Power transaction risk identification method and device, and computer program product

By using optical character recognition technology and machine learning to build a transaction risk identification model in power trading, key information in performance bonds can be automatically extracted, solving the problem of low intelligence level in power trading risk identification and achieving efficient and low-cost risk management.

CN118941390APending Publication Date: 2024-11-12STATE GRID BEIJING ELECTRIC POWER CO +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411080052.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Existing methods for identifying risks in electricity trading have a low level of intelligence, resulting in high human resource costs and an inability to effectively identify and manage electricity trading risks.

Method used

By extracting textual information from the target performance bond, and using optical character recognition (OCR) technology and machine learning to build a transaction risk identification model, key information can be automatically extracted and risk levels can be identified, reducing the need for human resources intervention.

Benefits of technology

It has enabled the intelligentization of the power trading risk identification process, reduced the waste of human resources, improved the efficiency and accuracy of trading risk identification, and reduced costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941390A_ABST
    Figure CN118941390A_ABST
Patent Text Reader

Abstract

The invention discloses a power transaction risk identification method and device and a computer program product. The method comprises the following steps: extracting target text information from a target performance warrant corresponding to a target power transaction; obtaining target power transaction information obtained by executing the target power transaction in the target time period; extracting key information in the target text information; and based on the key information and the target power transaction information, a transaction risk identification model is adopted to obtain a target risk identification result of the target power transaction, and the target risk identification result at least comprises a risk level of the target power transaction in the target time period. According to the invention, the technical problems of low intelligent level and high human resource investment cost of a power transaction risk identification method in the prior art are solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of power trading, and in particular to a method, device and computer program product for identifying power trading risks. Background Art

[0002] In order to strengthen transaction settlement management and prevent the risk of arrears in electricity bills, power sales companies provide default guarantees to power grid companies by signing performance bonds with banks. In order to reduce cooperation risks, it is necessary to identify the risks of power transactions for multiple power sales companies.

[0003] In related technologies, the risk identification of electricity transactions is based on manually entering performance bonds into electronic equipment, which is then manually calculated and analyzed to identify whether there are risks in the transaction. In this process, too much human resources are involved and the level of intelligence is low, leading to problems such as excessively high labor costs and waste of human resources.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a method, device and computer program product for identifying power trading risks, so as to at least solve the technical problems of low intelligence level and high human resource investment cost in the related art methods for identifying power trading risks.

[0006] According to one aspect of an embodiment of the present invention, a method for identifying risks in power trading is provided, comprising: extracting target text information from a target performance bond corresponding to a target power trading; acquiring target power trading information obtained by executing the target power trading in a target period; extracting key information from the target text information; and, based on the key information and the target power trading information, adopting a trading risk identification model to obtain a target risk identification result of the target power trading, wherein the target risk identification result includes at least the risk level of the target power trading in the target period; wherein the trading risk identification model is obtained through machine learning based on historical trading information, historical text information and corresponding historical risk identification results corresponding to multiple historical power trading, wherein the historical trading information is used to indicate power trading information respectively obtained by executing the multiple historical power trading, and the historical text information is extracted from historical performance bonds corresponding to the multiple historical power trading.

[0007] According to another aspect of an embodiment of the present invention, there is also provided an apparatus for identifying risks in power trading, comprising: a target text information extraction module, for extracting target text information from a target performance bond corresponding to a target power trading; an acquisition module, for acquiring target power trading information obtained by executing the target power trading in a target period; a key information extraction module, for extracting key information from the target text information; a target risk identification result determination module, for obtaining a target risk identification result of the target power trading based on the key information and the target power trading information, using a transaction risk identification model, wherein the target risk identification result includes at least the risk level of the target power trading in the target period; wherein the transaction risk identification model is obtained through machine learning based on historical transaction information, historical text information and corresponding historical risk identification results corresponding to multiple historical power transactions, wherein the historical transaction information is used to indicate the power trading information obtained by executing the multiple historical power transactions respectively, and the historical text information is extracted from the historical performance bonds corresponding to the multiple historical power transactions.

[0008] According to another aspect of an embodiment of the present invention, there is further provided a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the steps of any one of the methods for identifying power trading risks are implemented.

[0009] In an embodiment of the present invention, target text information is extracted from a target performance bond corresponding to a target power transaction; target power transaction information obtained by executing the target power transaction in a target period is obtained; key information in the target text information is extracted; based on the key information and the target power transaction information, a transaction risk identification model is used to obtain a target risk identification result of the target power transaction, wherein the target risk identification result at least includes the risk level of the target power transaction in the target period; wherein the transaction risk identification model is obtained by machine learning based on historical transaction information, historical text information and corresponding historical risk identification results corresponding to a plurality of historical power transactions, the historical transaction information is used to indicate the power transaction information obtained by executing the plurality of historical power transactions respectively, and the historical text information is extracted from the historical performance bonds corresponding to the plurality of historical power transactions, thereby achieving the purpose of intelligently extracting key information in the target performance bond in the process of power transaction risk identification, and intelligently obtaining the target risk identification result based on the transaction risk identification model, thereby achieving the technical effect of improving the intelligence level in the process of power transaction risk identification, and reducing the input cost of power transaction risk identification, thereby solving the technical problems of low intelligence level and high human resource input cost in the power transaction risk identification method in the related art. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0011] Figure 1 is a flow chart of a method for identifying power trading risks according to an embodiment of the present invention;

[0012] Figure 2 is a flow chart of an optional method for identifying power trading risks according to an embodiment of the present invention;

[0013] Figure 3 is a schematic diagram of a power trading risk identification device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0014] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0016] First, to facilitate understanding of the embodiments of the present invention, some terms or nouns involved in the present invention are explained below:

[0017] A performance bond is a letter of guarantee issued by a bank or insurance company to guarantee the performance of the buyer and seller's obligations in power transactions. A performance bond usually includes the responsibilities and obligations of the buyer and seller, as well as terms such as liquidated damages that may need to be paid in certain circumstances.

[0018] Optical Character Recognition (OCR) is a technology that uses optical equipment and computer technology to convert printed or handwritten character images into editable text information. Through optical character recognition technology, the text content in the document can be quickly and accurately identified, and automatic text recognition, text conversion and text editing can be achieved.

[0019] As the reform of the power market gradually deepens, commercial users enter the market in an all-round way, the floating range of power trading prices expands, and the operating risks of power sales companies increase exponentially. Many power sales companies with insufficient risk management capabilities, weak strength, and poor credit have begun to default, delist, and other adverse situations, which not only damages the interests of retail users, but also is not conducive to the development of the power market. Therefore, an effective power market performance risk prediction and identification method can warn of risks, take effective prevention and control measures in a timely manner, and reduce the probability of risk occurrence.

[0020] The performance risk of power sales companies is the most important risk in the power trading market, and the performance guarantee mechanism of power sales companies is a necessary means to maintain the stable and healthy development of the market. Before participating in power purchase transactions, power sales companies must submit a performance bond payable on demand or a performance insurance payable on demand to the power trading institution. The beneficiary or insured is the power grid company that conducts financial settlement with the power sales company. The performance bond or performance insurance amount is related to the credit rating of the power sales company and the historical transaction volume in the past year. During the execution of the power trading contract, if the power sales company fails to pay the relevant fees in full, there may be a performance risk. If the performance risk can be predicted, identified, and warned in advance, more effective management measures can be taken to prevent the occurrence of risks or reduce the probability of risks. Therefore, it is urgent to build a performance risk prediction and identification model for power sales companies. According to the basic situation, transaction situation, credit rating, performance bond or performance insurance of the power sales companies, the performance risk of power sales companies can be predicted and identified in real time. By taking corresponding prevention and management measures for different levels of performance risks, the probability of risk occurrence can be minimized to promote the healthy development of the power market.

[0021] Based on this, an embodiment of the present invention provides an embodiment of a method for identifying risks in electricity trading. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0022] Figure 1 is a flow chart of a method for identifying power trading risks according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0023] Step S102, extracting target text information from the target performance bond corresponding to the target power transaction;

[0024] Optionally, in the process of power trading, a performance bond is a form of document or guarantee. Power trading involves the purchase and sale of electricity or related energy products. The performance bond is signed before the power trading. The power trading data of the power sales company is identified once every three months. Based on the identification results, it is determined whether the power sales company has breached the contract. In the event of a breach of contract, the measures to be taken by the power sales company after the breach of contract can be proposed in accordance with the content of the performance bond.

[0025] Optionally, the target power transaction refers to the period during which power transactions currently exist (the current period is pushed forward three months). The target performance bond refers to the contract form used to guarantee the performance guarantee for the current power transaction period, wherein the target performance bond includes at least the guarantee amount and the guarantee time limit. Specifically, the target text information is the information required for subsequent transaction risk identification in the target performance bond. Based on the information in the target performance bond signed during the current power transaction period, transaction risk identification is carried out to effectively monitor the default behavior of power sales companies, thereby reducing the risks in power transactions.

[0026] In an optional embodiment, before the target performance bond extracts the target text information from the target performance bond corresponding to the target power transaction, the method also includes: obtaining an initial image of the target performance bond; preprocessing the initial image to obtain a preprocessed image; wherein the preprocessing includes at least one of the following: grayscale processing, binarization processing, and denoising processing; using an optical character recognition method to extract features of characters in the preprocessed image to obtain initial text information; post-processing the initial text information to obtain target text information, wherein the post-processing includes at least one of the following: error correction and character connection.

[0027] Optionally, based on the optical character recognition technology (OCR), the text is extracted from the image of the target performance bond obtained by taking a photo. The OCR technology ensures that the image of the target performance bond is accurately and comprehensively converted into processable text data based on multiple steps such as image preprocessing, text positioning, and text recognition. Efficiently realize the intelligent extraction of the content in the target performance bond into the electronic device. Effectively reduce the involvement of human resources, thereby solving the problem of waste of human resources.

[0028] Optionally, the paper version of the target performance bond is scanned or photographed and stored in an electronic device in the form of an image to obtain an initial image, and the initial image is subjected to grayscale processing, binarization processing and denoising processing to obtain a preprocessed image, thereby effectively removing the influence of various factors such as uneven lighting conditions, noise, blur, etc., making the preprocessed image clearer, so as to ensure that the quality of the preprocessed image is good when the text in the preprocessed image is subsequently extracted to improve the accuracy of target text information recognition.

[0029] Optionally, upload the preprocessed image to the selected OCR tool. When performing the OCR operation, you can set the language and other related parameters to ensure accurate target text information extraction results. Based on OCR technology, the conversion of image files to text data is realized, which effectively provides a basis for the subsequent use of the above target text information.

[0030] Step S104, obtaining target power transaction information obtained by executing the target power transaction during the target period;

[0031] Optionally, the pre-agreed transaction volume value can be obtained based on the target text information in the target performance bond, which can be used as a standard reference value. In order to determine whether there is a power trading risk during the target power trading period, the target power trading information generated by the target power trading during the target period can be further obtained, and the target power trading information is compared with the target text information to determine whether there is a risk in the target power trading, thereby effectively realizing intelligent power trading risk identification, reducing the involvement of human resources, and thus solving the problem of reducing waste of human resources.

[0032] Optionally, during the target power transaction, the cumulative settlement power and annual market-based transaction contract power of the power sales company reflect the market size of the power sales company and the transaction contracts that can be undertaken; the total market-based transaction contract power can reflect the insight of the overall market share; the number of effective complaints and the number of irregular information disclosure are directly related to the service quality and compliance of the power sales company; the number of transactions participated by the power sales company and the upper limit of sales power can reflect the market activity and scale control; the talent structure of the power sales company and the cumulative completion of the total renewable energy consumption responsibility weight can be related to the technical strength and environmental protection responsibility of the power sales company. The comprehensive analysis of the above characteristics helps to fully understand the operating status, market position and potential operating risks of the power sales company, and provides a strong basis for effective transaction risk identification and management. Among them, the target power transaction information required for transaction risk identification is shown in Table 1, which is a summary table of target power transaction information.

[0033] Table 1

[0034]

[0035]

[0036] Step S106, extracting key information from the target text information;

[0037] Optionally, there is a large amount of target text information in the target performance bond, which includes not only key information required for transaction risk identification but also information not required for transaction risk identification, such as the name and contact information of the signatory in the target performance bond. Therefore, it is necessary to extract the key information required for transaction risk identification in the target text information for use in subsequent transaction risk identification, thereby effectively improving the efficiency of subsequent transaction risk identification and realizing intelligent transaction risk identification.

[0038] In an optional embodiment, extracting key information from the target text information includes: based on preset matching rules, extracting multiple groups of identifiers from the target text information, wherein each group of identifiers includes a title in the target text information and a title level corresponding to the title; based on the title level in the multiple groups of identifiers, determining the hierarchical structures corresponding to the multiple groups of identifiers; and obtaining key information based on the information pointed to by the hierarchical structures corresponding to the multiple groups of identifiers.

[0039] Optionally, there are multiple title levels in the target performance bond, for example, there is a large title for the transaction information part, and the sub-titles below it include information such as the transaction amount, the transaction power required to fulfill the transaction, or the transaction time, and there is a large title for the trader information part, and the sub-titles below it include information such as the trader's name and the trader's contact information. Based on the multiple title levels existing in the target performance bond, the key information under the corresponding title level is obtained. Specifically, first, the target text information is sentence- and segmented, and special symbols, stop words and other information in the target text information are removed, so as to achieve the effect of reducing noise, and further process other possible labels and missing values ​​to ensure the integrity of the data. Secondly, based on the preset matching rules, the identifier in the target text information is determined, and then the hierarchical structure corresponding to the identifier is determined based on the identifier, and the key information under the corresponding hierarchical structure is effectively found, thereby realizing the extraction of key information in the target text information, and then realizing the intelligent identification of transaction risks and reducing the intervention of human resources.

[0040] Optionally, based on the obtained target text information, key information as shown in Table 2 in the target performance bond is extracted, wherein Table 2 is a key information data table in the target performance bond.

[0041] Table 2

[0042] Serial number Data attributes 1 Guarantee amount (yuan) 2 Validity end time

[0043] In an optional embodiment, key information is obtained based on the information pointed to by the hierarchical structures corresponding to the multiple groups of identifiers, including: constructing a topological relationship network with the hierarchical structures corresponding to the multiple groups of identifiers as nodes, wherein the topological relationship network is used to indicate the association relationship between the hierarchical structures corresponding to the multiple groups of identifiers; determining the title at a preset node in the topological relationship network based on the topological relationship network and the titles in the multiple groups of identifiers; determining the title at the preset node as a preset title, and using the text information corresponding to the preset title in the target text information as key information.

[0044] Optionally, based on the form of constructing a topological relationship network, the hierarchical structure is used as a node, and the preset title corresponding to the preset node is determined, so as to determine the text information corresponding to the preset title as the key information. Among them, the form of constructing a topological relationship network is a process of comprehensively acquiring all the contents in the target text information, so that when extracting key information later, the key information is extracted more accurately and the risk of missing key information is reduced. The above method not only improves the accuracy and completeness of the target text information extraction, but also provides more dimensional data support for the subsequent transaction risk identification.

[0045] Step S108, based on the key information and the target power transaction information, a transaction risk identification model is adopted to obtain the target risk identification result of the target power transaction, wherein the target risk identification result at least includes the risk level of the target power transaction in the target period; wherein the transaction risk identification model is based on historical transaction information, historical text information and corresponding historical risk identification results corresponding to multiple historical power transactions, and is obtained through machine learning, the historical transaction information is used to indicate the power transaction information obtained by executing multiple historical power transactions respectively, and the historical text information is extracted from the historical performance bonds corresponding to the multiple historical power transactions.

[0046] Optionally, the key information and target power transaction information obtained are used as input, and the transaction risk identification model is applied to identify the power transaction risk, and the target risk identification result corresponding to the target power transaction is output. Based on the target risk identification result, it is determined whether the power sales company has transaction risks. In the case of transaction risks, corresponding measures are proposed to the power sales company based on the content signed in the target performance bond. In this way, the default behavior of the power sales company is constrained or even avoided, which effectively reduces the risk in the power transaction process.

[0047] Optionally, the transaction risk identification model is obtained based on machine learning based on historical transaction information, historical text information and corresponding historical risk identification results (sample data sets) corresponding to multiple historical power transactions. In order to avoid the confusion of the above-mentioned multiple data information (historical transaction information, historical text information and historical risk identification results) corresponding to multiple historical transactions, and the difficulty in finding corresponding information during the application process, the multiple data information corresponding to the above-mentioned multiple historical transactions are integrated into a table, wherein the above-mentioned table at least includes the target power transaction information shown in Table 1, the key information in the target performance bond shown in Table 2, the sales company's turnover profit summary information shown in Table 3, and the target power transaction process data summary information shown in Table 4. Tables 3 and 4 are shown below.

[0048] Table 3

[0049]

[0050] Table 4

[0051]

[0052]

[0053] Optionally, before aggregating all the above information, it is also necessary to preliminarily confirm the historical risk identification results corresponding to multiple historical power transactions based on the information in all the above tables. Specifically, if the amount to be paid recorded in the historical performance bond is greater than the actual payment amount v, there is a non-compliance risk, and the characteristic value corresponding to the historical risk identification result is 1, otherwise it is 0; if the settlement amount between the power sales company and the power grid trading party during the identification period is less than the settlement amount between the power sales company and the power user, there is a settlement risk, and the characteristic value corresponding to the historical risk identification result is 1, otherwise it is 0; if the validity period of the historical performance bond does not match the power transaction contract period, there is a validity period mismatch risk, and the characteristic value corresponding to the historical risk identification result is 1, otherwise it is 0; if the information of the insured or beneficiary is inconsistent with the power transaction contract information, there is an information inconsistency risk, and the characteristic value corresponding to the historical risk identification result is 1, otherwise it is 0. The factors considered in the historical risk identification results are shown in Table 5, where Table 5 is a summary table of factors considered in the historical risk identification results.

[0054] Table 5

[0055]

[0056] Optionally, if the amount payable recorded in the historical performance bond is greater than the actual payment amount v, there is a risk of non-compliance. For power sales companies that are not participating in market transactions for the first time, the actual payment amount v is calculated based on the total electricity volume of historical transactions. The specific calculation method is as follows:

[0057] v = max{e1*0.8min / kWh, max{e2,e3}*5min / kWh} / 100

[0058] Among them, e1 represents the total electricity volume settled in market transactions in the past 12 months, in kilowatt-hours; e2 represents the total electricity volume settled in market transactions in the past two months, in kilowatt-hours; e3 represents the electricity volume settled in retail market transactions in the past two months, in kilowatt-hours.

[0059] For power sales companies participating in market transactions for the first time, the actual payment amount is calculated based on the estimated power sales in the remaining months of the transaction year. The specific calculation method is as follows:

[0060] v = e4*0.8 points / kWh / 100

[0061] Among them, e4 represents the total estimated electricity sales in the remaining months of the transaction year, in kilowatt-hours.

[0062] Optionally, based on all the contents in Tables 1 to 5 above, a summary table of information required for transaction risk identification is formed as shown in Table 6.

[0063] Table 6

[0064]

[0065] In an optional embodiment, when the historical transaction information includes the first historical transaction information and the second historical transaction information, the corresponding historical text information includes the first historical text information and the second historical text information, and the historical risk identification result includes the first historical risk identification result and the second historical risk identification result, before obtaining the target risk identification result of the target power transaction by adopting the transaction risk identification model based on the key information and the target power transaction information, the method also includes: using the first historical transaction information and the first historical text information as input and the first historical risk identification result as output, and using the gradient boosting decision tree algorithm for machine learning to obtain a first risk identification model; based on the second historical transaction information, the second historical text information and the second historical risk identification result, verifying the first risk identification model to obtain a verification result; when the verification result indicates that the first risk identification model has passed the verification, using the first risk identification model as the transaction risk identification model.

[0066] Optionally, a transaction risk identification model is constructed based on historical transaction information, historical text information and corresponding historical risk identification results, wherein the historical transaction information is the actual transaction information generated in the power transaction before the current power transaction, the historical text information is the key information extracted from the historical performance bond signed in the power transaction before the current power transaction, and the historical risk identification result is the risk identification result determined by considering the factors in Table 5 above.

[0067] Optionally, a gradient boosting decision tree algorithm (GBDT) is used to construct the first risk identification model through machine learning. The process of constructing the first risk identification model by the gradient boosting decision tree includes a training process, and the training process is an iterative process, which gradually optimizes the prediction performance of the first risk identification model by constructing multiple decision trees. GBDT is a regression algorithm, and classification tasks can also be achieved by adjusting the loss function.

[0068] Optionally, through the GBDT algorithm, a first risk identification model based on the GBDT algorithm is constructed to output a target risk identification result, wherein the target risk identification result is a risk level, which is divided into four categories: low risk, medium risk, high risk, and extremely high risk. This realizes the identification of performance risks in power transactions and helps the transaction process to more accurately understand the potential status of performance risks so as to adopt corresponding risk management strategies.

[0069] In an optional embodiment, the first historical transaction information and the first historical text information are used as input, and the first historical risk identification result is used as output, and a gradient boosting decision tree algorithm is used for machine learning to obtain a first risk identification model, including: determining multiple initial decision trees and a loss function; using a preset initial prediction result as a current prediction result; using a gradient boosting decision tree algorithm to determine a current decision tree based on the current prediction result and the loss function; starting from the first decision tree among the multiple initial decision trees, taking the multiple initial decision trees as current decision trees respectively, and iteratively training the multiple initial decision trees by cyclically performing the following operations until a predetermined termination condition is reached: training the current decision tree based on the loss function, the current prediction result, the first historical transaction information, and the first historical text information to obtain a target decision tree, and a first prediction result corresponding to the target decision tree; summing the current prediction result and the first prediction result to obtain a new current prediction result; and constructing a first risk identification model based on the multiple target decision trees obtained when the predetermined termination condition is reached, wherein the multiple target decision trees correspond one-to-one to the multiple initial decision trees.

[0070] Optionally, to construct a first risk identification model, first, a sample set needs to be constructed, namely, the first historical transaction information and the first historical text information as input x i =(x i1 ,x i2 ,…,x im ), and the first historical risk identification result as output y i Specifically, we construct a sample set S = {(x1, y1), (x2, y2), …, (x n ,y n)}, where n represents the number of samples, x i represents the characteristic value of the characteristic variable corresponding to the 5th to 29th rows in Table 5 for the i-th sample, y i Indicates the possible risk feature values ​​corresponding to rows 1 to 4 in Table 5 corresponding to the i-th sample.

[0071] Optionally, based on the sample set S constructed above, the GBDT algorithm is used to train the first risk identification model of the power sales company's performance, the number of iterations is T, L is the loss function, where the loss function uses the mean square error and the decision tree is used as a weak classifier. The GBDT algorithm process is as follows:

[0072] Step S1081, based on the current prediction result and the loss function, construct the current decision tree as an initial weak learner, and the specific construction is as follows:

[0073]

[0074] in, is the loss function, c is the loss function L(y i ,c) The current prediction result obtained when the value is the smallest.

[0075] Step S1082, after t rounds of iterations, t = 1, 2, ..., T, the initialized weak learner is updated to obtain a strong learner. The specific updating process is as follows:

[0076] Step S10821, based on the first residual value obtained by the loss function, update the relationship between the input (the first historical transaction information and the first historical text information) and the output (the first historical risk identification result) to obtain a new risk identification result r ti , calculated as follows:

[0077]

[0078] in, To find the partial derivative of the loss function, To find the partial derivative of the weak learner, i = 1, 2, ..., n is the number of samples in the sample set. Using the mean square error as the loss function, the above new risk identification result calculation formula is simplified to obtain the following formula:

[0079] r ti =-(f t-1 (x i )-y i )

[0080] Step S10822: Based on the above new risk identification result r ti , construct a new sample set, where the new sample set is S t ={(x1,rt1 ),(x2,r t2 ),…,(x n ,r tn )}, based on the sample set S t Fit the tth initial decision tree, take the t initial decision trees as the current decision trees respectively, and train the t initial decision trees by looping and iterating until a predetermined termination condition is reached, wherein the predetermined termination condition is that a predetermined number of iterations can be reached, and all decision trees are updated in sequence in each iteration.

[0081] Step S10823: Each initial decision tree includes K leaf nodes, where each leaf node corresponds to a leaf node region R tk , k = 1, 2, ..., K, based on the sample set S t , Leaf node area R tk And the loss function is used to determine the second residual value between the new risk identification result and the input (the first historical transaction information and the first historical text information), wherein the best fit between the new risk identification result and the input (the first historical transaction information and the first historical text information) is determined based on the second residual value, and the best fit is the complementary value of the residual value. The specific calculation is as follows:

[0082]

[0083] Among them, L is the loss function, β is the loss function L(r ti ,f t-1 (x i )+β) value is the smallest, and the current prediction result is obtained.

[0084] Step S10823: based on the second residual value obtained above, update the weak learner to obtain a strong learner. The strong learner is specifically as follows:

[0085]

[0086] Among them, I is the preset function, in x∈R tk In the case of tk )=1, otherwise I(x∈R tk )=0.

[0087] Step S1083: output the first transaction risk identification model, which is as follows:

[0088]

[0089] Optionally, based on the above-mentioned gradient boosting decision tree algorithm, multiple initial decision trees are constructed, each of which has multiple leaf nodes. The initial decision tree divides the input space into a number of non-overlapping leaf node regions through a series of split operations, and each leaf node region corresponds to a leaf node. These leaf node regions are determined based on the values ​​of the input features (the first historical transaction information and the first historical text information), that is, the difference in the input of the first transaction risk identification model will cause the above-mentioned input to be divided into different leaf nodes, thereby obtaining the output in the corresponding leaf node region.

[0090] In an optional embodiment, the second historical transaction information is a plurality of pieces, and based on the second historical transaction information, the second historical text information and the second historical risk identification result, the first risk identification model is verified to obtain a verification result, including: inputting the plurality of second historical transaction information and the plurality of second historical transaction information corresponding to the second historical text information into the first risk identification model respectively, to obtain the predicted risk identification results corresponding to the plurality of second historical transaction information respectively; based on the second historical risk identification results corresponding to the plurality of second historical transaction information respectively, and the predicted risk identification results corresponding to the plurality of second historical transaction information respectively, the precision rate of the first risk identification model is determined; wherein, The precision rate is used to indicate the proportion of the number of second historical risk identification results that match the predicted risk identification result among multiple second historical risk identification results, to the total number of multiple second historical risk identification results; based on the second historical risk identification results corresponding to the multiple second historical transaction information respectively, the multiple second historical transaction information respectively corresponds to the predicted risk identification results, and the recall rate of the first risk identification model is determined; wherein, the recall rate is used to indicate the proportion of the number of predicted risk identification results that match the second historical risk identification result among the multiple predicted risk identification results, to the total number of the multiple predicted risk identification results; based on the precision rate and the recall rate, the verification result is determined.

[0091] Optionally, based on the second historical risk identification results corresponding to the plurality of second historical transaction information, and the predicted risk identification results corresponding to the plurality of second historical transaction information, the precision rate of the first risk identification model is determined. l The specific calculation is as follows:

[0092]

[0093] Among them, l is the risk level, TP l Indicates that the actual risk is the lth type of risk, and the first transaction risk identification model predicts the output of the first historical risk identification result is also the number of samples of the lth type of risk, FP l It indicates the number of samples that are not actually the lth type of risk, but the first transaction risk identification model predicts and outputs the first historical risk identification result as the lth type of risk.

[0094] Optionally, based on the second historical risk identification results corresponding to the plurality of second historical transaction information, and the predicted risk identification results corresponding to the plurality of second historical transaction information, the recall rate of the first risk identification model is determined. l The specific calculation is as follows:

[0095]

[0096] Among them, FN l It indicates the number of samples that are actually the lth type of risk, but the first transaction risk identification model predicts and outputs the first historical risk identification result which is not the lth type of risk.

[0097] Optionally, the verification result is determined based on the precision and recall, where the verification result F1-score l The specific calculation is as follows:

[0098]

[0099] When the verification result is greater than the preset result threshold, it is determined that the first risk identification model has passed the verification, and the first risk identification model is used as the transaction risk identification model.

[0100] In an optional embodiment, based on key information and target power transaction information, a transaction risk identification model is used to obtain a target risk identification result of the target power transaction, including: inputting the key information and the target power transaction information into multiple target decision trees in the transaction risk identification model respectively, to obtain risk identification results corresponding to the multiple target decision trees respectively; determining weight values ​​corresponding to the multiple target decision trees respectively; based on the weight values ​​corresponding to the multiple target decision trees respectively, performing weighted calculations on the risk identification results corresponding to the multiple target decision trees respectively, to obtain a target risk identification result.

[0101] Optionally, the key information and the target power transaction information are respectively input into the transaction risk identification model to obtain the target risk identification result, wherein the target risk identification result specifically retains the low risk level, the medium risk level, the high risk level and the extremely high risk level. Specifically, the low risk level means that there are fewer potential risks, the contract terms are clear, the parties are reputable, and the relevant industry and market conditions are relatively stable. The probability of risk occurrence is low, and the impact on the guarantor and the beneficiary is limited. The medium risk level means that there is a certain performance risk, which may involve certain uncertainties, such as market fluctuations, economic changes or risks in specific industries. The performance capabilities of the parties are relatively good, but potential risks need to be monitored and managed more closely. The high risk level means that there are greater risks, which may involve market, policy or technological changes with high uncertainty. There are certain risks in the performance capabilities of the parties, and monitoring and response measures need to be strengthened to deal with possible risk events. The extremely high risk level means that there are great potential risks, which may be caused by extremely unstable market conditions or technical challenges. The performance capabilities of the parties face major uncertainties, and active risk management strategies and early warning mechanisms need to be adopted. Based on the target risk identification results, a level classification is performed, and based on the level classification, it is determined whether to continue to cooperate with the power sales company. The specific risk levels and feature comparison table are shown in Table 7, where Table 7 is a risk level and feature comparison table.

[0102] Table 7

[0103]

[0104] Through the above steps S102 to S108, the purpose of intelligently extracting key information in the target performance bond during the power trading risk identification process and intelligently obtaining the target risk identification results based on the trading risk identification model can be achieved, thereby achieving the technical effect of improving the intelligence level in the power trading risk identification process and reducing the investment cost of power trading risk identification, thereby solving the technical problems of low intelligence level and high human resource investment cost in the power trading risk identification method in related technologies.

[0105] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation mode: Figure 2 is a flow chart of an optional method for identifying power trading risks according to an embodiment of the present invention. Figure 2 As shown, the method includes:

[0106] Step S1, target text information extraction:

[0107] Optical character recognition (OCR) technology is used to efficiently extract the target text information from the image corresponding to the target performance bond. The process of extracting target text information using optical character recognition (OCR) technology involves image preprocessing, text positioning, and text recognition steps, effectively realizing the accurate and comprehensive conversion of images into processable text data.

[0108] Step S2, extract key information from the target performance bond:

[0109] By adopting the method of preset matching rules, key information is obtained from the target text information corresponding to the target performance bond, wherein the key information at least includes the identity information, validity period, guarantee amount and other contents of the insured.

[0110] Step S3, performance risk feature extraction:

[0111] Obtain the target power transaction information in the target power transaction, integrate the above key information and the target power transaction information, that is, organize them into the same table, so as to facilitate the subsequent power transaction risk identification process, quickly find the corresponding data, and integrate the power transaction process characteristics, performance risk characteristics, and basic credit characteristics.

[0112] Step S4, Identification of Power Transaction Performance Risks:

[0113] Through the gradient boosting decision tree algorithm (GBDT, Gradient Boosting Decision Tree), a transaction risk identification model based on the GBDT algorithm was constructed, and the risk levels were divided into four categories: low risk, medium risk, high risk, and extremely high risk. This realizes the identification of performance risks in the power trading process, helping decision makers to understand the potential status of performance risks more accurately so as to adopt corresponding risk management strategies.

[0114] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards in the relevant regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0115] In this embodiment, a power trading risk identification device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the terms "module" and "device" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0116] According to an embodiment of the present invention, there is also provided an embodiment of a device for implementing the above-mentioned power transaction risk identification method. Figure 3 is a schematic diagram of a structure of a power transaction risk identification device according to an embodiment of the present invention. Figure 3 As shown, the above-mentioned power transaction risk identification device includes: a target text information extraction module 300, an acquisition module 302, a key information extraction module 304 and a target risk identification result determination module 306, wherein:

[0117] A target text information extraction module 300 is used to extract target text information from a target performance bond corresponding to a target power transaction;

[0118] An acquisition module 302, connected to the target text information extraction module 300, is used to acquire target power transaction information obtained by executing the target power transaction in the target period;

[0119] The key information extraction module 304 is connected to the acquisition module 302 and is used to extract the key information in the target text information;

[0120] The target risk identification result determination module 306 is connected to the key information extraction module 304 and is used to obtain the target risk identification result of the target power transaction based on the key information and the target power transaction information by adopting the transaction risk identification model, wherein the target risk identification result at least includes the risk level of the target power transaction in the target period;

[0121] Among them, the transaction risk identification model is obtained through machine learning based on historical transaction information, historical text information and corresponding historical risk identification results corresponding to multiple historical power transactions. The historical transaction information is used to indicate the power transaction information obtained by executing multiple historical power transactions respectively, and the historical text information is extracted from the historical performance bonds corresponding to multiple historical power transactions.

[0122] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0123] It should be noted that the target text information extraction module 300, the acquisition module 302, the key information extraction module 304 and the target risk identification result determination module 306 correspond to steps S102 to S108 in the embodiment, and the examples and application scenarios implemented by the modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment. It should be noted that the modules as part of the device can be run in a computer terminal.

[0124] It should be noted that the optional or preferred implementation of this embodiment can refer to the relevant description in the embodiment, which will not be repeated here.

[0125] The above-mentioned power trading risk identification device may also include a processor and a memory. The above-mentioned target text information extraction module 300, acquisition module 302, key information extraction module 304 and target risk identification result determination module 306 are all stored in the memory as program modules, and the processor executes the above-mentioned program modules stored in the memory to realize corresponding functions.

[0126] The processor includes a kernel, which retrieves the corresponding program module from the memory. The kernel may be one or more. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. The memory includes at least one memory chip.

[0127] According to an embodiment of the present application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein when the program is executed, the device where the non-volatile storage medium is located is controlled to execute any of the above-mentioned power trading risk identification methods.

[0128] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group, and the non-volatile storage medium includes a stored program.

[0129] Optionally, when the program is running, the device where the non-volatile storage medium is located is controlled to perform the following functions: extract target text information from the target performance bond corresponding to the target power transaction; obtain the target power transaction information obtained by executing the target power transaction in the target time period; extract key information from the target text information; based on the key information and the target power transaction information, use a transaction risk identification model to obtain a target risk identification result of the target power transaction, wherein the target risk identification result includes at least the risk level of the target power transaction in the target time period; wherein the transaction risk identification model is obtained through machine learning based on historical transaction information, historical text information and corresponding historical risk identification results corresponding to multiple historical power transactions, wherein the historical transaction information is used to indicate the power transaction information obtained by executing multiple historical power transactions respectively, and the historical text information is extracted from historical performance bonds corresponding to multiple historical power transactions.

[0130] According to an embodiment of the present application, an embodiment of a processor is also provided. Optionally, in this embodiment, the processor is used to run a program, wherein the program executes any one of the above-mentioned power transaction risk identification methods when running.

[0131] According to an embodiment of the present application, an embodiment of a computer program product is also provided. Optionally, in this embodiment, the computer program product includes a computer program, and when the computer program is executed by a processor, the computer program implements any of the steps of the power trading risk identification method.

[0132] Optionally, the above-mentioned computer program product, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: extracting target text information from the target performance bond corresponding to the target power transaction; obtaining the target power transaction information obtained by executing the target power transaction in the target time period; extracting key information from the target text information; based on the key information and the target power transaction information, using a transaction risk identification model to obtain a target risk identification result for the target power transaction, wherein the target risk identification result includes at least the risk level of the target power transaction in the target time period; wherein the transaction risk identification model is obtained through machine learning based on historical transaction information, historical text information and corresponding historical risk identification results corresponding to multiple historical power transactions, wherein the historical transaction information is used to indicate the power transaction information obtained by executing multiple historical power transactions respectively, and the historical text information is extracted from historical performance bonds corresponding to multiple historical power transactions.

[0133] An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the program: extracting target text information from a target performance bond corresponding to a target power transaction; acquiring target power transaction information obtained by executing a target power transaction in a target time period; extracting key information from the target text information; and obtaining a target risk identification result of the target power transaction using a transaction risk identification model based on the key information and the target power transaction information, wherein the target risk identification result includes at least the risk level of the target power transaction in the target time period; wherein the transaction risk identification model is obtained through machine learning based on historical transaction information, historical text information, and corresponding historical risk identification results corresponding to multiple historical power transactions, wherein the historical transaction information is used to indicate power transaction information respectively obtained by executing multiple historical power transactions, and the historical text information is extracted from historical performance bonds corresponding to multiple historical power transactions.

[0134] The above sequence of the embodiments of the present invention is for description only and does not represent the superiority or inferiority of the embodiments.

[0135] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0136] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the above modules can be a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, modules or indirect coupling or communication connection of modules, which can be electrical or other forms.

[0137] The modules described above as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0138] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of software functional modules.

[0139] If the above-mentioned integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a non-volatile storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the various embodiments of the present invention. The aforementioned non-volatile storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0140] The above are only preferred embodiments of the present invention. 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 invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for identifying power trading risks, characterized in that: include: Extracting target text information from a target performance bond corresponding to a target power transaction; Acquire target power transaction information obtained by executing the target power transaction during the target period; Extracting key information from the target text information; Based on the key information and the target power transaction information, a transaction risk identification model is used to obtain a target risk identification result of the target power transaction, wherein the target risk identification result at least includes a risk level of the target power transaction in the target period; Among them, the transaction risk identification model is obtained through machine learning based on historical transaction information, historical text information and corresponding historical risk identification results corresponding to multiple historical power transactions. The historical transaction information is used to indicate the power transaction information obtained by executing the multiple historical power transactions respectively, and the historical text information is extracted from the historical performance bonds corresponding to the multiple historical power transactions.

2. The method according to claim 1, characterized in that Target Performance Bond Before extracting the target text information from the target performance bond corresponding to the target power transaction, the method further includes: obtaining an initial image of the target performance bond; Preprocessing the initial image to obtain a preprocessed image; wherein the preprocessing includes at least one of the following: grayscale processing, binarization processing, and denoising processing; Using an optical character recognition method to extract features of characters in the preprocessed image to obtain initial text information; The initial text information is post-processed to obtain the target text information, wherein the post-processing includes at least one of the following: error correction and character connection.

3. The method according to claim 2, characterized in that The step of extracting key information from the target text information includes: Extracting multiple groups of identifiers from the target text information based on a preset matching rule, wherein each group of identifiers in the multiple groups of identifiers includes a title in the target text information and a title level corresponding to the title; Based on the title levels in the multiple groups of identifiers, determining the hierarchical structures corresponding to the multiple groups of identifiers respectively; The key information is obtained based on the information pointed to by the hierarchical structures respectively corresponding to the multiple groups of identifiers.

4. The method according to claim 3, characterized in that The obtaining of the key information based on the information pointed to by the hierarchical structures respectively corresponding to the multiple groups of identifiers includes: Using the hierarchical structures corresponding to the multiple groups of identifiers as nodes, a topological relationship network is constructed, wherein the topological relationship network is used to indicate the association relationship between the hierarchical structures corresponding to the multiple groups of identifiers; Based on the topological relationship network and the titles in the multiple groups of identifiers, determining a title at a preset node in the topological relationship network; The title at the preset node is determined to be a preset title, and the text information corresponding to the preset title in the target text information is used as the key information.

5. The method according to any one of claims 1 to 4, characterized in that: In the case where the historical transaction information includes first historical transaction information and second historical transaction information, the corresponding historical text information includes first historical text information and second historical text information, and the historical risk identification result includes first historical risk identification result and second historical risk identification result, before obtaining the target risk identification result of the target power transaction by using the transaction risk identification model based on the key information and the target power transaction information, the method further includes: Taking the first historical transaction information and the first historical text information as input and the first historical risk identification result as output, a gradient boosting decision tree algorithm is used for machine learning to obtain a first risk identification model; Based on the second historical transaction information, the second historical text information, and the second historical risk identification result, verifying the first risk identification model to obtain a verification result; When the verification result indicates that the first risk identification model passes the verification, the first risk identification model is used as the transaction risk identification model.

6. The method according to claim 5, characterized in that The method uses the first historical transaction information and the first historical text information as input, takes the first historical risk identification result as output, and uses a gradient boosting decision tree algorithm to perform machine learning to obtain a first risk identification model, including: Determine multiple initial decision trees and loss functions; The preset initial prediction result is used as the current prediction result; Using the gradient boosting decision tree algorithm, based on the current prediction result and the loss function, determine the current decision tree; Starting from the first decision tree among the multiple initial decision trees, the multiple initial decision trees are respectively used as the current decision trees, and the multiple initial decision trees are iteratively trained by cyclically performing the following operations until a predetermined termination condition is reached: The current decision tree is trained based on the loss function, the current prediction result, the first historical transaction information, and the first historical text information to obtain a target decision tree and a first prediction result corresponding to the target decision tree; the current prediction result and the first prediction result are summed to obtain a new current prediction result; The first risk identification model is constructed based on a plurality of target decision trees obtained when the predetermined termination condition is reached, wherein the plurality of target decision trees correspond one-to-one to the plurality of initial decision trees.

7. The method according to claim 5, characterized in that The second historical transaction information is a plurality of pieces, and the first risk identification model is verified based on the second historical transaction information, the second historical text information, and the second historical risk identification result to obtain a verification result, including: Inputting a plurality of second historical transaction information and the second historical text information respectively corresponding to the plurality of second historical transaction information into the first risk identification model to obtain predicted risk identification results respectively corresponding to the plurality of second historical transaction information; Based on the second historical risk identification results respectively corresponding to the plurality of second historical transaction information, and the plurality of second historical transaction information respectively corresponding to the predicted risk identification results, determining the precision rate of the first risk identification model; wherein the precision rate is used to indicate the proportion of the number of second historical risk identification results that match the predicted risk identification result among the plurality of second historical risk identification results to the total number of the plurality of second historical risk identification results; Based on the second historical risk identification results respectively corresponding to the plurality of second historical transaction information, and the plurality of second historical transaction information respectively corresponding to the predicted risk identification results, determining the recall rate of the first risk identification model; wherein the recall rate is used to indicate the proportion of the number of predicted risk identification results that match the second historical risk identification result among the plurality of predicted risk identification results to the total number of the plurality of predicted risk identification results; The verification result is determined based on the precision rate and the recall rate.

8. The method according to claim 6, characterized in that The step of obtaining a target risk identification result of the target power transaction by using a transaction risk identification model based on the key information and the target power transaction information includes: Inputting the key information and the target power transaction information into a plurality of target decision trees in the transaction risk identification model respectively, and obtaining risk identification results corresponding to the plurality of target decision trees respectively; Determine the weight values ​​corresponding to the multiple target decision trees respectively; Based on the weight values ​​respectively corresponding to the multiple target decision trees, the risk identification results respectively corresponding to the multiple target decision trees are weightedly calculated to obtain the target risk identification result.

9. A device for identifying risks in power trading, characterized in that: include: A target text information extraction module, used to extract target text information from a target performance bond corresponding to a target power transaction; An acquisition module, used for acquiring target power transaction information obtained by executing the target power transaction during a target period; A key information extraction module, used to extract key information from the target text information; a target risk identification result determination module, configured to obtain a target risk identification result of the target power transaction based on the key information and the target power transaction information by adopting a transaction risk identification model, wherein the target risk identification result at least includes a risk level of the target power transaction in the target period; Among them, the transaction risk identification model is obtained through machine learning based on historical transaction information, historical text information and corresponding historical risk identification results corresponding to multiple historical power transactions. The historical transaction information is used to indicate the power transaction information obtained by executing the multiple historical power transactions respectively, and the historical text information is extracted from the historical performance bonds corresponding to the multiple historical power transactions.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for identifying power trading risks described in any one of claims 1 to 8 are implemented.