Generation method of target risk information of power retail market and electronic equipment

Abstract

The invention discloses a power retail market target risk information generation method and electronic equipment. According to the method, the risk factor extraction algorithm for the unstructured text is introduced, and dynamic weighted fusion is carried out on the risk factor extraction algorithm and the monitoring indexes based on the structured data, so that blind areas of traditional risk monitoring on soft risk factors such as policies and public opinions are effectively filled up, and the limitation of a single data source and a static analysis model is overcome; therefore, comprehensive risk information which is more complete in coverage dimension, higher in timeliness and higher in precision is generated, and more reliable data support is provided for risk identification and early warning of the power retail market.

Description

Technical Field

[0001] This invention relates to the technical field of electricity market risk monitoring, specifically to a method for generating target risk information in the electricity retail market and an electronic device thereof. Background Technology

[0002] In related technologies, with the deepening of power system reform, the electricity retail market is characterized by diversified players and complex transactions, and market risk monitoring has become a key link in ensuring the stable operation of the market.

[0003] In related technologies, there is a technical problem that the risk monitoring signals are not accurate enough because the multi-source heterogeneous data of the electricity retail market cannot be effectively integrated and the text risks cannot be dynamically quantified. Summary of the Invention

[0004] The technical problem this invention aims to solve is that, in related technologies, the failure to effectively integrate multi-source heterogeneous data from the electricity retail market and dynamically quantify textual risks leads to inaccurate risk monitoring signals. The purpose is to provide a method and electronic device for generating target risk information in the electricity retail market, thereby resolving the technical problem of inaccurate risk monitoring signals.

[0005] This invention is achieved through the following technical solution:

[0006] In a first aspect, the present invention provides a method for generating target risk information in the electricity retail market, comprising:

[0007] Obtain a first dataset and a second dataset; wherein the first dataset is structured data used to characterize the status of core elements in the electricity retail market, and the second dataset is unstructured data used to characterize the market environment and policy situation;

[0008] Based on a preset risk factor extraction algorithm, the second dataset is processed to obtain risk factor information for at least one risk type; wherein, the preset risk factor extraction algorithm is an algorithm used to identify risk events from unstructured text and quantify their degree of impact;

[0009] Based on the preset risk monitoring indicator system for the electricity retail market, the first dataset is processed to obtain structured risk information for at least one risk type.

[0010] The risk factor information and the structured risk information of the same risk type are weighted and fused to generate the target risk information of that risk type.

[0011] Further, the step of processing the second dataset based on a preset risk factor extraction algorithm to obtain risk factor information for at least one risk type includes:

[0012] Based on a pre-built dictionary of electricity market risk keywords, risk-related text content is identified from the second dataset; wherein, the text content includes risk keywords.

[0013] Using a pre-defined named entity recognition algorithm, key entities constituting the risk event are extracted from the identified text content;

[0014] Based on predefined risk impact calculation rules and the key entities, the degree of risk impact of the risk event is calculated;

[0015] Based on a preset sentiment analysis algorithm, the sentiment polarity of text content including the risk keywords is determined, and the corresponding risk impact direction coefficient is output.

[0016] By combining the risk impact degree of at least one of the aforementioned risk events and its corresponding risk impact direction coefficient, the risk factor information of the risk type is calculated.

[0017] Furthermore, the electricity market risk keyword dictionary is pre-constructed using the following method:

[0018] Based on a pre-set set of seed keywords for electricity market risks, a word embedding model is trained on a pre-acquired historical text corpus to calculate extended keywords that are semantically similar to the seed keywords.

[0019] The seed keywords and the extended keywords are merged to generate the electricity market risk keyword dictionary.

[0020] Furthermore, the step of calculating the degree of risk impact of the risk event based on predefined risk impact calculation rules and the key entity includes:

[0021] Based on the risk event type indicated by the key entity, the corresponding basic impact factor is queried from the pre-built event-risk knowledge base;

[0022] Based on the risk impact scope information indicated by the key entities, calculate the scope impact factor;

[0023] Based on the risk consequences information indicated by the key entities, severity impact factors are determined from a pre-built consequences-risk knowledge base;

[0024] The degree of risk impact of the risk event is calculated based on the product of the basic impact factor, the range impact factor, and the severity impact factor.

[0025] Furthermore, the preset risk monitoring indicator system for the electricity retail market includes multiple risk categories. Each risk category includes risk factor indicators and risk outcome indicators, and the indicators are defined for the overall market and individual market entities, respectively.

[0026] The step of processing the first dataset based on the preset electricity retail market risk monitoring indicator system to obtain structured risk information for at least one risk type includes:

[0027] Based on the target risk type, determine the corresponding risk factor indicators and risk outcome indicators from the indicator system;

[0028] Based on the first dataset, calculate the overall market indicator value and the individual market participant indicator value corresponding to the target risk type, respectively.

[0029] Based on a preset weighting scheme, the overall market indicator value and the individual market participant indicator value are weighted and fused to obtain the structured risk information of the target risk type.

[0030] Furthermore, the aforementioned risk categories include market power risk, price risk, operational risk, contract risk, performance risk, settlement risk, compliance risk, and policy risk;

[0031] Among them, the risk factor indicators corresponding to the market power risk include at least the market concentration index defined for the overall market and the price deviation rate index defined for individual market participants; the risk outcome indicators corresponding to the market power risk include at least the proportion of active electricity sales companies defined for the overall market and the total market share index defined for individual market participants.

[0032] Among them, the risk factor indicators corresponding to the price risk include at least a wholesale-retail price correlation indicator defined for the overall market and a wholesale-retail price correlation indicator for individual market participants; the risk outcome indicators corresponding to the price risk include at least a wholesale-retail price inversion rate indicator defined for the overall market and a wholesale-retail price inversion rate indicator for individual market participants.

[0033] The risk factor indicators corresponding to the operational risks include at least the wholesale market price volatility indicator defined for the overall market and the load forecast accuracy indicator defined for individual market participants; the risk outcome indicators corresponding to the operational risks include at least the proportion of loss-making electricity sales companies defined for the overall market and the user stickiness indicator defined for individual market participants.

[0034] Further, the step of weightedly fusing the risk factor information and the structured risk information of the same risk type to generate target risk information for that risk type includes:

[0035] Based on the risk type, a first weighting coefficient configured for the risk factor information and a second weighting coefficient configured for the structured risk information are obtained.

[0036] Calculate the product of the first weighting coefficient and the risk factor information to obtain the first weighted result;

[0037] Calculate the product of the second weighting coefficient and the structured risk information to obtain the second weighted result;

[0038] The first weighted result and the second weighted result are summed to generate the target risk information for this risk type.

[0039] Secondly, the present invention provides an apparatus for generating target risk information in the electricity retail market, comprising:

[0040] The acquisition module is used to acquire a first dataset and a second dataset; wherein the first dataset is structured data used to characterize the status of core elements in the electricity retail market, and the second dataset is unstructured data used to characterize the market environment and policy situation.

[0041] An extraction module is used to process the second dataset based on a preset risk factor extraction algorithm to obtain risk factor information for at least one risk type; wherein, the preset risk factor extraction algorithm is an algorithm used to identify risk events from unstructured text and quantify their impact.

[0042] The processing module is used to process the first dataset based on a preset electricity retail market risk monitoring indicator system to obtain structured risk information of at least one risk type.

[0043] The weighted fusion module is used to weightedly fuse the risk factor information and the structured risk information of the same risk type to generate target risk information of that risk type.

[0044] Thirdly, the present invention provides an electronic device, comprising: a memory, and one or more processors communicatively connected to the memory; the memory stores instructions executable by the one or more processors, the instructions being executed by the one or more processors to cause the one or more processors to implement the method described above.

[0045] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0046] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0047] This embodiment introduces a risk factor extraction algorithm for unstructured text and dynamically weights and integrates it with monitoring indicators based on structured data. This effectively fills the blind spots of traditional risk monitoring in terms of soft risk factors such as policies and feedback information, and overcomes the limitations of single data sources and static analysis models. As a result, it generates comprehensive risk information with more complete coverage, stronger timeliness, and higher accuracy, providing more reliable data support for risk identification and early warning in the electricity retail market. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0049] Figure 1 A flowchart illustrating a method for generating target risk information in the electricity retail market, as provided in the embodiments of this specification;

[0050] Figure 2 This is an architecture diagram of a method for generating target risk information in the electricity retail market, as provided in the embodiments of this specification.

[0051] Figure 3 A block diagram of an apparatus for generating target risk information in the electricity retail market, provided in the embodiments of this specification;

[0052] Figure 4 This is a block diagram of an electronic device provided in the embodiments of this specification. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0054] In related technologies, electricity retail market risk monitoring systems are often built upon a pre-defined data processing framework. In a possible technical solution, a typical architecture may include a data acquisition layer, a data processing layer, and a risk analysis layer. The data acquisition layer is primarily responsible for acquiring structured data from sources such as the electricity trading center database and the electricity sales company's operating system. The data processing layer cleans and standardizes this data, while the risk analysis layer performs calculations and analyses based on a pre-defined indicator system and evaluation model.

[0055] In related technologies, risk monitoring faces two major technical challenges. First, at the data level, there is an over-reliance on traditional structured data such as transaction prices and load data. Textual data containing important risk signals, such as policies and regulations, market feedback, and industry reports, are difficult to collect and quantify effectively due to their unstructured nature, resulting in a significant lack of risk data sources. Second, at the analysis level, the monitoring indicator systems used in related technologies are mostly statically designed. Their evaluation weights are often set once by domain experts based on historical experience, making it difficult to adapt to the rapid dynamic changes in electricity market policies, prices, and supply-demand relationships.

[0056] The aforementioned technical challenges collectively lead to a core technical problem: the inaccurate risk monitoring signals result from the failure to effectively integrate multi-source heterogeneous data from the electricity retail market and dynamically quantify textual risks. The root cause of this problem lies in the lack of a deep understanding of unstructured textual data and the absence of an analytical mechanism capable of adaptively adjusting to changes in the market environment.

[0057] The core inventive concept of this invention lies in: establishing a fusion processing mechanism for structured and unstructured data, and introducing a risk factor extraction algorithm to achieve dynamic quantification of textual risk, ultimately generating target risk information that can comprehensively and accurately reflect the market risk situation.

[0058] This embodiment provides a method for generating target risk information in the electricity retail market. The execution entity of the method can be a cloud server or server cluster, which can pull or receive data from multiple heterogeneous data sources such as decentralized electricity trading centers, electricity sales companies, and government websites. After completing all processing, the generated target risk information is provided to downstream regulatory agencies or electricity sales company users.

[0059] The method can be implemented by an edge computing device. In scenarios requiring low-latency processing of certain regions or the electricity market, the implementation entity can be an edge server deployed locally at an electricity trading center or a large electricity sales company.

[0060] The method can be implemented by user terminal equipment. Specifically, it can also be implemented by terminal equipment (such as high-performance workstations) used by regulatory personnel or analysts of electricity sales companies.

[0061] like Figure 1 and Figure 2 As shown, the method may include:

[0062] Step S12: Obtain the first dataset and the second dataset; wherein, the first dataset is structured data used to characterize the status of core elements in the electricity retail market, and the second dataset is unstructured data used to characterize the market environment and policy situation.

[0063] In this embodiment, the executing entity can obtain the raw data that forms the basis of the analysis from multiple sources.

[0064] In this embodiment, the first dataset can be structured data. The structured data can be represented as data that can be logically expressed in a two-dimensional table format and organized using a predefined data model (e.g., a relational database schema). Specifically, it can include: transaction data (e.g., date, electricity sales company identifier, electricity volume, price) obtained from the power trading center database via a JDBC / ODBC interface; operational data (e.g., user list, contract status, settlement information, financial indicators) obtained from the electricity sales company's internal system via an API interface; and credit rating data obtained from credit reporting agencies. The aforementioned data fields are clearly defined and of fixed types, reflecting the objective state of core market elements.

[0065] In this embodiment, the second dataset can be unstructured data. Unstructured data can be represented as data that does not have a predefined data model or is unsuitable for representation using a two-dimensional logical table in a database; it can be text data. Specifically, it can include text data collected from publicly available information sources such as government regulatory agency websites, news portals, and industry analysis reports using web crawling technology. This text data can specifically include policy and regulatory documents, industry management regulations, market-related news reports, listed company announcements, and social media feedback, etc. These texts collectively represent the dynamic trends of the external market environment and policy guidance.

[0066] Step S14: Based on a preset risk factor extraction algorithm, process the second dataset to obtain risk factor information for at least one risk type; wherein, the preset risk factor extraction algorithm is an algorithm used to identify risk events from unstructured text and quantify their impact.

[0067] In this embodiment, the preset risk factor extraction algorithm can be an algorithm that transforms unstructured text data into quantifiable risk factor information. Specifically, a pre-built dictionary of electricity market risk keywords can be used to scan the text. When a keyword from the dictionary appears in a sentence, risk event identification is triggered. Then, named entity recognition technology can be used to extract key entities from the sentence. The extracted key entities can collectively constitute a risk event. For example, "subject entity" (name of an electricity sales company), "regulatory entity" (name of a new regulation), and "numerical entity" (amount of fines or unit capacity) can be identified.

[0068] In a specific and possible implementation, the preset risk factor extraction algorithm can be a combination of a dictionary and a rule model. Specifically, firstly, a pre-built dictionary of electricity market risk keywords can be used for text scanning and pattern matching to identify potential risk-related segments. Then, combined with rule-based pattern matching or a conditional random field model, entities of predefined types are extracted from the segments. The quantification of their impact can rely on a pre-configured rule base that defines the mapping relationship between different entity combinations and numerical weights.

[0069] In a specific and possible implementation, the preset risk factor extraction algorithm can be a combination algorithm based on traditional machine learning models. Specifically, the algorithm can treat risk factor extraction as a classification and regression task. It can use feature engineering methods to extract linguistic features (e.g., word frequency, part-of-speech sequence) and domain features from the text, and then use classifiers such as support vector machines and random forests to classify risk events, while using models such as linear regression or decision trees to predict the numerical value of their impact.

[0070] In a specific and possible implementation, the preset risk factor extraction algorithm can be an end-to-end algorithm based on a deep learning model. For example, recurrent neural networks or long short-term memory networks can be used to capture sequence dependencies in text to identify risk events. Attention mechanisms can be used to focus on key information fragments in text to assist in entity extraction and impact assessment. Alternatively, a pre-trained language model can be used, which can be fine-tuned to simultaneously complete multiple sub-tasks such as event classification, entity recognition, and sentiment analysis.

[0071] In this embodiment, quantifying the degree of impact can be represented by assigning a numerical value representing the severity of each identified risk event. Specifically, based on the type of risk event, a basic impact factor can be obtained from a pre-built event-risk knowledge base. Then, a range impact factor can be calculated based on the scope of impact described in the text (e.g., the proportion of substandard unit capacity to total capacity). Next, a severity impact factor can be mapped from another pre-built consequence-risk knowledge base based on the severity of the consequences described in the text (e.g., rectification within a specified period or production shutdown). Finally, the above factors are multiplied to obtain the risk impact degree value of the event. Furthermore, the sentiment polarity (positive / negative) of the text containing the event can be determined using a sentiment analysis model and converted into a risk impact direction coefficient (e.g., +1 or -1).

[0072] In this embodiment, the risk factor information can be represented as a structured data record, which is a summary value obtained by combining the quantitative impact and direction of all relevant risk events for a certain risk type within a given time period. Specifically, it can be a timestamped numerical variable, the magnitude of which represents the overall activity level of the risk type within the corresponding period, while the positive or negative sign of the value represents the direction of the risk factor's impact on the overall market (e.g., a positive value indicates increased risk, and a negative value indicates decreased risk). Specifically, for the same risk type, all identified risk events belonging to that type can be traversed within a defined time period. The risk impact degree value calculated for each risk event is multiplied by the risk impact direction coefficient determined through sentiment analysis to obtain the directional impact value of the event. Finally, the directional impact values ​​of all events belonging to that risk type are summed, and the resulting algebraic sum is the risk factor information for that risk type within the current time period.

[0073] Step S16: Based on the preset risk monitoring indicator system for the electricity retail market, process the first dataset to obtain structured risk information for at least one risk type.

[0074] In this embodiment, this step transforms the raw structured data into standardized information representing specific risk dimensions. The pre-defined electricity retail market risk monitoring indicator system can be predefined, including a set of calculation rules for multiple risk types, each containing multiple specific monitoring indicators. Specifically, in terms of organizational structure, the electricity retail market risk monitoring indicator system can first be divided into eight major risk types according to the inherent attributes of the risk: market power risk, price risk, operational risk, contract risk, performance risk, settlement risk, compliance risk, and policy risk. Under each risk type, it is further divided into two observation dimensions: risk factor indicators and risk outcome indicators. Risk factor indicators can be used to monitor the preconditions and driving factors that may lead to the occurrence of risk, focusing on monitoring the causes of risk. Risk outcome indicators are used to measure the consequences and impacts that have actually manifested, focusing on monitoring the performance of risk. In defining each specific indicator, a macro-level observation perspective targeting the overall market and a micro-level observation perspective targeting a single market entity (e.g., electricity sales company, electricity user) can be set simultaneously, thus forming a three-dimensional indicator system that spans risk type, process stage, and observation level.

[0075] In terms of constituent elements, each specific monitoring indicator can have its data source, calculation logic, and unit of measurement clearly defined. For risk factor indicators under the market power risk type, there could be a market concentration indicator, which quantifies the market's monopolistic tendency by aggregating the market share data of various electricity retailers, calculating the sum of its squares, and multiplying it by a fixed coefficient. There could also be a price deviation rate indicator, which measures a retailer's potential for price manipulation by calculating the ratio of the difference between its retail price and its cost per kilowatt-hour to the retail price.

[0076] In this embodiment, the structured risk information can be represented as a quantitative representation of a certain risk type obtained by calculating, transforming, and aggregating the first dataset (i.e., structured data) through the rules in the preset electricity retail market risk monitoring indicator system.

[0077] Step S18: Weighted fusion of the risk factor information and the structured risk information of the same risk type to generate target risk information of that risk type.

[0078] In this embodiment, two types of risk information obtained from different data sources and different processing paths can be synthesized. The weighted fusion can be represented as assigning corresponding weight coefficients to the two types of information respectively, and then performing a linear combination.

[0079] Specifically, different weighting coefficient combinations can be pre-configured for different risk types (for example, for policy risks, textual risk factor information has a higher weight, and for operational risks, structured risk information has a higher weight). During processing, the corresponding weights can be read according to the current risk type, and the product of risk factor information and its weight, as well as the product of structured risk information and its weight, can be calculated separately. Then, the two product results are added together to finally obtain the target risk information for that risk type.

[0080] In this embodiment, the target risk information can be represented as the final, comprehensive risk quantification characterization generated by weighted fusion of structured risk information and risk factor information of the same risk type.

[0081] This embodiment introduces a risk factor extraction algorithm for unstructured text and dynamically weights and integrates it with traditional monitoring indicators based on structured data. This effectively fills the blind spots of traditional risk monitoring in terms of soft risk factors such as policies and feedback information, and overcomes the limitations of single data sources and static analysis models. As a result, it generates comprehensive risk information with more complete coverage, stronger timeliness, and higher accuracy, providing more reliable data support for risk identification and early warning in the electricity retail market.

[0082] It should be noted that although this implementation method uses risk management in the electricity retail market as its application scenario, it is essentially a typical technical solution that addresses a specific technical problem rather than an abstract business approach. Specifically, firstly, this implementation method designs and implements a complex data processing pipeline, involving the automated collection, cleaning, fusion, and storage of multi-source heterogeneous data (including structured data from databases and unstructured text data crawled from the web). This process relies on various underlying computer technologies such as JDBC / ODBC interfaces, web crawlers, and ETL. Secondly, this implementation method applies multiple Natural Language Processing (NLP) technologies, such as extracting key entities from text using named entity recognition algorithms, determining the polarity of feedback information from risk events using sentiment analysis algorithms, and constructing and expanding a domain dictionary based on word embedding models. Finally, this implementation method proposes a dynamic and quantifiable fusion calculation model that integrates risk factors obtained from text analysis with indicators calculated from structured data using a weighted algorithm to ultimately generate a comprehensive risk quantification value. This process involves mathematical operations such as weight allocation and product summation; essentially, it is a computer-executed technical process with clearly defined input-output and processing logic.

[0083] In some embodiments, the step of processing the second dataset based on a preset risk factor extraction algorithm to obtain risk factor information for at least one risk type includes:

[0084] Step S142: Based on the pre-built dictionary of electricity market risk keywords, identify risk-related text content from the second dataset; wherein the text content includes risk keywords.

[0085] In this embodiment, identifying risk-related text content from the second dataset based on a pre-built dictionary of electricity market risk keywords can be represented as scanning and filtering the text dataset using a pre-created vocabulary that includes terms from the field of electricity market risk as a matching criterion.

[0086] In this embodiment, the electricity market risk keyword dictionary may be a collection of a large number of words and phrases that are highly related to various risks in the electricity retail market, and risk keywords are the carriers of risk description.

[0087] In a specific and possible implementation plan, a keyword dictionary for electricity market risks can be constructed based on domain expert knowledge. This keyword dictionary can include core terms that directly represent risks, such as electricity price adjustments, market manipulation, credit defaults, and new regulatory rules, by summarizing historical risk events and policy texts.

[0088] In a specific and possible implementation, the electricity market risk keyword dictionary can also be a dynamically expanded derivative dictionary based on machine learning technology. The core terms mentioned above can be used as seed words, and word embedding models can be trained on a large-scale power industry corpus. By calculating semantic similarity, extended terms such as price linkage, energy efficiency standards, and load assessment that are highly related to the seed words but not covered by the initial dictionary can be automatically mined and included.

[0089] In a specific and possible implementation, sentences or paragraphs can be used as the basic scanning unit to check whether any one or more risk keywords from the electricity market risk keyword dictionary appear in their content. Once a match is detected, the sentence or paragraph is marked as risk-related text content.

[0090] Step S144: Extract the key entities constituting the risk event from the identified text content using a preset named entity recognition algorithm.

[0091] In this embodiment, the named entity recognition algorithm can be an algorithm used to identify and classify entities with corresponding meanings from text. That is, the model can understand the context and map strings in the text to predefined categories. These predefined categories (i.e., entity types) can be customized for the field of electricity risk. Specifically, they can include: risk subject entities (e.g., specific electricity sales company names, regulatory agency names), risk-related regulatory entities (e.g., the "Basic Rules for Electricity Market Operation"), risk numerical entities (e.g., 15%, a fine of one million yuan), risk consequence entities (e.g., production restrictions, rectification), as well as time entities, location entities, etc.

[0092] In this embodiment, the text content identified in step S142 can be processed using a preset named entity recognition algorithm to output entities that conform to predefined types and their category labels. For example, for the sentence: "A provincial energy bureau issued new regulations requiring units that fail to meet energy efficiency standards to rectify within a specified period," the preset named entity recognition algorithm can extract key entities such as: <provincial energy bureau, institution>, <new regulations, regulations>, <units that fail to meet energy efficiency standards, risk subjects>, <rectification within a specified period, consequences>, etc. These extracted key entities can constitute a structured record of risk events.

[0093] In one specific and possible implementation, the preset named entity recognition algorithm may be a rule-based named entity recognition algorithm.

[0094] In one specific and possible implementation, the preset named entity recognition algorithm can be a named entity recognition algorithm based on statistical machine learning. For example, it can be a conditional random field model.

[0095] In one specific and possible implementation, the preset named entity recognition algorithm can be a deep learning-based named entity recognition algorithm. For example, it could be a model combining a bidirectional long short-term memory network with a conditional random field, or a pre-trained model based on the Transformer architecture.

[0096] In this embodiment, the risk event can be represented as an independent factual description identified from unstructured text, consisting of one or more key entities, having complete semantics, and capable of influencing the risk status of the electricity retail market; specifically, it can be a structured data record.

[0097] In this embodiment, the key entities can be represented as basic semantic units constituting a risk event, specifically including: risk subject entities (market participants or regulatory agencies that bear or cause risks), regulatory and policy entities (laws, regulations, policy documents or industry standards), numerical parameter entities (quantitative information describing the scope of risk impact), consequence description entities (characterizing the possible or already caused state of the risk), time entities, and location entities, etc.

[0098] Step S146: Calculate the degree of risk impact of the risk event based on the predefined risk impact calculation rules and the key entity.

[0099] In this embodiment, the predefined risk impact calculation rule can be expressed as a logical criterion that maps entity information into numerical values. That is, it defines how to combine domain knowledge to quantitatively assess events. This risk impact calculation rule can be a multi-factor multiplication model. For example, the risk impact calculation rule can stipulate that the degree of impact of a risk event is determined by three factors: a basic impact factor, a scope impact factor, and a severity impact factor.

[0100] In a specific and possible implementation, firstly, based on the event type indicated by the risk-related regulatory entity or risk consequence entity extracted in step S144, a corresponding basic impact factor is retrieved from a pre-built event-risk knowledge base. This factor characterizes the baseline impact intensity of such events under normal circumstances. Secondly, using information such as the extracted risk numerical entity and risk subject entity, a scope impact factor is calculated. For example, by querying the database, the proportion of unit capacity affected by the event to the total capacity of the province is calculated. Next, based on the extracted risk consequence entity, a severity impact factor is determined from a pre-built consequence-risk knowledge base. This factor characterizes the severity level of the consequence. Finally, these three factors are multiplied together: Risk Impact Degree = Basic Impact Factor × Scope Impact Factor × Severity Impact Factor, thus obtaining a comprehensive, quantitative risk impact degree value.

[0101] Step S148: Based on a preset sentiment analysis algorithm, determine the sentiment polarity of the text content including the risk keywords and output the corresponding risk impact direction coefficient.

[0102] In this embodiment, a sentiment analysis algorithm can be used to analyze the narrative tone of a risk event and determine whether the risk event tends to increase or decrease the risk. The algorithm can identify and classify the sentiment tendencies expressed by text fragments containing risk keywords and convert them into quantitative parameters (i.e., risk influence direction coefficients) that characterize the direction of risk impact.

[0103] In one specific and possible implementation, the preset sentiment analysis algorithm can be a sentiment analysis algorithm based on a sentiment dictionary. This algorithm can determine the sentiment tendency by calculating the combined scores of sentiment words, degree adverbs, and negation words in the text.

[0104] In one specific and possible implementation, the preset sentiment analysis algorithm can be a machine learning-based sentiment analysis algorithm. This algorithm can extract n-gram features and syntactic features from the text, and then use classifiers such as support vector machines or Naive Bayes to perform sentiment classification.

[0105] In one specific and possible implementation, the preset sentiment analysis algorithm can be a deep learning-based sentiment analysis algorithm. For example, a recurrent neural network can be used to capture contextual information of a text sequence, or a pre-trained language model can be fine-tuned to achieve more accurate sentiment understanding.

[0106] In this embodiment, the risk impact direction coefficient can be represented as a numerical transformation result mapping sentiment polarity to the direction of risk impact. Specifically, when the sentiment analysis result determines that the text sentiment is negative, it indicates that the risk event may increase market risk, and a positive coefficient, such as +1, can be assigned to it; when the sentiment analysis result determines that the sentiment is positive, it indicates that the event may reduce or mitigate risk, and a negative coefficient, such as -1, can be assigned to it. When the sentiment analysis result determines that the sentiment is neutral or cannot be clearly determined, a neutral coefficient of 0 can be assigned. This risk impact direction coefficient is ultimately combined with the risk impact degree value to jointly determine the net impact direction and magnitude of a single risk event on the overall risk level.

[0107] Step S1410: Calculate the risk factor information of the risk type by combining the risk impact degree of at least one of the risk events and its corresponding risk impact direction coefficient.

[0108] In this embodiment, for a specific risk type, within a defined time period, the quantitative impact of all risk events belonging to that risk type can be aggregated and integrated to generate a single numerical indicator (risk factor information) that comprehensively reflects the overall activity level and direction of impact of that risk type.

[0109] Specifically, for each risk event, the risk impact level calculated in step S146 can be multiplied by the risk impact direction coefficient obtained in step S148. The result of this multiplication is a standardized risk contribution value for a single event, with a direction (positive or negative sign). Then, the contribution values ​​of all events belonging to the same risk type can be algebraically added together, and the final algebraic sum is the risk factor information for that risk type within this time period.

[0110] This implementation method constructs a streamlined, multi-stage text analysis pipeline to systematically transform the risk information implicit in unstructured text into standardized, calculable quantitative indicators. First, by combining a keyword dictionary with named entity recognition, the key elements of risk events are accurately captured and structured. Second, the degree of risk impact is calculated based on predefined rules, ensuring the objectivity and interpretability of the quantification process. Third, sentiment analysis is introduced to determine the direction of risk impact, enhancing the semantic integrity and dynamic representation capabilities of risk factor information. Finally, through comprehensive calculation of multiple events, quantitative information that comprehensively and accurately reflects the activity level and direction of impact of specific risk types is generated, providing a high-quality data foundation for subsequent risk fusion assessment.

[0111] In some implementations, the electricity market risk keyword dictionary is pre-constructed using the following method:

[0112] First, based on a pre-set set of seed keywords for electricity market risks, a word embedding model is trained on pre-acquired historical text corpus to calculate extended keywords that are semantically similar to the seed keywords.

[0113] In this embodiment, the preset set of seed keywords for electricity market risks can be a set of core terms predefined by domain experts based on their professional knowledge and historical experience. These core terms can be the most representative and fundamental terms among various types of electricity market risks.

[0114] In this embodiment, the word embedding model can be a natural language processing model that maps words to a high-dimensional vector space. It can represent each word as a dense vector (i.e., a word vector), such that words with similar meanings are also located close together in the vector space.

[0115] In this embodiment, a large amount of historical documents related to the electricity market (e.g., policy documents from previous years, industry research reports, market analysis news, etc.) can be collected and used as training data, which can be input into a selected word embedding model (e.g., Word2Vec, GloVe, etc.) for model training. Through this training process, the word embedding model can learn the specific semantics and relationships of vocabulary in the professional field of the electricity market.

[0116] In this implementation, a trained word embedding model can be used to find the semantic nearest neighbors of each keyword in the seed set in the vector space. Specifically, firstly, the word vector corresponding to the seed keyword can be obtained. Then, the semantic similarity between this word vector and the word vectors of all other words in the vocabulary can be calculated (using a metric such as cosine similarity). Finally, words with similarity exceeding a preset threshold are selected, or the top N words with the highest similarity are chosen. These selected words are the expanded keywords. For example, for the seed word "electricity reform," the model can automatically expand it to semantically highly related words such as "electricity system reform," "marketization reform," and "transmission and distribution price reform."

[0117] Then, the seed keywords and the extended keywords are merged to generate the electricity market risk keyword dictionary.

[0118] In this implementation, the automatically expanded vocabulary (extended keywords) can be integrated with the baseline vocabulary (seed keywords) defined by expert knowledge to form the final terminology library for risk identification. Specifically, the set of extended keywords obtained in the above steps can be combined with the initial set of seed keywords. Then, a deduplication operation is performed to remove duplicate words. Furthermore, manual review or rule-based filtering can be conducted to remove obviously irrelevant or noisy words, ensuring the professionalism and accuracy of the dictionary. Finally, a complete and structured dictionary of electricity market risk keywords is generated. This dictionary can be stored as a plain text file, with each line recording one keyword, or as a database table, which can include meta-information such as the words and their respective risk categories.

[0119] In some implementations, the step of calculating the degree of risk impact of the risk event based on predefined risk impact calculation rules and the key entity includes:

[0120] Step S1462: Based on the risk event type indicated by the key entity, query the corresponding basic impact factor from the pre-built event-risk knowledge base.

[0121] In this embodiment, the risk event type indicated by the key entity can be categorized by analyzing extracted entities (e.g., regulatory names, policy terms, or event descriptive phrases). For example, extracting the entity "Detailed Rules for the Management of New Energy Grid Connection" can indicate that the event type is policy change, and extracting the entity "Unit Shutdown" can indicate that the event type is supply-side shock.

[0122] In this embodiment, the pre-built event-risk knowledge base can be a database storing the mapping relationship between event types and basic impact factors. The construction of this knowledge base can be based on data analysis of historical events and their subsequent market impact (e.g., learned through regression analysis), or it can incorporate the experience and judgment of domain experts. That is, a standardized impact coefficient can be preset for each type of known risk event, representing the average impact level that such an event may cause under normal circumstances. The query can be expressed as retrieving the corresponding value from the knowledge base using the identified event type as an index. For example, the event-risk knowledge base can specify a basic impact factor of 0.8 for major policy release events and a basic impact factor of 0.3 for routine policy fine-tuning events.

[0123] Step S1464: Calculate the scope impact factor based on the risk impact scope information indicated by the key entity.

[0124] In this embodiment, numerical or descriptive information extracted from the text that characterizes the scope of influence can be used to calculate a factor representing the proportion or size of the standard affected entities.

[0125] The risk impact scope information indicated by the key entity may include extracted numerical entities and risk subject entities. For example, the text may include information involving 15% of the province's thermal power installed capacity or impacting three major electricity sales companies. The calculation of the scope impact factor can then be a process of converting the textual description into a specific proportion. For example, firstly, relevant total data can be obtained from external databases or internal records based on entity information. Next, the specific affected value is divided by this total data to obtain a proportion. For example, if the text indicates that the capacity of units failing energy efficiency standards is 500,000 kilowatts, then by querying the database, it can be found that the total installed capacity of thermal power in the province is 10 million kilowatts, and the calculated scope impact factor is 50 / 1000 = 0.05. This scope impact factor converts the absolute value of the impact into a relative proportion, making events of different scales comparable.

[0126] Step S1466: Based on the risk consequences information indicated by the key entity, determine the severity impact factor from the pre-built consequences-risk knowledge base.

[0127] In this embodiment, a numerical value representing the severity can be mapped from a database defining different severity levels of consequences, based on the event consequences described in the text. Specifically, the risk consequence information indicated by the key entity can be represented by the extracted consequence entities, such as fines, warnings, rectification within a specified period, production suspension, license revocation, etc. These terms directly describe the negative results that the event may cause.

[0128] In this embodiment, the pre-built consequence-risk knowledge base can be a database storing the correspondence between different consequence descriptions and their severity scores. The construction of this consequence-risk knowledge base can be based on the actual losses caused by different consequences in historical cases, or set by experts according to the penalties stipulated in regulations. For example, the consequence-risk knowledge base can specify a severity impact factor of 0.1 for warnings, 0.3 for fines, and 0.8 for production suspension. Therefore, the corresponding severity impact factor can be retrieved from the consequence-risk knowledge base using the identified consequence entity as the keyword. If multiple consequences are described in the text, the factor with the highest severity can be taken, or their average value can be calculated.

[0129] Step S1468: Calculate the risk impact degree of the risk event based on the product of the basic impact factor, the range impact factor, and the severity impact factor.

[0130] This implementation method constructs a multi-factor quantification calculation model to systematically transform key entity information extracted from text into objective risk impact values. First, by introducing an event-risk knowledge base and a consequence-risk knowledge base, domain expert experience and historical data are distilled into reusable quantitative standards, effectively eliminating the arbitrariness of subjective judgment and significantly improving the objectivity and consistency of risk quantification. Second, through the triple consideration of basic impact factors, scope impact factors, and severity impact factors, a comprehensive assessment of risk events is achieved, from their intrinsic attributes and breadth of impact to the depth of consequences, enabling the calculation results to more precisely and accurately reflect the actual impact of risk events. Finally, a product model is used for comprehensive calculation, ensuring that the synergistic effects between various impact dimensions are reasonably reflected, thereby outputting a risk impact value with clear physical meaning and comparability, providing a high-quality, interpretable data foundation for subsequent risk factor information synthesis.

[0131] In some implementations, the preset electricity retail market risk monitoring indicator system includes multiple risk categories, each risk category includes risk factor indicators and risk outcome indicators, and the indicators are defined for the overall market and individual market entities, respectively.

[0132] The step of processing the first dataset based on the preset electricity retail market risk monitoring indicator system to obtain structured risk information for at least one risk type includes:

[0133] Step S162: Based on the target risk type, determine the corresponding risk factor indicators and risk outcome indicators from the indicator system.

[0134] In this embodiment, the target risk type can be represented as the risk dimension currently being addressed, such as market power risk, price risk, or operational risk.

[0135] In this implementation, a target risk type (e.g., market power risk) can be received or specified as input. Then, a query and matching process is performed within the electricity retail market risk monitoring indicator system to retrieve all indicator items categorized under that risk type. The output can be two lists of indicators. One list may include all risk factor indicators related to the target risk type (e.g., HHI index, Lerner index), and the other list may include all risk outcome indicators related to the target risk type (e.g., proportion of active electricity retailers, industry excess profit margin).

[0136] Step S164: Based on the first dataset, calculate the overall market indicator value and the individual market participant indicator value corresponding to the target risk type.

[0137] In this embodiment, the required raw data fields can be extracted from the structured first dataset according to the specific index definition determined in step S162, and corresponding mathematical operations can be performed to generate quantitative results.

[0138] Specifically, first, macro-level data reflecting the overall market situation (e.g., total electricity trading volume, average profit margin of all electricity retailers, total number of users, etc.) can be obtained from the first dataset and calculated according to preset indicator formulas. For example, to calculate the overall HHI index of the market, market share data of all electricity retailers needs to be obtained and the sum of squares calculated. Then, for each specific market participant (e.g., a particular electricity retailer), individual data (e.g., the company's electricity sales volume, electricity consumption of its agents, its retail price, etc.) can be extracted from the first dataset and substituted into the individual indicator formulas for calculation. For example, to calculate the Lerner index of a particular electricity retailer, the company's retail price and marginal cost data need to be obtained. Finally, the output consists of two sets of values: one set of indicator values ​​representing the overall market risk situation, and the other set of indicator values ​​representing the individual risk situation of each market participant.

[0139] Step S166: Based on a preset weighting scheme, the overall market indicator value and the individual market entity indicator value are weighted and fused to obtain the structured risk information of the target risk type.

[0140] In this embodiment, the preset weighting scheme defines the rules for assigning importance coefficients to indicators from different sources. It can be statically set based on the experience and knowledge of domain experts (e.g., considering the HHI index, which reflects the overall market, to be more important than the market share of a single company, and thus assigning it a higher weight), or it can be a weighting model dynamically optimized through machine learning based on historical data.

[0141] Specifically, first, the weighting scheme configured for the current target risk type can be read. This scheme assigns corresponding weight coefficients to the selected overall market indicators and individual market participant indicators. Then, the calculated value for each indicator is multiplied by its corresponding weight coefficient to obtain a weighted result. Finally, all weighted results are summed to generate the final scalar value, which is the structured risk information for the target risk type. This value can be a comprehensive quantitative score, integrating information from both macroeconomic market conditions and microeconomic participant performance, thus providing a more comprehensive and balanced reflection of the overall level of this specific risk type.

[0142] This implementation method achieves refined and structured measurement of risks in the electricity retail market by constructing a multi-level indicator system that covers multiple risk categories, distinguishes between factors and results, and considers both the overall picture and individual situations, and by adopting a systematic data processing workflow. Specifically, firstly, by using risk factor-based and result-based indicators in tandem, it can simultaneously capture the pre-existing causes and actual manifestations of risks, enhancing the comprehensiveness and foresight of risk identification. Secondly, through dual-dimensional calculation of indicators for the overall market and individual market participants, it achieves correlation analysis between the macro-market situation and the behavior of micro-entities, improving the depth of risk insight. Finally, by weighting and fusing the multi-dimensional indicator values ​​through a preset weighting scheme, the resulting structured risk information can comprehensively reflect the overall situation of a specific risk type. Its output results are objective, comprehensive, and interpretable, providing a reliable decision-making basis for market risk management.

[0143] In some implementations, the multiple risk categories include market power risk, price risk, operational risk, contract risk, performance risk, settlement risk, compliance risk, and policy risk;

[0144] Among them, the risk factor indicators corresponding to the market power risk include at least the market concentration index defined for the overall market and the price deviation rate index defined for individual market participants; the risk outcome indicators corresponding to the market power risk include at least the proportion of active electricity sales companies defined for the overall market and the total market share index defined for individual market participants.

[0145] In this embodiment, the risk factor indicators corresponding to the market power risk may also include: the market share indicator of the top m companies defined for the overall market (Top-m share), the high selling price ratio indicator defined for individual companies in the market, the low selling price ratio indicator defined for individual companies in the market, and the same selling price ratio indicator defined for individual companies in the market, etc.

[0146] In this embodiment, the risk outcome indicators corresponding to the market power risk may also include: the proportion of new electricity sales companies actually conducting business as defined for the overall market, the proportion of users switching electricity sales companies as defined for the overall market, the industry excess profit rate as defined for the overall market, the relative market share as defined for individual market entities, and the user internal conversion rate as defined for individual market entities, etc.

[0147] The risk factor indicators corresponding to the price risk include at least a wholesale-retail price correlation indicator defined for the overall market and a wholesale-retail price correlation indicator for individual market participants; the risk outcome indicators corresponding to the price risk include at least a wholesale-retail price inversion rate indicator defined for the overall market and a wholesale-retail price inversion rate indicator for individual market participants.

[0148] In this embodiment, the risk factor indicators corresponding to the price risk may further include: a reasonableness indicator of the package structure defined for the overall market and an applicability indicator of the electricity user package defined for individual market participants.

[0149] In this embodiment, the risk outcome indicators corresponding to the price risk may further include: a deviation rate indicator between the retail package price and the catalog electricity price defined for the market as a whole, and a deviation rate indicator of the electricity user package price defined for individual market participants.

[0150] The risk factor indicators corresponding to the operational risks include at least the wholesale market price volatility indicator defined for the overall market and the load forecast accuracy indicator defined for individual market participants; the risk outcome indicators corresponding to the operational risks include at least the proportion of loss-making electricity sales companies defined for the overall market and the user stickiness indicator defined for individual market participants.

[0151] This implementation method clearly defines representative risk factor and risk outcome indicators under key categories such as market power risk, price risk, and operational risk, and ensures that each risk covers the monitoring perspectives of the overall market and individual market participants. This constructs a clear, comprehensive, and focused monitoring indicator system, providing a clear and operable basis for accurately extracting the comprehensive status of various risks from structured data, and significantly improving the systematicness and pertinence of risk identification.

[0152] In some implementations, the step of weightedly fusing the risk factor information and the structured risk information of the same risk type to generate target risk information for that risk type includes:

[0153] Step S182: Based on the risk type, obtain the first weight coefficient configured for the risk factor information and the second weight coefficient configured for the structured risk information.

[0154] In this embodiment, a first weight coefficient configured for the risk factor information and a second weight coefficient configured for the structured risk information can be obtained based on a pre-configured static weight table and the risk type. The pre-configured static weight table can store fixed values ​​for the first and second weight coefficients under different risk types. For example, for policy risks, which are highly sensitive to textual information, the first weight coefficient can be configured as 0.7 and the second weight coefficient as 0.3; while for operational risks, which rely more on structured data, the first weight coefficient can be configured as 0.4 and the second weight coefficient as 0.6.

[0155] In this embodiment, a first weighting coefficient configured for the risk factor information and a second weighting coefficient configured for the structured risk information can be obtained based on a dynamic calculation model and the risk type. This dynamic calculation model can dynamically adjust the weight allocation according to real-time data quality, market stage, or risk warning level.

[0156] Step S184: Calculate the product of the first weight coefficient and the risk factor information to obtain the first weighted result.

[0157] Step S186: Calculate the product of the second weighting coefficient and the structured risk information to obtain the second weighted result;

[0158] Step S188: Summate the first weighted result and the second weighted result to generate target risk information for this risk type.

[0159] In one specific implementation plan, a method for generating target risk information in the electricity retail market is provided, which may include:

[0160] S1. Acquisition and preprocessing of multi-source heterogeneous data.

[0161] 1. Multi-source data acquisition.

[0162] 1.1 Structured data acquisition.

[0163] For structured data (such as tabular data in relational databases like MySQL and Oracle) provided by power trading center databases and power sales company operation databases, this solution establishes a network connection with the target database through a pre-configured JDBC / ODBC database connector. Then, by executing standardized SQL queries, it incrementally extracts the required data tables or views periodically or in real-time. The extracted data is typically stored in rows and columns in memory or temporary files, containing structured fields such as numbers, dates, and text.

[0164] The collected structured data includes not only traditional transaction prices and load data, but also user credit ratings, contract status (such as whether it is nearing expiration), and settlement anomaly flags, laying a data foundation for building more comprehensive monitoring indicators for credit risk, contract risk, and settlement risk in the future.

[0165] 1.2 Unstructured data acquisition.

[0166] For unstructured information such as policy texts published by public information sources such as government websites and news portals, this solution uses web crawling technology for collection. The crawler program automatically sends HTTP / HTTPS requests to the target webpage and receives the returned HTML document.

[0167] This solution incorporates policy and regulatory texts, feedback news, and industry analysis reports into its risk monitoring data system. The focus of collection is on textual information related to electricity market reform, regulatory policies, and significant changes in market participants, as this data is crucial for assessing policy and market power risks. After collecting the original HTML documents, processing steps such as removing headers, footers, and advertising tags are performed to obtain plain text content.

[0168] For the specific structured and unstructured data items that need to be collected in the above two steps, please refer to the monitoring indicator system below, which will vary based on the needs of the monitoring indicator system.

[0169] The collection times of different data sources can be uniformly scheduled by a scheduler, and all collected data within a period can be transmitted to the next stage for preprocessing periodically.

[0170] 2. Data preprocessing.

[0171] 2.1 Data cleaning process.

[0172] For all data, detect null values (NULL) in the data. For numerical fields, fill them with the mean or median; for categorical text fields, use the mode or mark them as "unknown" separately.

[0173] Then, identify outliers beyond a reasonable range through statistical methods (such as standard score Z-Score, interquartile range IQR). For values that are clearly recording errors, perform smoothing or removal.

[0174] Then, uniformly convert date / time strings from different sources (such as 20231015, 2023-10-15) to the standard YYYY-MM-DD HH:MM:SS format; convert numerical strings (such as "1,000.50" with commas) to floating-point numbers; and then uniformly convert the character encoding of all text data to UTF-8 to prevent garbled characters.

[0175] 2.2 Preprocessing of unstructured text.

[0176] First, use Chinese word segmentation tools (such as Jieba, HanLP) to split the continuous text stream into a sequence of independent lexical units (Tokens); then, remove words that appear frequently in the text but have no actual meaning (such as "de", "le", "zai") to reduce the data dimension and noise; then label the词性 (such as noun, verb, adjective) for each word after word segmentation; finally, based on the词性 annotation results and a predefined entity dictionary, initially identify entities such as person names, organization names, place names, policy names, and regulation article numbers in the text.

[0177] 2.3 Label all data with data source tags and timestamps, and then use common keys (such as timestamps, company IDs) to associate structured data from different sources; the processed unstructured text data is associated with its corresponding release time and source website. Finally, all data is written into a unified and standardized target storage system to form a "historical dataset" for subsequent query and call.

[0178] S2. Construction and quantification of dynamic monitoring indicators.

[0179] 1. Risk factor extraction.

[0180] 1.1 Construction of a risk keyword dictionary.

[0181] First, experts in the field of electricity market risk initially compile a core keyword dictionary covering various risks based on historical experience. For example: "electricity price adjustment", "market access", "new regulatory rules", "implementation details" for policy risk; "market manipulation", "collusive bidding", "abuse of dominant position", etc. for market power risk; "debt default", "credit rating downgrade", "rupture of the capital chain", etc. for credit risk.

[0182] Then, using word embedding models (such as Word2Vec) or topic models (such as LDA), with core keywords compiled by experts as seed words, the system is trained on a large amount of historical text corpus to automatically discover and expand words that are semantically similar to these seed words, thereby forming a more comprehensive and dynamically evolving risk keyword expansion dictionary.

[0183] In this process, historical text corpora can be input into the model to obtain word vectors of different words in the corpus, forming a word vector library; then, the word vectors of the seed words and the word vector library are matched with cosine similarity, and the top-N (e.g., N is 20) high similarity words are selected for dictionary expansion.

[0184] 1.2 Risk event record.

[0185] In step S1, when performing text entity recognition on unstructured text, a Named Entity Recognition (NER) model with fine-tuning for the electricity retail sector is used, which can identify entities such as risk subjects (e.g., "a certain electricity sales company"), risk-related regulations (e.g., "Basic Rules for Electricity Market Operation"), and risk values ​​(e.g., "five percent") from sentences.

[0186] Then, when a risk keyword appears in an entity within a sentence, the risk event is identified, its impact is calculated, and the event is recorded: {risk type, risk subject, and impact level of the risk event}. This process transforms unstructured text into a structured record of risk events.

[0187] The formula for calculating the degree of risk impact is:

[0188] In the formula, It is the basic impact factor, which represents the baseline impact intensity that the corresponding type of risk of the event may cause under normal circumstances. It is an empirical value that is learned from historical data and correlates the event with the risk. It is the scope of impact factor, used to indicate the scope of the audience affected by the event, that is, what proportion of market entities will be affected, which is determined by risk values; The severity impact factor is used to represent the severity of the possible consequences of the event. It is achieved by semantically analyzing the "consequences" entity in the sentence and mapping it to the severity of a specific risk.

[0189] For example, “A province has introduced new energy efficiency standards for thermal power plants. Thermal power plants with 15% of their units failing to meet the standards must invest in upgrades, or they will face production restrictions.” The key risk words include the introduction of energy efficiency standards for thermal power plants, the identified risk event is the change in energy efficiency standards, and the risk subject is the non-compliant thermal power plants in the province.

[0190] The risk type of this risk event is market power risk, with a basic impact factor of 0.4 obtained from the event-risk knowledge base. The provincial power database is consulted to obtain the province's "total thermal power installed capacity" (A) and "total installed capacity of power plants with a current energy efficiency non-compliance rate > 15%" (B). The range impact factor (using installed capacity rather than the number of power plants) is calculated by dividing B by A. Then, the severity of investment in renovation funds is determined to be 0.3, and the severity of production restrictions is determined to be 0.9, with the maximum value of 0.9 taken. Finally, the risk impact level is calculated as 0.4 * B / A * 0.9 = 0.36B / A.

[0191] Among them, the event-risk knowledge base and the consequence-risk severity knowledge base are constructed and iteratively optimized in advance by using machine learning (such as regression analysis) on historical events and their subsequent market performance.

[0192] Specifically, risk events are identified by calculating the matching similarity between the entity composition vector of the sentence containing the risk keyword and the event name vector in the event-risk knowledge base.

[0193] 1.3 Sentiment Analysis and Risk Polarity Quantification

[0194] Sentiment analysis models (such as BERT-based classification models) are used to determine the sentiment polarity of sentences or paragraphs containing risk keywords, outputting a positive, negative, or neutral result, along with a confidence score. Then, the sentiment polarity is converted into a quantitative parameter representing the direction of risk impact. For example, in a news report about an upcoming policy, if the sentiment analysis is negative, the policy's risk factors are considered to have a positive impact (i.e., increasing risk), and a positive coefficient (e.g., +1) is assigned; if it is positive, a negative coefficient (e.g., -1, indicating reduced risk) is assigned; and neutral is 0.

[0195] 1.4 Calculation of Risk Factor Values

[0196] By combining the risk impact level and sentiment quantification parameters output from the above steps, a periodic (e.g., daily) risk factor quantification value is generated for each type of risk.

[0197] For each risk type r, its risk factor value F(r,t) over a time period t can be calculated using the following formula: ;

[0198] In the formula, This represents the comprehensive risk factor value of risk type r within period t. This value is a dimensionless relative value; its absolute value reflects the activity level of the risk, and the positive or negative sign indicates the direction of the risk's impact. i represents the i-th risk event belonging to risk type r identified within period t. This represents the emotional polarity coefficient of the i-th event.

[0199] 2. Calculation of monitoring indicators.

[0200] 2.1 Construction of monitoring indicator system.

[0201] This scheme combines market risk monitoring indicator design theory, distinguishes between process and outcome indicators, and designs an indicator system for monitoring risks in the electricity retail market, starting from the overall market and individual entities.

[0202] I. Market Power Risk Monitoring Indicators

[0203] (1) Risk Factors

[0204] 1) Market share of the top m companies (Top-m share)

[0205] Market share directly reflects a company's position in the market for its goods and services. A higher market share indicates stronger business operations and competitiveness.

[0206] 2) Market Concentration (HHI Index)

[0207] A higher HHI index indicates a greater likelihood of market abuse in the electricity retail market.

[0208] The expression for the HHI index is as follows:

[0209]

[0210] In the formula, This indicates the overall size of the electricity retail market in a certain region; This represents the size of the i-th electricity retail company in the electricity retail market; This represents the share of the i-th electricity retail company in the overall electricity retail market; n represents the number of competitors in a certain region's electricity retail market.

[0211] 3) Price deviation rate (Lerner index)

[0212] The price deviation rate is the degree to which a price increase caused by market forces deviates from marginal cost. It measures the degree of market monopoly by comparing the actual market price with the ideal competitive price. The specific function expression for the price deviation rate (Lerner index) is as follows:

[0213]

[0214] In the formula, This indicates the retail price of a standard retail package from iPower, the electricity company in that region; This indicates the average cost per kilowatt-hour for users represented by the electricity sales company under this package model; The magnitude of this indicator reflects whether electricity retail company i has the motivation to exert market power. Substituting typical prices and costs in the electricity retail industry into the above formula yields the overall price deviation rate indicator for the retail market.

[0215] 4) High selling price ratio

[0216] The high-price ratio refers to the ratio of the electricity sales price of a power company's standard retail package to the highest market price for that standard package. The calculation formula is as follows:

[0217]

[0218] In the formula, The percentage of high selling prices indicated by the electricity sales company reflects whether the company is engaging in price gouging. This indicates the electricity price for the company's standard retail package. This indicates the highest market price for electricity sold under the standard package.

[0219] 5) Low selling price ratio

[0220] The low selling price ratio refers to the ratio of the electricity sales price of a power company's standard retail package to the lowest market selling price for that standard package. The calculation formula is as follows:

[0221]

[0222] In the formula, This indicates the electricity sales company's low selling price ratio, reflecting whether the company has engaged in malicious competition by suppressing electricity prices. This indicates the electricity price for the company's standard retail package. This indicates the lowest market price for standard package electricity.

[0223] 6) Price same ratio

[0224] The price-match ratio refers to the ratio of the number of units sold by electricity retailer i under its standard retail package model at the same price as those offered by other electricity retailers for the same standard package, to the total number of units sold under that standard package. The calculation formula is as follows:

[0225]

[0226] In the formula, This indicates the percentage of the same standard retail package price as that of the electricity sales company i, reflecting whether the electricity sales company has engaged in price collusion with other electricity sales companies. This indicates the quantity of electricity sold at the same price for a standard retail package by this electricity sales company as other electricity sales companies offering the same standard package; n represents the total quantity of this standard package priced in the market.

[0227] (2) Risk outcome category

[0228] 1) Proportion of active electricity sales companies

[0229] This refers to the proportion of electricity sales companies participating in market transactions within the statistical period to the total number of registered electricity sales companies. The higher this proportion, the lower the market power risk. The calculation formula is as follows:

[0230]

[0231] 2) Proportion of business actually conducted by newly entered electricity sales companies

[0232] The proportion of newly entered electricity sales companies that actually conduct business during the assessment period reflects the barriers to entry and the intensity of competition in the electricity sales market. The calculation formula is as follows:

[0233]

[0234] 3) The proportion of users switching electricity sales companies

[0235] The percentage of users who switch electricity retailers refers to the percentage of all users in a regional retail market who switch electricity retailers within a certain period.

[0236]

[0237] In the formula, This indicates that the user has changed the frequency of the electricity sales company; This indicates the number of users in the region who switched electricity retailers within a year. This represents the total number of users in the region. This reflects the competitive vitality of the retail market.

[0238] 4) Industry excess profit margin (Bain Index)

[0239] The Bain Index represents the industry's excess profit margin. If an industry consistently has excess profits, it indicates that the market has monopolistic power. The higher the Bain Index, the stronger the monopolistic power.

[0240]

[0241] in

[0242]

[0243] In the formula, V represents the industry's economic profit; V represents the industry's total investment. 'i' represents industry accounting profit, and 'i' represents the socially evaluated rate of return on investment. The excess profit margin can also be calculated for an individual electricity sales company using the Bain & Company index formula.

[0244] 5) Total market share

[0245] This refers to the proportion of electricity sold by a single electricity sales company to the total electricity sold through its agency.

[0246]

[0247] In the formula, This represents the total market share of electricity sales company i; This indicates the amount of electricity sold by electricity company i in this region.

[0248] 6) Relative market share

[0249] The relative market share indicator refers to the market share of a single electricity sales company relative to its three largest competitors. It is the ratio of the electricity sales volume handled by a single company to the total sales volume handled by the three largest competitors in the market. The calculation formula is as follows:

[0250]

[0251] In the formula, This represents the relative market share of electricity sales company i. For the market share of electricity sales company i, This is the sum of the market share of the three largest electricity retailers besides retailer i. For example, if a retailer i has a market share of 30%, and its three largest competitors have market shares of 20%, 10%, and 10% respectively, then the retailer i's relative market share is 30% ÷ 40% = 75%. If each of the four retailers i has a market share of 25%, then the retailer i's relative market share is 33%.

[0252] 7) Internal user conversion rate

[0253] The user internal switching rate, which represents the proportion of users who, without changing their electricity sales company, switched their electricity sales plan or package within the same company, directly reflects users' willingness to respond to electricity sales competition and indirectly reflects the efficiency with which competition motivates electricity sales companies to optimize their service plans and packages. The calculation formula is as follows:

[0254]

[0255] II. Price Risk Monitoring Indicators

[0256] Price risk may originate from the electricity retail operation stage, the electricity wholesale market, etc., and both electricity sales companies and electricity users may be bearers of price risk. The price risk monitoring indicator system for the electricity retail market is designed as follows:

[0257] (1) Risk Factors

[0258] 1) Correlation between wholesale and retail prices

[0259] Wholesale market transactions significantly influence the behavior of electricity retailers in the retail market. The wholesale-retail price correlation examines the transmission of price signals from the wholesale market to the retail market across the entire market. A higher value indicates a smoother transmission of wholesale market prices to the retail market, and lower price risk on the electricity retail side. The calculation formula is as follows:

[0260]

[0261] In the formula, The correlation coefficient between retail market prices and wholesale market prices measures the degree of correlation between wholesale and retail prices. for and covariance; , These represent the variances of retail market prices and wholesale market prices, respectively.

[0262] The formulas for calculating covariance and variance are as follows:

[0263]

[0264]

[0265] , Let $i$ be the average retail market price and the average wholesale market price for the i-th statistical analysis, respectively. , These represent the average prices in the retail market and the average prices in the wholesale market, respectively.

[0266] 2) Reasonableness of package structure

[0267] Electricity retail charges include wholesale electricity costs, transmission and distribution costs, government funds and agency fees, electricity sales costs, and retail profits. By monitoring the deviations of each component's proportion from the ideal proportion in electricity retail charges, the rationality of the electricity retail charge structure can be reflected. Since transmission and distribution costs, government funds and agency fees, and electricity sales costs are relatively fixed, it is only necessary to monitor the proportions of wholesale costs and retail profits in the electricity charge to determine the rationality of the charge structure. The calculation formula is as follows:

[0268]

[0269] In the formula, This value indicates the rationality of the electricity retail pricing structure; the higher the value, the more unreasonable the package structure. , These are the wholesale electricity cost and retail profit on the electricity sales side; These are the ideal percentages of electricity wholesale costs and retail profits, respectively, set by the market regulatory and operating agency. This represents the total retail electricity cost.

[0270] 3) Correlation between wholesale and retail prices of electricity sales companies

[0271] The calculation method for this indicator is the same as that for the wholesale-retail price correlation indicator, but the selected values ​​are the wholesale market price and retail market price of a certain electricity sales company. The higher the value, the lower the price risk of the electricity sales company.

[0272] 4) The rationality of the retail package structure of the electricity sales company

[0273] The calculation method for this indicator is the same as that for the rationality of the package structure, but the selected values ​​are the wholesale cost and retail profit of a certain electricity sales company. The higher the value, the more unreasonable the retail package structure of the electricity sales company is.

[0274] 5) Applicability of electricity user packages

[0275] Electricity user package suitability is a qualitative indicator, referring to whether the package signed by an electricity user matches their electricity load characteristics and risk preferences. This indicator can be obtained through surveys; the higher the suitability, the lower the price risk for the electricity user.

[0276] (2) Risk outcome category

[0277] 1) Wholesale-retail inversion rate

[0278] The wholesale-retail price inversion rate refers to the percentage of electricity sales companies participating in the electricity sales market where the difference between the average wholesale price and the average retail price is negative within the statistical period. A higher percentage indicates a more frequent occurrence of wholesale-retail price inversions, and a higher likelihood of price risk for the electricity sales companies. The calculation formula is as follows:

[0279]

[0280] 2) Deviation rate between retail package price and catalog electricity price

[0281] By comparing the deviations of different retail package prices from the catalog electricity price, the price level of the retail packages can be reflected. The greater the deviation, the lower the electricity price for the user, and the lower the price risk. The calculation formula is as follows:

[0282]

[0283] In the formula, For package The price This indicates the catalog electricity price.

[0284] 3) Year-on-year growth rate of retail package prices

[0285] The year-on-year growth rate represents the rate of change in the retail package price this period compared to the same period last year. It reflects the change in the retail package price level. The higher the value, the more upward the retail package price trend, and the greater the price risk for electricity users. The calculation formula is as follows:

[0286]

[0287] In the formula, Indicates package In the period price level, Indicates the number of the same period (e.g., year).

[0288] 4) Wholesale-retail inversion rate of electricity sales companies

[0289] The wholesale-retail price inversion rate for an electricity sales company refers to the proportion of months in which the average monthly wholesale market price and the average monthly retail price are negative, within a statistical period, out of the total number of months in the statistical period. The higher this value, the greater the price risk for the electricity sales company. The calculation formula is as follows:

[0290] 5) User package price deviation rate

[0291] The deviation rate between a user's package price and the market average price for similar users measures the level of a particular electricity user's retail package price compared to other similar users. The higher this value, the more the user's retail package price exceeds the average level, and the greater the user's price risk. The calculation formula is as follows:

[0292]

[0293] 6) Guaranteed retail price level

[0294] With the withdrawal of electricity retailers, electricity users may be provided with guaranteed electricity sales services at a guaranteed retail price by a minimum-price retail company. The higher the guaranteed retail price, the greater the price risk for the electricity user. The minimum retail price can be determined by comparing the price the electricity user pays for the guaranteed electricity sales service with the original package price.

[0295] III. Operational Risk Monitoring Indicators

[0296] Operational risks may originate from electricity sales companies or electricity users, but the electricity sales companies are the risk bearers. Operational risk monitoring needs to be approached from both the perspectives of the overall market and market participants, examining risk factors and their outcomes. The operational risk monitoring indicator system for the electricity retail market is designed as follows:

[0297] (1) Risk Factors

[0298] 1) Wholesale market price volatility

[0299] Wholesale market price volatility reflects the degree of fluctuation in wholesale (spot) market prices and measures the ease with which electricity retailers can participate in wholesale (spot) market bidding. Higher wholesale market price volatility means greater difficulty for electricity retailers to participate in the wholesale market and higher operational risks. This price volatility is calculated based on market prices over a period of time, using the following formula:

[0300]

[0301] In the formula, The number of samples within the statistical period. For the wholesale market Electricity prices at any given time This represents the average wholesale market price during the statistical period. The standard deviation of the sample measures the dispersion of the data and represents the volatility of wholesale market prices.

[0302] 2) Changes in electricity consumption by retail market users

[0303] The change in electricity consumption by retail market users refers to the rate of change in the total electricity consumption of retail market users within a statistical period compared to the total electricity consumption of the previous period. The greater the change, the more difficult it is for electricity sales companies to predict, and the higher the potential for operational risks. The calculation formula is as follows:

[0304]

[0305] 3) Is there any price risk?

[0306] Price risks can lead to losses for electricity sales companies, causing them to fall into operational risks. Therefore, whether an electricity sales company faces price risks should be considered a qualitative indicator for monitoring operational risk factors.

[0307] 4) Load forecasting accuracy

[0308] Load forecasting accuracy measures whether a power sales company in the market has a grasp of actual electricity consumption and its ability to predict and analyze electricity consumption trends. A higher load forecasting accuracy indicates stronger business capabilities of the power sales company. The calculation formula is as follows:

[0309]

[0310] 5) Changes in electricity consumption by contracted users

[0311] The change in electricity consumption by contracted users refers to the rate of change in the total electricity consumption of a power sales company's contracted users within a statistical period compared to the total electricity consumption of the previous period. The larger this value, the more difficult it is for the power sales company to predict. The calculation formula is as follows:

[0312]

[0313] 6) Business Diversity

[0314] Business diversity refers to the number of business types actually carried out by a power sales company within a statistical period, including power purchase and sale, power engineering services, energy efficiency management services, demand-side response services, services for new business entities, green factory / industrial park services, green electricity and green certificate services, and carbon asset management services. The more types of business a power sales company actually carries out, the stronger its competitiveness and the lower its operational risk.

[0315]

[0316] (2) Risk outcome category

[0317] 1) Percentage of electricity sales companies subject to deviation assessment

[0318] The percentage of electricity retailers subject to deviation assessment refers to the proportion of electricity retailers subject to deviation assessment in the retail market to the total number of electricity retailers actually conducting transactions within the statistical period. The higher this percentage, the lower the accuracy of electricity purchase and sale by the retailers in the retail market, and the higher the operational risk. The calculation formula is as follows:

[0319]

[0320] 2) Proportion of loss-making electricity sales companies

[0321] The proportion of loss-making electricity retailers refers to the ratio of the number of loss-making electricity retailers in the retail market to the number of electricity retailers actually conducting transactions within a statistical period. A higher proportion indicates a higher operational risk in the electricity retail market. The calculation formula is as follows:

[0322]

[0323] 3) Industry cost-profit ratio

[0324] The industry cost-profit ratio refers to the ratio of total profit to total cost in the electricity sales industry. It reflects the operating efficiency of the electricity sales industry; the higher the value, the lower the operating risk in the electricity sales market. The calculation formula is as follows:

[0325] 4) Revenue generation capability of trading business

[0326] Revenue generation capability refers to the ability of a power sales company to generate income through market transactions. It is measured by the company's net profit; the higher the value, the stronger the company's revenue generation capability and the lower its operational risk. The calculation formula is as follows:

[0327]

[0328] 5) Proportion of losing months

[0329] The percentage of months with operating losses refers to the proportion of months in which a power sales company incurs losses out of the total number of months within a statistical period. The higher this value, the worse the company's operating condition and the higher its operating risk. The calculation formula is as follows:

[0330]

[0331] 6) User stickiness

[0332] Customer stickiness refers to the renewal rate of retail customers of a power sales company. The higher the value, the higher the customer satisfaction with the power sales company's services, and the lower the company's operating risk. The calculation formula is as follows:

[0333]

[0334] IV. Contract Risk Monitoring Indicators

[0335] The indicator system for monitoring contract risks in the electricity retail market is designed as follows:

[0336] (1) Risk Factors

[0337] 1) Have value-added service contract templates been developed?

[0338] As user needs become more diversified and electricity sales companies offer a wider range of services, there are more and more scenarios where both parties sign value-added service contracts. Developing value-added service contract templates can help users reduce contract risks.

[0339] 2) Is the electricity sales company's contract management standardized?

[0340] Does the electricity sales company sign contracts according to the template specifications, file contracts as required, and properly keep paper contracts for the past three years?

[0341] (2) Risk outcome category

[0342] 1) The proportion of contracts that do not clearly define liability for deviations and dispute resolution clauses

[0343] The proportion of contracts lacking clearly defined liability for deviations and dispute resolution clauses refers to the percentage of retail contracts with such clauses out of the total number of retail contracts within a statistical period. A higher proportion indicates more prevalent irregularities in contract signing in the market and a higher contract risk. The calculation formula is as follows:

[0344]

[0345] 2) Frequency of contract disputes

[0346] The frequency of contract disputes refers to the number of times a contract dispute occurs between an electricity sales company and an electricity user within a statistical period due to reasons such as unclear liability for deviations in the contract. Market supervision and operation agencies can set warning numbers for the frequency of contract disputes.

[0347] 3) The proportion of contracts with breaches of trust by electricity sales companies

[0348] Contract breaches include discrepancies between transaction volume and price on paper contracts and those on the trading platform, contracts lacking an official seal or signature of the legal representative, and failure to provide original paper contracts despite using paper contracts. This indicator measures the likelihood of contract problems occurring for a power sales company. It represents the proportion of contracts with breaches of contract to the company's total retail contracts within a statistical period. A higher proportion indicates less standardized contract management by the power sales company and a higher risk to users. The calculation formula is as follows:

[0349]

[0350] V. Performance Risk Monitoring Indicators

[0351] The primary bearer of performance risk is the electricity user, and the source is the electricity sales company. The ability of the electricity sales company to provide services according to the retail contract is crucial to the safe and stable electricity supply for users. To avoid performance risk, electricity users should carefully select the electricity sales company when signing the contract. Therefore, performance risk monitoring indicators target electricity sales companies, with a focus on risk factor indicators. The performance risk monitoring indicator system for the retail market is as follows:

[0352] (1) Risk Factors

[0353] 1) Are there any policy risks?

[0354] Significant policy changes could lead to widespread defaults by electricity sales companies, increasing the likelihood that users will face performance risks due to policy risks.

[0355] 2) Does the electricity sales company face operational risks?

[0356] The operational risks of electricity sales companies are a significant cause of performance risks. The more operational risks an electricity sales company faces, the greater the likelihood of performance risks for electricity users after signing a contract with it.

[0357] 3) Debt-to-asset ratio of electricity sales companies

[0358] The debt-to-equity ratio of an electricity sales company measures its solvency and is an important financial indicator. A higher debt-to-equity ratio indicates higher financial risk and a greater likelihood of contractual obligations being missed by electricity users. The calculation formula is as follows:

[0359]

[0360] 4) Whether the electricity sales company continues to meet the registration requirements.

[0361] If a power sales company fails to pass the verification process for consistently meeting registration requirements, it must rectify the non-compliance. During the rectification period, its trading qualification will be suspended, which may affect its performance of obligations to electricity users. Whether a power sales company continues to meet registration requirements needs to be examined:

[0362] Accuracy of registration information: In accordance with the relevant requirements for market member access management, the key registration information of market entities is complete and accurate.

[0363] Timeliness of Registration Information Changes: Market entities must apply for changes to their registration information to the power trading center within 5 working days. Failure to do so within 5 working days will be considered a delay in the timeliness of the registration information change.

[0364] Annual Public Notice: When market entities change their registration information, they should apply for the change to the power trading center within 5 working days. If more than 5 working days are exceeded, the registration information change will be considered untimely.

[0365] 5) Is the electricity sales company's technical support system complete?

[0366] A technical support system is a prerequisite for an electricity sales company to participate in electricity trading. The completeness of the technical support system needs to be examined from the following aspects:

[0367] Disaster recovery capabilities of the electricity sales company's business information system;

[0368] According to the "Administrative Measures for Electricity Sales Companies", the technical support system of electricity sales companies should have functions such as transaction quotation, information reporting, contract signing, and customer service.

[0369] Ministry of Industry and Information Technology (MIIT) ICP filing status;

[0370] Status of the system's emergency response mechanism;

[0371] Internet access status;

[0372] System network security protection capabilities.

[0373] 6) Does the electricity sales company have any serious breaches of trust?

[0374] Serious breaches of trust include violating credit commitments, having serious negative credit records, failing to pay performance guarantee certificates in a timely and sufficient manner, failing to pay related fees in a timely and sufficient manner, and posing serious contract risks. If a power sales company has serious breaches of trust, the likelihood of performance risks faced by power users after signing a contract with them is greater.

[0375] VI. Settlement Risk Monitoring Indicators

[0376] The retail market settlement risk monitoring indicator system is designed as follows:

[0377] (1) Risk Factors

[0378] 1) Frequency of archival issues

[0379] Record-related issues include the timely synchronization or errors in electricity users' transaction records, marketing records, and settlement records. This indicator refers to the frequency of such problems occurring among electricity users within the statistical period; the higher the frequency, the higher the probability of settlement risk for market participants. Market regulatory and operational agencies can set corresponding warning frequencies.

[0380] 2) Percentage of users with incorrect profiles

[0381] The percentage of users with incorrect records refers to the proportion of electricity users with incorrect records within a statistical period to the total number of users. A higher value indicates a higher market settlement risk. The calculation formula is as follows:

[0382]

[0383] 3) Do electricity users update information in a timely manner?

[0384] If electricity users experience policy adjustments, changes in registration information, or changes in electricity sales companies, timely updates to their registration information and transaction records can effectively prevent settlement risks.

[0385] (2) Risk outcome category

[0386] 1) Number of complaints due to settlement issues

[0387] This refers to the number of complaints received from market entities regarding settlement issues within a statistical period. The higher the number of complaints, the higher the settlement risk in the retail market. Market supervision and operation agencies can set corresponding warning thresholds.

[0388] VII. Compliance Risk Monitoring Indicators

[0389] The retail market compliance risk monitoring indicator system is designed as follows:

[0390] (1) Risk Factors

[0391] 1) Is there a risk of market power imbalance?

[0392] Unfair competition by electricity retailers is a significant manifestation of regulatory violations. Therefore, when market power risks exist in the retail market, the probability of market players facing compliance risks increases.

[0393] 2) Retail user real-person authentication rate

[0394] Retail user real-person authentication rate refers to the proportion of users who have completed real-person authentication in the retail market within a statistical period. A higher proportion indicates a more standardized market operation, more effective market supervision, and lower compliance risks for market participants. The calculation formula is as follows:

[0395]

[0396] 3) Accuracy rate of retail user registration information

[0397] Retail user registration information verification includes company name, legal representative, and unified social credit code. The accuracy rate of retail user registration information refers to the proportion of retail users whose registration information verification items are all accurate within the statistical period, out of the total number of retail market users. The higher this value, the lower the compliance risk for the market entity. The calculation formula is as follows:

[0398]

[0399] 4) Does the electricity sales company have sound systems and standardized management?

[0400] This refers to whether a power sales company has sound rules and regulations, a complete organizational structure, and standardized internal control management systems and processes. If these are in place, the probability of compliance risks occurring for the power sales company and its contracted users is lower.

[0401] (2) Risk outcome category

[0402] 1) Frequency of incidents involving leakage of user personal information and identity theft

[0403] This refers to the frequency of incidents involving the leakage of user personal information and identity theft within the statistical period. The higher the frequency, the higher the probability of compliance risks for market entities. Market supervision and operation agencies can set corresponding warning frequencies.

[0404] 2) Frequency of violations and penalties

[0405] This refers to the total frequency of events in which market entities are penalized for violations during the statistical period. The higher the frequency, the higher the probability of compliance risks for market entities. Market supervision and operation agencies can set corresponding warning frequencies.

[0406] 3) Fraud complaint rate

[0407] The fraud complaint rate refers to the proportion of electricity sales companies that have received complaints from electricity users due to false information, price fraud, or other reasons within a statistical period, out of the total number of electricity sales companies participating in market transactions. The higher this ratio, the more non-compliant electricity sales companies there are, and the higher the compliance risk for market participants. The calculation formula is as follows:

[0408]

[0409] 4) Number of times the electricity sales company has been penalized for violations

[0410] This refers to the number of times a power sales company is penalized for violations during the statistical period. The higher the frequency, the higher the probability of compliance risk for that power sales company.

[0411] 5) Number of complaints against the electricity sales company

[0412] This refers to the number of times a power sales company is complained about by electricity users for fraud or other misconduct within a statistical period. The higher the frequency, the higher the probability of compliance risk for that power sales company.

[0413] VIII. Policy Risk Monitoring Indicators

[0414] Policy risk monitoring needs to take into account both changes in electricity market rules and adjustments in macroeconomic policies. The indicator system for monitoring policy risks in the retail market is designed as follows:

[0415] (1) Risk Factors

[0416] 1) Macroeconomic Prosperity Early Warning Index

[0417] The Macroeconomic Prosperity Early Warning Index released by the National Bureau of Statistics is a composite prosperity index reflecting the operation of my country's macroeconomy, developed by the Macroeconomic Monitoring and Early Warning Research Group of the National Information Center. It includes 10 early warning indicators: industrial production index, fixed asset investment, total retail sales of consumer goods, total import and export volume, fiscal revenue, industrial enterprise profits, disposable income of residents, various loans from financial institutions, money supply M2, and consumer price index. The index divides the economic operation into five levels, with 100 being the ideal level. A "red light" indicates an overheated economy, a "yellow light" indicates a slightly overheated economy, a "green light" indicates normal economic operation, a "light blue light" indicates a slightly sluggish economy, and a "blue light" indicates an excessively sluggish economy.

[0418] When the economy is booming, businesses are highly motivated to produce, and electricity consumption is likely to increase; when the economy is sluggish, businesses are less motivated to produce or even face survival challenges, and electricity consumption is likely to decrease.

[0419] 2) Whether a new or revised version of the electricity market rules has been issued.

[0420] Changes in electricity market rules are one source of policy risk. When new or revised versions of electricity market rules are issued, the probability of policy risks for market participants increases.

[0421] 3) Industry attributes of electricity users

[0422] Electricity users are affected by macroeconomic policies to varying degrees due to their different industry attributes. Some industries belong to basic industries, with relatively inelastic electricity demand, and their electricity consumption is relatively stable or exhibits relatively regular fluctuations in the long term. Some industries belong to outdated industries, high-energy-consuming and high-polluting industries, and may be gradually phased out due to industrial policies; conversely, some industries belong to high-tech industries and strategic emerging industries, encouraged and supported by national policies, and their production and operation may continue to expand.

[0423] (2) Risk outcome category

[0424] 1) Changes in user electricity consumption affected by macroeconomic policies

[0425] This indicator measures the impact of policy changes on electricity consumers. The higher the value, the higher the market policy risk. The calculation formula is as follows:

[0426]

[0427] In the formula, Indicates the time point when the policy was released. This indicates the time from the policy's release to the end of the statistical period. This refers to electricity users in the retail market. Time to Cumulative electricity consumption at any given time This refers to electricity users in the retail market. Time to The cumulative electricity consumption at any given moment.

[0428] 2) Adaptability of the profit model of electricity sales companies

[0429] The adaptability of a power sales company's profit model refers to whether the company's current operating model and business scope meet and adapt to the requirements of new or revised electricity market rules. If they meet or adapt well, the power sales company has a lower policy risk.

[0430] 3) Change in revenue of electricity sales companies after policy implementation

[0431] The rate of change in electricity sales company revenue directly measures the impact of policy announcements on the company's operations. A higher value indicates a more significant impact from the policy and a greater policy risk. It is important to note that comparing revenue changes before and after the policy change requires selecting the same time period. The calculation formula is as follows:

[0432]

[0433] In the formula, Indicates the time point when the policy was released. This indicates the time from the policy's release to the end of the statistical period. This refers to the electricity sales company itself Time to Cumulative earnings at any given moment This refers to the electricity sales company itself Time to Cumulative earnings at any given moment.

[0434] 2.2 Indicator Integration

[0435] Based on the formulas in the above indicator system, various indicators can be calculated based on the structured data obtained in step S1.

[0436] Then these indicators are categorized and processed:

[0437] For indicators that can be measured by high-frequency, near-real-time, and objective structured data, such as price risk, operational risk, and settlement risk, the corresponding primary and secondary indicators can be used to calculate the structured data indicators of this type of risk as comprehensive monitoring indicators.

[0438] For the remaining risk types, their driving factors may include a large amount of soft information that is difficult to directly represent with structured data, such as policy changes, market sentiment, public opinion, and corporate reputation. Alternatively, their structured data may be lagging or limited, requiring unstructured data to provide early signals. In such cases, a comprehensive monitoring indicator can be obtained by combining the structured indicators of this type of risk with the risk factor values ​​mentioned above.

[0439] For the market as a whole (and the same applies to individual markets), the calculation process for a structured indicator of a type of risk (a broad category, such as market power risk) is as follows:

[0440] 1) Assign weights to all market indicators for this type of risk. The sum of all indicator weights is 1. Specifically, the weights are preset based on expert experience, with higher weights for indicators that better reflect market risk.

[0441] 2) The coefficient of indicators positively correlated with market risk is set to 1, and the coefficient of indicators negatively correlated with market risk is set to -1.

[0442] 3) Set an ideal value for each indicator, so as to... Based on this, and combined with coefficients and weights, a weighted calculation is performed to obtain a structured indicator of the corresponding type of risk for the overall market.

[0443] Comprehensive monitoring indicators are calculated by combining structured indicators and risk factor values. The calculation formula is as follows:

[0444]

[0445] in, For risk factor values, Structured indicators; , The weighting coefficients used to balance the importance of indicators from two different sources can be set through training with historical data or expert experience. Note: When calculating risk factor values ​​for individual companies, only risk events when that company is the risk subject are considered.

[0446] like Figure 3 As shown in the figure, according to an embodiment of the present invention, an apparatus for generating target risk information in the electricity retail market is provided, comprising:

[0447] The acquisition module is used to acquire a first dataset and a second dataset; wherein the first dataset is structured data used to characterize the status of core elements in the electricity retail market, and the second dataset is unstructured data used to characterize the market environment and policy situation.

[0448] An extraction module is used to process the second dataset based on a preset risk factor extraction algorithm to obtain risk factor information for at least one risk type; wherein, the preset risk factor extraction algorithm is an algorithm used to identify risk events from unstructured text and quantify their impact.

[0449] The processing module is used to process the first dataset based on a preset electricity retail market risk monitoring indicator system to obtain structured risk information of at least one risk type.

[0450] The weighted fusion module is used to weightedly fuse the risk factor information and the structured risk information of the same risk type to generate target risk information of that risk type.

[0451] According to an embodiment of the present invention, an electronic device is provided; please refer to... Figure 4 The electronic device in this embodiment may include one or more of the following components: a processor, a network interface, memory, non-volatile memory, and one or more application programs, wherein the one or more application programs may be stored in non-volatile memory and configured to be executed by one or more processors, and the one or more programs are configured to perform the methods as described in the foregoing method embodiments.

[0452] According to embodiments of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a computer, causes the computer to perform the method described in any of the above embodiments.

[0453] According to embodiments of the present invention, a computer program product comprising instructions is also provided, which, when executed by a computer, cause the computer to perform a method in any of the above embodiments.

[0454] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating target risk information in the electricity retail market, characterized in that, include: Obtain a first dataset and a second dataset; wherein the first dataset is structured data used to characterize the status of core elements in the electricity retail market, and the second dataset is unstructured data used to characterize the market environment and policy situation; Based on a preset risk factor extraction algorithm, the second dataset is processed to obtain risk factor information for at least one risk type; wherein, the preset risk factor extraction algorithm is an algorithm used to identify risk events from unstructured text and quantify their degree of impact; Based on the preset risk monitoring indicator system for the electricity retail market, the first dataset is processed to obtain structured risk information for at least one risk type. The risk factor information and the structured risk information of the same risk type are weighted and fused to generate the target risk information of that risk type.

2. The method according to claim 1, characterized in that, The step of processing the second dataset based on a preset risk factor extraction algorithm to obtain risk factor information for at least one risk type includes: Based on a pre-built dictionary of electricity market risk keywords, risk-related text content is identified from the second dataset; wherein, the text content includes risk keywords. Using a pre-defined named entity recognition algorithm, key entities constituting the risk event are extracted from the identified text content; Based on predefined risk impact calculation rules and the key entities, the degree of risk impact of the risk event is calculated; Based on a preset sentiment analysis algorithm, the sentiment polarity of text content including the risk keywords is determined, and the corresponding risk impact direction coefficient is output. By combining the risk impact degree of at least one of the aforementioned risk events and its corresponding risk impact direction coefficient, the risk factor information of the risk type is calculated.

3. The method according to claim 2, characterized in that, The dictionary of keywords related to electricity market risks is pre-constructed using the following method: Based on a pre-set set of seed keywords for electricity market risks, a word embedding model is trained on a pre-acquired historical text corpus to calculate extended keywords that are semantically similar to the seed keywords. The seed keywords and the extended keywords are merged to generate the electricity market risk keyword dictionary.

4. The method according to claim 2, characterized in that, The step of calculating the degree of risk impact of the risk event based on predefined risk impact calculation rules and the key entity includes: Based on the risk event type indicated by the key entity, the corresponding basic impact factor is queried from the pre-built event-risk knowledge base; Based on the risk impact scope information indicated by the key entities, calculate the scope impact factor; Based on the risk consequences information indicated by the key entities, severity impact factors are determined from a pre-built consequences-risk knowledge base; The degree of risk impact of the risk event is calculated based on the product of the basic impact factor, the range impact factor, and the severity impact factor.

5. The method according to claim 1, characterized in that, The preset risk monitoring indicator system for the electricity retail market includes multiple risk categories. Each risk category includes risk factor indicators and risk outcome indicators, and the indicators are defined for the overall market and individual market entities, respectively. The step of processing the first dataset based on the preset electricity retail market risk monitoring indicator system to obtain structured risk information for at least one risk type includes: Based on the target risk type, determine the corresponding risk factor indicators and risk outcome indicators from the indicator system; Based on the first dataset, calculate the overall market indicator value and the individual market participant indicator value corresponding to the target risk type, respectively. Based on a preset weighting scheme, the overall market indicator value and the individual market participant indicator value are weighted and fused to obtain the structured risk information of the target risk type.

6. The method according to claim 5, characterized in that, The aforementioned risk categories include market power risk, price risk, operational risk, contract risk, performance risk, settlement risk, compliance risk, and policy risk. Among them, the risk factor indicators corresponding to the market power risk include at least the market concentration index defined for the overall market and the price deviation rate index defined for individual market participants; the risk outcome indicators corresponding to the market power risk include at least the proportion of active electricity sales companies defined for the overall market and the total market share index defined for individual market participants. Among them, the risk factor indicators corresponding to the price risk include at least a wholesale-retail price correlation indicator defined for the overall market and a wholesale-retail price correlation indicator for individual market participants; the risk outcome indicators corresponding to the price risk include at least a wholesale-retail price inversion rate indicator defined for the overall market and a wholesale-retail price inversion rate indicator for individual market participants. The risk factor indicators corresponding to the operational risks include at least the wholesale market price volatility indicator defined for the overall market and the load forecast accuracy indicator defined for individual market participants; the risk outcome indicators corresponding to the operational risks include at least the proportion of loss-making electricity sales companies defined for the overall market and the user stickiness indicator defined for individual market participants.

7. The method according to claim 1, characterized in that, The step of weightedly fusing the risk factor information and the structured risk information of the same risk type to generate target risk information for that risk type includes: Based on the risk type, a first weighting coefficient configured for the risk factor information and a second weighting coefficient configured for the structured risk information are obtained. Calculate the product of the first weighting coefficient and the risk factor information to obtain the first weighted result; Calculate the product of the second weighting coefficient and the structured risk information to obtain the second weighted result; The first weighted result and the second weighted result are summed to generate the target risk information for this risk type.

8. A device for generating target risk information in the electricity retail market, characterized in that, include: The acquisition module is used to acquire a first dataset and a second dataset; wherein the first dataset is structured data used to characterize the status of core elements in the electricity retail market, and the second dataset is unstructured data used to characterize the market environment and policy situation. An extraction module is used to process the second dataset based on a preset risk factor extraction algorithm to obtain risk factor information for at least one risk type; wherein, the preset risk factor extraction algorithm is an algorithm used to identify risk events from unstructured text and quantify their impact. The processing module is used to process the first dataset based on a preset electricity retail market risk monitoring indicator system to obtain structured risk information of at least one risk type. The weighted fusion module is used to weightedly fuse the risk factor information and the structured risk information of the same risk type to generate target risk information of that risk type.

9. An electronic device, characterized in that, include: A memory, and one or more processors communicatively connected to the memory; The memory stores instructions that can be executed by the one or more processors to cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 7.

Images