A method and system for evaluating and managing green financial products

Abstract

The present application discloses a method and system for evaluating and managing green financial products. The method and system utilize blockchain technology to reconstruct the credit system for evaluating and managing green financial products, and utilize big data and artificial intelligence technologies to realize risk prediction and real-time monitoring of green financial products. The inefficient, high-cost, and morally hazard-containing manual management method of green financial products is rebuilt into a consensus management system that automatically calls upon artificial intelligence analysis results and in which all stakeholders (corresponding to the various structural nodes in this application) participate. This improves the management efficiency of green financial products and reduces management costs, and establishes a more reliable green financial product management method and credit system, thereby ensuring that green finance better serves the needs of carbon neutrality and carbon peak.

Description

Technical Field

[0001] The present application relates to the field of financial information management, and in particular to a method and system for evaluating and managing green financial products. Background Art

[0002] Green finance aims to promote the green transformation of the financial system and encourage more funds to be invested in green industry projects. In particular, for the current decision-making and deployment with carbon peak and carbon neutrality as the goals, if we want to achieve environmental goals with carbon peak and carbon neutrality as the core, fiscal funds alone are far from enough. We also need to make full use of green financial products and mobilize social capital to invest in low-carbon and other green industries to promote the development of green industries and serve the construction of ecological civilization and ecological goals such as carbon peak and carbon neutrality.

[0003] However, how to ensure that all raised funds are used for designated green industry projects, whether relevant green industry projects comply with relevant laws and regulations and green standards, whether environmental, social and corporate governance (ESG) risks can be effectively managed and avoided during the implementation of green industry projects, and whether green industry projects have truly generated environmental benefits such as carbon emission reduction, etc., are issues that need to be addressed in the management of green financial products. In particular, after the decision-making and deployment of environmental goals related to carbon peak and carbon neutrality, new requirements have been further put forward for the high-quality development of green finance. Whether a project supported by a green financial product can achieve carbon emission reduction effects should be effectively measured, monitored and reported.

[0004] Traditional green financial products are often not managed or poorly managed due to the high cost and low efficiency of manual management, making it difficult to ensure management quality. In addition, manual management methods cannot avoid moral risks and technical errors, resulting in the failure to establish a green financial credit system. Therefore, building a green financial credit system to achieve efficient and low-cost green financial product management has become a technical problem that needs to be solved in this field. Summary of the invention

[0005] To this end, this application discloses a method and system for evaluating and managing green financial products, which is used to use blockchain technology to build an efficient and low-cost green financial credit system, and combine artificial intelligence and big data technology to achieve real-time, online and efficient management of green financial products.

[0006] The specific technical solutions are as follows:

[0007] A method for evaluating and managing green financial products, applied to a system for evaluating and managing green financial products, the system comprising at least: a plurality of institutional nodes corresponding to a plurality of participating institutions, an application service layer, a blockchain service layer, and a risk prediction service; the method comprises:

[0008] The institutional node submits its own institution’s business data and evidence data on financial products to the system;

[0009] The risk prediction service performs risk prediction on the financial product based on at least artificial intelligence processing for the submission event of the institutional node, and feeds back the risk prediction result to the institutional node;

[0010] The institutional node enters corresponding risk assessment data of the financial product into the blockchain corresponding to the blockchain service layer based on the risk prediction result;

[0011] The blockchain service layer processes the business data and the evidence data on the blockchain, and triggers consensus processing among participating institutions based on smart contracts after the data are uploaded to the blockchain;

[0012] If a consensus is reached, the blockchain service layer will certify the financial product as a green financial product; if no consensus is reached, the financial product will be certified as a non-green financial product; and the green certification result will be uploaded to the blockchain;

[0013] The application service layer generates report information based on the on-chain data and the green certification results.

[0014] Optionally, the risk forecast includes at least one of an environmental, social and corporate governance (ESG) risk forecast, an environmental benefit risk forecast, an extreme climate and meteorological event risk forecast, and a transition risk forecast;

[0015] The risk prediction service performs ESG risk prediction on the financial product based on at least artificial intelligence processing for the submission event of the institutional node, including:

[0016] For each forecast entity corresponding to the financial product, by counting historical penalty information of each forecast entity, predicting the penalty situation of the forecast entity for each penalty type in the penalty type set in a future predetermined time period, and obtaining a first penalty prediction sub-result corresponding to each penalty type of the forecast entity;

[0017] Using the constructed ESG risk prediction artificial intelligence model, predict the penalty situation of each prediction entity for each penalty type in the penalty type set in a future predetermined time period, and obtain the second penalty prediction sub-result of the prediction entity corresponding to each penalty type;

[0018] The penalty prediction result corresponding to each penalty type of the prediction entity is determined according to the first penalty prediction sub-result and the second penalty prediction sub-result corresponding to each penalty type of the prediction entity.

[0019] Optionally, in the above method:

[0020] Predicting the environmental benefits of the financial product based on a pre-built artificial intelligence model for environmental benefits prediction; the artificial intelligence model for environmental benefits prediction is implemented using a polynomial regression method;

[0021] Based on a pre-built artificial intelligence model for predicting extreme climate and meteorological event risks, predict the extreme climate and meteorological event risks of the financial product; the artificial intelligence model for predicting extreme climate and meteorological event risks adopts a logistic regression method and is implemented by converting into multiple binary classification problems;

[0022] Based on a pre-built transformation risk prediction artificial intelligence model, the transformation risk of the financial product is predicted; the transformation risk prediction artificial intelligence model is implemented using a polynomial regression method.

[0023] Optionally, the authentication of financial products based on the consensus mechanism includes at least one of the following aspects:

[0024] Intelligent judgment of green attributes, intelligent assessment of environmental benefits and climate risk prediction;

[0025] in:

[0026] Intelligent judgments on the green attributes of financial products include:

[0027] Obtaining the project information of the financial product uploaded by the corresponding institution node; identifying the project content in the project information; determining whether there is an entry matching the project content in the predetermined green certification standard; if so, determining that the financial product has green attributes;

[0028] Intelligent assessment of the environmental benefits of financial products includes:

[0029] The environmental benefits of the projects corresponding to the financial products are calculated using a multi-dimensional data intelligent model; the process of calculating the environmental benefits of the projects using the multi-dimensional data intelligent model includes: when the corresponding institutional node cannot provide the parameters required for the environmental benefit calculation under the traditional model, the environmental benefits of the projects are calculated using the following alternative calculation formula:

[0030] P CO2 =W×t×α

[0031] Among them, P CO2 represents the amount of carbon dioxide emission reduction, W represents the installed capacity, t represents the number of available hours of wind power generation in the project area, and α represents the electricity emission factor applicable to the project area; the system obtains W by identifying and extracting the project information of the financial products uploaded by the corresponding institution node, and obtains t and α by identifying and extracting the project information of the financial products uploaded by the corresponding institution node, and matching the extracted results with the pre-built corresponding database;

[0032] Climate risk forecasts for financial products include:

[0033] For the risk of extreme climate and meteorological events: identify and extract the project information of financial products uploaded by the corresponding institutional nodes to obtain the region where the project is located; determine the various types of meteorological events corresponding to the region where the project is located; quantify the sensitivity of the project to different types of meteorological events in the region, and obtain the climate risk index of the project;

[0034] Regarding transition risks: calculate the carbon costs that may be incurred by a project or enterprise; assess transition risks based on the calculated carbon costs.

[0035] Optionally, the forecast entity corresponding to the financial product includes at least one of the following: an enterprise associated with the financial product and requiring risk forecasting, an industry to which the enterprise belongs, and a region where the enterprise is located;

[0036] The penalty types in the penalty type concentration include at least one of the following: fine; temporary seizure or revocation of license; confiscation of illegal gains; administrative detention; order to suspend production and rectify; order to suspend production, business or close down; and other penalties.

[0037] Among them, if the penalty type set includes multiple penalty types, the ESG risk prediction artificial intelligence model includes multiple binary classification models corresponding to various penalty types, and each binary classification model is used to determine the probability of penalty of the predicted entity in the corresponding penalty type in a predetermined time period in the future.

[0038] Optionally, determining the penalty prediction result corresponding to each penalty type of the prediction entity according to the first penalty prediction sub-result and the second penalty prediction sub-result corresponding to each penalty type of the prediction entity includes:

[0039] According to the weights respectively configured for the statistical prediction method and the model prediction method, a weighted operation is performed on the first penalty prediction sub-result and the second penalty prediction sub-result corresponding to each penalty type of the prediction entity to obtain the penalty prediction result corresponding to each penalty type of the prediction entity.

[0040] Optionally, the system further includes a storage service;

[0041] The blockchain service layer processes the business data and the evidence data on the blockchain, including:

[0042] The blockchain service layer uploads the business data to the blockchain;

[0043] The blockchain service layer performs hash operation on the evidence data, uploads the obtained hash result to the blockchain, and stores the evidence data through the storage service.

[0044] Optionally, the application service layer generates report information based on the on-chain data and the green certification result, including:

[0045] The application service layer generates an MRV report including core business data and green certification results.

[0046] Optionally, before generating the MRV report, the method further includes:

[0047] After the green financial products are issued, MRV tracking and management shall be carried out on the green financial products during their life cycle.

[0048] Optionally, the above method further includes:

[0049] After the issuance of the financial product is completed, each institutional node shall submit or modify the business data and evidence data corresponding to its own institution in each duration of the financial product as required;

[0050] The blockchain service layer monitors nodes according to the term of the financial product. When monitoring that the institutional node submits or modifies business data and / or evidence data, the submitted or modified business data and / or evidence data are uploaded to the blockchain, and after being uploaded to the blockchain, the consensus processing step of each participating institution based on the smart contract is triggered to re-certify whether the financial product is green or not;

[0051] Among them, the system generates report information based on the corresponding on-chain data and green certification results in each life period of the financial product, and the system can maintain all data and consensus results corresponding to the entire life cycle of the financial product.

[0052] Optionally, the system assigns a corresponding role to each institutional node by establishing a role account system, and adds corresponding system permissions and blockchain permissions to each role;

[0053] Based on its corresponding role, each institutional node submits or modifies data within its authority to the system, and confirms data within its authority during consensus processing.

[0054] Optionally, the blockchain is a consortium chain.

[0055] A system for evaluating and managing green financial products, comprising: multiple institutional nodes corresponding to multiple participating institutions, an application service layer, a blockchain service layer, and a risk prediction service; wherein:

[0056] The institution node is used to submit the business data and evidence data of the institution itself on the financial products to the system;

[0057] The risk prediction service is used to predict the risk of the financial product based on at least artificial intelligence processing for the submission event of the institutional node, and to feed back the risk prediction result to the institutional node;

[0058] The institutional node is further used to input corresponding risk assessment data of the financial product into the blockchain corresponding to the blockchain service layer based on the risk prediction result;

[0059] The blockchain service layer is used to: process the business data and the evidence data on the blockchain, and trigger consensus processing among participating institutions based on smart contracts after the data are uploaded; if a consensus is reached, the financial product is certified as a green financial product; if no consensus is reached, the financial product is certified as a non-green financial product; and the green certification result is uploaded to the blockchain;

[0060] The application service layer is used to generate report information based on the uplink data and the authentication result.

[0061] Optionally, the system further includes a storage service;

[0062] When the blockchain service layer processes the business data and the evidence data on the blockchain, it is specifically used to:

[0063] The business data is put on the blockchain; the evidence data is processed by hash operation, the obtained hash result is put on the blockchain, and the evidence data is stored through the storage service, wherein all the on-chain data can be traced.

[0064] Optionally, the application service layer is further used to provide at least one of the following functions:

[0065] Data collection, user management, bond management, authority management and back-end system management;

[0066] The blockchain service layer is used to provide at least one of the following functions:

[0067] Node management, evidence management, security management, contract management and node monitoring.

[0068] Optionally, the risk prediction service is an independent application system module that can be dynamically plugged in and out of the system.

[0069] According to the above scheme, the method and system for evaluating and managing green financial products disclosed in this application use blockchain technology to reconstruct the credit system for evaluating and managing green financial products, and at the same time use big data and artificial intelligence technology to realize risk prediction and real-time monitoring of green financial products, and rebuild the inefficient, high-cost, and morally hazard-containing manual management method of green financial products into a consensus management system that automatically calls the results of artificial intelligence analysis and participates in all stakeholders (corresponding to the various structural nodes in this application), thereby improving the management efficiency of green financial products and reducing management costs, and establishing a more reliable green financial product management method and credit system, thereby ensuring that green finance better serves the needs of carbon neutrality and carbon peak. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0071] Figure 1 This is a system architecture diagram of a system for evaluating and managing green financial products provided in an embodiment of the present application;

[0072] Figure 2 It is an optional deployment method of the system for evaluating and managing green financial products provided in the embodiment of the present application;

[0073] Figure 3 This is a flow chart of a method for evaluating and managing green financial products provided in an embodiment of the present application;

[0074] Figure 4 It is a data business flow chart of financial product management provided by an embodiment of the present application;

[0075] Figure 5 This is another flow chart of the method for evaluating and managing green financial products provided in the embodiment of the present application;

[0076] Figure 6 It is a structural diagram of the network structure of MLP;

[0077] Figure 7 This is a schematic diagram of the input and output of a neuron;

[0078] Figure 8 It is a common activation function graph. DETAILED DESCRIPTION

[0079] For the sake of reference and clarity, the relevant terms / technical terms used below are summarized and explained as follows:

[0080] Green financial products: including green bonds, green credit, green funds, green trusts, impact investments, etc.;

[0081] Blockchain: From a technological perspective, blockchain involves many scientific and technological issues such as mathematics, cryptography, the Internet, and computer programming; from an application perspective, in simple terms, blockchain is a distributed shared ledger and database with the characteristics of decentralization, immutability, full traceability, traceability, collective maintenance, and openness and transparency. These characteristics ensure the "honesty" and "transparency" of blockchain and lay the foundation for blockchain to create trust. The rich application scenarios of blockchain are basically based on the ability of blockchain to solve the problem of information asymmetry and achieve collaborative trust and concerted action among multiple entities.

[0082] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0083] Traditional green financial products are often not managed or poorly managed due to the high cost and low efficiency of manual management, making it difficult to ensure management quality. In addition, manual management cannot avoid moral risks and technical errors, resulting in the failure to establish a green financial credit system. The inventors have found that the traditional green financial product management method has at least the following defects:

[0084] 1) Insufficient information collaboration

[0085] In the traditional model, the parties involved in green financial products mainly rely on offline communication to exchange information, which is inefficient and timeless, and is prone to information transmission errors and loss. There is no effective response to changes in information from a certain party, and there is a lack of cross-verification, which may lead to the situation where funds are not invested in green industry projects, or there are violations in the projects or errors in the calculation of the environmental benefits of the projects. Such an inefficient model makes it difficult to avoid risks and cannot form a high-quality credit system.

[0086] 2) Information tracing is difficult and accountability is difficult

[0087] The issuance and management of a green financial product requires the participation of many stakeholders, and each stakeholder often continuously updates the information provided. The previous and subsequent versions are sometimes very different, and if risks occur, it is difficult to hold anyone accountable.

[0088] 3) Difficult to adapt to long-term management needs

[0089] Green financial products often have a long life cycle, usually 1-5 years, and require continuous tracking, management, and disclosure. Traditional manual management methods often make it difficult to ensure the continuity of data for continuous tracking. If personnel adjustments occur, it is likely to cause project management failure, ultimately leading to the failure of green finance goals.

[0090] 4) ESG risks are not effectively managed

[0091] Returns and risks are two sides of financial products, and risk monitoring and assessment are also very important. Taking project environmental risk as an example, if a green industry project invested by a green financial product has environmental violations, it may not achieve the "green" effect, but instead contribute to environmental violations. On the other hand, environmental violations may cause financial institutions that provide green financial products to be punished by regulatory authorities. The main reason for this problem is that the number of financial institutions is limited, and they do not have professional environmental backgrounds. It is difficult to query green industry projects and related corporate environmental violations, and it is even more impossible to predict environmental risks.

[0092] 5) MRV is inefficient

[0093] MRV (Monitoring-Reporting-Verification) includes the MRV of the environmental benefits of green financial products and the MRV of ESG risks. Under the traditional manual model, due to high labor costs and poor efficiency, tracking management is often carried out once a year, but the implementation and risk status of the project cannot be managed throughout the year, which brings great management risks to green financial products. In particular, if ESG risks cannot be identified in a timely manner, the relevant financial institutions may be punished.

[0094] In order to solve at least some of the above-mentioned defects of the traditional management method, this application proposes a method and system for evaluating and managing green financial products. The method for evaluating and managing green financial products disclosed in this application is specifically a method for evaluating and managing green financial products based on blockchain, big data and artificial intelligence technology, which can be applied to the system for evaluating and managing green financial products disclosed in this application.

[0095] The following first describes the system to which the method of the present application is applicable, namely, the system for evaluating and managing green financial products disclosed in the present application.

[0096] The system includes multiple institutional nodes corresponding to multiple participating institutions, application service layer, blockchain service layer and risk prediction service; each of the above components has the following functions:

[0097] The above-mentioned institutional nodes are used to submit the business data and evidence data of their own institutions on financial products to the system;

[0098] The above-mentioned risk prediction service is used to predict the risk of financial products based on at least artificial intelligence processing for the submission event of the institution's node, and to feedback the risk prediction results to the institution's node;

[0099] The above-mentioned institutional node is also used to input the corresponding risk assessment data of the financial product into the blockchain corresponding to the blockchain service layer based on the risk prediction results;

[0100] The above-mentioned blockchain service layer is used to: process the submitted business data and evidence data on the blockchain, and trigger the consensus processing of the participating institutions based on the smart contract after the blockchain is uploaded; if a consensus is reached, the financial product is certified as a green financial product; if no consensus is reached, the financial product is certified as a non-green financial product; and the green certification results (certification results when consensus is reached and when no consensus is reached) are uploaded to the blockchain;

[0101] The application service layer is used to generate report information based on the uploaded data and the authentication result.

[0102] In other implementations of the embodiments of the present application, the green financial product evaluation and management system of the present application also includes a storage service, wherein the blockchain service layer is specifically used to:

[0103] Put the submitted business data on the blockchain;

[0104] The submitted evidence data is hashed, the obtained hash result is uploaded to the blockchain, and the evidence data is stored through the storage service.

[0105] Next, a specific implementation of the system for evaluating and managing green financial products disclosed in this application is further provided.

[0106] See also Figure 1 The system architecture of the green financial product evaluation and management system provided by the company mainly includes the terminal layer, application service layer and blockchain service layer from the perspective of technical framework. Figure 1 As shown, the system also includes storage services such as data storage and file storage, as well as risk prediction services based on artificial intelligence. Among them:

[0107] (I) Terminal layer

[0108] The terminal layer serves the institutional nodes of all participants in financial products, including but not limited to financial institutions, green industry project owners, law firms, accountants, local governments, regulatory agencies, green assessment agencies, certification agencies, etc.

[0109] (II) Application Service Layer and Blockchain Service Layer

[0110] The application service layer and the blockchain service layer constitute the consensus system of the green financial product evaluation and management system. The consensus system uses SaaS+BaaS to provide services to each participating node. Among them, SaaS (Software-as-a-Service) refers to providing software services through the network, and BaaS (Blockchain-as-a-Service) refers to embedding the blockchain framework into the cloud computing platform, using the deployment and management advantages of the cloud service infrastructure to provide developers with a convenient and high-performance blockchain ecological environment and ecological supporting services, and support developers' business expansion and operation support blockchain open platform.

[0111] Optionally, in the embodiment of the present application, SaaS is built using a microservice architecture and deployed on a private cloud or private server. It can be used to provide basic services such as data collection, bond management, data management, and report generation to institutional node users, and customized service development can be carried out according to actual user needs.

[0112] BaaS uses distributed storage, is based on blockchain services, and is deployed on private clouds or private servers. It can be used to provide node users with services such as data on-chain, smart contract consensus mechanism, node management, node monitoring, and can also provide customized development services.

[0113] Both SaaS and BaaS can independently provide services to green finance customers, i.e. the corresponding institutional nodes in this application. SaaS helps customers complete customized business logic development and complete data on-chain work. If the customer system already has a business logic application module and only needs to provide blockchain services such as data on-chain, BaaS can be used directly.

[0114] (III) Storage services such as data storage and file storage

[0115] Non-chain business data is stored in the database server or file server through this storage service. In actual applications, the data storage service can be expanded into a big data architecture according to the data scale business requirements.

[0116] (IV) Risk prediction services based on artificial intelligence

[0117] In the embodiments of the present application, preferably, the risk prediction service is specifically an ESG risk prediction service, which can be used to provide artificial intelligence risk prediction and basic data analysis services.

[0118] In implementation, preferably, the service can be implemented as a separate application system module independent of other components in the system to support dynamic plug-in and dynamic expansion of the risk prediction service in the system.

[0119] The risk prediction service conducts risk analysis, assessment and prediction of financial products based on the basic information of the financial products (such as the name of the issuing company, the geographical location of the issuing company, the industry to which it belongs, etc.).

[0120] See also Figure 2 When deploying the system, you can deploy a private cloud / private server on the client according to actual needs, or directly provide services to customers through a browser based on the public cloud.

[0121] It should be noted that the functional role and deployment of each component in the above-mentioned specific implementation of the system disclosed in this application are only illustrative and non-restrictive descriptions of the embodiments of this application. During implementation, based on the examples provided in this application, adaptive deformation / variant applications can be carried out in combination with actual needs or actual technical environment, which are all within the scope of protection of this application.

[0122] Next, based on the above-mentioned system for evaluating and managing green financial products, the implementation process of the method for evaluating and managing green financial products disclosed in this application will be described.

[0123] See also Figure 3 A flowchart of a method for evaluating and managing green financial products is provided, which specifically includes:

[0124] Step 301: The institutional node submits its own institutional business data and evidence data on financial products to the green financial product assessment and management system.

[0125] Specific, combined with reference Figure 4 Each institutional node can log in to the system through the system page and submit the institution’s own business data and evidence data for financial products to the system.

[0126] Among them, business data includes the data provided by the institution to which the institution node belongs in response to the corresponding business logic of the financial product, and evidence data includes some evidence / proof information required for the business (the business within the authority of the current institution in the entire business logic covered by the financial product), which can be any type of file form such as documents, pictures / photos, including but not limited to obtained land ownership certificates, relevant certification documents issued by the government, legal permission documents, etc.

[0127] During the issuance stage of financial products, the issuer's institutional node can submit its own business data and evidence data on financial products to the green financial product assessment and management system for use in risk prediction, green certification and issuance of financial products.

[0128] In this step 301, it mainly refers to the issuer's institutional node submitting relevant data to the green financial product evaluation and management system to realize the issuance of financial products.

[0129] Step 302: The risk prediction service performs risk prediction on financial products based on at least artificial intelligence processing for the submission event of the institution node, and feeds back the risk prediction results to the institution node.

[0130] Among them, for the event that the issuer's institutional node submits business data and evidence data to the system, the risk prediction service performs ESG risk prediction for this financial product based on the statistical processing and artificial intelligence processing of the historical ESG data of the prediction entity, and feeds back the ESG risk prediction results to the issuer's institutional node. At the same time, the risk prediction results are stored on the risk prediction service side for other business-related participating institutions to obtain. The statistical processing and artificial intelligence processing of the risk prediction service can specifically use the historical ESG data in the ESG database as the basis for data analysis and statistics, and as the sample basis for the AI ​​(Artificial Intelligence) risk analysis model that is trained and optimized through the update and iteration of ESG data.

[0131] Furthermore, in order to address the drawback that ESG risks are not effectively managed in traditional methods, this application implements real-time risk monitoring of green financial products by providing an "ESG risk prediction service" artificial intelligence engine and an ESG risk database, and predicts relevant ESG risks in advance by establishing a risk prediction model, which can be provided to relevant financial institutions for business reference and prevention of related risks.

[0132] When the risk prediction service performs risk prediction, the prediction entities targeted include the issuing enterprise (e.g., company), the industry to which the issuing enterprise belongs, and the region (e.g., city) where the issuing enterprise is located. That is, the risk prediction service performs ESG risk prediction for financial products based on these three dimensions. The specific implementation process of ESG risk prediction will be described in detail below with an embodiment.

[0133] In addition, optionally, in addition to the ESG risk prediction results, the relevant ESG historical data of the predicted entity can be fed back to the institutional node as needed, or the relevant ESG historical data of the predicted entity can be stored on the public service end provided by the risk prediction service, so as to better assist the participating institutions in making objective risk judgments on this financial product.

[0134] Step 303: The institution node enters corresponding risk assessment data into the blockchain corresponding to the blockchain service layer based on the risk prediction result.

[0135] The ESG risk prediction results fed back to the institutional nodes are used to assist the nodes in making objective risk judgments and entering data into the system. For example, assisting the nodes in entering into the system the objective judgment results on whether a financial product has risks or the approximate degree of risk.

[0136] Step 304: The blockchain service layer processes the submitted business data and evidence data on the blockchain, and after the data is uploaded to the blockchain, consensus is reached among the participating institutions based on the smart contract.

[0137] In the embodiment of the present application, the blockchain used is specifically a consortium chain. The system organizes each institution's nodes into groups on the consortium chain according to different financial products, and the nodes in each group are responsible for one financial product. The system can dynamically manage the grouping of each institution's nodes according to actual business needs.

[0138] This application adopts the alliance chain method to abandon the trustless interest relationship under the complete openness of the public chain, and establishes a strong trust relationship through the mechanism of sharing the ledger of specific nodes (such as green financial product issuers, banks and other financial institutions, law firms, green certification agencies and non-governmental organizations NGOs, etc.). All parties involved can perform their duties and verify and track the compliance of green financial products, the green attributes of fundraising projects, environmental benefits, ESG management, and information disclosure through visual shared data. Reduce information asymmetry from the system design level, so that business processes can be reconstructed based on shared ledgers, generate more voucher data and peer information through online data sharing, and streamline business processes and reduce central links on this basis to achieve overall efficiency improvement. More importantly, through the ESG database and ESG risk prediction artificial intelligence model, ESG risks can be monitored in real time and predicted in advance, which greatly improves the efficiency, reduces costs, and improves accuracy for the management of green financial products.

[0139] For the submitted business data and evidence data, the blockchain service layer processes them on the alliance chain.

[0140] Optionally, the consortium chain on-chain processing is specifically implemented as follows:

[0141] Put the submitted business data on the alliance chain;

[0142] The submitted evidence data is hashed and the hash result is uploaded to the alliance chain. The evidence data is stored and processed through the storage service of the green financial product evaluation and management system. For example, the evidence file can be saved in the file server.

[0143] After the data is uploaded to the chain, the smart contract will be executed, automatically triggering the consensus processing of all participating institutions based on the smart contract.

[0144] Step 305: If a consensus is reached, the blockchain service layer will certify the financial product as a green financial product.

[0145] Step 306: If no consensus is reached, the blockchain service layer will certify the financial product as a non-green financial product.

[0146] The consensus-based certification of financial products includes at least one of the following aspects: compliance judgment, intelligent judgment of green attributes, intelligent assessment of environmental benefits, fund management, ESG risk prediction and climate risk prediction; if the first to fourth certifications are all passed, the system will automatically certify the financial product as a green financial product, and the third, fifth and sixth certifications will serve as annotation items for recording and disclosing relevant information on the financial products, so as to provide a basis for regulatory management and investment decision-making by investment institutions.

[0147] in:

[0148] (I) Compliance judgment

[0149] Compliance judgment mainly refers to the judgment on the status of the approval and verification documents obtained for the project. The embodiment of this application is based on the documents uploaded by the relevant institution nodes to make compliance judgments on financial products.

[0150] 2. Intelligent judgment of green attributes

[0151] The judgment of the green attributes of financial products mainly refers to whether the projects corresponding to the financial products meet the relevant standards. Taking green bond certification as an example, the projects supported by green bonds should meet the requirements of the "Green Bond Support Project Catalogue".

[0152] The prior art generally uses manual methods to judge the green attributes of financial products. In order to overcome the shortcomings of low efficiency and poor accuracy of the manual method, the embodiments of the present application use intelligent methods to achieve automatic judgment of green attributes.

[0153] Specifically, the intelligent judgment of the green attributes of financial products in the embodiment of the present application includes:

[0154] 1) Obtain the project information of financial products uploaded by the corresponding institutional nodes;

[0155] 2) identifying the project content in the project information;

[0156] 3) Determine whether there is an item in the predetermined green certification standard that matches the project content; if so, determine that the financial product has green attributes.

[0157] The following examples illustrate:

[0158] The system can specifically obtain the project names of financial products uploaded by relevant nodes (including in compliance documents), such as "50MW wind power plant construction project in Zhuhai City, Guangdong Province", and perform semantic analysis on the project name, identifying "Zhuhai City, Guangdong Province" as the project location and "wind power plant construction" as the project content. It also compares the two keywords "wind power generation" and "construction" with the specific entries in the "Catalogue of Green Bond Supported Projects", and identifies the corresponding entry in the "Catalogue of Green Bond Supported Projects" as "3.2.2.1 Construction and operation of wind power generation facilities". Based on this, it is judged that the project meets the requirements of the "Catalogue of Green Bond Supported Projects" and can be identified as a green bond in terms of green attributes.

[0159] 3. Intelligent evaluation of environmental benefits

[0160] In green financial products, it is necessary to calculate the environmental benefits of the project, such as the carbon emission reduction of a wind power plant. The calculation principle is to multiply the online power generation of the wind power plant by the emission factor of the local power grid. That is, in the traditional mode, the two parameters of "online power generation of the wind power plant" and "local power grid emission factor" uploaded by the node need to be used to calculate the environmental benefits of the project.

[0161] However, the inventors have found through research that in actual applications, the data uploaded by the nodes is often not comprehensive enough, such as the lack of any or all of the above parameters, which may often lead to situations where environmental benefits cannot be calculated according to the traditional model.

[0162] In order to solve the problem of calculating the environmental benefits of the project in this case, the embodiment of the present application proposes to use the following alternative calculation formula to calculate the environmental benefits of the project:

[0163] P CO2 =W×t×α

[0164] Among them, P CO2 represents the amount of carbon dioxide emission reduction, W represents the installed capacity, t represents the number of available hours for wind power generation in the project area (e.g., the number of available hours for wind power generation per year in the project area), and α represents the electricity emission factor applicable to the project area; the system obtains W by identifying and extracting the project information of the financial products uploaded by the corresponding institution nodes, and obtains t and α by identifying and extracting the project information of the financial products uploaded by the corresponding institution nodes, and matching the extracted results with the pre-built corresponding database.

[0165] Specifically, in order to solve the problem of calculating the environmental benefits of the project in the above situation, the embodiment of the present application pre-creates a "regional power grid emission factor library" and a "regional renewable resource database" in the system, and uses a multi-dimensional data intelligent model for calculation based on the created database.

[0166] Still taking the aforementioned "50MW wind power plant construction project in Zhuhai City, Guangdong Province" as an example, assuming that the node cannot provide the project's annual online power consumption and the local power grid emission factor, the project's environmental benefits cannot be calculated in the traditional mode. However, the above-mentioned backup calculation formula provided in the embodiment of this application can still automatically calculate the environmental benefits based on the project name information.

[0167] The process of automatically calculating environmental benefits based on project name information using the above-mentioned alternative calculation formula is as follows:

[0168] Based on the project name, the system identified that the project was located in Zhuhai City, Guangdong Province, and then called the "Regional Power Grid Emission Factor Library". Based on the regional / regional data matching results, it identified that the applicable emission factor for Zhuhai City, Guangdong Province is 0.6894CO 2 / MWh. Based on the project name, the system identified the project content as "wind power plant construction". The system matched the keyword "wind power generation" and called the "wind power generation available hours database" in the "regional renewable resource database", and determined that the annual wind power generation available hours in Zhuhai City, Guangdong Province is 1841. When the system sets the wind power project to have unknown power generation, the project emission reduction calculation formula uses P CO2 =W×t×α (i.e., carbon dioxide emission reduction = installed capacity × available hours × electricity emission factor). Based on this, it is calculated that the annual carbon emission reduction of the 50MW wind power plant construction project in Zhuhai City, Guangdong Province is 63459.27tCO 2 e.

[0169] The embodiment of the present application provides a methodological model for calculating environmental benefits through the above-mentioned functional design, which can make the calculation of environmental benefits automatic or semi-automatic, greatly reduce the requirements for the integrity of data uploaded by nodes, greatly improve the efficiency of environmental benefit calculation, and realize the automation and intelligence of environmental benefit calculation.

[0170] Subsequently, during the life of green finance products, the environmental benefits of green finance products will continue to be verified and managed as a quantitative indicator. If, after a later review, the carbon emission reductions generated by green finance products do not reach the carbon emission reductions promised and confirmed during the previous certification, corresponding penalties will be imposed according to the strategy, such as reducing the interest rate of green finance products. Otherwise, if the target is reached, the interest rate of green finance products will remain unchanged.

[0171] In addition, in an embodiment of the present application, the system also provides an environmental benefit risk prediction AI model to assist relevant institutional nodes in completing green certification of environmental benefits. Unlike the environmental benefit assessment mentioned above, the assessment is specifically implemented in the early stage of the project based on the uploaded project content, while the risk prediction of environmental benefits based on the AI ​​model is specifically based on the actual operation process of the project. For details, please refer to the relevant description of AI risk prediction of environmental benefits below.

[0172] (IV) Fund Management

[0173] The protection core that does not belong to the embodiments of the present application will not be described in detail.

[0174] (V) ESG risk prediction

[0175] For details, please refer to the relevant treatment of ESG risk assessment and prediction below.

[0176] 6. Climate risk prediction

[0177] Climate risks include risks of extreme climate and meteorological events and transition risks.

[0178] The risks of extreme climate and meteorological events mainly refer to the risks such as sea level rise and extreme meteorological events that may be caused by global climate change; the transition risks mainly refer to the risks brought about by enterprises or projects adapting to policy and market transitions under national and international climate change issues such as carbon peak and carbon neutrality.

[0179] In the embodiment of the present application, the forecast processing of the climate of the financial product includes:

[0180] Among them, the prediction of the risk of extreme climate and meteorological events:

[0181] 1) Identify and extract the project information of the financial products uploaded by the corresponding institutional nodes to obtain the region where the project is located;

[0182] 2) Determine the types of meteorological events corresponding to the project area;

[0183] 3) By quantifying the sensitivity of the project to different types of meteorological events in the region, the climate risk index of the project is obtained.

[0184] Specifically, the embodiment of the present application performs semantic recognition on the project name of the financial product uploaded by the corresponding node, determines the region / area where the project is located, and matches the determined region / area where the project is located with the pre-built "climate risk database" to identify possible meteorological events such as typhoons, droughts, saltwater tides, freezing, extreme high temperatures, extreme precipitation, and possible risks of sea level rise in the project location. On this basis, the system automatically gives a climate risk index for the project by further quantifying the sensitivity of different types of projects to different meteorological events, which is used as a reference for decision-making on investment and financing projects.

[0185] In addition, in response to the risks of extreme climate and meteorological events, the system also provides an AI model for risk prediction of extreme climate and meteorological events to assist relevant institutional nodes in completing green certification of extreme climate and meteorological event risks. Similarly, the prediction method for the risks of extreme climate and meteorological events provided above is mainly to realize the risk prediction of extreme climate and meteorological events based on the project content uploaded by the node in the early stage of the project, while the risk prediction of extreme climate and meteorological events based on the AI ​​model is specifically based on the actual operation process of the project. For details, please refer to the relevant description of AI prediction of the risk of extreme climate and meteorological events below.

[0186] Prediction of transformation risks:

[0187] 1) Calculate the carbon costs that a project or enterprise may incur;

[0188] 2) Assess transition risks based on calculated carbon costs.

[0189] In response to the transition risk, the embodiment of the present application pre-builds a "carbon price database" by real-time monitoring of carbon prices in multiple regional carbon trading markets and multiple carbon trading pilot areas, and making trend forecasts. When judging the transition risk of an enterprise or project, the constructed database is used as a basis to calculate the carbon cost that the enterprise may incur through the enterprise's carbon emissions and the local carbon price, thereby evaluating the relevant financial risks, etc., and achieving the purpose of evaluating the transition risk accordingly.

[0190] In addition, for transformation risks, the system also provides a risk prediction AI model for transformation risks to assist relevant institutional nodes in completing green certification of transformation risks. Similarly, the prediction method for transformation risks provided above is mainly based on the project content uploaded by the node in the early stage of the project (such as extracting the project location from the uploaded project name, etc.) to achieve risk prediction of transformation risks, while the risk prediction of transformation risks based on the AI ​​model is specifically based on the actual operation process of the project. For details, please refer to the relevant description of risk AI prediction of transformation risks below.

[0191] Step 307: The blockchain service layer uploads the green certification result to the blockchain.

[0192] If the logic of the smart contract is met, that is, a consensus is reached, the financial product is certified as a green financial product, and the certification result is uploaded to the alliance chain; conversely, if it does not meet the logic of the smart contract, it indicates that no consensus has been reached, and the financial product is accordingly identified as a non-green financial product, and the result is also uploaded to the alliance chain.

[0193] Step 308: The application service layer generates report information based on the on-chain data and green certification results.

[0194] After the smart contract runs, the application service layer will automatically generate an MRV report including core business data and green certification results, etc., to explain the core business data and green certification results, etc.

[0195] Among them, the generation of the MRV report is based on the MRV tracking and management of green financial products. Therefore, in the embodiment of the present application, after the green financial product is issued, the MRV tracking and management of the green financial product will continue during each life period of the green financial product.

[0196] According to the above scheme, the method and system for evaluating and managing green financial products disclosed in this application use blockchain technology to reconstruct the credit system for evaluating and managing green financial products, and at the same time use big data and artificial intelligence technology to realize risk prediction and real-time monitoring of green financial products, and rebuild the inefficient, high-cost, and morally hazard-containing manual management method of green financial products into a consensus management system that automatically calls the results of artificial intelligence analysis and participates in all stakeholders (corresponding to the various structural nodes in this application), thereby improving the management efficiency of green financial products and reducing management costs, and establishing a more reliable green financial product management method and credit system, thereby ensuring that green finance better serves the needs of carbon neutrality and carbon peak.

[0197] In an optional embodiment of the present application, if Figure 5 The flowchart of the method for evaluating and managing green financial products shown in the figure, the method for evaluating and managing green financial products disclosed in this application also includes the following processing:

[0198] Step 309: After the issuance of the financial product is completed, each institutional node shall submit or modify the business data and evidence data corresponding to its own institution during each life cycle of the financial product as required.

[0199] The issuance and management of a green financial product requires the participation of many stakeholders, and each stakeholder often continuously updates the information provided. Therefore, after the issuance of a financial product is completed, each institutional node can submit or modify the business data and evidence data corresponding to its own institution during each life cycle of the financial product as required.

[0200] In order to solve the problem of insufficient information collaboration among the participating parties, the embodiment of the present application establishes a role account system in the system through corresponding technologies, and assigns corresponding roles to each stakeholder (institutional node) based on the role account system, and adds corresponding system permissions and blockchain permissions to each role. On this basis, each institutional node confirms the information that should be confirmed within its authority based on its corresponding role (such as the relevant consensus confirmation involved in consensus processing), uploads the materials that should be uploaded within its authority, and stores the materials in the system. The uploaded information is automatically confirmed, shared and agreed upon through the model established by the smart contract system, and the green financial product evaluation results are automatically generated.

[0201] Step 310: The blockchain service layer monitors nodes according to the term of the financial product. When it is monitored that the institutional node submits or modifies business data and / or evidence data, the submitted or modified business data and / or evidence data is uploaded to the blockchain, and after being uploaded to the blockchain, the above-mentioned consensus processing steps of the participating institutions based on the smart contract are triggered to re-certify whether the financial product is green or not.

[0202] The blockchain service layer is responsible for smart contract management, node management, node monitoring and evidence management. When a certain institution node modifies the business data or evidence files on the chain on the client side according to the term of the financial product, the system will process the modified data on the chain again, where the modified business data is uploaded to the alliance chain, the evidence data is hashed and the HASH result is uploaded to the chain, and the evidence data is saved in the file server through the storage service. In addition, the system fully stores each submission or modification of data during each life cycle of the financial product, so as to ensure the storage of data throughout the life cycle of the financial product and the traceability of future data.

[0203] In response to the traceability issue, the embodiments of the present application further utilize technical features such as blockchain timestamps and distributed accounting that are difficult to tamper with to design the evidence storage function of green financial products, so that any information confirmation and data upload can be traced, thereby ensuring traceability and accountability.

[0204] In addition, for each data submission or modification event by any related party of the financial product, a smart contract is executed, automatically triggering a consensus process among all participating institutions based on the smart contract to re-green-certify the financial product, achieve multi-party data consensus and cross-verification, ensure data authenticity, and reduce various risks caused by errors.

[0205] In addition, in order to address the problem of low efficiency of traditional MRV, this application can retrieve ESG risk database information online in real time, so as to monitor ESG risks quickly and efficiently, and can automatically generate MRV and related MRV reports in each life cycle of financial products through smart contracts based on the information uploaded or modified by relevant participants, greatly improving the efficiency of MRV. The system also uses multi-party data to verify the model to ensure the authenticity and accuracy of MRV results.

[0206] Until the last period of a financial product expires, the life cycle of the financial product ends on the system. The system will maintain all data and consensus results submitted / modified by all participants during the entire life cycle of the financial product. In order to address the problem of difficulty in adapting to long-term management needs, the embodiment of the present application uses a blockchain system design to unify the data format and input standards, which can support calls at any time, thereby ensuring data consistency over a long period of time to facilitate long-term data management.

[0207] The following describes in detail the risk prediction service in the system of the present application and the risk prediction implementation process based on the service.

[0208] In the embodiments of the present application, the risk prediction services include ESG risk prediction services, environmental benefit risk prediction services, extreme climate and meteorological event risk prediction services, and transition risk prediction services, which are based on corresponding AI prediction models respectively.

[0209] in:

[0210] (I) ESG risk prediction

[0211] The ESG risk prediction service constructs an AI risk analysis model in advance through analysis of non-financial data of various companies (e.g., companies across the country) and artificial intelligence learning. It also makes environmental risk predictions about companies’ penalties in future time periods through statistical analysis of historical penalty records of the companies to be predicted and model-based AI risk prediction.

[0212] During implementation, non-financial data of enterprises can be collected through crawler systems, compliant data sources and other channels, and through system data cleaning or combined with manual data cleaning, the basic data required for AI learning and data statistical analysis can be finally obtained.

[0213] Correspondingly, the risk prediction service is mainly divided into two modules: statistical module and model module.

[0214] The two modules predict the penalties of the issuer (company) of the financial product, the industry to which the issuer belongs, and the region (e.g., the city) in the future, and predict the possibility of being punished by various types of penalties at a given future time. There are three corresponding input modes for risk prediction, namely, enterprise, industry, and region. In essence, enterprises, industries, and regions are each a prediction entity, and the output includes but is not limited to the probability of the predicted entity being punished in the following seven types of penalties:

[0215] fine;

[0216] Suspend or revoke a license;

[0217] Confiscation of illegal gains;

[0218] Administrative detention;

[0219] Order to suspend production for rectification;

[0220] Order to stop production, business or close down;

[0221] Other penalties.

[0222] When making ESG risk predictions, the statistical module measures three specific aspects: previous penalties on the company, previous penalties on the industry in which the company is located, and previous penalties on the region in which the company is located. Through these three aspects of measurement, we can ultimately obtain penalty forecasts for each predicted entity corresponding to various types of penalties in future time periods.

[0223] For the model module, since each predicted entity may correspond to multiple penalty types, multiple independent binary classification models are used to solve the risk prediction problem corresponding to each penalty type. Taking the above-mentioned seven penalty types as an example, the model module uses seven independent binary classification models to solve the problem. Each model does not interfere with each other. The significance of each model is to predict the penalty probability of the predicted entity in the corresponding penalty type.

[0224] Therefore, for each input, the statistical module will give a predicted probability value, and the model module will also give a probability value. Finally, the contribution weighted values ​​of the two modules configured in the configuration file are used to obtain the final probability of the predicted entity corresponding to each penalty type. The probability of the predicted entity being punished corresponding to each penalty type is used as the risk prediction result of the predicted entity.

[0225] More specifically, risk prediction of financial products can be achieved through the following processing:

[0226] 1) For each forecast entity corresponding to the financial product, by collecting historical penalty information of each forecast entity, predict the penalty situation of the corresponding forecast entity for each penalty type in the penalty type set in the future predetermined time period, and obtain the first penalty prediction sub-result of each penalty type corresponding to the forecast entity;

[0227] 2) Using the constructed ESG risk prediction artificial intelligence model, predict the penalty situation of each predicted entity for each penalty type in the penalty type set in the future predetermined time period, and obtain the second penalty prediction sub-result of the predicted entity corresponding to each penalty type;

[0228] 3) According to the weights respectively configured for the statistical prediction method and the model prediction method, a weighted operation is performed on the first penalty prediction sub-result and the second penalty prediction sub-result corresponding to each penalty type of the prediction entity to obtain the penalty prediction result corresponding to each penalty type of the prediction entity.

[0229] Among them, when selecting a model to build an AI risk analysis model, as a binary classification problem, naive Bayes, logistic regression, decision tree, random forest, SVM (Support Vector Machine), neural network, etc. can be used to implement the binary classification model corresponding to each risk type (penalty type) in this application.

[0230] The inventors took into account that as the amount of data increases, the effect of the deep neural network model will obviously exceed the traditional machine learning model, and neural networks are a major trend in artificial intelligence processing and have strong fitting capabilities. Based on this, the embodiment of the present application preferably adopts an MLP (Multi-Layer Perception) model to solve the binary classification problem of the present application and achieve classification.

[0231] The following is an introduction to the MLP model:

[0232] Neural networks are simulations and simplifications of biological neurons, which are composed of dendrites, cell bodies, axons, etc. Dendrites are the input end of the cell body, which receive nerve impulses from all around; axons are the output end of the cell body, which transmit nerve impulses to other neurons. Biological neurons have two states: excitation and inhibition. When the stimulus received is higher than a certain threshold, it will enter the excited state and transmit nerve impulses from the axon, otherwise there will be no nerve impulses.

[0233] The basic structure of the multilayer perceptron MLP can be obtained based on the biological neuron model. The most typical MLP includes three layers: input layer, hidden layer and output layer. The different layers of the MLP neural network are fully connected (that is, any neuron in the previous layer is connected to all neurons in the next layer). Neural networks have three basic elements: weights, biases and activation functions.

[0234] in:

[0235] Weight: The strength of the connection between neurons is represented by the weight, and the size of the weight represents the size of the possibility;

[0236] Bias: Bias is set to correctly classify samples and is an important parameter in the model, which ensures that the output value calculated by the input cannot be activated arbitrarily;

[0237] Activation function: It plays a role of nonlinear mapping, which can limit the output amplitude of neurons to a certain range, generally between (-1 to 1) or (0 to 1). The most commonly used activation function is the Sigmoid function, which can map numbers from (-∞, +∞) to the range of (0 to 1).

[0238] MLP, also known as Full-connection Neural Network, is combined with Figure 6 The three-layer network structure including input layer - hidden layer - output layer is described as follows:

[0239] The input is x=(x 1 , x 2 ..., x N ), after multiple hidden layers and the final output layer, we get y = (y 1 ,y...,y M ).

[0240] like Figure 7 As shown, the input of each neuron is a i , parameter is w i , the output of the neuron is a, σ is the activation function, common ones are sigmod, tanh, etc. The purpose of the activation function is to introduce nonlinearity and increase the expression ability of the neural network:

[0241]

[0242] For more information about common activation functions, see Figure 8 .

[0243] In classification problems, the number of neurons in the output layer is equal to the number of categories.

[0244] The following describes in detail the binary classification model constructed based on the MLP model in the embodiment of the present application from the aspects of data and features, training, and prediction:

[0245] a) Data and Features

[0246] For the penalty data of each company, industry, and city, this embodiment converts it into multiple binary classification problems, and correspondingly multiple binary classification models need to be trained. For the binary classification problem faced by this application, this embodiment of the application sets the number of neurons in the output layer to 1, and uses the sigmoid activation function to compress the output of the neuron to [0, 1].

[0247] Take the classification problem of "fine" as an example. Let the feature be x and the label be y (fine is 1, no fine is 0), then we can get many (x, y) pairs. Divide them into training, validation, and test sets. Train the network parameters in the training set, validate them in the validation set, and finally test them in the test set to get the optimal parameters.

[0248] Specifically, in one example, a total of 15 features are used, including: 'industry', 'subIndustry', 'econ_kind', 'status', 'regist_capi', 'province_final', 'city_final', 'year', 'season', 's_0', 's_1', 's_2', 's_3', 's_4', 's_5', each feature corresponding to the industry, sub-industry, company type, status, capital, province, city, year to be predicted, quarter to be predicted, and air status 0-5.

[0249] Since it is inevitable that there may be some missing data and dirty data in the collected data, it is necessary to preprocess the collected data before obtaining the above features:

[0250] When predicting company penalties, this embodiment first queries the company's basic information table and the company's regional table to obtain the company's basic information, normalizes the pattern used, and fills in the blanks (a reserved category) for category information (such as industry) to improve accuracy. For number information (such as registered capital), the mean of all training data is used as filler. When predicting industry and regional penalties, only the industry / region is used as a feature, and the remaining features are replaced by the mean of the entire data. When predicting industry and regional penalties, only the industry / region is used as a feature, and the remaining features are replaced by the mean of the entire data.

[0251] Considering that there is no air quality data for later years and air quality is a very important pattern, when a company is queried, the model will obtain the average air quality of the city based on the city where the company is located (or the city specified by the model). If the air quality of the city is not available in the data, the mean of all training data will be used as the input of the air quality data.

[0252] s_0–s_5 correspond to the following five states

[0253] {'Excellent': 0, 'Good': 1, 'Mild pollution': 2, 'Moderate pollution': 3, 'Heavy pollution': 4, 'Severe pollution': 5}

[0254] In order to improve the accuracy, this embodiment counts the state values ​​of each air quality on a quarterly basis, and specifically takes the probability of each state among all states as the input of the air quality data.

[0255] b) Training

[0256] For each (x, y) pair, the embodiment of the present application converts the feature x=(x 1 , x 2 ..., x k ), representing x features of the data (i.e., industry, capital, etc.), firstly vectorize each feature, and use embedding technology to map each feature into a k-dimensional vector. For example, for the feature of "industry" with a total of V values ​​(representing "manufacturing", "construction", etc.), initialize a (V, k) matrix, and according to the index corresponding to the actual value of the "industry" feature, check the index-th row of the (V, k) matrix to obtain a k-dimensional vector, which is the vector of this specific "industry" feature. For numerical features, this embodiment adopts a bucketing method to map continuous features into different buckets according to the threshold, and then perform embedding. Embedding technology can greatly enhance the expressive power of the network.

[0257] After mapping each feature into an embedding, the embodiment of the present application converts x=(x 1 , x 2 ..., x k ) are concatted together. Assuming that the embedding of each feature is k-dimensional, there are 15*k inputs. Initialize the two-layer MLP network, which are (15*k, 40) and (40, 1). The activation function of the first layer is tanh, and the second layer is sigmoid. After forward propagation through the network, the predicted probability y of the classification can be calculated in the last layer. logit , the penalty of the data is known, and there is y label , so according to the cross entropy, we can get the classification loss function:

[0258] loss=-{(y label log(y logit )+(1-y label )log(1-y logit ))}.

[0259] Through the back-propagation algorithm, the MLP and embedding parameters can be finally trained (updated) to obtain the binary classification model corresponding to each penalty type required by this application.

[0260] c) Prediction

[0261] After completing the model training, for new data, the embodiment of the present application inputs it into seven binary classification models ('fine', 'temporary seizure and revocation of license', 'confiscation of illegal gains', 'administrative detention', 'order to suspend production and rectify', 'order to suspend production, suspend business, close down', 'other penalties'). In each model, embedding mapping is performed first, and then logits are obtained through MLP. If logits>0.5, it is predicted that the predicted entity will be punished with the corresponding penalty type. If logits<0.5, it is predicted that the predicted entity will not be punished with the corresponding penalty type.

[0262] 2. Environmental Benefit Risk Prediction

[0263] In terms of environmental benefit risk prediction, the embodiments of the present application specifically provide AI prediction of carbon emission reduction.

[0264] Among them, predicting carbon emissions is a regression problem, and the prediction model uses the "polynomial regression" method. Assume that the collected data is D = (x, y), where x is the feature and y is the label (carbon emissions). Assume x = (x 1 , x 2 , ..., x n ) represent "available hours for wind power generation", "wind weather data", "regional data", "time data" and "installed capacity" respectively. Input numerical features directly, id discrete features (such as mapping each province in the regional data to 1-34, representing different provinces), and initialize weights w T , then the total input z is z = w T x=w 0 x 0 +w 1 x 1 +…+w n x n , initialization parameters αβ 1 β 2 , we can get y pred =α+β 1 z+β 2 z 2 ,y pred is the predicted carbon emissions, based on the actual carbon emissions y label , through the MSE (mean square error) loss function loss = ∑ (y label-y pred ) 2 , combined with the gradient descent method, the optimal parameter w can be trained T αβ 1 β 2 .

[0265] During actual prediction, carbon emissions can be obtained by simply inputting the features into the model.

[0266] (III) Risk prediction service for extreme climate and meteorological events

[0267] The risk of extreme climate and meteorological events is a binary classification problem. It predicts whether there will be extreme weather risks. The model features are "time" and "region". It predicts whether there will be "typhoons" and "drought risks", etc. It can be transformed into multiple binary classification problems. Considering that there are fewer features, the logistic regression method can be used.

[0268] Take the prediction of "whether there is a typhoon" as an example. Assume that the collected data is D = (x, y), where x is the feature and y is the label (1 for typhoon and 0 for no typhoon). 1 , x 2 , ..., x n ), representing features such as "time" and "region". Similarly, the numerical features are directly input, the discrete features are idized, and the weights w are initialized. T , then the total input z is z = w T x=w 0 x 0 +w 1 x 1 +…+w n x n ,according to You can get y pred , that is, whether there is a "typhoon", according to the actual typhoon situation y label , through the Cross Entropy loss function loss = -{(y label log(y pred )+(1-y label )log(1-y pred ))}, combined with the gradient descent method, training obtains the optimal parameter w T .

[0269] In actual prediction, just input the features into the model and get y pred , if y pred If it is greater than 0.5, then there is a “typhoon risk” forecast, otherwise there is no risk.

[0270] (IV) Transformation risk prediction

[0271] Like the previous model for predicting carbon emissions, predicting carbon costs is also a regression problem. This embodiment still uses the "polynomial regression" method.

[0272] Assume that the data collected by the system is D = (x, y), where x is the feature and y is the label (carbon cost). 1 , x 2 , ..., x n ), representing "carbon price", "carbon emissions", "enterprise location", "enterprise industry" and "time". Input numerical features directly, id discrete features (such as mapping the same "enterprise industry" to the same id, mapping "manufacturing industry" to 1, and "construction industry" to 2), and initialize weights w T , then the total input z is z = w T x=w 0 x 0 +w 1 x 1 +…+w n x n , initialization parameters αβ 1 β 2 , we can get y pred =α+β 1 z+β 2 z 2 ,y pred is the predicted carbon cost, based on the actual carbon cost y label , through the MSE loss function loss = ∑(y label -y pred ) 2 , combined with the gradient descent method, the optimal parameter w can be found T αβ 1 β 2 .

[0273] During actual prediction, the carbon cost can be obtained by simply inputting the features into the model.

[0274] In summary, the method and system for evaluating and managing green financial products disclosed in this application can produce at least the following technical effects compared with the manual management of traditional green financial products:

[0275] 1) A decentralized management model has established a more efficient, high-quality, and low-cost green financial product management trust system;

[0276] 2) Multi-party data consensus and cross-verification ensure data authenticity and reduce the risk of errors and moral risks;

[0277] 3) All data is stored in a distributed manner, which reduces the trouble caused by information loss in the certification and tracking management of green financial products;

[0278] 4) Use blockchain technology to manage core business data on the chain. The data on the chain can be traced, which provides conditions for possible accountability and makes green financial product certification more effective;

[0279] 5) Automated management mode, with additional tracking and management of MRV functions, lays the foundation for the expansion of green financial products, such as impact investment or environmental benefit asset trading;

[0280] 6) Automatically generate reports, unify and standardize certification report standards, ensure the authenticity of report data, and reduce enterprise management costs;

[0281] 7) It can automatically call other databases, such as the ESG database, to provide compatibility for the multi-dimensional management of green financial products;

[0282] 8) Realize ESG risk prediction and greatly reduce the risks of green financial products.

[0283] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0284] For the convenience of description, the above system or device is described by dividing it into various modules or units according to its functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0285] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0286] Finally, it should be noted that, in this article, relational terms such as first, second, third and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0287] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for evaluating and managing green financial products, characterized in that: A system for green financial product evaluation and management, the system comprising at least: a plurality of institutional nodes corresponding to a plurality of participating institutions, an application service layer, a blockchain service layer and a risk prediction service; the method comprising: The institutional node submits its own institution’s business data and evidence data on financial products to the system; The risk prediction service performs risk prediction on the financial product based on at least artificial intelligence processing for the submission event of the institutional node, and feeds back the risk prediction result to the institutional node; The institutional node enters corresponding risk assessment data of the financial product into the blockchain corresponding to the blockchain service layer based on the risk prediction result; The blockchain service layer processes the business data and the evidence data on the blockchain, and triggers consensus processing among participating institutions based on smart contracts after the data are uploaded to the blockchain; If a consensus is reached, the blockchain service layer will certify the financial product as a green financial product; if no consensus is reached, the financial product will be certified as a non-green financial product; and the green certification result will be uploaded to the blockchain; The application service layer generates an MRV report including core business data and green certification results based on the on-chain data and the green certification results; The authentication of financial products based on the consensus mechanism includes the following aspects: Intelligent judgment of green attributes and intelligent evaluation of environmental benefits; in: Intelligent judgments on the green attributes of financial products include: Obtaining the project information of the financial product uploaded by the corresponding institution node; identifying the project content in the project information; determining whether there is an entry matching the project content in the predetermined green certification standard; if so, determining that the financial product has green attributes; Intelligent assessment of the environmental benefits of financial products includes: The environmental benefits of the projects corresponding to the financial products are calculated using a multi-dimensional data intelligent model; in the process of calculating the environmental benefits of the projects using a multi-dimensional data intelligent model: when the corresponding institutional node cannot provide the parameters required for calculating the carbon dioxide emission reduction in the traditional model, the system identifies the region or area where the project is located based on the project name, calls the pre-created "regional power grid emission factor library", and identifies the emission factors applicable to the region or area where the project is located based on the regional or regional data matching results. The system identifies the project content as "wind power plant construction" based on the project name, matches the keyword "wind power generation", calls the "wind power generation available hours database" in the pre-created "regional renewable resource database", and determines the annual wind power generation available hours. When the system sets the wind power project to have an unknown power generation, the following backup calculation formula is used to calculate the carbon dioxide emission reduction of the project: P CO2 =W×t×α Among them, P CO2 represents the carbon dioxide emission reduction, W represents the installed capacity, t represents the available hours of wind power generation in the project area, and α represents the electricity emission factor applicable to the project area; the system obtains W by identifying and extracting the project information of the financial products uploaded by the corresponding institution node, and obtains t and α by matching the extracted results with the corresponding pre-built database, the pre-built database includes a regional power grid emission factor library and a regional renewable resource database, and the regional renewable resource database includes a database of available hours of wind power generation; The risk forecasts include environmental, social and corporate governance (ESG) risk forecasts, environmental benefit risk forecasts, extreme climate and meteorological event risk forecasts and transition risk forecasts; The risk prediction service performs ESG risk prediction on the financial product based on at least artificial intelligence processing for the submission event of the institutional node, including: For each forecast entity corresponding to the financial product, by counting historical penalty information of each forecast entity, predicting the penalty situation of the forecast entity for each penalty type in the penalty type set in a future predetermined time period, and obtaining a first penalty prediction sub-result corresponding to each penalty type of the forecast entity; Using the constructed ESG risk prediction artificial intelligence model, predict the penalty situation of each prediction entity for each penalty type in the penalty type set in a future predetermined time period, and obtain the second penalty prediction sub-result of the prediction entity corresponding to each penalty type; Determine the penalty prediction result corresponding to each penalty type of the prediction entity according to the first penalty prediction sub-result and the second penalty prediction sub-result corresponding to each penalty type of the prediction entity; Among them, each penalty type corresponds to its own ESG risk prediction artificial intelligence model, and the ESG risk prediction artificial intelligence model is a multi-layer perceptron MLP model. The number of neurons in the output layer of the MLP model is 1, and the input data of the MLP model includes air quality data; The forecast entity corresponding to the financial product includes: the enterprise associated with the financial product that needs risk forecasting, the industry to which the enterprise belongs, and the region where the enterprise is located; The penalty types in the penalty type set include at least one of the following: fine; temporary seizure or revocation of license; confiscation of illegal gains; administrative detention; order to suspend production for rectification; order to suspend production, business or close down; other penalties; Among them, if the penalty type set includes multiple penalty types, the ESG risk prediction artificial intelligence model includes multiple binary classification models corresponding to various penalty types, and each binary classification model is used to determine the probability of penalty of the predicted entity in the corresponding penalty type in a predetermined time period in the future.

2. The method according to claim 1, characterized in that: Predicting the environmental benefits of the financial product based on a pre-built artificial intelligence model for environmental benefits prediction; the artificial intelligence model for environmental benefits prediction is implemented using a polynomial regression method; Based on a pre-built artificial intelligence model for predicting extreme climate and meteorological event risks, predict the extreme climate and meteorological event risks of the financial product; the artificial intelligence model for predicting extreme climate and meteorological event risks adopts a logistic regression method and is implemented by converting into multiple binary classification problems; Based on a pre-built transformation risk prediction artificial intelligence model, the transformation risk of the financial product is predicted; the transformation risk prediction artificial intelligence model is implemented using a polynomial regression method.

3. The method according to claim 1, characterized in that Determining the penalty prediction result corresponding to each penalty type of the prediction entity according to the first penalty prediction sub-result and the second penalty prediction sub-result corresponding to each penalty type of the prediction entity comprises: According to the weights respectively configured for the statistical prediction method and the model prediction method, a weighted operation is performed on the first penalty prediction sub-result and the second penalty prediction sub-result corresponding to each penalty type of the prediction entity to obtain the penalty prediction result corresponding to each penalty type of the prediction entity.

4. The method according to claim 1, characterized in that: The system also includes a storage service; The blockchain service layer processes the business data and the evidence data on the blockchain, including: The blockchain service layer uploads the business data to the blockchain; The blockchain service layer performs hash operation on the evidence data, uploads the obtained hash result to the blockchain, and stores the evidence data through the storage service.

5. The method according to claim 1, characterized in that Before generating the MRV report, it also includes: After the green financial products are issued, MRV tracking and management shall be carried out on the green financial products during their life cycle.

6. The method according to claim 1, characterized in that Also includes: After the issuance of the financial product is completed, each institutional node shall submit or modify the business data and evidence data corresponding to its own institution in each duration of the financial product as required; The blockchain service layer monitors nodes according to the term of the financial product. When monitoring that the institutional node submits or modifies business data and / or evidence data, the submitted or modified business data and / or evidence data are uploaded to the blockchain, and after being uploaded to the blockchain, the consensus processing step of each participating institution based on the smart contract is triggered to re-certify whether the financial product is green or not; Among them, the system generates report information based on the corresponding on-chain data and green certification results in each life period of the financial product, and the system can maintain all data and consensus results corresponding to the entire life cycle of the financial product.

7. The method according to claim 6, characterized in that The system establishes a role account system, assigns a corresponding role to each institutional node, and adds corresponding system permissions and blockchain permissions to each role; Based on its corresponding role, each institutional node submits or modifies data within its authority to the system, and confirms data within its authority during consensus processing.

8. The method according to any one of claims 1 to 7, characterized in that: The blockchain is a consortium chain.

9. A system for evaluating and managing green financial products, characterized in that: include: Multiple institutional nodes, application service layer, blockchain service layer and risk prediction service corresponding to multiple participating institutions; among which: The institution node is used to submit the business data and evidence data of the institution itself on the financial products to the system; The risk prediction service is used to predict the risk of the financial product based on at least artificial intelligence processing for the submission event of the institutional node, and to feed back the risk prediction result to the institutional node; The institutional node is further used to input corresponding risk assessment data of the financial product into the blockchain corresponding to the blockchain service layer based on the risk prediction result; The blockchain service layer is used to: process the business data and the evidence data on the blockchain, and trigger consensus processing among participating institutions based on smart contracts after the data are uploaded; if a consensus is reached, the financial product is certified as a green financial product; if no consensus is reached, the financial product is certified as a non-green financial product; and the green certification result is uploaded to the blockchain; The application service layer is used to generate an MRV report including core business data and green certification results based on the on-chain data and the certification results; The authentication of financial products based on the consensus mechanism includes the following aspects: Intelligent judgment of green attributes and intelligent evaluation of environmental benefits; in: Intelligent judgments on the green attributes of financial products include: Obtaining the project information of the financial product uploaded by the corresponding institution node; identifying the project content in the project information; determining whether there is an entry matching the project content in the predetermined green certification standard; if so, determining that the financial product has green attributes; Intelligent assessment of the environmental benefits of financial products includes: The environmental benefits of the projects corresponding to the financial products are calculated using a multi-dimensional data intelligent model; in the process of calculating the environmental benefits of the projects using a multi-dimensional data intelligent model: when the corresponding institutional node cannot provide the parameters required for calculating the carbon dioxide emission reduction in the traditional model, the system identifies the region or area where the project is located based on the project name, calls the pre-created "regional power grid emission factor library", and identifies the emission factors applicable to the region or area where the project is located based on the regional or regional data matching results. The system identifies the project content as "wind power plant construction" based on the project name, matches the keyword "wind power generation", calls the "wind power generation available hours database" in the pre-created "regional renewable resource database", and determines the annual wind power generation available hours. When the system sets the wind power project to have an unknown power generation, the following backup calculation formula is used to calculate the carbon dioxide emission reduction of the project: P CO2 =W×t×α Among them, P CO2 represents the carbon dioxide emission reduction, W represents the installed capacity, t represents the available hours of wind power generation in the project area, and α represents the electricity emission factor applicable to the project area; the system obtains W by identifying and extracting the project information of the financial products uploaded by the corresponding institution node, and obtains t and α by matching the extracted results with the corresponding pre-built database, the pre-built database includes a regional power grid emission factor library and a regional renewable resource database, and the regional renewable resource database includes a database of available hours of wind power generation; The risk forecasts include environmental, social and corporate governance (ESG) risk forecasts, environmental benefit risk forecasts, extreme climate and meteorological event risk forecasts and transition risk forecasts; The risk prediction service performs ESG risk prediction on the financial product based on at least artificial intelligence processing for the submission event of the institutional node, including: For each forecast entity corresponding to the financial product, by counting historical penalty information of each forecast entity, predicting the penalty situation of the forecast entity for each penalty type in the penalty type set in a future predetermined time period, and obtaining a first penalty prediction sub-result corresponding to each penalty type of the forecast entity; Using the constructed ESG risk prediction artificial intelligence model, predict the penalty situation of each prediction entity for each penalty type in the penalty type set in a future predetermined time period, and obtain the second penalty prediction sub-result of the prediction entity corresponding to each penalty type; Determine the penalty prediction result corresponding to each penalty type of the prediction entity according to the first penalty prediction sub-result and the second penalty prediction sub-result corresponding to each penalty type of the prediction entity; Among them, each penalty type corresponds to its own ESG risk prediction artificial intelligence model. The ESG risk prediction artificial intelligence model is a multi-layer perceptron MLP model. The number of neurons in the output layer of the MLP model is 1, and the input data of the MLP model includes air quality data.

10. The system according to claim 9, characterized in that The system also includes a storage service; When the blockchain service layer processes the business data and the evidence data on the blockchain, it is specifically used to: The business data is put on the blockchain; the evidence data is processed by hash operation, the obtained hash result is put on the blockchain, and the evidence data is stored through the storage service, wherein all the on-chain data can be traced.

11. The system according to claim 9, characterized in that The application service layer is also used to provide at least one of the following functions: Data collection, user management, bond management, authority management and back-end system management; The blockchain service layer is used to provide at least one of the following functions: Node management, evidence management, security management, contract management and node monitoring.

12. The system according to claim 9, characterized in that The risk prediction service is an independent application system module and can be dynamically plugged in and out of the system.

Images