vectors corresponding to the candidate governance technologies contained in each candidate technology combination chain and the interaction correction vectors corresponding to adjacent governance technologies are aggregated to obtain the combined governance effect index data after interaction correction. The combined governance effect index data after the interaction correction is updated to update the combined chain evaluation matrix, and the applicability assessment result of the combined governance technology chain is updated according to the updated combined chain evaluation matrix.

10. An ecological environment governance technology applicability assessment system, characterized in that, Includes at least one service node; The service node includes a storage unit and a computing unit; the storage unit is used to store program code; the computing unit is used to run the program code to perform the method according to any one of claims 1 to 9.

Images