Power market subject risk early warning method

Abstract

The invention relates to the technical field of power market risk management, in particular to a power market subject risk early warning method. According to the method, macroscopic and microscopic data are combined, the risk of the power market is analyzed and evaluated by utilizing multidimensional analysis, accurate risk early warning signals are provided, and market subjects are helped to identify potential high-risk transactions in time. A support vector machine model is adopted to evaluate microscopic transaction data in real time, and the risk level is dynamically adjusted, so that market fluctuation can be monitored in real time, an early warning response mechanism can be triggered, and high-risk transactions can be intervened in time. Through a multi-level risk identification and evaluation mechanism, illegal transaction subjects are marked and tracked, the power market transaction compliance is ensured, the market supervision efficiency and transparency are improved, and the market stability is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of power market risk management, and specifically to a risk warning method for power market entities. Background Art

[0002] With the gradual opening and development of the power market, the trading mechanism in the power market has become more complex, and the unpredictability and risk of market fluctuations have increased, resulting in more complex risk challenges for market entities. In this context, how to effectively identify and control potential risks in the power market has become an important issue in power market supervision and management.

[0003] Although there are already some technical solutions for power market risk assessment in the market, most of them focus on macro-level data analysis or traditional regression analysis methods, and have the following deficiencies:

[0004] The risk warning mechanisms of the existing technologies mostly rely on regular analysis and static models, lacking the ability of real-time monitoring and rapid response to dynamic market changes. For the power market environment with rapid changes and complexity, the existing warning mechanisms are often difficult to cope with emergencies or abnormal fluctuations.

[0005] Secondly, the existing technologies often only focus on macro market indicators and lack in-depth analysis of individual trading behaviors. Even for the analysis methods based on big data, it is often impossible to accurately capture the characteristics of individual high-risk transactions, resulting in the inability to effectively identify potential high-risk transactions in the market.

[0006] In view of the above problems, it is necessary to propose a risk warning method for power market entities. Summary of the Invention

[0007] The purpose of the present invention is to solve the problems existing in the background art and propose a risk warning method for power market entities.

[0008] The purpose of the present invention can be achieved through the following technical solutions:

[0009] A risk warning method for power market entities includes the following steps:

[0010] Step 1: Data collection and preprocessing;

[0011] Access the data recorder and power market transaction records to obtain the original power market data therein, including power transaction data, power company operation and maintenance data, policy and regulation data, and market price fluctuation data.

[0012] As a preferred mode of the present invention, perform structured data extraction on the collected original power market data, and the specific process is:

[0013] Extract microtransaction data from the original power market data to obtain information on the buyer and seller in the power market data, including the transaction number i of each transaction, the transaction volume v(i), the name of the buyer buyer(i), the name of the seller seller(i), the transaction price p(i), the settlement time t(y), and the transaction type symbol type(i).

[0014] Among them, the specific value of the transaction type symbol corresponds to a preset transaction type, and the transaction types include power spot transactions and futures transactions.

[0015] Extract macrotransaction data from the original power market data. At preset time intervals, obtain the average market transaction price P_Ave(t), the highest price P_Max(t), and the lowest price P_Min(t) of power transactions; obtain the real-time power transaction volume V(t), market demand Dem(t), and market supply Sup(t) at each moment t; where t is the timestamp, that is, the collection moment of each data.

[0016] As a preferred embodiment of the present invention, data preprocessing is performed, including microtransaction data preprocessing and macrotransaction data preprocessing.

[0017] The specific process of microdata preprocessing is as follows:

[0018] Taking power spot transactions as the object, standardize the microtransaction data, calculate the means μ11, μ12, and μ13 of the transaction volume, price, and time of all power spot transactions in the power market, and calculate the standard deviations σ11, σ12, and σ13 of the transaction volume, price, and time of all power spot transactions in the power market.

[0019] For the transaction number i with the transaction type of power spot transaction, generate a standardized microtransaction data vector:

[0020] Taking power futures transactions as the object, standardize the microtransaction data, calculate the means μ21, μ22, and μ23 of the transaction volume, price, and time of all power futures transactions in the power market, and calculate the standard deviations σ21, σ22, and σ23 of the transaction volume, price, and time of all power futures transactions in the power market.

[0021] For the transaction number i with the transaction type of power futures transaction, generate a standardized microtransaction data vector:

[0022] The specific process of macrodata preprocessing is as follows:

[0023] Smooth the macrotransaction data changing along the timestamp t through Gaussian filtering, and the filtering function is: Where x is the input variable, i.e., the collected macroscopic transaction data, including the market transaction average price, the highest price, the lowest price, the real-time electricity trading volume, the market demand, and the market supply collected at each moment; where σ is the preset standard deviation, which determines the smoothness of the Gaussian filtering result. Where h(x) is the output value after noise reduction processing by the Gaussian filtering function.

[0024] Collect the market transaction average price, the highest price, the lowest price, the real-time electricity trading volume, the market demand, and the market supply of electricity transactions at each preset time interval t after data noise reduction, and generate the macroscopic transaction data vector H(t) at time t.

[0025] H(t) = [P_Ave(t), P_Max(t), P_Min(t), V(t), Dem(t), Sup(t)] T 。

[0026] Step 2: Macro feature extraction and risk level identification;

[0027] Extract features and identify risks for the macroscopic transaction data vectors at each moment through the mean reversion model, evaluate the macroscopic risk level contained in the behavior of the electricity price returning to the long-term average value, and obtain the corresponding risk factors.

[0028] Establish a mean reversion model for the macroscopic transaction data for each piece of macroscopic transaction data after data noise reduction. The model function is: H(t) = H(t - 1) + θ[μ - H(t - 1)] + ε(t); where H(t) is the macroscopic transaction data vector, H(t) = [P_Ave(t), P_Max(t), P_Min(t), V(t), Dem(t), Sup(t)] T

[0029] Where θ is the mean reversion vector, which controls the regression speed of the macroscopic transaction data and reflects the overall adjustment ability of the market. θ = [θ1, θ2, θ3, θ4, θ5, θ6]. Where θ1, θ2, θ3, θ4, θ5, θ6 are the regression coefficients of the market transaction average price, the highest price, the lowest price, the real-time electricity trading volume, the market demand, and the market supply respectively, and are used to quantitatively evaluate the speed at which the corresponding macroscopic transaction data returns to its average level in the long term. The larger the regression coefficient, the faster the corresponding macroscopic transaction data returns to the long-term mean.

[0030] Where μ is the long-term average value vector, μ = [μ1, μ2, μ3, μ4, μ5, μ6], and the components μ1, μ2, μ3, μ4, μ5, μ6 correspond to the means of the historical data of the market transaction average price, the highest price, the lowest price, the real-time electricity trading volume, the market demand, and the market supply respectively, representing the regression target of the macroscopic transaction data.

[0031] Where ε(t) is the error vector, and ε(t) = θ = [θ1(t), θ2(t), θ3(t), θ4(t), θ5(t), θ6(t)]. The components θ1(t), θ2(t), θ3(t), θ4(t), θ5(t), θ6(t) are all random perturbation variables that conform to the Gaussian distribution, corresponding to the random perturbations of the average transaction price, the highest price, the lowest price, the real-time electricity trading volume, the market demand, and the market supply at time t respectively, representing the abnormal fluctuations in the market that cannot be explained by known data.

[0032] Every preset statistical period T, input the macro transaction data vector H(t) at each time t collected in the previous statistical period T into the mean reversion model of the macro transaction data, and solve the mean reversion vector θ, the error vector ε(t) at each time t, and the long-term average vector μ in the current statistical period T.

[0033] As a preferred embodiment of the present invention, every preset statistical period T, based on the mean reversion vector, the long-term average vector, and the error vector at each time obtained through the mean reversion model, perform feature extraction and analysis, evaluate the macro risk level, and identify the factors that may affect the risks of the electricity market entities. The specific process is as follows:

[0034] Perform feature extraction and analysis on the long-term average vector and the mean reversion vector, and calculate the deviation fluctuation eigenvalue Deviation of the market in the previous statistical period T through the preset formula

[0035] Perform feature extraction and analysis on the error vector, and calculate the irrational fluctuation eigenvalue Volatility of the market in the previous statistical period T through the preset formula

[0036] As a preferred embodiment of the present invention, calculate the risk factor Risk through the comprehensive risk factor formula Risk = w1×Deviation + w2×Volatility. According to the specific value of the risk factor, perform risk classification, and further divide the macro risk level of the market into different levels. The specific process is as follows:

[0037] If the risk factor Risk is less than the first preset threshold RiskMin, it is determined that the electricity market is in the macro low-risk level, the market tends to be stable, the regression process is faster and the random fluctuations are less;

[0038] If the risk factor Risk is less than the second preset threshold RiskMax and greater than or equal to the first preset threshold RiskMax, it is determined that the electricity market is in the macro medium-risk level, and there are certain fluctuations in the electricity market, and the regression process is delayed.​​

[0039] If the risk factor Risk is greater than or equal to the second preset threshold RiskMax, it is determined that the power market is in a macro high-risk level, the power market fluctuates violently, the price deviates far from the long-term average, and the regression speed is slow.

[0040] Step 3: Micro risk assessment;

[0041] Use the support vector machine model to conduct risk assessment on the micro transaction data of the power market.

[0042] Obtain the standardized micro transaction data vector, including the micro transaction data vector for power spot transactions: and the micro transaction data vector for power futures transactions:

[0043] Take the components in the standardized micro transaction data vector as input parameters, extract features and conduct risk assessment on the micro transaction data through the support vector machine model, evaluate the risk level of the micro transaction, and match it with the risk factor to obtain a transaction risk score representing the risk degree of the micro transaction.

[0044] The decision function of the support vector machine model is: f[M(i)] = sign(w T M(i)+b); where w is the weight vector of the support vector machine and b is the bias term; where f[M(i)] is the output value, that is, the transaction risk score output by the support vector machine model; the value range of f[M(i)] is from 0 to 100, the higher the value of the output value f[M(i)], the higher the transaction risk of this transaction i; the lower the value of the output value f[M(i)], the lower the transaction risk of this transaction i;

[0045] Train the support vector machine model to obtain a trained support vector machine model. The specific process is as follows:

[0046] Collect the micro transaction data vectors corresponding to power spot transactions and power futures transactions that have been manually determined as illegal transactions, and mark them, making the corresponding marked output value y(i) equal to 100. Use them as the training data of the support vector machine to conduct model training, record the output value f[M(i)] obtained by the operation and processing of the micro transaction data vectors corresponding to power spot transactions and power futures transactions that have been manually determined as illegal transactions through the support vector machine model, and make the loss function of the training process: Calculate the weight vector w and bias term b when the minimum loss function is obtained through the steepest descent method, and bring them back to the support vector machine model for risk prediction of newly input micro transaction data vectors.

[0047] As a preferred embodiment of the present invention, the trained support vector machine model is used to perform arithmetic processing on all micro-transaction data vectors to obtain the output value f[M(i)] corresponding to each micro-transaction data vector M(i), that is, the transaction risk score of transaction i.

[0048] Step Four: Early Warning Signal Matching and Generation;

[0049] Based on the risk level obtained from the arithmetic analysis of the macro-transaction data and the transaction risk score obtained from the arithmetic analysis of the micro-transaction data, matching is performed to generate corresponding early warning signals.

[0050] If the macro risk level is low, then obtain all micro-transaction data vectors M(i) whose transaction risk scores are greater than the preset threshold M1, and mark them as early warning transactions;

[0051] If the macro risk level is medium, then obtain all micro-transaction data vectors M(i) whose transaction risk scores are greater than the preset threshold M2, and mark them as early warning transactions;

[0052] If the macro risk level is high, then obtain all micro-transaction data vectors M(i) whose transaction risk scores are greater than the preset threshold M3, and mark them as early warning transactions;

[0053] Where M1, M2, and M3 are all preset transaction risk score thresholds, and M3 > M2 > M1.

[0054] For all micro-transaction data vectors M(i) marked as early warning transactions, obtain their corresponding transaction parties, that is, the buyer name buyer(i) and the seller name seller(i), and mark them as early warning transaction parties.

[0055] Every preset statistical period T, count the number of times all buyer names buyer(i) and seller names seller(i) are marked as early warning transaction parties.

[0056] For a buyer or seller whose number of times marked as an early warning transaction party exceeds the preset upper limit of the number of times, generate a corresponding early warning signal.

[0057] Step Five: Early Warning Response and Feedback Execution;

[0058] Once the early warning signal is triggered, start the early warning response mechanism, conduct a market behavior investigation on the transaction party that triggered the early warning signal, and require it to provide further transaction explanations for a more in-depth review of the transaction behavior. Continuously monitor the transaction party that has triggered the early warning signal to track whether there are other transaction violations.

[0059] As a preferred embodiment of the present invention, a detailed risk report is generated, and the content of the report includes: the type of warning signal triggered, the warning transactions involved and the information of the transaction parties, including the names of the buyer and seller, the trading volume and the trading price, as well as the risk assessment result and the subsequent countermeasures.

[0060] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0061] 1. The present invention combines data at the macro and micro levels and evaluates market risks through multi-dimensional analysis (including the average market transaction price, trading volume, market demand and supply, etc.). Through the combined analysis of macro market risks and micro transaction risks, it can provide more accurate risk warning signals for electricity market entities and help market participants identify potential high-risk transactions in a timely manner;

[0062] 2. The present invention uses a support vector machine model to evaluate micro transaction data in real time and dynamically adjusts the risk level in combination with macro market changes. This technical solution can monitor the risk fluctuations of the market in real time, trigger corresponding warning response mechanisms according to different risk levels, and intervene in high-risk transaction entities in a timely manner to reduce the uncertainty of the market;

[0063] 3. The present invention can mark and track illegal transaction entities through a multi-level risk identification and assessment mechanism to ensure that the trading behaviors in the electricity market comply with the regulations. At the same time, through means such as risk reports and market behavior investigations, the efficiency and transparency of market supervision are enhanced, and the stability and fairness of the market are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] For the convenience of those skilled in the art to understand, the present invention will be further described below in conjunction with the accompanying drawings:

[0065] Figure 1 is the method flow chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0067] Please refer to Figure 1 As shown, a method for warning risks of electricity market entities includes the following steps:

[0068] Step 1: Data collection and preprocessing;

[0069] Access the data recorder and power market transaction records to obtain the original power market data therein, including power transaction data, power company operation and maintenance data, policy and regulation data, and market price fluctuation data.

[0070] Furthermore, perform structured data extraction on the collected original power market data. The specific process is as follows:

[0071] Extract microtransaction data from the original power market data to obtain the buyer and seller information in the power market data, including the transaction number i of each transaction, the trading volume v(i), the buyer's name buyer(i), the seller's name seller(i), the transaction price p(i), the settlement time t(y), and the transaction type symbol type(i).

[0072] Among them, the specific value of the transaction type symbol corresponds to a preset transaction type, and the transaction types include power spot trading and futures trading.

[0073] Extract macrotransaction data from the original power market data. At preset time intervals, obtain the average market transaction price P_Ave(t), the highest price P_Max(t), and the lowest price P_Min(t) of power transactions; obtain the real-time power trading volume V(t), market demand Dem(t), and market supply Sup(t) at each moment t; where t is the timestamp, that is, the collection moment of each data.

[0074] Furthermore, perform data preprocessing, including microtransaction data preprocessing and macrotransaction data preprocessing.

[0075] The specific process of microdata preprocessing is as follows:

[0076] Taking power spot trading as the object, standardize the microtransaction data, calculate the means μ11, μ12, and μ13 of the trading volume, price, and time of all power spot transactions in the power market, and calculate the standard deviations σ11, σ12, and σ13 of the trading volume, price, and time of all power spot transactions in the power market.

[0077] For the transaction number i with the transaction type of power spot trading, generate a standardized microtransaction data vector:

[0078] Taking power futures trading as the object, standardize the microtransaction data, calculate the means μ21, μ22, and μ23 of the trading volume, price, and time of all power futures transactions in the power market, and calculate the standard deviations σ21, σ22, and σ23 of the trading volume, price, and time of all power futures transactions in the power market.

[0079] For the transaction number i with the transaction type of power futures trading, generate a standardized micro-transaction data vector:

[0080] The specific process of macro-data preprocessing is as follows:

[0081] Smooth the macro-transaction data changing along the time stamp t through Gaussian filtering, and the filtering function is: where x is the input variable, that is, the collected macro-transaction data, including the market transaction average price, highest price, lowest price, real-time power trading volume, market demand, and market supply of power trading collected at each moment; where σ is the preset standard deviation, which determines the smoothness of the Gaussian filtering result. And h(x) is the output value after noise reduction processing by the Gaussian filtering function.

[0082] Collect the market transaction average price, highest price, lowest price, real-time power trading volume, market demand, and market supply of power trading at each preset time interval t after data noise reduction, and generate a macro-transaction data vector H(t) at time t,

[0083] H(t) = [P_Ave(t), P_Max(t), P_Min(t), V(t), Dem(t), Sup(t)] T .

[0084] Step 2: Macro-feature extraction and risk level identification;

[0085] Extract features and identify risks for the macro-transaction data vector at each moment through a mean reversion model, evaluate the macro-risk level contained in the behavior of the power price returning to the long-term average value, and obtain the corresponding risk factors.

[0086] Establish a mean reversion model for the macro-transaction data for each macro-transaction data after data noise reduction. The model function is: H(t) = H(t - 1) + θ[μ - H(t - 1)] + ε(t); where H(t) is the macro-transaction data vector, and H(t) = [P_Ave(t), P_Max(t), P_Min(t), V(t), Dem(t), Sup(t)] T

[0087] where θ is the mean reversion vector, which controls the reversion speed of the macro trading data and reflects the overall adjustment ability of the market. θ = [θ1, θ2, θ3, θ4, θ5, θ6]. Among them, θ1, θ2, θ3, θ4, θ5, θ6 are the reversion coefficients of the average transaction price, highest price, lowest price, real-time electricity trading volume, market demand, and market supply in the market, respectively, and are used to quantitatively evaluate the speed at which the corresponding macro trading data reverts to its average level in the long term. The larger the reversion coefficient, the faster the corresponding macro trading data reverts to the long-term mean.

[0088] where μ is the long-term average value vector, μ = [μ1, μ2, μ3, μ4, μ5, μ6], and the components μ1, μ2, μ3, μ4, μ5, μ6 correspond to the means of the historical data of the average transaction price, highest price, lowest price, real-time electricity trading volume, market demand, and market supply in the market, respectively, representing the reversion target of the macro trading data.

[0089] where ε(t) is the error vector, ε(t) = θ = [θ1(t), θ2(t), θ3(t), θ4(t), θ5(t), θ6(t)]. The components θ1(t), θ2(t), θ3(t), θ4(t), θ5(t), θ6(t) are all random perturbation variables that conform to the Gaussian distribution, corresponding to the random perturbations of the average transaction price, highest price, lowest price, real-time electricity trading volume, market demand, and market supply at time t, respectively, representing the abnormal fluctuations in the market that cannot be explained by the known data.

[0090] Every preset statistical period T, the macro trading data vector H(t) at each moment t collected within the previous statistical period T is input into the mean reversion model of the macro trading data to solve the mean reversion vector θ, the error vector ε(t) at each moment t, and the long-term average value vector μ within the current statistical period T.

[0091] Furthermore, every preset statistical period T, based on the mean reversion vector, long-term average value vector, and error vectors at each moment obtained through the mean reversion model, feature extraction and analysis are performed to evaluate the macro risk level and identify the factors that may affect the risks of electricity market entities. The specific process is as follows:

[0092] Perform feature extraction and analysis on the long-term average value vector and the mean reversion vector, and calculate the deviation fluctuation eigenvalue Deviation of the market within the previous statistical period T through a preset formula

[0093] ​It should be noted that the deviation fluctuation eigenvalue represents the difference between the mean reversion speed of each macro trading data and its long-term mean. The larger the deviation fluctuation eigenvalue, the more difficult it is for the regression speed of the macro trading data to match the regression target of the macro trading data, and the more difficult it is to maintain dynamic balance. It indicates that the market price regression process in this statistical period is less obvious and there may be higher risks.

[0094] Extract and analyze the features of the error vector through a preset formula Calculate the irrational volatility eigenvalue Volatility of the market in the previous statistical period T.

[0095] It should be noted that the irrational volatility eigenvalue represents the unexplained abnormal fluctuations in each macro trading data. These fluctuations mainly come from irrational factors in the market or external interferences, such as sudden policy changes, market sentiment changes, natural disasters, etc. These factors are often unable to be predicted or explained by traditional market analysis methods, so they are manifested as random disturbances in market data. The higher the irrational volatility eigenvalue, the more uncertainties in the market, and the greater the risks faced by market players in the power market.

[0096] Furthermore, calculate the risk factor Risk through the comprehensive risk factor formula Risk = w1×Deviation + w2×Volatility. According to the specific value of the risk factor, conduct risk classification, and further divide the macro risk level of the market into different levels. The specific process is as follows:

[0097] If the risk factor Risk is less than the first preset threshold RiskMin, it is determined that the power market is in the macro low-risk level, the market tends to be stable, the regression process is fast and the random fluctuations are less;

[0098] If the risk factor Risk is less than the second preset threshold RiskMax and greater than or equal to the first preset threshold RiskMax, it is determined that the power market is in the macro medium-risk level, and there are certain fluctuations in the power market, and the regression process is delayed.

[0099] If the risk factor Risk is greater than or equal to the second preset threshold RiskMax, it is determined that the power market is in the macro high-risk level, the power market fluctuates violently, the price deviates far from the long-term mean, and the regression speed is slow.

[0100] Step 3: Micro risk assessment;

[0101] Use the support vector machine model to conduct risk assessment on the micro trading data of the power market.

[0102] Obtain the standardized micro trading data vector, including the micro trading data vector of the power spot trading with the trading type: And the micro - transaction data vector with the transaction type being power futures trading:

[0103] Take the components in the standardized micro - transaction data vector as input parameters, and through the support vector machine model, extract features and evaluate risks for the micro - transaction data, evaluate the risk level of the micro - transaction, and match it with the said risk factors to obtain a transaction risk score representing the risk degree of the micro - transaction.

[0104] The decision function of the support vector machine model is: f[M(i)] = sign(w T M(i)+b); where w is the weight vector of the support vector machine, and b is the bias term; where f[M(i)] is the output value, that is, the transaction risk score output by the support vector machine model; the numerical range of f[M(i)] is from 0 to 100, the higher the value of the output value f[M(i)], the higher the transaction risk of this transaction i; the lower the value of the output value f[M(i)], the lower the transaction risk of this transaction i.

[0105] It should be noted that the goal of the support vector machine model is to find a hyperplane that maximizes the classification margin and determine the optimal w and b through the training data.

[0106] Train the support vector machine model to obtain a trained support vector machine model. The specific process is as follows:

[0107] Collect the micro - transaction data vectors corresponding to power spot transactions and power futures transactions that have been manually determined as illegal transactions, and mark them, making the corresponding marked output value y(i) equal to 100. Use them as the training data of the support vector machine to carry out model training, record the output value f[M(i)] obtained after the operation and processing of the micro - transaction data vectors corresponding to power spot transactions and power futures transactions that have been manually determined as illegal transactions by the support vector machine model, and let the loss function of the training process be: Calculate the weight vector w and the bias term b when obtaining the minimum loss function through the steepest descent method, and bring them back to the support vector machine model for risk prediction of newly input micro - transaction data vectors.

[0108] Furthermore, perform operation and processing on all micro - transaction data vectors by the trained support vector machine model to obtain the output value f[M(i)] corresponding to each micro - transaction data vector M(i), that is, the transaction risk score of transaction i.

[0109] Step Four: Early - warning signal matching and generation;

[0110] Match the risk level obtained from the operation and analysis of macro trading data with the trading risk score obtained from the operation and analysis of micro trading data to generate corresponding warning signals.

[0111] If the macro risk level is low, obtain the micro trading data vectors M(i) whose trading risk scores are greater than the preset threshold M1, and mark them as warning transactions;

[0112] If the macro risk level is medium, obtain the micro trading data vectors M(i) whose trading risk scores are greater than the preset threshold M2, and mark them as warning transactions;

[0113] If the macro risk level is high, obtain the micro trading data vectors M(i) whose trading risk scores are greater than the preset threshold M3, and mark them as warning transactions;

[0114] Where M1, M2, and M3 are all preset trading risk score thresholds, and M3 > M2 > M1.

[0115] For all micro trading data vectors M(i) marked as warning transactions, obtain their corresponding trading entities, namely the buyer name buyer(i) and the seller name seller(i), and mark them as warning trading entities.

[0116] Every preset statistical period T, count the number of times all buyer names buyer(i) and seller names seller(i) are marked as warning trading entities.

[0117] For a buyer or seller whose number of times marked as a warning trading entity exceeds the preset upper limit of the number of times, generate a corresponding warning signal.

[0118] Step Five: Warning Response and Feedback Execution;

[0119] Once the warning signal is triggered, activate the warning response mechanism, conduct a market behavior investigation on the trading entity that triggered the warning signal, and require it to provide further trading explanations for a more in-depth review of trading behaviors. Continuously monitor the trading entity that has triggered the warning signal to track whether it has other trading violations.

[0120] Furthermore, generate a detailed risk report, and the report content includes: the type of warning signal triggered, information on warning transactions and trading entities involved, including the names of buyers and sellers, trading volumes, and trading prices, as well as risk assessment results and subsequent countermeasures.

[0121] It should be understood that the terms "comprising" and "including" as used in the specification and claims of this disclosure indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0122] It should also be understood that the terms used in this disclosure specification are for the purpose of describing particular embodiments only and are not intended to limit this disclosure. As used in this disclosure specification and the claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms. It should further be understood that the term "and / or" as used in this disclosure specification and the claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations;

[0123] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the present invention to only the specific embodiments. Obviously, many modifications and variations can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

[0124] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the present invention to only the specific embodiments. Obviously, many modifications and variations can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A risk early warning method for power market entities, characterized in that, It includes the following steps: Step 1, data collection and preprocessing; Access the data recorder and power market transaction records to obtain the original power market data therein, including power transaction data, power company operation and maintenance data, policy and regulation data, and market price fluctuation data, and perform structured data extraction and preprocessing on the collected original power market data to obtain a microtransaction data vector and a macrotransaction data vector; Step 2, macro feature extraction and risk level identification; Perform feature extraction and risk identification on the macrotransaction data vector at each moment through a mean reversion model, evaluate the macro risk level contained in the behavior of the electricity price returning to the long-term average value, and obtain the corresponding risk factor; Step 3, micro risk assessment; Use a support vector machine model to perform risk assessment on the microtransaction data of the power market; Step 4, early warning signal matching and generation; Match the risk level obtained from the operation and analysis of the macrotransaction data with the transaction risk score obtained from the operation and analysis of the microtransaction data, and generate the corresponding early warning signal; Step 5, early warning response and feedback execution; Once the early warning signal is triggered, start the early warning response mechanism, conduct a review of the trading behavior according to the early warning signal, continuously monitor the trading entities that have triggered the early warning signal, and track whether they have other trading violations.

2. The risk warning method for a power market entity according to claim 1, wherein The specific process of performing structured data extraction on the collected original power market data is as follows: Extract microtransaction data from the original power market data to obtain the buyer and seller information in the power market data, including the transaction number i of each transaction, the trading volume v(i), the buyer name buyer(i), the seller name seller(i), the transaction price p(i), the settlement time t(y), and the transaction type symbol type(i); Among them, the specific value of the transaction type symbol corresponds to a preset transaction type, and the transaction types include power spot trading and futures trading; Extract macrotransaction data from the original power market data. At preset time intervals, obtain the average market transaction price P_Ave(t), the highest price P_Max(t), and the lowest price P_Min(t) of the power transaction; obtain the real-time power trading volume V(t), market demand Dem(t), and market supply Sup(t) at each moment t; where t is the time stamp, that is, the collection moment of each data.

3. The risk warning method for power market entities according to claim 2, characterized in that, The specific process of preprocessing the collected original power market data is as follows: The specific process of microdata preprocessing is as follows: Taking power spot trading as the object, perform standardization processing on the microtransaction data, calculate the mean values μ11, μ12, and μ13 of the trading volume, price, and time of all power spot transactions in the power market, and calculate the standard deviations σ11, σ12, and σ13 of the trading volume, price, and time of all power spot transactions in the power market; For the transaction number i with the transaction type being the electricity spot transaction, generate the standardized micro-transaction data vector: Taking power futures trading as the object, standardize the micro trading data, calculate the means μ21, μ22, and μ23 of the trading volume, price, and time of all power futures transactions in the power market, and calculate the standard deviations σ21, σ22, and σ23 of the trading volume, price, and time of all power futures transactions in the power market; Generate a standardized micro-transaction data vector for the transaction number i with the transaction type of electricity futures trading: The specific process of macro data preprocessing is as follows: Smoothing the macro transaction data that varies along the time stamp t through Gaussian filtering, and the filtering function is: where x is the input variable, that is, the collected macro transaction data, including the market transaction average price, highest price, lowest price, real-time electricity transaction volume, market demand, and market supply collected at each moment; where σ is the preset standard deviation, which determines the smoothing degree of the Gaussian filtering result; where h(x) is the output value after noise reduction processing by the Gaussian filtering function; Collect the market transaction average price, highest price, lowest price, real-time power trading volume, market demand, and market supply of power transactions at each preset time interval t after data denoising, and generate the macro trading data vector H(t) at time t. H(t) = [P_Ave(t), P_Max(t), P_Min(t), V(t), Dem(t), Sup(t)] T .

4. The risk warning method for power market entities according to claim 1, characterized in that The specific mean reversion model is as follows: For each macro trading data after data noise reduction, a mean regression model of macro trading data is established, and the model function is: H(t) = H(t - 1) + θ[μ - H(t - 1)] + ε(t); where H(t) is the macro trading data vector, and H(t) = [P_Ave(t), P_Max(t), P_Min(t), V(t), Dem(t), Sup(t)] T Where θ is the mean reversion vector, which controls the regression speed of macro trading data and reflects the overall adjustment ability of the market; θ = [θ1, θ2, θ3, θ4, θ5, θ6]; where θ1, θ2, θ3, θ4, θ5, θ6 are the regression coefficients of the market transaction average price, highest price, lowest price, real-time power trading volume, market demand, and market supply respectively, and are used to quantitatively evaluate the speed at which the corresponding macro trading data returns to its average level in the long term; the larger the regression coefficient, the faster the corresponding macro trading data returns to the long-term mean. Where μ is the long-term average value vector, μ = [μ1, μ2, μ3, μ4, μ5, μ6], and the components μ1, μ2, μ3, μ4, μ5, μ6 correspond to the means of the historical data of the market transaction average price, highest price, lowest price, real-time power trading volume, market demand, and market supply respectively, representing the regression target of macro trading data. Where ε(t) is the error vector, ε(t) = θ = [θ1(t), θ2(t), θ3(t), θ4(t), θ5(t), θ6(t)]; The components θ1(t), θ2(t), θ3(t), θ4(t), θ5(t), θ6(t) are all random perturbation variables that conform to the Gaussian distribution, corresponding to the random perturbations of the market transaction average price, highest price, lowest price, real-time power trading volume, market demand, and market supply at time t respectively, representing the abnormal fluctuations in the market that cannot be explained by the known data.

5. The risk warning method for power market entities according to claim 4, wherein The specific process of evaluating the macro risk level contained in the behavior of the electricity price returning to the long-term average value is as follows: Every preset statistical period T, input the macro trading data vector H(t) at each time t collected in the previous statistical period T into the mean reversion model of macro trading data, and solve the mean reversion vector θ, the error vector ε(t) at each time t, and the long-term average value vector μ in this statistical period T; Every preset statistical period T, based on the mean reversion vector, long-term average value vector, and error vectors at each time obtained through the mean reversion model, perform feature extraction and analysis, evaluate the macro risk level, and identify the factors that may affect the risks of power market participants. The specific process is as follows: Extract and analyze the features of the long-term average vector and the mean reversion vector, and calculate the deviation volatility eigenvalue Deviation of the market within the previous statistical period T through a preset formula ​ Extract and analyze the features of the error vector, and calculate the irrational volatility eigenvalue Volatility of the market in the previous statistical period T through a preset formula ​ Calculate the risk factor Risk through the comprehensive risk factor formula Risk = w1 × Deviation + w2 × Volatility; perform risk classification according to the specific values of the risk factor.

6. The risk early warning method for power market entities according to claim 1, wherein The specific process of risk classification according to the specific values of risk factors is as follows: If the risk factor Risk is less than the first preset threshold RiskMin, it is determined that the electricity market is in the macro low-risk level, the market tends to be stable, the regression process is relatively fast and the random fluctuations are less; If the risk factor Risk is less than the second preset threshold RiskMax and greater than or equal to the first preset threshold RiskMax, it is determined that the electricity market is in the macro medium-risk level, there are certain fluctuations in the electricity market, and there is a delay in the regression process; If the risk factor Risk is greater than or equal to the second preset threshold RiskMax, it is determined that the electricity market is in the macro high-risk level, the electricity market fluctuates violently, the price deviates far from the long-term average value, and the regression speed is slow.

7. A risk warning method for power market entities according to claim 1, characterized in that The specific process of using the support vector machine model to evaluate the risk of the micro transaction data in the electricity market is as follows: Take the components in the standardized micro transaction data vector as input parameters, extract features and evaluate the risk of the micro transaction data through the support vector machine model, evaluate the risk level of the micro transaction, and match it with the above-mentioned risk factor to obtain the transaction risk score representing the risk degree of the micro transaction; The decision function of the support vector machine model is: f[M(i)] = sign(w T M(i) + b); where w is the weight vector of the support vector machine and b is the bias term; where f[M(i)] is the output value, that is, the transaction risk score output by the support vector machine model; the numerical range of f[M(i)] is from 0 to 100, and the higher the value of the output value f[M(i)], the higher the transaction risk of this transaction i; The lower the value of the output f[M(i)], the lower the transaction risk of this transaction i; Train the support vector machine model to obtain a trained support vector machine model; Perform arithmetic processing on all micro transaction data vectors using the trained support vector machine model to obtain the output value f[M(i)] corresponding to each micro transaction data vector M(i), that is, the transaction risk score of transaction i.

8. A risk warning method for power market entities according to claim 1, characterized in that, The specific process of training the support vector machine model is as follows: Collect the micro-transaction data vectors corresponding to the electricity spot transactions and electricity futures transactions that have been manually determined to be illegal transactions, and mark them. Let the corresponding marked output value y(i) be 100, and use it as the training data of the support vector machine to train the model. Record the output value f[M(i)] obtained by the operation of the micro-transaction data vectors corresponding to the electricity spot transactions and electricity futures transactions that have been manually determined to be illegal transactions through the support vector machine model. Let the loss function of the training process be: Calculate the weight vector w and the bias term b when the minimum loss function is obtained by the steepest descent method, and bring them back to the support vector machine model for risk prediction of newly input micro-transaction data vectors.

9. A risk early warning method for power market entities according to claim 1, characterized in that, The specific process of generating the corresponding warning signal is as follows: If the macro risk level is low, obtain all micro transaction data vectors M(i) whose transaction risk scores are greater than the preset threshold M1, and mark them as warning transactions; If the macro risk level is medium, obtain all micro transaction data vectors M(i) whose transaction risk scores are greater than the preset threshold M2, and mark them as warning transactions; If the macro risk level is high, obtain all micro transaction data vectors M(i) whose transaction risk scores are greater than the preset threshold M3, and mark them as warning transactions; Where M1, M2, and M3 are all preset transaction risk score thresholds, and M3 > M2 > M1; For all micro transaction data vectors M(i) marked as warning transactions, obtain their corresponding transaction parties, that is, the buyer name buyer(i) and the seller name seller(i), and mark them as warning transaction parties; Every preset statistical period T, count the number of times all buyer names buyer(i) and seller names seller(i) are marked as warning transaction parties; For the buyer or seller whose number of times marked as a warning transaction party exceeds the preset upper limit, generate the corresponding warning signal.

Images