Three-party environmental protection management system suitable for industrial park

Abstract

The invention relates to the technical field of environmental protection steward management systems, and provides a three-party environmental protection management system suitable for an industrial park, and the system comprises an admission site selection module which is used for carrying out the analysis according to the basic information of an enterprise and the regulations of the industrial park, and generating an admission study and judgment report; the environment credit evaluation module is used for evaluating and grading the environmental protection management conditions of the settled enterprises, generating enterprise grading lists and carrying out directional information pushing; the steward assessment module is used for making a patrol work plan, recording patrol results and generating an evaluation report for a third-party service unit; the complaint processing module is used for receiving the complaint information, generating rectification suggestions according to the complaint information and generating a settlement report; and the report generation module is used for summarizing the result content of each module to generate a staged analysis report. Through the technical scheme, the problems of low model training efficiency and insufficient rectification suggestion reliability in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental steward management systems, and in particular to a three-party environmental management system suitable for industrial parks. Background Art

[0002] With the deepening development of industrial upgrading and ecological environmental protection policies in various industries across the country, mainstream enterprises have stepped out of the extensive development stage and developed in the direction of resource conservation and environmental friendliness. The requirements for environmental protection management in various regions are constantly increasing. At present, the environmental protection awareness of enterprises has been greatly improved, but due to factors such as their own scale or insufficient personnel capacity and insufficient professional level, the improvement of environmental protection management is still difficult. At the same time, the environmental protection departments of industrial parks also lack certain professionalism and openness in solving environmental pollution incidents and complaints. Therefore, the "environmental steward" work is being actively and extensively carried out at the district, village, park, and enterprise levels. As a new model of industry service, it can start from the perspective of the park and service units, strive to improve the quality of environmental steward services, improve the ability of enterprises to solve problems, and reduce the management costs of the park. It plays an important role in improving the comprehensive service quality and technical level of environmental management and meeting the needs of parks and enterprises for environmental management.

[0003] However, third-party environmental protection services have not yet completely gotten rid of the constraints of paper verification sheets. A large number of people are needed to support offline services, and the service efficiency is low. There are problems such as untimely processing and unreasonable suggestions for the verification and rectification of complaint information. In the prior art, although there are also methods that use a variety of neural network models to assist in the processing of third-party environmental protection services or complaint information, since the verification of complaint information itself requires consideration of a large number of factors, that is, a large number of parameters, according to actual conditions, different complaint information not only has different priorities, but also corresponds to different rectification suggestions. When multiple models are used for training to judge the priority of complaint information and output rectification suggestions, there is not only parameter redundancy, but also the need to train each model and adjust parameters separately, which increases the model training time. It also causes different tasks to be unable to share learned information with each other, resulting in poor consistency in the output of model results, thereby reducing the reliability of rectification suggestions. Summary of the invention

[0004] The present invention proposes a three-party environmental management system suitable for industrial parks, which solves the problems of low model training efficiency and insufficient reliability of rectification suggestions in related technologies.

[0005] The technical solution of the present invention is as follows:

[0006] The present invention proposes a tripartite environmental management system suitable for industrial parks, comprising:

[0007] The access site selection module is used to match and analyze the basic information of the enterprise with the regulations of the industrial park and generate an access assessment report; the enterprise includes the enterprise to be settled or the enterprise already settled;

[0008] The environmental credit evaluation module is used to evaluate and grade the environmental management status of settled enterprises based on their information, generate a graded list of enterprises, and push targeted information to settled enterprises with different ratings based on the graded list of enterprises;

[0009] The housekeeper assessment module is used to formulate inspection work plans, record inspection results, and conduct assessment and evaluation of third-party service units based on the inspection results and generate evaluation reports;

[0010] The complaint handling module is used to receive complaint information, generate rectification suggestions based on the complaint information, generate a complaint event response after the complaint is resolved, and generate a completion report after the complainant's satisfaction evaluation;

[0011] The complaint handling module includes:

[0012] A judgment unit, the judgment unit includes a natural language understanding neural network model, the natural language understanding neural network model includes a first branch and a second branch, the first branch and the second branch share the same input layer, text encoder, structured data encoder and fusion layer, the first branch also includes a fully connected layer and an output layer, and the second branch also includes a decoder;

[0013] The first branch of the natural language understanding neural network model is used to determine the risk level of complaint information based on the basic information of the enterprise and its surrounding environment and the complaint information; the second branch of the natural language understanding neural network model is used to output rectification suggestions based on the basic information of the enterprise and its surrounding environment, complaint information, risk level and incident supplementary information;

[0014] Among them, the input layer is used to obtain data and divide the data into structured data and unstructured data according to the data structure;

[0015] The text encoder uses the BERT model to encode unstructured data and obtain the feature vector of unstructured data;

[0016] The structured data encoder includes a categorical variable branch and a continuous variable branch, the categorical variable branch is composed of an embedding layer, the continuous variable branch is composed of a fully connected layer, and the structured data encoder is used to encode structured data;

[0017] The structured data is divided into categorical variables and continuous variables according to the data type. The categorical variable branch is used to encode the categorical variables in the structured data to obtain the embedded vector of the categorical variable. The continuous variable branch is used to encode the continuous variables in the structured data to obtain the feature vector of the continuous variable.

[0018] Among them, the calculation formula of the embedding vector of the categorical variable is as follows:

[0019] e multi =[E1(x1),E2(x2),...,E h (x h )];

[0020] Among them, e multi represents the embedding vector of the categorical variable, x h represents the hth categorical variable, E h represents the embedding matrix corresponding to the h-th categorical variable;

[0021] The formula for calculating the eigenvector of a continuous variable is as follows:

[0022] y multi =[σ(W1q1+b),σ(W2q2+b),...,σ(W v q v +b)];

[0023] Among them, y multi Characteristic vector representing continuous variables, q v represents the vth continuous variable, W v represents the weight matrix corresponding to the vth continuous variable, b represents the bias vector, and σ represents the activation function;

[0024] The fusion layer is used to concatenate the output of the text encoder and the output of the structured data encoder in the feature dimension to obtain the final data feature vector or the completed data feature vector;

[0025] The fully connected layer is used to obtain the probability of complaint information belonging to each risk level based on the final data feature vector;

[0026] The output layer is used to output the risk level with the highest probability as the risk level of the complaint information;

[0027] The decoder uses the T5 model to decode the feature vector of the completed data and output rectification suggestions;

[0028] The report generation module is connected with the access site selection module, environmental credit evaluation module, steward assessment module and complaint handling module, and is used to summarize the results of each module to generate a phased analysis report.

[0029] Furthermore, it also includes: enterprise side, third-party side and park side;

[0030] The enterprise end is connected with the access site selection module, environmental credit evaluation module, and steward assessment module. The enterprise logs in through the enterprise end of the management system and performs corresponding operations;

[0031] The third-party terminal is connected with the access site selection module, environmental credit evaluation module, housekeeper assessment module, complaint handling module, and report generation module. The third-party service unit logs in through the third-party terminal of the management system and performs corresponding operations;

[0032] The park end is connected to the access site selection module, environmental credit evaluation module, housekeeper assessment module, complaint handling module, and report generation module. The industrial park logs in through the park end of the management system and performs corresponding operations.

[0033] Furthermore, the admission site selection module includes:

[0034] The site selection data acquisition unit is used to obtain the project type, industry type of the proposed project and supplementary information of the enterprise through the enterprise end, and obtain the regulations of the industrial park through the system database;

[0035] An access assessment unit is connected to the site selection data acquisition unit, and the access assessment unit includes a feedforward neural network model;

[0036] The feedforward neural network model is used to compare the enterprise's project type and the industry type of the proposed project with the regulations of the industrial park, and output the enterprise's access determination result;

[0037] If the enterprise's access determination result is no, the research and determination process ends; if the enterprise's access determination result is yes, the site selection plot is screened according to the industry type of the enterprise's proposed project;

[0038] The feedforward neural network model is also used to compare the supplementary information of the enterprise with the regulations of the industrial park, and output site selection reference suggestions and access assessment reports based on the site selection plot.

[0039] Furthermore, the environmental credit evaluation module includes:

[0040] An evaluation data acquisition unit is used to obtain information of settled enterprises through a third party terminal and determine the impact factor based on the information of settled enterprises;

[0041] A calculation unit, connected to the evaluation data acquisition unit, for calculating the environmental credit score of the settled enterprise according to the influencing factors and preset weights;

[0042] A grading unit, connected to the calculation unit, is used to compare the environmental credit scores of settled enterprises with preset scoring levels, determine the evaluation grades of settled enterprises, and generate an enterprise grading list;

[0043] The push unit is connected to the evaluation data acquisition unit and the grading unit, and is used to push information to settled enterprises with different ratings through the enterprise side according to the enterprise grading list and the information of settled enterprises.

[0044] Furthermore, the butler assessment module includes:

[0045] The inspection plan unit is used to obtain the inspection period, inspection frequency and inspection requirements through the park end, generate an inspection work plan, obtain the number of inspection teams, inspection team members, historical inspection conditions and enterprise classification list through the third-party end, and perform intelligent scheduling according to the inspection work plan;

[0046] The inspection results unit is connected to the inspection plan unit and is used to obtain the basic information of the settled enterprises, environmental protection procedures, pollution prevention and control facilities, environmental management, on-site inspection forms and on-site photos through the third-party terminal, and form inspection results; among them, the inspection results are one file for each enterprise;

[0047] The assessment and evaluation unit is connected to the inspection result unit, and is used to determine the completion rate, accuracy rate and rectification rate of the inspection results based on the inspection results, obtain the satisfaction evaluation of the settled enterprises on the third-party service units through the enterprise side, and generate an evaluation report based on the completion rate, accuracy rate, rectification rate and satisfaction evaluation of the inspection results.

[0048] Furthermore, the complaint handling module also includes:

[0049] The complaint information acquisition unit is connected to the judgment unit and is used to obtain the complaint information through the third-party terminal and the park terminal, and retrieve the basic information of the enterprise and the surrounding environment within the preset range of the complaint point according to the complaint information;

[0050] A comparison unit, connected to the complaint information acquisition unit, is used to compare the complaint information with previous complaint information. If the previous complaint information contains the same type of complaint events as the complaint information, the previous complaint information is referenced to determine the rectification suggestions for the complaint information;

[0051] The task sorting unit is connected to the comparison unit and the judgment unit, and is used to sort the complaint information from high to low according to the risk level, and send the sorting results to the park end.

[0052] Furthermore, the complaint handling module also includes:

[0053] A supplementary information acquisition unit, connected to the judgment unit, for acquiring supplementary event information through a park terminal or a third-party terminal;

[0054] The completion report generation unit is connected to the complaint information acquisition unit and the judgment unit, and is used to obtain the response to the complaint event and the complainant's satisfaction evaluation through the park end or the third-party end, and generate a completion report based on the complaint information, rectification suggestions, the response to the complaint event and the complainant's satisfaction evaluation.

[0055] Furthermore, the report generation module includes:

[0056] Data aggregation unit, used to receive access assessment reports, enterprise classification lists, evaluation reports and completion reports;

[0057] The report generation unit is connected to the data aggregation unit and is used to generate periodic analysis reports based on the access assessment report, enterprise classification list, evaluation report and completion report, and send them to the third-party end.

[0058] The working principle and beneficial effects of the present invention are:

[0059] 1. The present invention sets two branches in the natural language understanding neural network model to perform different task learning respectively, and the two branches share feature extraction layers such as text encoders and structured data encoders. Through joint training, the model can share information between multiple tasks, thereby improving the performance of each task, and reducing the neural network layers required for the model structure, reducing the number of parameters, reducing computing overhead and memory usage, and improving model training efficiency. Since the two tasks are trained in the same model, they share information, reducing the problem of inconsistency in the outputs of different tasks, thereby improving the reliability and accuracy of the rectification suggestions finally output.

[0060] 2. The present invention carries out pollution source control in the industrial park from the enterprise access and site selection stage, to the enterprise environmental credit rating and third-party steward assessment, implements the supervision of the pollution process in the industrial park, and then to the complaint handling, results management and improvement suggestions after the pollution occurs, promotes the end-of-pipe treatment of pollution in the industrial park, and realizes a multi-dimensional management closed loop of the entire process of environmental protection on the scale of the industrial park, which is conducive to improving the overall environmental management efficiency; in addition, the present invention realizes the evaluation of the work of the third-party service unit by the enterprise through the assessment module, and the industrial park supervises the process of the work of the third-party service unit, which provides an objective basis for the industrial park to evaluate the service level of the third-party service unit and promotes the further standardization of the services of the third-party service unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0062] Figure 1 A schematic diagram of the structure of a three-party environmental management system applicable to an industrial park provided by an embodiment of the present invention;

[0063] Figure 2 A schematic diagram of the structure of an access site selection module provided by an embodiment of the present invention;

[0064] Figure 3 A schematic diagram of the structure of an environmental credit evaluation module provided by an embodiment of the present invention;

[0065] Figure 4A schematic diagram of the structure of a housekeeper assessment module provided by an embodiment of the present invention;

[0066] Figure 5 A schematic diagram of the structure of a complaint handling module provided by an embodiment of the present invention;

[0067] Figure 6 A structural diagram of a report generation module provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0068] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] Example 1

[0070] like Figure 1 to Figure 6 As shown, this embodiment proposes a three-party environmental management system suitable for industrial parks.

[0071] In this embodiment, a tripartite environmental management system applicable to an industrial park includes: an access site selection module, an environmental credit evaluation module, a housekeeper assessment module, a complaint handling module and a report generation module.

[0072] The access site selection module is used to match and analyze the basic information of the enterprise with the regulations of the industrial park and generate an access assessment report;

[0073] The environmental credit evaluation module is used to evaluate and grade the environmental management status of settled enterprises based on their information, generate a graded list of enterprises, and push targeted information to settled enterprises with different ratings based on the graded list of enterprises;

[0074] The housekeeper assessment module is used to formulate inspection work plans, record inspection results, and conduct assessment and evaluation of third-party service units based on the inspection results and generate evaluation reports;

[0075] The complaint handling module is used to receive complaint information, generate rectification suggestions based on the complaint information, generate a complaint event response after the complaint is resolved, and generate a completion report after the complainant's satisfaction evaluation;

[0076] The report generation module is connected with the access site selection module, environmental credit evaluation module, steward assessment module and complaint handling module, and is used to summarize the results of each module to generate a phased analysis report.

[0077] In this embodiment, a three-party environmental management system applicable to an industrial park further includes: an enterprise end, a third-party end, and a park end;

[0078] The enterprise side is connected to the access site selection module, environmental credit evaluation module, and steward assessment module. The enterprise logs in through the enterprise side of the management system and performs corresponding operations.

[0079] The third-party end is connected to the access site selection module, environmental credit evaluation module, housekeeper assessment module, complaint handling module, and report generation module. The third-party service unit logs in through the third-party end of the management system and performs corresponding operations.

[0080] The park end is connected to the access site selection module, environmental credit evaluation module, housekeeper assessment module, complaint handling module, and report generation module. The industrial park logs in through the park end of the management system and performs corresponding operations.

[0081] In this embodiment, the system supports the login of the environmental protection management department, third-party service units and enterprises of the industrial park. After logging in, GPS positioning punch-in and shooting functions can be realized. The park end has the authority to manage the input data of the enterprise end and the third-party end, enterprise lists, inspection plans, special plans, collaborative work, enterprise zoning and grading, environmental protection inspection case analysis / environmental protection policy interpretation / professional skills training push, complaint event distribution and settlement, third-party service quality evaluation, report generation, etc. The third-party end has the authority to manage enterprise information filling, enterprise environmental management level evaluation, complaint and rectification problem closure, etc. The enterprise end has the authority to input access unit information and view results, existing problems and case / policy / training push, rectification response, third-party service satisfaction evaluation, etc.

[0082] In this embodiment, the admission site selection module includes:

[0083] The site selection data acquisition unit is used to obtain the project type of the enterprise, the industry type of the planned project and the supplementary information of the enterprise through the enterprise side, and obtain the regulations of the industrial park through the system database.

[0084] Specifically, enterprises include enterprises that are planned to be settled or enterprises that have already settled. Project types include new projects, reconstruction projects, expansion projects, relocation projects, etc. The industry type of the proposed project can be based on the industry classification in the "National Economic Industry Classification". For example, for the reconstruction, expansion, relocation and other project types of settled enterprises, it is also necessary to conduct access assessment through the access site selection module.

[0085] Furthermore, the supplementary information of the enterprise may include address, business area, estimated scale, main production process, main raw and auxiliary material usage, main equipment, raw and auxiliary material usage containing volatile organic compounds, types of pollutants involved, types of energy used, etc. The provisions of the industrial park include the laws, regulations, policy documents, planning plans and other information applicable to the current industrial park.

[0086] An access assessment unit is connected to the site selection data acquisition unit, and the access assessment unit includes a feedforward neural network model;

[0087] The feedforward neural network model is used to compare the enterprise's project type and the industry type of the planned project with the regulations of the industrial park, and output the enterprise's access judgment result.

[0088] Specifically, by referring to the laws, regulations, policy documents, planning programs and other information applicable to the region (such as the negative list of industrial access, the ecological environment access list and the "Guidelines for Industrial Structure Adjustment" and other documents), determine whether the industry type of the proposed project is on the access list. If it is not on the access list, the enterprise's access judgment result is no, and the analysis process ends; if it is on the access list, the enterprise's access judgment result is yes, and the site selection plots that can be used for site selection are screened out according to the industry type of the enterprise's proposed project and highlighted.

[0089] The feedforward neural network model is also used to compare the supplementary information of the enterprise with the regulations of the industrial park, and output site selection reference suggestions and access assessment reports based on the site selection plot.

[0090] Specifically, the access assessment report includes a summary of the basic information of the enterprise, the comparison results of the access list with the control requirements, the types of projects that need to go through environmental protection procedures, the process, time requirements and precautions. Among them, the types of environmental protection procedures mainly include environmental impact assessment reports, pollution discharge permits, completion acceptance reports, emergency plans for sudden environmental incidents, etc. For example, the site selection reference suggestions can be distinguished by color, for example, the recommended site selection plot is displayed in green, the optional site selection plot is displayed in yellow, and the prohibited area is displayed in red.

[0091] Furthermore, when insufficient information exists to make a judgment or the judgment fails, a third-party service provider can intervene to assist in the judgment.

[0092] For example, the industry that the enterprise intends to enter is "Forty-seven, ecological protection and environmental governance industry, general industrial solid waste (including sewage treatment sludge) and construction waste disposal and comprehensive utilization". According to the "List of Industrial Project Ecological Environmental Access Catalog", in the Class I atmospheric zone of this district, "it is prohibited to build hazardous waste (excluding medical waste) utilization and disposal, medical waste disposal, harmless treatment of dead and diseased animals, general industrial solid waste (including sewage treatment sludge), construction waste disposal and comprehensive utilization projects. Except for simple collection, storage and transfer", if the main production process is entered as "storage and transfer" in the supplementary data type, the site selection range restriction of the Class I atmospheric zone will be cancelled, and according to the "List of Classification Management of Fixed Pollution Sources" and the "List of Classification Management of Environmental Impact Assessment of Construction Projects", it is concluded that the environmental impact assessment report form needs to be processed and included in the key management of pollutant discharge permits. After that, according to the environmental impact assessment of the park planning, it is determined that the park's leading industry includes the environmental governance industry and has corresponding public supporting facilities. The land where the park is located is displayed as a green recommendation, other optional areas are yellow, and the environmental protection area and the surrounding areas of sensitive points are red prohibited areas.

[0093] In this embodiment, the environmental credit evaluation module includes:

[0094] The evaluation data acquisition unit is used to obtain the information of the settled enterprises through a third party terminal and determine the impact factor based on the information of the settled enterprises.

[0095] Specifically, the information of settled enterprises may include basic information of settled enterprises, environmental protection procedures information, status of pollution prevention and control facilities, environmental management status, on-site verification sheets and on-site photos, etc. Among them, environmental protection procedures information includes environmental impact assessment reports, approval documents, acceptance materials, monitoring reports, solid waste disposal materials, emergency plans, pollution discharge permits, etc. Influencing factors include energy consumption per unit of GDP, pollutant emission reduction rate, carbon dioxide emission reduction rate and problem rectification rate.

[0096] The calculation unit is connected to the evaluation data acquisition unit and is used to calculate the environmental credit score of the settled enterprises according to the influencing factors and preset weights.

[0097] Specifically, the preset weights of each influencing factor can be set by the environmental management department of the industrial park, and the environmental credit score of the settled enterprises can be calculated according to the weighted coefficient method. For example, the calculation formula of the weighted score of the influencing factor is: weighted score of the influencing factor = ∑ (value of the influencing factor - the set standard value of the influencing factor) / the set standard value of the influencing factor × the weight of the influencing factor. The environmental credit score of the settled enterprise is obtained according to the sum of the weighted scores of all influencing factors.

[0098] The grading unit is connected to the calculation unit and is used to compare the environmental credit scores of the settled enterprises with the preset scoring levels, determine the evaluation grades of the settled enterprises, and generate an enterprise grading list.

[0099] Specifically, the environmental credit score of the settled enterprise is compared with the preset score level. When the environmental credit score of the settled enterprise is greater than the positive preset score, the evaluation level of the settled enterprise is positive. When the environmental credit score of the settled enterprise is less than the negative preset score, the evaluation level of the settled enterprise is negative. When the environmental credit score of the settled enterprise is less than or equal to the positive preset score and greater than or equal to the negative preset score, the settled enterprise is not evaluated. Based on the evaluation level of the settled enterprises, a list of enterprise levels is generated, wherein the list of enterprise levels only includes settled enterprises with positive evaluation levels and settled enterprises with negative evaluation levels.

[0100] For example, the list management refers to the conditions included in the management measures for the positive list of regional ecological environment supervision and law enforcement. According to the list, relevant environmental protection inspection and punishment cases are pushed to enterprises in the park, and the latest environmental protection policy interpretations and special environmental protection training are carried out. Information on relevant policy cases is obtained through the network search and related data capture function. The data acquisition channels include the official websites of the Ministry of Ecology and Environment and the Environmental Protection Bureau, public account tweets, news media, etc. Through key fields and image and text recognition, the captured data is processed, similar data is merged and key information is extracted.

[0101] The push unit is connected to the evaluation data acquisition unit and the grading unit, and is used to push information to settled enterprises with different ratings through the enterprise side according to the enterprise grading list and the basic information of the settled enterprises.

[0102] Specifically, the push unit pushes information to settled enterprises with different ratings according to the enterprise classification list and the basic information of settled enterprises, where the basic information of settled enterprises may include industry category, type of pollutants involved in emission, address, etc. For example, for settled enterprises with positive evaluation rating, the focus is on the interpretation of applicable green policies, and enterprises are encouraged to actively obtain the qualification for applying for special support funds; for settled enterprises with negative evaluation rating, the focus is on compliance management training and supervision and punishment case push, so as to clear the environmental protection management vein of enterprises.

[0103] In this embodiment, the housekeeper assessment module includes:

[0104] The inspection plan unit is used to obtain the inspection period, inspection frequency and inspection requirements through the park end, generate an inspection work plan, obtain the number of inspection teams, inspection team members, historical inspection situations and enterprise classification list through the third-party end, and perform intelligent scheduling according to the inspection work plan.

[0105] Specifically, the park side formulates a phased inspection work plan by setting the inspection period, inspection frequency and inspection requirements for each management level. In addition, according to the special inspection requirements, a list of enterprises to be inspected can be screened and sent to the third-party side to carry out special inspections. The third-party side enters the number of inspection teams, team members, and historical inspections in the inspection unit, and performs intelligent scheduling based on the existing enterprise zoning and grading information in the park. The park side can view the recent arrangements and actual completion status of the inspection in this module to achieve process supervision of the third-party inspection work. For example, the park side publishes special inspection time limits, lists, inspection requirements, etc. on the system. The system arranges inspection personnel based on existing historical data, and after the inspection personnel confirm, feedback is given to the park side to predict whether the inspection results meet the set requirements.

[0106] The inspection results unit is connected to the inspection plan unit and is used to obtain the basic information of the settled enterprises, environmental protection procedures, pollution control facilities, environmental management, on-site inspection forms and on-site photos through a third-party terminal, and form inspection results; among them, the inspection results are one file for each enterprise.

[0107] Specifically, if the park side and the third party side find environmental problems during the inspection process, they can log in to the system through the mobile terminal, locate the problem points, take pictures, describe the problems and upload them. According to the problem classification, they are automatically included in each module for processing. Each inspection team conducts daily inspections of enterprises according to the inspection plan, and puts forward rectification suggestions based on the environmental protection problems verified on site. After the rectification of the enterprise is implemented, the relevant certificates are uploaded to complete the rectification closed loop. The enterprise can log in to the system through the enterprise side to view the rectification problems and suggestions fed back by the inspection. After the rectification is completed, the rectification content can be photographed for rectification closed-loop response. The system recognizes the submitted pictures (i.e., on-site photos) through the convolutional neural network model, and compares the standard rectification template to determine whether the rectification requirements are met. When the rectification requirements are met, the rectification closed loop is completed.

[0108] Furthermore, the management levels of enterprises in the zone will be further divided according to the inspection results. According to information such as the enterprise's industry category, type and amount of pollutants discharged, violation penalty records, overdue rectification status, etc., they will be specifically classified into key supervision units, general supervision units and simplified supervision units.

[0109] The assessment and evaluation unit is connected to the inspection result unit, and is used to determine the completion rate, accuracy rate and rectification rate of the inspection results based on the inspection results, obtain the satisfaction evaluation of the settled enterprises on the third-party service units through the enterprise side, and generate an evaluation report based on the completion rate, accuracy rate, rectification rate and satisfaction evaluation of the inspection results.

[0110] Specifically, the assessment benchmark can be set by retrieving the situation of similar areas / projects or past projects, and the weights of various influencing factors of the assessment and evaluation can be determined. The influencing factors include the inspection frequency of enterprises at each management level, file update status, new enterprise screening rate, special frequency, training frequency and complaint status. Among them, the completion rate of the inspection results refers to the proportion of the number of enterprises that have completed the inspection within the preset period to the total number of enterprises to be inspected; the accuracy rate of the inspection results refers to the automatic correlation and identification of the information reported on the on-site verification form with the graphic information of one enterprise and one file as the benchmark, to obtain the accuracy rate of environmental protection information reported by the third-party service unit; the rectification rate of the inspection results refers to the proportion of enterprise problems that have completed closed-loop rectification to the total number of problems found; the satisfaction evaluation includes the satisfaction evaluation of the settled enterprises from the three aspects of the professionalism, response speed and service attitude of the third-party service unit. The calculation formula for the score of each indicator is as follows:

[0111]

[0112] Where P is the indicator score, which includes the total completion rate, total accuracy rate, total rectification rate and total satisfaction evaluation within the preset period, j represents the sub-items included in each indicator, such as routine inspection, special inspection, complaint inspection, etc., m represents the total number of sub-items included in each indicator, and p j Represents the score of the jth sub-item.

[0113]

[0114] Where i represents the i-th category under the sub-item, n represents the total number of categories under the sub-item, and s i represents the actual value of the i-th category, k i represents the weight of the i-th category, S i represents the target value of the ith category, and K represents the weight of the sub-item in the total number of indicators. i ≤S i , that is, when the actual value of the i-th category is greater than the target value of the i-th category, the target value of the i-th category is calculated as the actual value of the i-th category.

[0115] For example, the completion rate of previous projects in this area is 98%, the weight of routine inspections in the total completion rate is 60%, the weight of special inspections in the total completion rate is 20%, and the weight of complaint inspections in the total completion rate is 20%. The classification of this routine inspection includes key supervision units, general supervision units and simplified supervision units. The actual number of inspections of key supervision units, general supervision units and simplified supervision units is 29 / 30, 92 / 90, and 165 / 150, and the weight ratio is 3:2:1, accounting for The weight of the total completion rate is 60%, so the completion rate of this routine inspection is [(29×3+90×2+150×1) / (30×3+90×2+150×1)]×60%=59.57%. The scores of the remaining sub-items can be obtained in the same way. The total completion rate score of this cycle is obtained by summing up the scores of each sub-item. The total completion rate score is compared with the assessment benchmark of previous projects. The assessment evaluation of the third-party service unit is obtained based on the total completion rate, total accuracy rate, total rectification rate and total satisfaction evaluation within the period.

[0116] In this embodiment, the complaint handling module includes:

[0117] The complaint information acquisition unit is used to obtain complaint information through the third-party terminal and the park terminal, and retrieve the basic information of enterprises and service units within the preset range of the complaint point based on the complaint information.

[0118] Specifically, the basic information of enterprises and surrounding environment within the preset range of the complaint point includes operation and pollution discharge information of surrounding industrial enterprises, catering, auto repair and other service units, construction information, road traffic information, river information and climate information.

[0119] The comparison unit is connected to the complaint information acquisition unit and is used to compare the complaint information with the previous complaint information. If the previous complaint information contains the same type of complaint events as the complaint information, the rectification suggestions for the complaint information are determined by referring to the previous complaint information.

[0120] Specifically, the comparison scope of past complaint information includes historical complaint events in the same area, complaint events of the same type, unresolved complaint events, and events from the same complaint source, etc. The comparison content of past complaint information includes the source of the complaint, the complainant, the time period of the event, the content of the event, etc.

[0121] A judgment unit connected to the complaint information acquisition unit, the judgment unit including a natural language understanding neural network model;

[0122] Specifically, the natural language understanding neural network model includes a first branch and a second branch. The first branch and the second branch share the same input layer, text encoder, structured data encoder and fusion layer. The first branch also includes a fully connected layer and an output layer, and the second branch also includes a decoder. The first branch of the natural language understanding neural network model is used to determine the risk level of the complaint information based on the basic information and complaint information of the enterprise and its surrounding environment.

[0123] Among them, the input layer is used to obtain primary data, which includes basic information and complaint information of enterprises and service units. The primary data is divided into structured data and unstructured data according to the data structure; structured data is data with standardized format and predefined model, such as various parameters, classification levels and other information, and unstructured data is data that does not have standardized format and predefined model, such as text description and other information.

[0124] The text encoder uses the BERT model to encode unstructured data and obtain the feature vector of unstructured data;

[0125] The structured data encoder includes a categorical variable branch and a continuous variable branch, the categorical variable branch is composed of at least one embedding layer, the continuous variable branch is composed of at least one fully connected layer, and the structured data encoder is used to encode the structured data;

[0126] The structured data are divided into categorical variables and continuous variables according to the data type. The categorical variables may include the degree of environmental sensitivity, the type of pollution discharge in the pollution incident, the degree of pollution discharge hazard, the risk level of surrounding enterprises, etc. The continuous variables may include the complaint rate, the number of complaints, the distance between the complaint point and the residential area, etc.

[0127] The categorical variable branch is used to encode the categorical variables in the structured data to obtain the embedded vector of the categorical variable, and the continuous variable branch is used to encode the continuous variables in the structured data to obtain the feature vector of the continuous variable;

[0128] Among them, the calculation formula of the embedding vector of the categorical variable is as follows:

[0129] e multi =[E1(x1),E2(x2),...,E h (x h )];

[0130] Among them, e multi represents the embedding vector of the categorical variable, x h represents the hth categorical variable, E h represents the embedding matrix corresponding to the h-th categorical variable;

[0131] The formula for calculating the eigenvector of a continuous variable is as follows:

[0132] y multi =[σ(W1q1+b),σ(W2q2+b),...,σ(W v q v +b)];

[0133] Among them, y multi Characteristic vector representing continuous variables, q v represents the vth continuous variable, W v represents the weight matrix corresponding to the vth continuous variable, b represents the bias vector, σ represents the activation function, and the activation function can be the ReLU function;

[0134] The fusion layer is used to concatenate the output of the text encoder and the output of the structured data encoder in the feature dimension to obtain the final data feature vector;

[0135] The fully connected layer is used to obtain the probability of complaint information belonging to each risk level based on the final data feature vector;

[0136] The output layer is used to output the risk level with the highest probability as the risk level of the complaint information.

[0137] Specifically, the natural language understanding neural network model is used to conduct a comprehensive analysis of information such as the environmental sensitivity of the incident location, the complaint rate, the type of pollution discharge, the degree of pollution hazard, the risk level of surrounding enterprises, etc., to determine the risk level of the complaint information.

[0138] For example, there have been multiple complaints of pungent odors at locations upwind from residential areas. Judging from the distance of the complaint points from the residential areas, the environmental sensitivity level is ranked second priority. The number of complaints is 16, with a complaint rate of 42%. The pollution incident is a pungent odor, and the type of pollution discharge is air-related. There has been no on-site investigation and monitoring, and the degree of pollution hazard is unknown. There are no enterprises with odor emission processes in the surrounding area, and the risk level is low at level three.

[0139] The task sorting unit is connected to the comparison unit and the judgment unit, and is used to sort the complaint information from high to low according to the risk level, and send the sorting results to the park end.

[0140] Furthermore, the industrial park decides whether to forward the complaint to a third party and handle it together with the third-party service unit. The person in charge of the inspection team of the area where the complaint is located will work with the park's environmental management department to verify the complaint on site and make recommendations for rectification based on the same type of complaint handling methods in the database and on-site feedback.

[0141] In this embodiment, the complaint handling module also includes:

[0142] The supplementary information acquisition unit is used to acquire the supplementary information of the event through the park end or the third-party end.

[0143] Among them, the supplementary information of the event includes pollutant type, pollutant emission amount, accident risk level, event status (such as continuous pollution discharge), etc.

[0144] The judgment unit is also connected to the supplementary information acquisition unit. The second branch of the natural language understanding neural network model is used to output rectification suggestions based on the basic information of the enterprise and its surrounding environment, complaint information, risk level and incident supplementary information;

[0145] Among them, the input layer is used to obtain primary data, event supplementary information and risk level, and divides the primary data, event supplementary information and risk level into structured data and unstructured data according to the data structure;

[0146] The text encoder uses the BERT model to encode unstructured data and obtain the feature vector of unstructured data;

[0147] The structured data encoder includes a categorical variable branch and a continuous variable branch, the categorical variable branch is composed of an embedding layer, the continuous variable branch is composed of a fully connected layer, and the structured data encoder is used to encode structured data;

[0148] The structured data is divided into categorical variables and continuous variables according to the data type. The categorical variable branch is used to encode the categorical variables in the structured data to obtain the embedded vector of the categorical variable. The continuous variable branch is used to encode the continuous variables in the structured data to obtain the feature vector of the continuous variable.

[0149] The fusion layer is used to concatenate the output of the text encoder and the output of the structured data encoder in the feature dimension to obtain the completed data feature vector;

[0150] The decoder uses the T5 model to decode the feature vector of the completed data and output rectification suggestions.

[0151] The training process of the natural language understanding neural network model is as follows:

[0152] Label known complaint information, basic information and supplementary information of enterprises and service units, and corresponding rectification suggestions, mark the corresponding risk levels, and build a data set;

[0153] Preprocess the data in the dataset, including missing value filling, normalization, etc.;

[0154] The complaint information, basic information and supplementary information of enterprises and service units in the preprocessed data set are used as input, and the risk level and rectification suggestions are used as output for cyclic training until the loss function of the natural language understanding neural network model is minimized or the maximum number of iterations is reached.

[0155] Furthermore, the loss functions of the first branch and the second branch of the natural language neural network model both adopt the cross entropy loss function, and the loss function of the natural language neural network model is the weighted sum of the loss function of the first branch and the loss function of the second branch.

[0156] The completion report generation unit is connected to the complaint information acquisition unit and the judgment unit, and is used to obtain the response to the complaint event and the complainant's satisfaction evaluation through the park end or the third-party end, and generate a completion report based on the complaint information, rectification suggestions, the response to the complaint event and the complainant's satisfaction evaluation.

[0157] In this embodiment, the report generation module includes:

[0158] The data aggregation unit is used to receive access assessment reports, enterprise classification lists, evaluation reports and completion reports.

[0159] The report generation unit is connected to the data aggregation unit and is used to generate periodic analysis reports based on the access assessment report, enterprise classification list, evaluation report and completion report, and send them to the third-party end.

[0160] Specifically, the interim analysis report can be in the form of daily, weekly, quarterly and annual analysis reports, special analysis reports, environmental inspection analysis reports, etc. For example, the content of daily and weekly reports can include the situation of the proposed settled enterprises, inspection situation, complaint handling situation, etc. The content of the quarterly report can include the summary of daily and weekly reports, interim work summary and environmental management evaluation results, etc. The content of the annual report can include a basic summary of the park, the work of environmental stewards, the implementation of environmental laws and regulations, environmental benefits, improvement measures and work plans for the next year, etc. Among them, the improvement measures are based on the environmental protection goals that need to be achieved at each stage of the industrial park, and put forward short-term, medium-term and long-term rectification opinions from the aspects of enterprise access and site selection, environmental credit evaluation, steward assessment, complaint handling, etc.

[0161] The present invention sets two branches in a natural language understanding neural network model to perform different task learning respectively, and the two branches share feature extraction layers such as a text encoder and a structured data encoder. Through joint training, the model can share information between multiple tasks, thereby improving the performance of each task, and reducing the neural network layers required for the model structure, reducing the number of parameters, reducing computing overhead and memory usage, and improving model training efficiency. Since the two tasks are trained in the same model, they share information, reducing the problem of inconsistency in the outputs of different tasks, thereby improving the reliability and accuracy of the rectification suggestions finally outputted.

[0162] The present invention carries out pollution source control in the industrial park from the enterprise access and site selection stage, to the enterprise environmental credit rating and third-party steward assessment, implements the supervision of the pollution process in the industrial park, and then to the complaint handling, results management and improvement suggestions after the pollution occurs, promotes the end-of-pipe treatment of pollution in the industrial park, and realizes a multi-dimensional management closed loop of the entire process of environmental protection on the scale of the industrial park, which is conducive to improving the overall environmental protection management efficiency.

[0163] The present invention enables enterprises to evaluate the work of third-party service units through the assessment module, and the industrial park supervises the process of the work of the third-party service units, providing an objective basis for the industrial park to evaluate the service level of the third-party service units and promoting further standardization of the services of the third-party service units.

[0164] The present invention reduces the reliance of existing park environmental management system functions on hardware monitoring, applies neural network algorithm models to enterprise access and site selection auxiliary analysis, enterprise information database image and text recognition, and complaint information analysis, effectively reducing enterprise environmental management costs and improving the park's environmental management level; through network data capture, environmental supervision cases and environmental policy interpretation information push are realized, breaking the past passive environmental management and negative environmental response of enterprises, broadening the channels for enterprises to participate in environmental management, and improving environmental protection awareness and enthusiasm; relying on neural network algorithm models and network data capture to realize the application of each module function, taking the park end as the management core, strengthening the interactive feedback between industrial parks and enterprises, industrial parks and third-party service units, and third-party service units and enterprises, which is conducive to coordinating the synchronization of environmental information and management of the three parties, and further improving the overall environmental management efficiency.

Claims

1. A tripartite environmental management system suitable for industrial parks, characterized in that: include: The access site selection module is used to match and analyze the basic information of the enterprise with the regulations of the industrial park and generate an access assessment report; wherein the enterprise includes the enterprise to be settled in or the enterprise already settled in; The environmental credit evaluation module is used to evaluate and grade the environmental management status of settled enterprises based on their basic information, generate a graded list of enterprises, and push information to settled enterprises with different ratings based on the graded list of enterprises; The housekeeper assessment module is used to formulate inspection work plans, record inspection results, and conduct assessment and evaluation of third-party service units based on the inspection results and generate evaluation reports; The complaint handling module is used to receive complaint information, generate rectification suggestions based on the complaint information, generate a complaint event response after the complaint problem is resolved, and generate a completion report after the complainant's satisfaction evaluation; The complaint handling module includes: A judgment unit, wherein the judgment unit includes a natural language understanding neural network model, wherein the natural language understanding neural network model includes a first branch and a second branch, wherein the first branch and the second branch share the same input layer, text encoder, structured data encoder, and fusion layer, wherein the first branch further includes a fully connected layer and an output layer, and the second branch further includes a decoder; The first branch of the natural language understanding neural network model is used to determine the risk level of the complaint information based on the basic information of the enterprise and its surrounding environment and the complaint information; the second branch of the natural language understanding neural network model is used to output rectification suggestions based on the basic information of the enterprise and its surrounding environment, the complaint information, the risk level and the supplementary information of the event; The input layer is used to obtain data and divide the data into structured data and unstructured data according to the data structure; The text encoder adopts a BERT model to encode the unstructured data to obtain a feature vector of the unstructured data; The structured data encoder comprises a categorical variable branch and a continuous variable branch, the categorical variable branch is composed of an embedding layer, the continuous variable branch is composed of a fully connected layer, and the structured data encoder is used to encode the structured data; The structured data is divided into categorical variables and continuous variables according to data types, the categorical variable branch is used to encode the categorical variables in the structured data to obtain an embedded vector of the categorical variable, and the continuous variable branch is used to encode the continuous variables in the structured data to obtain a feature vector of the continuous variable; Among them, the calculation formula of the embedding vector of the categorical variable is as follows: and multi =[E1(x1),E2(x2),...,E h (x h )]; Among them, e multi represents the embedding vector of the categorical variable, x h represents the hth categorical variable, E h represents the embedding matrix corresponding to the h-th categorical variable; The formula for calculating the eigenvector of a continuous variable is as follows: y multi =[σ(W1q1+b),σ(W2q2+b),...,σ(W v q v +b)]; Among them, y multi Characteristic vector representing continuous variables, q v represents the vth continuous variable, W v represents the weight matrix corresponding to the vth continuous variable, b represents the bias vector, and σ represents the activation function; The fusion layer is used to splice the output of the text encoder and the output of the structured data encoder in the feature dimension to obtain a final data feature vector or a completed data feature vector; The fully connected layer is used to obtain the probability that the complaint information belongs to each risk level according to the final data feature vector; The output layer is used to output the risk level with the highest probability as the risk level of the complaint information; The decoder adopts a T5 model to decode the feature vector of the completed data and output rectification suggestions; The report generation module is connected with the access site selection module, the environmental credit evaluation module, the steward assessment module and the complaint handling module, and is used to summarize the results of each module to generate a phased analysis report.

2. According to claim 1, a three-party environmental management system suitable for industrial parks is characterized in that: Also includes: Enterprise side, third-party side and park side; The enterprise end is connected to the access site selection module, the environmental credit evaluation module, and the steward assessment module. The enterprise logs in through the enterprise end of the management system and performs corresponding operations; The third-party terminal is connected to the access site selection module, the environmental credit evaluation module, the housekeeper assessment module, the complaint handling module, and the report generation module. The third-party service unit logs in through the third-party terminal of the management system and performs corresponding operations; The park end is connected to the access site selection module, the environmental credit evaluation module, the housekeeper assessment module, the complaint handling module, and the report generation module. The industrial park logs in through the park end of the management system and performs corresponding operations.

3. According to claim 2, a three-party environmental management system suitable for industrial parks is characterized in that: The access site selection module includes: A site selection data acquisition unit, used to acquire the project type, industry type of the proposed project and supplementary information of the enterprise through the enterprise terminal, and acquire the regulations of the industrial park through the system database; An access assessment unit connected to the site selection data acquisition unit, wherein the access assessment unit includes a feedforward neural network model; The feedforward neural network model is used to compare the project type of the enterprise and the industry type of the planned project with the regulations of the industrial park, and output the access determination result of the enterprise; If the result of the enterprise's access determination is no, the determination process ends; if the result of the enterprise's access determination is yes, the site selection plot is screened out according to the industry type of the enterprise's proposed project; The feedforward neural network model is also used to compare the supplementary information of the enterprise with the regulations of the industrial park, and output a site selection reference suggestion and an access assessment report based on the site selection plot.

4. According to claim 2, a three-party environmental management system suitable for industrial parks is characterized in that: The environmental credit evaluation module includes: An evaluation data acquisition unit, used to acquire basic information of settled enterprises through the third-party terminal, and determine the impact factor according to the basic information of settled enterprises; A calculation unit, connected to the evaluation data acquisition unit, for calculating the environmental credit score of the settled enterprise according to the influencing factors and preset weights; A grading unit, connected to the calculation unit, for comparing the environmental credit scores of settled enterprises with preset scoring levels, determining the evaluation grades of settled enterprises, and generating an enterprise grading list; The push unit is connected to the evaluation data acquisition unit and the grading unit, and is used to push information to the settled enterprises with different ratings through the enterprise terminal according to the enterprise grading list and the basic information of the settled enterprises.

5. According to claim 2, a three-party environmental management system suitable for industrial parks is characterized in that: The housekeeper assessment module includes: The inspection plan unit is used to obtain the inspection period, inspection frequency and inspection requirements through the park end, generate an inspection work plan, obtain the number of inspection teams, inspection team members, historical inspection conditions and enterprise classification list through the third-party end, and perform intelligent scheduling according to the inspection work plan; An inspection result unit is connected to the inspection plan unit, and is used to obtain basic information, environmental protection procedures, pollution prevention and control facilities, environmental management, on-site inspection forms and on-site photos of settled enterprises through the third-party terminal, and form inspection results; wherein the inspection results are one file for each enterprise; An assessment and evaluation unit is connected to the inspection result unit, and is used to determine the completion rate, accuracy rate and rectification rate of the inspection results based on the inspection results, obtain the satisfaction evaluation of the settled enterprises on the third-party service unit through the enterprise side, and generate an evaluation report based on the completion rate, accuracy rate, rectification rate and satisfaction evaluation of the inspection results.

6. A three-party environmental management system suitable for industrial parks according to claim 2, characterized in that: The complaint handling module also includes: A complaint information acquisition unit, connected to the judgment unit, is used to acquire complaint information through the third-party terminal and the park terminal, and retrieve basic information of enterprises and surrounding environment within a preset range of the complaint point according to the complaint information; a comparison unit connected to the complaint information acquisition unit, and used to compare the complaint information with previous complaint information, and if the previous complaint information contains a complaint event of the same type as the complaint information, determine a rectification suggestion for the complaint information with reference to the previous complaint information; The task sorting unit is connected to the comparison unit and the judgment unit, and is used to sort the complaint information from high to low according to the risk level, and send the sorting result to the park end.

7. A three-party environmental management system suitable for industrial parks according to claim 6, characterized in that: The complaint handling module also includes: A supplementary information acquisition unit, connected to the judgment unit, and configured to acquire event supplementary information through the park terminal or the third-party terminal; A completion report generating unit is connected to the complaint information acquiring unit and the judging unit, and is used to obtain a response to the complaint event and a satisfaction evaluation of the complainant through the park end or the third-party end, and to generate a completion report based on the complaint information, the rectification suggestions, the response to the complaint event and the satisfaction evaluation of the complainant.

8. The three-party environmental management system applicable to industrial parks according to claim 2 is characterized in that: The report generation module includes: A data aggregation unit, used to receive the access assessment report, the enterprise classification list, the evaluation report and the completion report; A report generation unit is connected to the data aggregation unit and is used to generate a phased analysis report based on the access assessment report, the enterprise classification list, the evaluation report and the completion report, and send it to the third-party terminal.

Images