Method and device for modeling multi-dimensional risk preference stochastic utility of power users

Abstract

This invention discloses a method and apparatus for modeling the stochastic utility of multidimensional risk preferences of power users. The method includes: generating a set of multidimensional standardized risk indicators; subsequently extracting user-side feature parameters and generating a corresponding set of multidimensional risk adjustment capability indicators through composite calculation; endogenously coupling the above two sets with the heterogeneous risk aversion coefficient to be estimated into the model, calculating utility gains and losses, and constructing a multidimensional risk utility function by combining basic utility and stochastic disturbance terms; deriving the multivariate discrete choice probability and establishing a log-likelihood function based on this; iteratively optimizing using an extreme value estimation algorithm to output the optimal estimate of the risk aversion coefficient to be estimated. The technical solution of this application quantifies the multidimensional risks of the power market, endogenously couples them with user adjustment capabilities to construct a risk utility function, and improves the prediction accuracy of users' actual decision-making behavior through optimization algorithms.

Description

Technical Field

[0001] This invention relates to the technical field of power system data processing, and in particular to a method and apparatus for modeling the stochastic utility of multidimensional risk preferences of power users. Background Technology

[0002] With the ongoing deepening of power market reforms and the advancement of dual-carbon goals, the variety of electricity trading products is constantly increasing, and users are facing an increasingly complex market environment with risks exhibiting multi-dimensional, strongly coupled, and dynamic characteristics. Against this backdrop, modeling and analysis techniques targeting electricity users' risk preferences have received widespread attention. Existing risk preference modeling schemes can, to some extent, provide quantitative guidance for users' trading behavior, helping them mitigate basic market volatility risks.

[0003] However, existing technologies still have several shortcomings when facing the complex electricity market trading environment. First, existing modeling often focuses on single price risk, failing to cover emerging risk dimensions such as supply reliability, contract flexibility, and green premiums, resulting in significant blind spots in risk characterization. Second, existing models do not intrinsically couple user risk adjustment capabilities with risk preferences, ignoring the impact of heterogeneity among different users in load regulation, reserve configuration, and trading rights on risk decision-making behavior, leading to significant deviations between model outputs and actual user behavior. Third, existing utility functions often employ a deterministic expected utility framework, failing to effectively capture the random disturbances and heterogeneity in user risk decision-making, resulting in insufficient explanatory power and predictive accuracy for user decision-making behavior. Fourth, existing model solving algorithms are prone to slow convergence, getting trapped in local optima, and large parameter estimation biases when dealing with multi-layered heterogeneous parameters, making them unsuitable for solving multi-dimensional and heterogeneous user risk preference parameters. Therefore, how to properly address these issues has become an urgent problem for the industry. Summary of the Invention

[0004] This invention provides a method and apparatus for modeling the stochastic utility of multidimensional risk preferences of power users, which endogenously couples multidimensional risk with user adjustment capabilities and improves the prediction accuracy of user decision-making behavior through precise algorithm solution.

[0005] According to a first aspect of the present invention, a method for modeling the stochastic utility of multidimensional risk preferences of electricity users is provided, the method comprising: Acquire electricity market transaction data, power grid operation data, and user-side characteristic data. Calculate the price fluctuation conditional variance, power shortage probability, expected power shortage amount, contract rigidity constraint, unit contract adjustment cost, green premium rate conditional variance, and absorption responsibility gap cost using the electricity market transaction data and the power grid operation data, respectively. Then, perform extreme value standardization processing on the above calculation results to generate a multi-dimensional standardized risk indicator set. Based on the risk dimensions included in the multi-dimensional standardized risk indicator set, load response parameters, emergency reserve parameters, market transaction permission parameters, and new energy consumption parameters are extracted from the user-side feature data. Through the composite ratio and weight allocation of the corresponding dimensions, a multi-dimensional risk adjustment capability indicator set corresponding to each risk dimension is generated. The generated set of multidimensional standardized risk indicators, the set of multidimensional risk adjustment capability indicators, and the introduced heterogeneous risk aversion coefficient are substituted into the absolute risk aversion model for endogenous coupling calculation to obtain the utility loss and utility gain of each dimension. These are then combined with the basic electricity utility to generate a deterministic utility term, and then superimposed with a random disturbance utility term that follows an extreme value distribution to construct a multidimensional risk utility function. Based on the constructed multidimensional risk utility function, the multivariate discrete choice probability of users when facing different decision options is derived. The corresponding log-likelihood function is established with the goal of maximizing the multivariate discrete choice probability. The log-likelihood function is iteratively optimized using an extreme value estimation algorithm containing adaptive step size adjustment and Monte Carlo sampling. The optimal estimate of the heterogeneous risk aversion coefficient to be estimated is calculated and output.

[0006] In one embodiment, the extreme value standardization process of the above calculation results to generate a multi-dimensional standardized risk indicator set includes: Based on the electricity market transaction data, the period price return series is calculated. The price volatility conditional variance of the period price return series is fitted according to the generalized autoregressive conditional heteroscedasticity first-order model. The price volatility conditional variance is divided by the maximum conditional variance value in the sample period to output the standardized price volatility risk index of the first dimension. Based on the power grid operation data, the probability of power shortage between the available power supply capacity of the power grid to the user's node and the user's rigid power demand is calculated. The expected amount of power shortage is obtained by integrating the probability density function. The probability of power shortage and the expected amount of power shortage (divided by the maximum expected value) are weighted and summed according to the regional power grid structure characteristics to output a standardized supply reliability risk index for the second dimension. Based on the electricity market transaction data, the rigidity of the electricity contracts held by users and the unit contract adjustment cost caused by electricity hedging in the spot market are calculated. The rigidity and the unit contract adjustment cost after dividing by the maximum cost value are weighted and compounded according to the market transaction rules to output a standardized contract flexibility risk index in the third dimension. The conditional variance of the green premium rate is calculated based on the green electricity trading price in the electricity market trading data. The cost of the consumption responsibility gap caused by the difference between the renewable energy consumption responsibility volume and the actual green electricity volume is calculated. The two are divided by their corresponding maximum values ​​and then weighted according to policy weights to output a standardized green premium risk indicator of the fourth dimension. The risk indicators output from the above four dimensions together constitute the multi-dimensional standardized risk indicator set.

[0007] In one embodiment, generating a set of multi-dimensional risk adjustment capability indicators corresponding to each risk dimension includes: Extract the maximum adjustable load power and the rated available capacity of the energy storage device from the load response parameters, calculate their ratio to the user's total electricity demand during the cycle, and perform a weighted calculation based on the load characteristic weights to obtain the risk adjustment capability index corresponding to the price fluctuation dimension. Extract the rated backup capacity of self-provided power supply and emergency energy storage from the emergency backup parameters, as well as the maximum interruptible power in rigid loads, and calculate their proportion with the total rigid power demand. Based on the user's power supply guarantee level weight, perform weighted summation to obtain the risk adjustment capability index corresponding to the supply reliability dimension. Extract the number of electricity markets for which the user has trading rights and the scale of working capital reserves used for electricity trading from the market trading permission parameters. Calculate the ratio of the number of markets to the total number of tradable markets in the region and the ratio of the capital reserves to the expected total electricity cost. Apply the weights of the trading system to perform weighted merging to obtain the risk adjustment capability index corresponding to the contract flexibility dimension. Extract the total self-generated and self-consumed electricity of distributed renewable energy and the maximum electricity saving that can be achieved through energy-saving renovation from the new energy consumption parameters. Calculate the ratio of the two in the total electricity demand and weight them according to the configuration weight to obtain the risk adjustment capability index corresponding to the green premium dimension. Finally, aggregate the indicators calculated from each dimension to generate the multi-dimensional risk adjustment capability index set.

[0008] In one embodiment, the construction of the multi-dimensional risk utility function includes: For each risk dimension, the product of the heterogeneous risk aversion coefficient to be estimated, the standardized risk index of the corresponding dimension, and the result of subtracting the risk adjustment capacity index of the corresponding dimension from 1 is multiplied and the opposite number is taken as the exponent of the natural constant e to obtain the exponent term. Then, the exponent term is subtracted from the base value 1 to obtain the single-dimensional risk utility loss of the dimension. Extract the risk dimension attention weights that are intrinsically determined by the user's own attributes, and then sum the calculated single-dimensional risk utility losses of all single dimensions according to the risk dimension attention weights to obtain the total risk utility loss of multiple dimensions. The basic electricity utility is obtained by multiplying the marginal utility of a unit of electricity consumption by the scale of electricity demand. The basic electricity utility is obtained by multiplying the attention weight of each risk dimension with the corresponding risk adjustment capability index, summing the results, and then multiplying by the marginal utility gain brought by the unit comprehensive adjustment capability. The deterministic utility term is obtained by subtracting the multidimensional total risk utility loss from the basic electricity utility and adding the adjustment capability utility gain. The deterministic utility term is then added to the random disturbance utility term, which follows an independent and identically distributed Gumbel extreme value distribution, to generate the multidimensional risk utility function.

[0009] In one embodiment, outputting the optimal estimate of the heterogeneity risk aversion coefficient to be estimated includes: After obtaining the multidimensional risk utility function, the deterministic utility term of each decision option is extracted, a multivariate Logit probability model is constructed to which the probability of a user choosing a specific decision option follows, and the selection probability output by the multivariate Logit probability model is multiplied and accumulated with the logarithm of the actual selection indicator variable to generate the log-likelihood function of the overall sample. The Bayesian hierarchical model framework is applied to deconstruct all the parameters to be estimated contained in the log-likelihood function, and to divide them into a population mean parameter layer that reflects the average level of all users and follows a normal prior distribution, and an individual heterogeneity variance parameter layer that reflects the differences between individuals and follows an inverse gamma prior distribution. When maximizing the log-likelihood function, Newton iteration is used to calculate the gradient and Hessian matrix, and the step size coefficient of Newton iteration is adaptively adjusted according to the change in the likelihood function of the previous and next iterations. During the iterative optimization calculation, a Gibbs sampling mechanism is embedded to alternately update the population mean parameter layer and the individual heterogeneity variance parameter layer. If the potential scaling factor of the Markov chain formed by sampling is less than the set discrimination threshold, it indicates that the chain has converged, and the mean of the posterior distribution result is extracted to derive the optimal estimate.

[0010] In one embodiment, it also includes: The individual risk aversion coefficient in the derived optimal estimate is compared with the risk-neutral benchmark coefficients of four dimensions based on the average conditions of the regional electricity market to classify users’ single-dimensional risk preference types under the dimensions of price, supply, contract and green. Based on the risk dimension attention weight of each dimension, the individual risk aversion coefficients of the four dimensions are weighted and accumulated to output a comprehensive risk aversion coefficient. The comprehensive risk aversion coefficient is then compared with the comprehensive benchmark coefficient generated by weighting all single-dimensional benchmark coefficients to determine the user's final comprehensive risk preference type. A stochastic differential evolution equation with mean regression velocity coefficient and long-term equilibrium value parameters is constructed, and a standard Wiener process increment to characterize the impact of market random disturbances is superimposed on the stochastic differential evolution equation. The optimal estimated value is substituted into the stochastic differential evolution equation as the initial evolution state for time-series recursive calculation, and the time-varying dynamic risk aversion coefficient trajectory curve representing the evolution of user's future characteristics is output.

[0011] According to a second aspect of the present invention, a device for modeling the stochastic utility of multidimensional risk preferences of electricity users is provided, comprising: The acquisition module is used to acquire electricity market transaction data, power grid operation data, and user-side characteristic data. It calculates the price fluctuation conditional variance, power shortage probability, expected power shortage amount, contract rigidity constraint degree, unit contract adjustment cost, green premium rate conditional variance, and absorption responsibility gap cost through the electricity market transaction data and the power grid operation data, respectively. It then performs extreme value standardization processing on the above calculation results to generate a multi-dimensional standardized risk indicator set. The generation module extracts load response parameters, emergency reserve parameters, market transaction permission parameters, and new energy consumption parameters from the user-side feature data based on the risk dimensions included in the multi-dimensional standardized risk indicator set. It then calculates the multi-dimensional risk adjustment capability indicator set corresponding to each risk dimension through the composite ratio and weight allocation of the corresponding dimensions. The construction module takes the generated set of multi-dimensional standardized risk indicators, the set of multi-dimensional risk adjustment capability indicators, and the introduced heterogeneous risk aversion coefficient to be estimated, and substitutes them into the absolute risk aversion model to perform endogenous coupling calculations to obtain the utility loss and utility gain of each dimension. Then, it combines them with the basic electricity utility to generate a deterministic utility term, and then superimposes a random disturbance utility term that follows an extreme value distribution to construct a multi-dimensional risk utility function. The output module, based on the constructed multidimensional risk utility function, derives and generates the multivariate discrete choice probability of the user when facing different decision options. It establishes a corresponding log-likelihood function with the goal of maximizing the multivariate discrete choice probability, and iteratively optimizes the log-likelihood function using an extreme value estimation algorithm containing adaptive step size adjustment and Monte Carlo sampling. Finally, it calculates and outputs the optimal estimate of the heterogeneous risk aversion coefficient to be estimated.

[0012] According to a third aspect of the present invention, an electronic device is provided, comprising: a communication interface, a processor, and a memory; The memory is used to store program instructions, which, when executed by the processor connected to the memory via the communication interface, implement any of the above-described methods for modeling the stochastic utility of multidimensional risk preferences of power users.

[0013] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored, which, when executed by a computer (e.g., a processor in the computer), implement any of the above-described methods for modeling the stochastic utility of multidimensional risk preferences of electricity users.

[0014] In summary, this invention provides a method and apparatus for modeling the stochastic utility of multidimensional risk preferences of electricity users. The method includes: acquiring electricity market transaction data, power grid operation data, and user-side characteristic data; calculating the conditional variance of price fluctuations, probability of power shortage, expected power shortage, contract rigidity, unit contract adjustment cost, conditional variance of green premium rate, and cost of absorption responsibility gap using the electricity market transaction data and the power grid operation data, respectively; and performing extreme value standardization on the above calculation results to generate a multidimensional standardized risk indicator set; based on the risk dimensions included in the multidimensional standardized risk indicator set, extracting load response parameters, emergency reserve parameters, market transaction authority parameters, and renewable energy absorption parameters from the user-side characteristic data; and generating multidimensional risk adjustment parameters corresponding to each risk dimension by calculating the composite ratio and weight allocation of the corresponding dimensions. The system employs a set of capability indicators. The generated set of multi-dimensional standardized risk indicators, the set of multi-dimensional risk adjustment capability indicators, and the introduced heterogeneous risk aversion coefficient are substituted into an absolute risk aversion model for endogenous coupling calculations to obtain the utility loss and utility gain for each dimension. These are then combined with the basic electricity consumption utility to generate a deterministic utility term, which is then superimposed with a random disturbance utility term following an extreme value distribution to construct a multi-dimensional risk utility function. Based on the constructed multi-dimensional risk utility function, the system derives the multivariate discrete choice probabilities of users facing different decision-making schemes. A corresponding log-likelihood function is established with the goal of maximizing these multivariate discrete choice probabilities. The log-likelihood function is iteratively optimized using an extreme value estimation algorithm incorporating adaptive step size adjustment and Monte Carlo sampling to calculate and output the optimal estimate of the heterogeneous risk aversion coefficient. This technical solution endogenously couples multi-dimensional risks in the electricity market with user adjustment capabilities, constructing a risk utility function containing a random disturbance term, thus realistically depicting the user decision-making mechanism under complex environments. Meanwhile, an improved optimization algorithm combining Monte Carlo sampling is adopted, which effectively overcomes the problem of easily getting trapped in local optima when solving complex parameters, and improves the convergence stability and behavior prediction accuracy of the model.

[0015] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and drawings.

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 A flowchart of a method for modeling the stochastic utility of multidimensional risk preferences of electricity users, provided as an embodiment of the present invention; Figure 2 A flowchart of another method for modeling the stochastic utility of multidimensional risk preferences of electricity users, provided as an embodiment of the present invention; Figure 3 A flowchart illustrating another method for modeling the stochastic utility of multidimensional risk preferences of electricity users, provided as an embodiment of the present invention; Figure 4 A flowchart illustrating another method for modeling the stochastic utility of multidimensional risk preferences of electricity users, provided as an embodiment of the present invention; Figure 5 A flowchart illustrating another method for modeling the stochastic utility of multidimensional risk preferences of electricity users, provided as an embodiment of the present invention; Figure 6 A flowchart illustrating another method for modeling the stochastic utility of multidimensional risk preferences of electricity users, provided as an embodiment of the present invention; Figure 7 A structural diagram of a stochastic utility modeling device for multidimensional risk preferences of electricity users provided as an embodiment of the present invention; Figure 8 This is a structural diagram of an electronic device provided as an embodiment of the present invention. Detailed Implementation

[0019] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are intended only to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0020] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0021] like Figure 1 As shown, this invention provides a method for modeling the stochastic utility of multidimensional risk preferences of electricity users. This method includes: In step S11, electricity market transaction data, power grid operation data, and user-side characteristic data are acquired. Price fluctuation conditional variance, power shortage probability, expected power shortage amount, contract rigidity constraint, unit contract adjustment cost, green premium rate conditional variance, and absorption responsibility gap cost are calculated using the electricity market transaction data and the power grid operation data, respectively. The above calculation results are then subjected to extreme value standardization processing to generate a multi-dimensional standardized risk indicator set. In step S12, based on the risk dimensions included in the multi-dimensional standardized risk indicator set, load response parameters, emergency reserve parameters, market transaction permission parameters, and new energy consumption parameters are extracted from the user-side feature data. The multi-dimensional risk adjustment capability indicator set corresponding to each risk dimension is generated by calculating the composite ratio and weight allocation of the corresponding dimensions. In step S13, the generated set of multi-dimensional standardized risk indicators, the set of multi-dimensional risk adjustment capability indicators, and the introduced heterogeneous risk aversion coefficient to be estimated are substituted into the absolute risk aversion model for endogenous coupling operation to obtain the utility loss and utility gain of each dimension. These are then combined with the basic electricity utility to generate a deterministic utility term, and then superimposed with a random disturbance utility term that follows an extreme value distribution to construct a multi-dimensional risk utility function. In step S14, based on the constructed multidimensional risk utility function, the multivariate discrete choice probability of the user when facing different decision options is derived and generated. The corresponding log-likelihood function is established with the goal of maximizing the multivariate discrete choice probability. The log-likelihood function is iteratively optimized and solved using an extreme value estimation algorithm containing adaptive step size adjustment and Monte Carlo sampling. The optimal estimate of the heterogeneous risk aversion coefficient to be estimated is calculated and output.

[0022] In one embodiment, the system performs steps to quantify and construct an indicator system for a multi-dimensional risk environment. Specifically, the system acquires electricity market transaction data, grid operation data, and user-side characteristic data. Based on this data, a series of underlying parameters characterizing market risk are calculated using the electricity market transaction data and the grid operation data, including price volatility conditional variance, power shortage probability, expected power shortage amount, contract rigidity, unit contract adjustment cost, green premium rate conditional variance, and absorption responsibility gap cost. To eliminate the differences in the dimensions and physical meaning of the above-mentioned risk parameters and avoid data bias in subsequent calculations, extreme value standardization processing is performed on the calculation results, thereby generating a multi-dimensional standardized risk indicator set. This multi-dimensional standardized risk indicator set systematically covers four core risk dimensions: price, supply, contract, and green.

[0023] After quantifying the risk environment, the system proceeds to the stage of extracting and generating indicators for risk adjustment capabilities. Based on the risk dimensions included in the multi-dimensional standardized risk indicator set, the system extracts various heterogeneous parameters from the user-side feature data, specifically covering load response parameters, emergency reserve parameters, market transaction permission parameters, and renewable energy consumption parameters. For each of these extracted parameters, the system calculates a multi-dimensional risk adjustment capability indicator set corresponding to each risk dimension through composite ratios and weight allocations for each dimension. The system calculates the ratio of load response parameters to total demand to determine the adjustment capability in the price volatility dimension; calculates the reserve capacity ratio based on emergency reserve parameters to determine the adjustment capability in the supply reliability dimension; calculates the share ratio based on market transaction permission parameters to determine the adjustment capability in the contract flexibility dimension; and calculates the electricity ratio based on renewable energy consumption parameters to determine the adjustment capability in the green premium dimension. The resulting multi-dimensional risk adjustment capability indicator set objectively and accurately characterizes the differentiated defense and adjustment endowments of users when facing multi-dimensional risks.

[0024] The system executes the core endogenous coupling and utility function construction steps. It takes the multi-dimensional standardized risk index set, the multi-dimensional risk adjustment capability index set, and the introduced heterogeneous risk aversion coefficient generated in the previous steps, and substitutes them into the absolute risk aversion model for deep endogenous coupling calculation. During the endogenous coupling calculation, for each risk dimension, the system multiplies the estimated heterogeneous risk aversion coefficient, the corresponding dimension's standardized risk index, and the result of subtracting the corresponding dimension's risk adjustment capability index from 1, takes the negative of the product, and uses it as the exponent of the natural constant e for power operation to obtain an exponent term. Then, it subtracts the exponent term from the baseline value 1 to derive the single-dimensional risk utility loss for each dimension. The system uses the risk dimension attention weights to perform a weighted summation to calculate the total multi-dimensional risk utility loss, and simultaneously calculates the adjustment capability utility gain. Further, by subtracting the total multi-dimensional risk utility loss from the basic electricity consumption utility and adding the adjustment capability utility gain, a deterministic utility term is generated. Based on this, a random perturbation utility term that follows an independent and identically distributed Gumbel extreme value distribution is superimposed to construct a multidimensional risk utility function that fully represents the user's multidimensional preferences.

[0025] Based on the constructed multidimensional risk utility function, the system enters the crucial model solving and parameter output stage. The system extracts the deterministic utility term from the multidimensional risk utility function and derives the multivariate discrete choice probabilities for users facing different decision options. To achieve accurate identification of model parameters, the system establishes a corresponding log-likelihood function with the goal of maximizing the multivariate discrete choice probabilities. In solving this function, an extreme value estimation algorithm incorporating adaptive step size adjustment and Monte Carlo sampling is used for iterative optimization. Specifically, the system applies a Bayesian hierarchical model framework to deconstruct the parameters into a population mean parameter layer and an individual heterogeneity variance parameter layer. Newton iteration is used to calculate the gradient and Hessian matrix, and the step size coefficient is adaptively adjusted based on the change in the likelihood function. During the iterative optimization calculation, a Gibbs sampling mechanism is embedded to alternately update the two-layer parameters. If the potential scaling factor of the Markov chain formed by sampling is less than a set discrimination threshold, it indicates that the chain has converged. At this point, the system extracts the mean of the posterior distribution results, calculates and outputs the optimal estimate of the heterogeneity risk aversion coefficient to be estimated.

[0026] After successfully outputting the optimal estimate of the heterogeneous risk aversion coefficient, the system further extends to the definition of risk preference types and the prediction of its temporal evolution. The system compares the individual risk aversion coefficients in the derived optimal estimate with the regional neutral benchmark coefficient to classify the user's single-dimensional risk preference type in each dimension. Simultaneously, it outputs a comprehensive risk aversion coefficient through a weighted summation operation and compares it with the comprehensive benchmark coefficient to determine the user's final comprehensive risk preference type. To characterize the dynamic features of preferences, the system constructs a stochastic differential evolution equation with mean regression velocity coefficients and long-term equilibrium value parameters, and superimposes standard Wiener process increments onto this equation. The optimal estimate is used as the initial evolutionary state and substituted into the stochastic differential evolution equation for temporal recursive calculation, accurately outputting a time-varying dynamic risk aversion coefficient trajectory curve representing the user's future characteristic evolution.

[0027] The technical solution is as follows: 1. Construction of a multi-dimensional power market risk quantification system In the context of the electricity market, users face risks that are multi-dimensional and coupled, and a single-dimensional risk profile cannot fully cover the actual decision-making scenarios of users. This invention first constructs a four-dimensional risk quantification system covering price volatility risk, supply reliability risk, contract flexibility risk, and green premium risk. Each risk dimension is characterized by quantifiable and real-time updated indicators, achieving standardized measurement of risk levels and providing basic input for the subsequent construction of utility functions.

[0028] 1.1 Quantification of Price Volatility Risk Price volatility risk refers to the uncertainty in electricity costs for users caused by random fluctuations in electricity prices and ancillary service prices in the electricity market. This invention uses the conditional variance of price volatility to characterize the level of price volatility risk. It achieves dynamic fitting of price volatility through a generalized autoregressive conditional heteroscedasticity first-order model, capturing the leptokurtic peaks, heavy tails, and volatility clustering characteristics of the price series, and obtaining a standardized price volatility risk index.

[0029] The formula for defining a price return series is as follows:

[0030] In the formula, Indicates the first Electricity market price yield over a given period; Indicates the first The electricity market clearing price for a given period; Indicates the first The electricity market clearing price for a given period.

[0031] The formula for fitting the conditional variance of price fluctuations is as follows:

[0032] In the formula, For the first The conditional variance of price fluctuations over a period of time serves as a core quantitative indicator of price volatility risk. This is a constant term in the price fluctuation model; For ARCH term coefficients, reflecting the degree of impact of previous period price shocks on current period volatility; The coefficients of the GARCH term reflect the degree of sustained influence of the previous period's volatility on the current period's volatility; For the first Conditional variance of price fluctuations over a period of time.

[0033] The formula for the standardized price volatility risk indicator is as follows:

[0034] In the formula, For the first The price volatility risk indicator after time period standardization has a value range of 0 to 1; This represents the maximum value of the conditional variance of price fluctuations within the sample period.

[0035] 1.2 Quantification of Supply Reliability Risk Supply reliability risk refers to the risk that users' electricity demand cannot be met due to factors such as power system outages, line congestion, and limited generator output. This invention uses a composite index of power shortage probability and expected power shortage amount to characterize the level of supply reliability risk, and combines real-time power grid operation data with historical fault data to achieve accurate quantification of supply reliability risk.

[0036] The formula for calculating the probability of power outages during a given period is as follows:

[0037] In the formula, For the first The probability of power shortage on the user side during a given period serves as a fundamental indicator of supply reliability risk. For the first The available power supply capacity of the power grid to the user's node during a given time period; For the first The rigid electricity demand of users during certain time periods.

[0038] The formula for calculating the expected amount of power shortage is as follows:

[0039] In the formula, For the first Users' expected power consumption during the specified time period; For the first Available power supply capacity during the period The probability density function; This is the integral variable representing the available power supply capacity.

[0040] The formula for the standardized supply reliability risk index is as follows:

[0041] In the formula, For the first The supply reliability risk index after time period standardization has a value range of 0 to 1; The weighting coefficients for the probability of power shortage are determined based on the characteristics of the power grid structure in the user's area; This represents the expected maximum amount of power shortage within the sample period.

[0042] 1.3 Quantification of Contract Flexibility Risk Contractual flexibility risk refers to the risk that rigid constraints on terms such as electricity allocation, price adjustments, and contract transfers in long-term electricity contracts and contracts for difference (CFDs) prevent users from adjusting their contract positions according to market changes and electricity demand. This invention uses a composite index of contract adjustment costs and contract rigidity to characterize the level of contractual flexibility risk, enabling horizontal comparability of risks under different contract structures.

[0043] The formula for calculating the rigidity of a contract is as follows:

[0044] In the formula, The rigidity of the user's current electricity contracts; The total number of time periods during which users are allowed to adjust the contracted electricity volume within the contract period; This represents the total number of periods in the contract period.

[0045] The formula for calculating unit contract adjustment costs is as follows:

[0046] In the formula, Adjusting the cost per unit of contracted electricity for users; The total amount of penalty fees payable for adjusting the contracted battery capacity for users; The total transaction costs for users to hedge contract electricity in the spot market; Adjust the total contracted electricity volume for the user.

[0047] The formula for the standardized contract flexibility risk indicator is as follows:

[0048] In the formula, For the first The contract flexibility risk indicator after time period standardization has a value range of 0 to 1; The weighting coefficient for the rigidity of the contract is determined based on the electricity market contract trading rules; This represents the maximum unit contract adjustment cost within the sample period.

[0049] 1.4 Quantification of Green Premium Risk Green premium risk refers to the risk of excess costs incurred by users when purchasing green electricity and green certificates to fulfill their renewable energy consumption obligations and achieve carbon neutrality goals, due to fluctuations in the price difference between green electricity and conventional thermal power, and adjustments to the weighting of consumption obligations. This invention uses a composite index of the volatility of the green premium rate and the cost of the consumption obligation gap to characterize the level of green premium risk, adapting to the development trend of the electricity market under dual carbon objectives.

[0050] The formula for calculating the green premium rate is as follows:

[0051] In the formula, For the first Green premium rate for the period; For the first Green electricity trading prices for specific time periods; For the first The regular on-grid price for thermal power or the spot clearing price in the electricity market during a given period.

[0052] The formula for calculating the volatility of the green premium rate is as follows:

[0053] In the formula, For the first The conditional variance of the green premium rate over a given period reflects the volatility of the green premium. This is the constant term in the green premium volatility model; These are the ARCH coefficients of the green premium model; These are the GARCH coefficients of the green premium model; For the first Conditional variance of the green premium rate over a given period.

[0054] The formula for calculating the cost of absorbing the responsibility gap is as follows:

[0055] In the formula, For the first The cost of the renewable energy consumption liability gap for users in a given time period; For the first The user's renewable energy consumption responsibility electricity volume approved by the regulatory authority for the specified period; For the first The sum of the green electricity already held by the user during the time period and the electricity converted from the green certificate; For the first The market trading price of green certificates for specific time periods.

[0056] The formula for the standardized green premium risk indicator is as follows:

[0057] In the formula, For the first The green premium risk indicator after time period standardization has a value range of 0 to 1; The weighting coefficient for the volatility of the green premium rate is determined based on regional renewable energy consumption policies; This represents the maximum value of the conditional variance of the green premium rate within the sample period; This represents the maximum cost of absorbing the responsibility gap within the sample period.

[0058] 2. Quantitative Model of User's Multi-Dimensional Risk Adjustment Capability Different types of electricity users exhibit significant heterogeneity in terms of electricity consumption characteristics, load adjustability potential, energy storage configuration, contract trading rights, and green energy consumption channels, leading to fundamental differences in their risk adjustment capabilities across different risk dimensions. A user's risk adjustment capability directly determines their willingness to bear and avoid risk. This invention constructs quantitative models of user risk adjustment capability for each of the four risk dimensions, achieving a standardized measurement of users' heterogeneous adjustment capabilities and providing endogenous variables on the user side for constructing the risk utility function.

[0059] 2.1 Quantification of Price Volatility Risk Adjustment Capability Users' ability to adjust to price fluctuation risks primarily stems from the demand response potential of their interruptible and transferable loads, as well as the charging and discharging regulation capabilities of their energy storage devices. By adjusting electricity consumption periods and power consumption, they can hedge against the risk of increased costs due to price fluctuations. This invention uses a composite index of adjustable load ratio and energy storage regulation capacity to characterize users' ability to adjust to price fluctuation risks.

[0060] The formula for calculating price volatility risk tolerance is as follows:

[0061] In the formula, For the first This is an indicator of a user's ability to adjust to price volatility risk, with a value ranging from 0 to 1. The weighting coefficient for adjustable load in price risk adjustment is determined based on the user's load characteristics; For the first The maximum adjustable load power for each user type, including the adjustment limits for interruptible load and transferable load; For the first Total electricity demand of user class within a period; For the first The rated available capacity of the energy storage device configured by the user.

[0062] 2.2 Quantification of Supply Reliability Risk Adjustment Capability A user's ability to adjust to supply reliability risks primarily stems from the reserve capacity of their backup power sources and emergency energy storage devices, as well as the interruptibility of rigid loads. When grid power is interrupted, core power needs can be met through backup power sources, reducing losses from power outages. This invention uses a composite index of the backup power reserve ratio and the interruptibility ratio of rigid loads to characterize a user's ability to adjust to supply reliability risks.

[0063] The formula for calculating supply reliability risk adjustment capability is as follows:

[0064] In the formula, For the first The indicator of a user's ability to adjust to supply reliability risks ranges from 0 to 1. The weighting coefficient for the backup capacity of self-contained power supply in the adjustment of supply reliability risk is determined based on the user's power supply guarantee level. For the first Rated backup capacity of self-contained power supply and emergency energy storage configured for user-type applications; For the first The maximum interruptible power in rigid loads of user-type users; For the first Total rigid electricity demand of user groups within a given period.

[0065] 2.3 Quantification of Contract Flexibility and Risk Adjustment Capabilities A user's ability to adjust for contract flexibility risk primarily stems from their trading authority in the electricity market, contract trading channels, and the scale of their capital reserves. This can be achieved through multi-market contract combinations, contract transfers, and hedging transactions, reducing the adjustment costs associated with rigid contract constraints. This invention employs a composite indicator of trading market coverage and capital reserve levels to characterize a user's ability to adjust for contract flexibility risk.

[0066] The formula for calculating contract flexibility risk adjustment capability is as follows:

[0067] In the formula, For the first This is an indicator of a user's ability to adjust for contract flexibility risks, with a value ranging from 0 to 1. The weighting coefficient for market coverage in contract flexibility risk adjustment is determined based on the electricity market trading system; For the first The number of electricity markets in which users have trading rights, including medium- and long-term markets, spot markets, ancillary services markets, and green electricity markets; This refers to the total number of tradable markets within the regional electricity market system. For the first The size of working capital reserves available for electricity trading for this type of user; For the first Total expected electricity cost for this user group over the period.

[0068] 2.4 Quantification of Green Premium Risk Adjustment Capability Users' ability to mitigate the risk of green premiums primarily stems from their distributed renewable energy installed capacity, their ability to lock in long-term green electricity contracts, and the potential for reduced electricity demand resulting from energy-saving retrofits. They can hedge against the risks of green premium fluctuations and increased grid connection costs by self-consuming green electricity, locking in premiums through long-term contracts, and reducing total electricity consumption. This invention uses a composite index of the proportion of distributed renewable energy self-consumption and the potential for energy-saving retrofits to characterize users' ability to mitigate the risk of green premiums.

[0069] The formula for calculating the risk mitigation capability of the green premium is as follows:

[0070] In the formula, For the first The indicator of a user's ability to adjust for green premium risk ranges from 0 to 1. The weighting coefficient for the self-generated and self-consumed electricity of distributed renewable energy in the adjustment of green premium risk is determined based on the user's renewable energy configuration. For the first Total self-generated and self-consumed electricity within the installed capacity of distributed renewable energy for similar users during the installation period; For the first The maximum energy savings that users can achieve through energy-saving renovations.

[0071] 3. Construction of a multidimensional risk utility function based on stochastic utility theory The core idea of ​​stochastic utility theory is that a decision-maker's utility consists of an observable deterministic utility component and an unobservable stochastic utility component, effectively characterizing the heterogeneity and uncertainty in the decision-making process. Existing technologies for constructing utility functions for electricity users often employ a deterministic expected utility framework, which fails to effectively capture the stochastic disturbances and heterogeneity characteristics in user risk decision-making and does not incorporate multi-dimensional risks and user adaptability into the endogenous variables of the utility function. This invention, based on stochastic utility theory, uses four dimensions of risk level and the user's corresponding risk adaptability as core inputs, introduces a user risk aversion coefficient, and constructs a multi-dimensional risk utility function containing deterministic and stochastic utility terms, achieving a precise characterization of user risk decision-making behavior.

[0072] The basic framework formula for the stochastic utility function is as follows:

[0073] In the formula, For the first Class of users in the first Total risk utility over the period; For the first Class of users in the first The deterministic risk utility over a given period can be quantified using observable risk levels, adjustment capabilities, and user attribute variables. For the first Class of users in the first The random utility term for a given time period is used to characterize unobservable random disturbances in user risk decisions and follows an independent and identically distributed Gumbel extreme value distribution.

[0074] The deterministic risk utility term is a core component of a user's total utility, consisting of three parts: the user's basic electricity consumption utility, the utility loss caused by multi-dimensional risks, and the utility gain brought by risk adjustment capabilities. The utility loss caused by risks is characterized using a constant absolute risk aversion framework, introducing the user's risk aversion coefficient to reflect the degree of the user's aversion to risk. The utility gain brought by risk adjustment capabilities is positively correlated with the user's adjustment capabilities in the corresponding dimension, reflecting the user's ability to hedge risks and improve utility levels through their own adjustment behaviors.

[0075] The formula for the utility loss function of one-dimensional risk is as follows:

[0076] In the formula, For the first Class of users in the first Time period, the The utility loss resulting from each risk dimension The values ​​range from 1 to 4, corresponding to price volatility risk, supply reliability risk, contract flexibility risk, and green premium risk, respectively. For the first Class of users targeting the first The risk aversion coefficient for each risk dimension is greater than 0. The larger the coefficient, the higher the user's aversion to the risk of that dimension. For the first Class of users in the first Time period, the Standardized risk indicators for each risk dimension; For the first Class of users targeting the first Indicators of adjustment capacity across risk dimensions.

[0077] The formula for the multidimensional total risk utility loss function is as follows:

[0078] In the formula, For the first Class of users in the first Total risk utility loss over a period of time across multiple dimensions; For the first Class of users on the first The attention weight of each risk dimension satisfies The weight value is intrinsically determined by factors such as the user's electricity consumption attributes, industry characteristics, and policy constraints.

[0079] Basic electricity utility is the fundamental utility a user obtains from meeting their production and living needs through electricity consumption. It is positively correlated with the scale of the user's electricity demand and serves as the benchmark for user utility. The formula for basic electricity utility is as follows:

[0080] In the formula, For the first Class of users in the first Basic electricity consumption efficiency during a given time period; For the first The marginal utility of a user's electricity demand is determined based on the user's electricity value attribute. The marginal utility of an industrial user is determined by the output value of a unit of electricity consumption, while the marginal utility of a residential user is determined by the value of the convenience of electricity use in daily life. For the first Class of users in the first Electricity demand during a given time period.

[0081] The utility gain from risk adjustment capability reflects how users, by actively utilizing their risk adjustment capabilities, can not only reduce utility losses caused by risks but also gain additional utility by optimizing electricity consumption behavior and trading strategies. This gain is positively correlated with the user's overall adjustment capability. The formula for the utility gain is as follows:

[0082] In the formula, For the first Class of users in the first The utility gain resulting from the risk-modifying ability over a period of time; For the first The marginal utility gain brought about by the comprehensive adjustment capabilities of user units is determined based on the user's risk management capabilities and market participation level.

[0083] The complete formula for the deterministic risk utility term obtained by integration is as follows:

[0084] In the formula, For the first Class of users in the first The deterministic risk utility term for a given period is composed of the basic electricity utility minus the total risk utility loss, plus the utility gain brought by the regulation capacity, which fully characterizes the impact of observable factors on the user's risk utility.

[0085] 4. Construction of a User Heterogeneous Risk Preference Model A user's risk preference is essentially determined by the values ​​and distribution characteristics of their risk aversion coefficients across different risk dimensions. Significant heterogeneity exists in the risk aversion coefficients of different user types, and the risk aversion coefficient of the same user also changes dynamically across different time periods and market environments. This invention categorizes users' risk preferences based on the value range of their risk aversion coefficients and constructs a dynamic evolution model for these coefficients. This enables the static classification and dynamic updating of user risk preferences, comprehensively characterizing the heterogeneous risk preference features of users.

[0086] For each risk dimension, based on the value of the user's risk aversion coefficient, the user's risk preference is divided into three categories: risk-averse, risk-neutral, and risk-seeking. Then, by combining the comprehensive risk aversion coefficient of the four dimensions, the user's comprehensive risk preference type is obtained, thus achieving a refined classification of the user's risk preference.

[0087] The criteria for classifying single-dimensional risk preference types are as follows: When At that time, the first Class of users in the first The risk dimension is risk-averse; the higher the risk aversion coefficient, the greater the degree of risk aversion. At that time, the first Class of users in the first Each risk dimension is risk-neutral; when At that time, the first Class of users in the first The risk dimension is risk-preference oriented.

[0088] In the formula, For the first The risk-neutral benchmark coefficient for each risk dimension is calibrated based on the average risk level of the regional electricity market and the decision-making behavior characteristics of overall users.

[0089] The formula for calculating the overall risk aversion coefficient is as follows:

[0090] In the formula, For the first Class of users in the first The overall risk aversion coefficient for a given period reflects the user's overall risk preference level.

[0091] The criteria for classifying comprehensive risk preferences are as follows: When At that time, the user is a comprehensive risk-averse type; when At that time, the user was classified as having a neutral overall risk profile; when At that time, the user had a comprehensive risk appetite.

[0092] In the formula, The comprehensive risk-neutral benchmark coefficient is obtained by weighted averaging of the risk-neutral benchmark coefficients of the four risk dimensions.

[0093] Users' risk aversion coefficient is not fixed but dynamically evolves with factors such as market risk levels, self-regulation capabilities, changes in electricity demand, and policy adjustments. This invention constructs a dynamic evolution model of the risk aversion coefficient with mean reversion characteristics to capture the time-varying characteristics of users' risk preferences and achieve dynamic updates to risk preferences. The formula for the dynamic evolution model is as follows:

[0094] In the formula, For the first Class of users targeting the first Time-varying risk aversion coefficient for each risk dimension; The mean regression velocity coefficient reflects how quickly the risk aversion coefficient regresses to the long-term equilibrium level. For the first Class of users targeting the first The long-term equilibrium value of the risk aversion coefficient for each risk dimension; Volatility is the risk aversion coefficient; This is an increment of the standard Wiener process, used to characterize the impact of random changes in the market environment on the risk aversion coefficient.

[0095] 5. Adaptive Maximum Likelihood Estimation Algorithm Based on Bayesian Hierarchical Framework The multi-dimensional risk preference model constructed in this invention includes multiple parameters to be estimated, such as user heterogeneity risk aversion coefficient, risk dimension attention weight, and marginal utility coefficient. Furthermore, the random utility term follows a Gumbel extreme value distribution. Conventional maximum likelihood estimation algorithms are prone to slow convergence, getting trapped in local optima, and large parameter estimation biases when dealing with multi-level heterogeneous parameters. This invention designs an adaptive maximum likelihood estimation algorithm that integrates a Bayesian hierarchical framework. The Bayesian hierarchical framework characterizes the heterogeneous distribution of user parameters, and the Newton iteration method with adaptive step size adjustment optimizes the likelihood function solution process. Combined with Markov chain Monte Carlo simulation, the globally optimal parameter estimation is achieved, improving both the accuracy and efficiency of parameter estimation.

[0096] The algorithm consists of three core steps: likelihood function construction, Bayesian hierarchical prior distribution setting, and parameter solving using adaptive Newton iteration combined with Markov chain Monte Carlo.

[0097] 5.1 Construction of Likelihood Function Based on stochastic utility theory, the probability of a user choosing among different decision options is determined by the magnitude of their total utility value. In the context of electricity market risk decision-making, user decision-making behavior manifests as the selection of different risk management options, different trading contracts, and different electricity consumption strategies, with each option corresponding to a total utility value. When the stochastic utility term follows an independent and identically distributed Gumbel extreme value distribution, the probability of a user choosing a particular option follows the form of a multivariate Logit model. Based on this, the likelihood function of the model is constructed as the objective function for parameter estimation.

[0098] The formula for the probability of a user choosing a decision option is as follows:

[0099] In the formula, For the first Class of users in the first Time period selection The probability of each decision option; For the first Class of users in the first Time period selection The deterministic utility term corresponding to each decision option; The total number of decision options available to the user; This is the index variable for the decision-making scheme.

[0100] The formula for the log-likelihood function of the sample is as follows:

[0101] In the formula, Let be the log-likelihood function of the model, and the objective of parameter estimation is to maximize this log-likelihood function; This is the set of parameters to be estimated for the model, including all parameters to be estimated such as risk aversion coefficient, risk dimension weights, and marginal utility coefficients. This represents the total number of users in the sample. This represents the total number of time periods in the sample. As an indicator variable, when the first Class of users in the first The time period was actually selected as the first When there are multiple decision options, the value is 1; otherwise, the value is 0.

[0102] 5.2 Bayesian Hierarchical Prior Distribution Specification To address the heterogeneity of user parameters, this invention employs a Bayesian hierarchical framework, dividing the parameters to be estimated into two levels: overall distribution parameters and individual heterogeneous parameters. A reasonable prior distribution is set for the parameters at each level, and the range of parameter values ​​is constrained by prior information, thereby reducing the bias in parameter estimation and simultaneously characterizing the heterogeneous distribution of parameters among different users.

[0103] For any individual parameter in the set of parameters to be estimated Its hierarchical structure formula is as follows:

[0104]

[0105]

[0106] In the formula, For the first Individual heterogeneity parameters of user class; This represents the overall mean of individual parameters, reflecting the average level of all user parameters; The population variance of individual parameters reflects the degree of heterogeneity of parameters among users; , The hyperparameters of the prior distribution of the population mean are calibrated based on prior empirical data from the electricity market. , The hyperparameters of the inverse gamma prior distribution of the population variance are calibrated based on prior empirical data from the electricity market.

[0107] 5.3 Parameter Solving Using Adaptive Newton Iteration Combined with Markov Chain Monte Carlo Method To address the problem that the conventional Newton iteration method has a fixed step size and is prone to getting trapped in local optima, this invention designs an adaptive step-size Newton iteration method for gradient optimization of the log-likelihood function. At the same time, it combines the Gibbs sampling algorithm in Markov chain Monte Carlo simulation to achieve global sampling and solution of Bayesian hierarchical parameters, balancing the accuracy and efficiency of the solution.

[0108] The iterative formula for adaptive Newton iteration is as follows:

[0109] In the formula, For the first The parameter estimates for the next iteration; For the first The parameter estimates for the next iteration; For the first The adaptive step size coefficient for each iteration ranges from 0 to 1 and is dynamically adjusted according to the change in the likelihood function. For the first The Hessian matrix of the log-likelihood function at the nth iteration; For the first The gradient vector of the log-likelihood function at the nth iteration.

[0110] The update rule for the adaptive step size coefficient is as follows: when hour, ; when hour, .

[0111] In the formula, For the first The adaptive step size coefficient for each iteration increases the step size to improve the solution speed when the likelihood function value increases. When the likelihood function value decreases, the step size decreases to avoid iteration divergence and ensure stable convergence of the iteration process.

[0112] Based on a Bayesian hierarchical framework, the Gibbs sampling algorithm is used to alternately sample population distribution parameters and individual heterogeneity parameters. When the Markov chain converges, the posterior distribution of the parameter to be estimated is obtained, and the mean of the posterior distribution is taken as the optimal estimate of the parameter. The formula for the conditional posterior distribution is as follows:

[0113]

[0114]

[0115] In the formula, The conditional posterior probability distribution for the parameter; Sample observation data of user decision-making behavior; This is the sample likelihood distribution based on individual parameters.

[0116] The convergence criterion of the algorithm is as follows: when the potential scaling factor of the Markov chain is less than 1.1, the sampling process is considered to have converged, the iteration is stopped, and the optimal estimate of the parameters is output.

[0117] In summary, the method provided in this embodiment achieves deep interconnection between risk environment data and user adjustment characteristics. Through rigorous data standardization, precise endogenous coupling operations, and a high-order iterative optimization algorithm incorporating Monte Carlo sampling, the scientific validity of the construction of the multi-dimensional risk utility function and the accuracy of the optimal estimate of the heterogeneous risk aversion coefficient are ensured.

[0118] In one embodiment, such as Figure 2 As shown, the extreme value standardization process for the above calculation results to generate a multi-dimensional standardized risk index set includes the following steps S21-S24: In step S21, the period price return series is calculated based on the electricity market transaction data. The price volatility conditional variance of the period price return series is fitted according to the generalized autoregressive conditional heteroscedasticity first-order model. The price volatility conditional variance is divided by the maximum conditional variance value in the sample period to output the standardized price volatility risk index of the first dimension. In step S22, the probability of power shortage between the available power supply capacity of the power grid to the user's node and the user's rigid power demand is calculated based on the power grid operation data. The expected amount of power shortage is obtained by integrating the probability density function. The probability of power shortage and the expected amount of power shortage after dividing by the maximum expected value are weighted and summed according to the regional power grid structure characteristics to output the standardized supply reliability risk index of the second dimension. In step S23, the rigidity of the electricity contract held by the user and the unit contract adjustment cost caused by electricity hedging in the spot market are calculated based on the electricity market transaction data. The rigidity and the unit contract adjustment cost after dividing by the maximum cost value are weighted and compounded according to the market transaction rules to output a standardized contract flexibility risk index of the third dimension. In step S24, the conditional variance of the green premium rate is calculated based on the green electricity trading price in the electricity market trading data, and the cost of the consumption responsibility gap caused by the difference between the renewable energy consumption responsibility volume and the actual green electricity volume is calculated. After dividing both by their corresponding maximum values, the standardized green premium risk index of the fourth dimension is output by weighted composite according to policy weights. The risk indicators output from the above four dimensions together constitute the multi-dimensional standardized risk index set.

[0119] In one embodiment, based on the acquired electricity market transaction data, a detailed solution is performed to address the volatility characteristics of the electricity price environment. The system retrieves electricity spot prices from historical trading periods and calculates a time-period price return series. The system applies a generalized autoregressive conditional heteroscedasticity (GHP) first-order model to fit the time-period price return series, thereby calculating the conditional variance of price volatility, which accurately reflects price uncertainty. To eliminate dimensional differences, the system normalizes the conditional variance of price volatility by dividing it by the maximum conditional variance value within the sample period, thus outputting a standardized price volatility risk indicator for the first dimension. This indicator can accurately characterize the economic impact of drastic fluctuations in electricity spot market prices on electricity consumers.

[0120] This system provides a concrete implementation path for risk quantification on both the supply and transaction fulfillment sides. On the supply side, based on collected grid operation data, the system calculates the grid's available power supply capacity to the user's node and compares it with the user's rigid electricity demand. Through probabilistic modeling, the system calculates the probability of power shortage under insufficient grid supply and calculates the expected power shortage amount by integrating the probability density function. Further, the system weights and sums the power shortage probability with the expected power shortage amount (divided by the maximum expected value) according to the regional grid structure characteristics reflecting network topology, thereby outputting a standardized supply reliability risk index for the second dimension. On the transaction fulfillment side, based on electricity market transaction data, the system evaluates various forward contracts held by the user, calculates the rigidity of the user's electricity contracts, and simultaneously calculates the unit contract adjustment cost due to deviation assessments, etc. Subsequently, the system weights and combines the rigidity with the unit contract adjustment cost (divided by the maximum cost value) according to preset market transaction rules, thereby outputting a standardized contract flexibility risk index for the third dimension.

[0121] The system quantifies and aggregates low-carbon premium risks under the new power system framework. It delves into green electricity trading prices within the electricity market transaction data to calculate the conditional variance of the green premium rate. Simultaneously, it calculates the cost of the renewable energy consumption gap arising from the difference between the renewable energy consumption responsibility volume and the actual green electricity held. The system standardizes both the conditional variance of the green premium rate and the consumption gap cost by dividing them by their respective maximum values, and then performs a weighted composite calculation based on the policy weights set by current low-carbon power policies, thereby outputting a fourth-dimensional standardized green premium risk indicator. The first-dimensional standardized price volatility risk indicator, the second-dimensional standardized supply reliability risk indicator, the third-dimensional standardized contract flexibility risk indicator, and the fourth-dimensional standardized green premium risk indicator, together constitute the multi-dimensional standardized risk indicator set. This multi-dimensional standardized risk indicator set comprehensively and without blind spots covers heterogeneous external risk sources at the price, supply, contract, and low-carbon levels.

[0122] In one embodiment, such as Figure 3 As shown, the generation of a set of multi-dimensional risk adjustment capability indicators corresponding to each risk dimension includes the following steps S31-S34: In step S31, the maximum adjustable load power and the rated available capacity of the energy storage device are extracted from the load response parameters, and their ratios to the total electricity demand during the user's cycle are calculated respectively. The risk adjustment capability index corresponding to the price fluctuation dimension is obtained by weighting the load characteristic weights. In step S32, the rated backup capacity of self-provided power supply and emergency energy storage and the maximum interruptible power in rigid load are extracted from the emergency backup parameters. Their proportions with the total rigid power demand are calculated respectively. Based on the user's power supply guarantee level weight, a weighted sum is performed to obtain the risk adjustment capability index corresponding to the supply reliability dimension. In step S33, the number of electricity markets for which the user has trading rights and the scale of working capital reserves for electricity trading are extracted from the market trading permission parameters. The share ratio of the number of markets to the total number of tradable markets in the region and the share ratio of the capital scale to the expected total cost of electricity are calculated respectively. The weighted combination of the trading system weights is applied to obtain the risk adjustment capability index corresponding to the contract flexibility dimension. In step S34, the total self-generated and self-consumed electricity of distributed renewable energy and the maximum electricity saving that can be achieved through energy-saving renovation are extracted from the new energy consumption parameters. The ratio of the two in the total electricity demand is calculated and weighted according to the configuration weight to obtain the risk adjustment capability index corresponding to the green premium dimension. Finally, the indicators calculated from each dimension are aggregated to generate the multi-dimensional risk adjustment capability index set.

[0123] In one embodiment, the system extracts corresponding feature parameters from the user-side feature data based on the risk dimensions included in the multi-dimensional standardized risk indicator set, to implement deep endogenous capability quantification. Specifically, the system focuses on the price volatility dimension, extracting load response parameters from the user-side feature data, including maximum adjustable load capacity and rated available capacity of energy storage devices. The system calculates the ratio of the maximum adjustable load capacity to the user's total electricity demand during the cycle, and the ratio of the rated available capacity of energy storage devices to the user's total electricity demand during the cycle. After obtaining these two ratios, the system performs a weighted calculation based on load characteristic weights, thereby deriving a risk adjustment capability index corresponding to the price volatility dimension. This index can accurately characterize the adjustment space for users to utilize load elasticity and energy storage buffers to mitigate spot price risks.

[0124] This provides a concrete implementation path for quantifying risk mitigation capabilities in terms of supply reliability and contract flexibility. For supply reliability, the system extracts emergency backup parameters from the user-side feature data, specifically covering the rated backup capacity of self-provided power and emergency energy storage, as well as the maximum interruptible power volume in rigid loads. The system calculates the proportion of these parameters to the total rigid electricity demand, and performs a weighted summation based on the user's power supply guarantee level weight to derive a risk mitigation capability index corresponding to the supply reliability dimension. For contract flexibility, the system extracts the number of electricity markets with which the user has trading rights and the scale of working capital reserves used for electricity trading from the market trading permission parameters. The system calculates the share ratio of the number of markets to the total number of tradable markets in the region and the share ratio of the capital scale to the expected total electricity cost, and applies weighted merging using the trading system weights to derive a risk mitigation capability index corresponding to the contract flexibility dimension.

[0125] The system quantifies and aggregates the capabilities of the green premium dimension involving low-carbon quotas. It extracts the total self-consumption of distributed renewable energy from the renewable energy consumption parameters and the maximum energy savings achievable through energy-saving retrofits, calculating the ratio of these two values ​​to the total electricity demand. The system then performs a weighted summation based on configuration weights to derive a risk adjustment capability index corresponding to the green premium dimension. The system organically aggregates the risk adjustment capability indices calculated in the above steps for price volatility, supply reliability, contract flexibility, and green premium dimensions, formally generating the multi-dimensional risk adjustment capability index set. This multi-dimensional risk adjustment capability index set comprehensively and objectively outlines the differentiated defensive endowments of electricity users facing multi-dimensional external market risks.

[0126] In one embodiment, such as Figure 4As shown, the construction of the multi-dimensional risk utility function includes the following steps S41-S44: In step S41, for each risk dimension, the product of the heterogeneous risk aversion coefficient to be estimated, the standardized risk index of the corresponding dimension, and the result of subtracting the risk adjustment capability index of the corresponding dimension from 1 is multiplied and the opposite number is taken as the exponent of the natural constant e to obtain the exponent term. Then, the exponent term is subtracted from the base value 1 to obtain the single-dimensional risk utility loss of the dimension. In step S42, the risk dimension attention weights determined by the user's own attributes are extracted, and the single-dimensional risk utility loss of all single dimensions is weighted and summed according to the risk dimension attention weights to obtain the total risk utility loss of multiple dimensions. In step S43, the basic electricity utility is obtained by multiplying the marginal utility of unit electricity consumption by the scale of electricity demand. The risk dimension attention weight of each dimension is multiplied and summed with the corresponding risk adjustment capability index, and then multiplied by the marginal utility gain brought by the unit comprehensive adjustment capability to obtain the adjustment capability utility gain. In step S44, the deterministic utility term is obtained by subtracting the multidimensional total risk utility loss from the basic electricity utility and adding the adjustment capability utility gain. Then, the deterministic utility term is added to the random disturbance utility term that follows an independent and identically distributed Gumbel extreme value distribution to generate the multidimensional risk utility function.

[0127] In one embodiment, the system performs refined utility loss quantification for each of the four core risk dimensions: price, supply, contracts, and green risks. In the specific computational pipeline, the pre-processing steps calculate the estimated heterogeneous risk aversion coefficient, the standardized risk index corresponding to the current processing dimension, and the result of subtracting the corresponding dimension's risk adjustment capability index from a baseline value of 1. These three are then rigorously multiplied, and the result is inversely multiplied. The system uses this inverse as the exponent of the natural constant e to perform a power operation, thus obtaining the exponent term. Next, the system further subtracts the exponent term from the baseline value of 1 to rigorously derive and obtain the single-dimensional risk utility loss for that dimension. After traversing all risk dimensions and calculating a set of single-dimensional risk utility losses, the system extracts the risk dimension attention weights intrinsically determined by the user's own attributes from the database. Using these risk dimension attention weights, the system performs a rigorous weighted summation operation on all the calculated single-dimensional risk utility losses to accurately derive the multi-dimensional total risk utility loss that comprehensively represents the combined impact of external risks.

[0128] While quantifying the total risk utility loss across multiple dimensions, this embodiment simultaneously advances the calculation of positive utility gains from user-side proactive adjustment behavior and the quantification of basic production utility. The system acquires the user's production and operational status, multiplies the preset marginal utility per unit of electricity consumption by the user's electricity demand scale, thereby obtaining the basic electricity utility. This utility term represents the benchmark value that the user can obtain by consuming electricity under a completely ideal, risk-free, and unadjusted state. To accurately characterize the economic compensation or utility improvement brought about by the user leveraging their own adjustment endowment, the system performs a multiplication operation and summation operation on each dimension, multiplying the risk dimension attention weight with the corresponding risk adjustment capability index. After obtaining this comprehensive adjustment characterization value, it is multiplied by the marginal utility gain brought by the unit comprehensive adjustment capability, and finally, the adjustment capability utility gain is rigorously calculated. Through this rigorous cross-multiplication and summation logic, the system successfully digitally maps the heterogeneous load elasticity and multidimensional external preferences of the user side, quantifying the positive contribution of differentiated risk-resistant behavior to total utility.

[0129] The system performs comprehensive utility superposition and perturbation injection to construct the function. It retrieves the basic electricity utility calculated in the preceding steps, subtracts the quantified multi-dimensional total risk utility loss, and simultaneously adds the adjustment capability utility gain. Through algebraic summation of these three utility indicators, the deterministic utility term is formally generated. This deterministic utility term theoretically and logically reflects the net utility output after the interplay of risk suppression and adjustment promotion in the observable dimension. To further capture the random psychological perturbations and irrational decision-making characteristics of electricity users in actual discrete choice decisions caused by unobservable factors, the system introduces a random perturbation utility term following an independent and identically distributed Gumbel extreme value distribution. The system sums the deterministic utility term and the random perturbation utility term to formally generate the multi-dimensional risk utility function. This function not only achieves the endogenous coupling of risk and capability but also, through the superposition of random terms, provides the model with the derivation basis for multivariate Logit discrete choice probabilities.

[0130] In one embodiment, such as Figure 5 As shown, the process of outputting the optimal estimate of the heterogeneity risk aversion coefficient includes the following steps S51-S53: In step S51, after obtaining the multidimensional risk utility function, the deterministic utility term of each decision option is extracted, a multivariate Logit probability model is constructed to which the probability of the user choosing a specific decision option follows, and the selection probability output by the multivariate Logit probability model is multiplied and accumulated with the logarithm of the actual selection indicator variable to generate the log-likelihood function of the overall sample. In step S52, the Bayesian hierarchical model framework is applied to deconstruct all the parameters to be estimated contained in the log-likelihood function, and to divide the population mean parameter layer, which reflects the average level of all users and follows a normal prior distribution, and the individual heterogeneity variance parameter layer, which reflects the differences between individuals and follows an inverse gamma prior distribution. In step S53, when maximizing the log-likelihood function, Newton iteration is used to calculate the gradient and Hessian matrix, and the step size coefficient of Newton iteration is adaptively adjusted according to the change in the likelihood function of the previous and next iterations. During the iterative optimization calculation, a Gibbs sampling mechanism is embedded to alternately update the population mean parameter layer and the individual heterogeneity variance parameter layer. If the potential scaling factor of the sampling to form a Markov chain is detected to be less than the set discrimination threshold, it indicates that the chain has converged, and the mean of the posterior distribution result is extracted to derive the optimal estimate.

[0131] In one embodiment, during execution, the system extracts the deterministic utility term of each decision option after obtaining the multidimensional risk utility function. Based on classical discrete choice theory, the system rigorously constructs a multivariate Logit probability model that follows the probability of a user choosing a specific decision option using the extracted deterministic utility term. This model can scientifically transform abstract multidimensional utility values ​​into selection probabilities between 0 and 1. To comprehensively evaluate the model's goodness of fit to the overall sample, the system further performs a logarithmic multiplication operation on the selection probabilities output by the multivariate Logit probability model and the actual selection indicator variable reflecting the user's historical decision trajectory, and rigorously accumulates the product of all sample data, thereby generating the log-likelihood function of the overall sample.

[0132] To address the objective function constructed above, this embodiment introduces advanced statistical inference theory to handle the multidimensional heterogeneity of the parameters. The system applies a Bayesian hierarchical model framework to systematically deconstruct all the parameters to be estimated contained in the log-likelihood function. Guided by this hierarchical framework, the system scientifically divides the complex parameters to be estimated into a population mean parameter layer that reflects the average level of all users and follows a normal prior distribution, and an individual heterogeneity variance parameter layer that reflects the differences between individuals and follows an inverse gamma prior distribution. By performing this rigorous two-layer deconstruction operation, the model can not only accurately capture the commonalities of the overall macroeconomic risk preferences in the electricity market, but also accurately preserve the micro-level individual differences among different electricity users.

[0133] After parameter deconstruction, the system formally enters the iterative computation phase of maximizing the log-likelihood function. To overcome the shortcomings of traditional estimation algorithms, such as getting trapped in local optima and slow convergence, the system applies Newton's iteration to calculate the gradient and Hessian matrix, using second-order derivative information to accurately guide the optimization search direction. Simultaneously, the system dynamically monitors the optimization process and adaptively adjusts the step size coefficient of Newton's iteration based on the change in the likelihood function between iterations, ensuring the stability and efficiency of the search path in the complex high-dimensional parameter space. During the iterative optimization computation, the system embeds a Gibbs sampling mechanism, using conditional probability distributions to alternately update the population mean parameter layer and the individual heterogeneity variance parameter layer. Throughout the dynamic iteration and sampling update process, the system continuously performs convergence checks. If the potential scaling factor of the Markov chain formed by sampling is less than a set discrimination threshold, it indicates chain convergence, and the system immediately stops iterative updates. At this point, the system extracts the mean of the posterior distribution results that satisfy the convergence condition and rigorously derives the optimal estimate of the heterogeneity risk aversion coefficient to be estimated.

[0134] In one embodiment, such as Figure 6 As shown, it also includes the following steps S61-S63: In step S61, the individual risk aversion coefficient in the derived optimal estimate is compared with the risk-neutral benchmark coefficients of the four dimensions based on the average conditions of the regional electricity market to classify the user's single-dimensional risk preference type under the dimensions of price, supply, contract and green. In step S62, the individual risk aversion coefficients on the four dimensions are weighted and accumulated based on the risk dimension attention weights of each dimension to output a comprehensive risk aversion coefficient. The comprehensive risk aversion coefficient is then compared with the comprehensive benchmark coefficient generated by weighting all single-dimensional benchmark coefficients to determine the user's final comprehensive risk preference type. In step S63, a stochastic differential evolution equation with mean regression velocity coefficient and long-term equilibrium value parameters is constructed, and a standard Wiener process increment for characterizing the impact of market random disturbances is superimposed on the stochastic differential evolution equation. The optimal estimated value is substituted into the stochastic differential evolution equation as the initial evolution state for time-series recursive calculation, and the time-varying dynamic risk aversion coefficient trajectory curve representing the evolution of user's future characteristics is output.

[0135] In one embodiment, after obtaining the optimal estimate of the heterogeneous risk aversion coefficient, the system further performs fine-grained classification and static profiling of electricity users' risk preferences. Specifically, the system extracts the individual risk aversion coefficient from the derived optimal estimate. The system retrieves pre-set risk-neutral benchmark coefficients for four dimensions, calibrated based on the average conditions of the regional electricity market. With complete data, the system rigorously compares the extracted individual risk aversion coefficients with the aforementioned risk-neutral benchmark coefficients one by one. By analyzing the direction and degree of deviation of the individual risk aversion coefficients relative to the risk-neutral benchmark coefficients, the system accurately classifies users' single-dimensional risk preference types under the dimensions of price, supply, contract, and green factors. For example, it objectively defines whether a user is risk-averse, risk-neutral, or risk-seeking under a specific single dimension, thus providing a solid underlying basis for multi-dimensional and refined risk characterization.

[0136] Having completed the classification of single-dimensional risk preferences, the system then proceeds to the stage of defining the comprehensive risk preference type. To accurately measure the relative importance of each single-dimensional risk in the user's overall decision-making behavior, the system invokes the risk dimension attention weights used in the preceding steps. Based on these weights, the system scientifically weights and accumulates the individual risk aversion coefficients across the four dimensions, using this as a rigorous mathematical method to output the comprehensive risk aversion coefficient. Simultaneously, the system utilizes the same weight allocation logic to perform weighted generation operations on all single-dimensional benchmark coefficients, thereby obtaining a comprehensive benchmark coefficient for macro-level reference. The system then compares the calculated comprehensive risk aversion coefficient with the comprehensive benchmark coefficient. Based on the numerical relationship of the comparison results, the system ultimately defines the user's final comprehensive risk preference type, thus achieving a systematic elevation and global identification from micro-level single-dimensional characteristics to macro-level comprehensive preferences.

[0137] After establishing the user's current static comprehensive risk preference type, the system further performs dynamic trajectory prediction calculations to fully capture the temporal evolution characteristics of preferences under complex market environment fluctuations. The system constructs a stochastic differential evolution equation with mean regression velocity coefficients and long-term equilibrium value parameters. This equation effectively characterizes the inherent physical and economic mechanisms by which preferences converge towards a long-term rational state. To accurately reflect the uncertainties of the external environment, the system rigorously superimposes standard Wiener process increments, used to characterize the impact of random market disturbances, into the stochastic differential evolution equation. After the entire equation framework is constructed, the system directly uses the aforementioned optimal estimate as the initial evolutionary state and rigorously substitutes it into the stochastic differential evolution equation for time-series recursive calculations. By continuously recursively solving for the dispersion magnitude within a set future time window, the system calculates and outputs a time-varying dynamic risk aversion coefficient trajectory curve representing the user's future characteristic evolution. This time-varying dynamic risk aversion coefficient trajectory curve provides a highly forward-looking digital decision-making basis for dynamic adjustment of power trading strategies, long-term risk management, and stable supply and demand interaction on the system side.

[0138] In one embodiment, Figure 7 This is a block diagram illustrating a stochastic utility modeling apparatus for multidimensional risk preferences of electricity users, according to an exemplary embodiment. Figure 7 As shown, the stochastic utility modeling device for multidimensional risk preferences of electricity users includes an acquisition module 71, a generation module 72, a construction module 73, and an output module 74.

[0139] The acquisition module 71 is used to acquire electricity market transaction data, power grid operation data and user-side characteristic data. It calculates the price fluctuation conditional variance, power shortage probability, expected power shortage amount, contract rigidity constraint degree, unit contract adjustment cost, green premium rate conditional variance and absorption responsibility gap cost through the electricity market transaction data and the power grid operation data, respectively. It then performs extreme value standardization processing on the above calculation results to generate a multi-dimensional standardized risk indicator set. The generation module 72 extracts load response parameters, emergency reserve parameters, market transaction permission parameters, and new energy consumption parameters from the user-side feature data based on the risk dimensions included in the multi-dimensional standardized risk indicator set. It then calculates the multi-dimensional risk adjustment capability indicator set corresponding to each risk dimension through the composite ratio and weight allocation of the corresponding dimensions. The construction module 73 takes the generated set of multi-dimensional standardized risk indicators, the set of multi-dimensional risk adjustment capability indicators, and the introduced heterogeneous risk aversion coefficient to be estimated, and substitutes them into the absolute risk aversion model to perform endogenous coupling calculations to obtain the utility loss and utility gain of each dimension. It then combines these with the basic electricity utility to generate a deterministic utility term, and then superimposes a random disturbance utility term that follows an extreme value distribution to construct a multi-dimensional risk utility function. The output module 74, based on the constructed multidimensional risk utility function, derives and generates the multivariate discrete choice probability of the user when facing different decision options. With the goal of maximizing the multivariate discrete choice probability, it establishes a corresponding log-likelihood function. An extreme value estimation algorithm containing adaptive step size adjustment and Monte Carlo sampling is used to iteratively optimize the log-likelihood function, calculate and output the optimal estimate of the heterogeneous risk aversion coefficient to be estimated.

[0140] The acquisition module 71, generation module 72, construction module 73, and output module 74 included in the block diagram of the multidimensional risk preference stochastic utility modeling device for electricity users are controlled to execute the multidimensional risk preference stochastic utility modeling method for electricity users described in any of the above embodiments.

[0141] like Figure 5 As shown, the present invention provides an electronic device 800, which includes: a communication interface, a processor 801, and a memory 802; The memory 802 stores program instructions. When executed by the processor 801, which is connected to the memory 802 via the communication interface, the program instructions acquire electricity market transaction data, grid operation data, and user-side characteristic data. Using the electricity market transaction data and the grid operation data, the processor calculates the conditional variance of price fluctuations, the probability of power shortage, the expected amount of power shortage, the rigidity of contracts, the unit contract adjustment cost, the conditional variance of the green premium rate, and the cost of the absorption responsibility gap. The calculation results are then subjected to extreme value standardization to generate a multi-dimensional standardized risk indicator set. Based on the risk dimensions included in the multi-dimensional standardized risk indicator set, the processor extracts load response parameters, emergency reserve parameters, market transaction authority parameters, and renewable energy absorption parameters from the user-side characteristic data. These parameters are then calculated using the corresponding composite ratios and weight allocations to generate parameters corresponding to each risk dimension. A corresponding set of multi-dimensional risk adjustment capability indicators is generated. The generated set of multi-dimensional standardized risk indicators, the set of multi-dimensional risk adjustment capability indicators, and the introduced heterogeneous risk aversion coefficient are substituted into the absolute risk aversion model for endogenous coupling calculation to obtain the utility loss and utility gain of each dimension. These are then combined with the basic electricity utility to generate a deterministic utility term, and then superimposed with a random disturbance utility term that follows an extreme value distribution to construct a multi-dimensional risk utility function. Based on the constructed multi-dimensional risk utility function, the multivariate discrete choice probability of the user when facing different decision options is derived. A corresponding log-likelihood function is established with the goal of maximizing the multivariate discrete choice probability. The log-likelihood function is iteratively optimized using an extreme value estimation algorithm containing adaptive step size adjustment and Monte Carlo sampling to calculate and output the optimal estimate of the heterogeneous risk aversion coefficient.

[0142] This invention provides a computer-readable storage medium storing computer program instructions. When executed by a processor, the computer program instructions acquire electricity market transaction data, power grid operation data, and user-side characteristic data. Using the electricity market transaction data and the power grid operation data, the invention calculates price fluctuation conditional variance, power shortage probability, expected power shortage volume, contract rigidity constraint, unit contract adjustment cost, green premium rate conditional variance, and absorption responsibility gap cost, respectively. The calculation results are then subjected to extreme value standardization to generate a multi-dimensional standardized risk indicator set. Based on the risk dimensions included in the multi-dimensional standardized risk indicator set, load response parameters, emergency reserve parameters, market transaction authority parameters, and renewable energy absorption parameters are extracted from the user-side characteristic data. These parameters are then calculated using the corresponding composite ratios and weight allocations to generate parameters corresponding to each risk dimension. A multi-dimensional risk adjustment capability index set is generated. The generated multi-dimensional standardized risk index set, the multi-dimensional risk adjustment capability index set, and the introduced heterogeneous risk aversion coefficient are substituted into an absolute risk aversion model for endogenous coupling calculations to obtain the utility loss and utility gain for each dimension. These are then combined with the basic electricity utility to generate a deterministic utility term, which is then superimposed with a random disturbance utility term following an extreme value distribution to construct a multi-dimensional risk utility function. Based on the constructed multi-dimensional risk utility function, the multivariate discrete choice probability of a user facing different decision options is derived. A corresponding log-likelihood function is established with the goal of maximizing the multivariate discrete choice probability. The log-likelihood function is iteratively optimized using an extreme value estimation algorithm containing adaptive step size adjustment and Monte Carlo sampling to calculate and output the optimal estimate of the heterogeneous risk aversion coefficient.

[0143] It should be understood that the specific features, operations, and details described above regarding the method of the present invention can also be similarly applied to the apparatus and system of the present invention, or vice versa. Furthermore, each step of the method of the present invention described above can be performed by a corresponding component or unit of the apparatus or system of the present invention.

[0144] It should be understood that the various modules / units of the device of the present invention can be implemented wholly or partially through software, hardware, firmware, or a combination thereof. Each module / unit can be embedded in the processor of a computer device in hardware or firmware form or independent of the processor, or it can be stored in the memory of a computer device in software form for the processor to call to execute the operation of each module / unit. Each module / unit can be implemented as an independent component or module, or two or more modules / units can be implemented as a single component or module.

[0145] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores computer instructions executable by the processor, which, when executed by the processor, instruct the processor to perform steps of the methods of embodiments of the present invention. The computer device can be broadly categorized as a server, terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, memory, network interface, communication interface, etc., connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing, and / or control capabilities. The memory of the computer device may include a non-volatile storage medium and internal memory. The non-volatile storage medium may store an operating system, computer programs, etc. The internal memory can provide an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface and communication interface of the computer device can be used to connect and communicate with external devices via a network. When the computer program is executed by the processor, it performs the steps of the methods of the present invention.

[0146] This invention can be implemented as a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, causes the steps of the methods of embodiments of the invention to be performed. In one embodiment, the computer program is distributed across multiple network-coupled computer devices or processors, such that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, may be executed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations may be executed by one or more computer devices or processors, and one or more other method steps / operations may be executed by one or more other computer devices or processors. One or more computer devices or processors may execute a single method step / operation, or execute two or more method steps / operations.

[0147] It will be understood by those skilled in the art that the method steps of the present invention can be performed by a computer program instructing related hardware, such as a computer device or processor. The computer program may be stored in a non-transitory computer-readable storage medium, and its execution causes the steps of the present invention to be performed. Depending on the context, any references herein to memory, storage, databases, or other media may include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state drive, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0148] The technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification, provided that such combination does not contain contradictions.

[0149] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for modeling the stochastic utility of multidimensional risk preferences of electricity users, characterized in that, include: Acquire electricity market transaction data, power grid operation data, and user-side characteristic data. Calculate the price fluctuation conditional variance, power shortage probability, expected power shortage amount, contract rigidity constraint, unit contract adjustment cost, green premium rate conditional variance, and absorption responsibility gap cost using the electricity market transaction data and the power grid operation data, respectively. Then, perform extreme value standardization processing on the above calculation results to generate a multi-dimensional standardized risk indicator set. Based on the risk dimensions included in the multi-dimensional standardized risk indicator set, load response parameters, emergency reserve parameters, market transaction permission parameters, and new energy consumption parameters are extracted from the user-side feature data. Through the composite ratio and weight allocation of the corresponding dimensions, a multi-dimensional risk adjustment capability indicator set corresponding to each risk dimension is generated. The generated set of multidimensional standardized risk indicators, the set of multidimensional risk adjustment capability indicators, and the introduced heterogeneous risk aversion coefficient are substituted into the absolute risk aversion model for endogenous coupling calculation to obtain the utility loss and utility gain of each dimension. These are then combined with the basic electricity utility to generate a deterministic utility term, and then superimposed with a random disturbance utility term that follows an extreme value distribution to construct a multidimensional risk utility function. Based on the constructed multidimensional risk utility function, the multivariate discrete choice probability of users when facing different decision options is derived. The corresponding log-likelihood function is established with the goal of maximizing the multivariate discrete choice probability. The log-likelihood function is iteratively optimized using an extreme value estimation algorithm containing adaptive step size adjustment and Monte Carlo sampling. The optimal estimate of the heterogeneous risk aversion coefficient to be estimated is calculated and output.

2. The method for modeling the stochastic utility of multidimensional risk preferences of electricity users as described in claim 1, characterized in that, The process of standardizing the above calculation results to generate a multi-dimensional standardized risk index set includes: Based on the electricity market transaction data, the period price return series is calculated. The price volatility conditional variance of the period price return series is fitted according to the generalized autoregressive conditional heteroscedasticity first-order model. The price volatility conditional variance is divided by the maximum conditional variance value in the sample period to output the standardized price volatility risk index of the first dimension. Based on the power grid operation data, the probability of power shortage between the available power supply capacity of the power grid to the user's node and the user's rigid power demand is calculated. The expected amount of power shortage is obtained by integrating the probability density function. The probability of power shortage and the expected amount of power shortage (divided by the maximum expected value) are weighted and summed according to the regional power grid structure characteristics to output a standardized supply reliability risk index for the second dimension. Based on the electricity market transaction data, the rigidity of the electricity contracts held by users and the unit contract adjustment cost caused by electricity hedging in the spot market are calculated. The rigidity and the unit contract adjustment cost after dividing by the maximum cost value are weighted and compounded according to the market transaction rules to output a standardized contract flexibility risk index in the third dimension. The conditional variance of the green premium rate is calculated based on the green electricity trading price in the electricity market trading data. The cost of the consumption responsibility gap caused by the difference between the renewable energy consumption responsibility volume and the actual green electricity volume is calculated. The two are divided by their corresponding maximum values ​​and then weighted according to policy weights to output a standardized green premium risk indicator of the fourth dimension. The risk indicators output from the above four dimensions together constitute the multi-dimensional standardized risk indicator set.

3. The method for modeling the stochastic utility of multidimensional risk preferences of electricity users as described in claim 1, characterized in that, The generation of a multi-dimensional risk adjustment capability indicator set corresponding to each risk dimension includes: Extract the maximum adjustable load power and the rated available capacity of the energy storage device from the load response parameters, calculate their ratio to the user's total electricity demand during the cycle, and perform a weighted calculation based on the load characteristic weights to obtain the risk adjustment capability index corresponding to the price fluctuation dimension. Extract the rated backup capacity of self-provided power supply and emergency energy storage from the emergency backup parameters, as well as the maximum interruptible power in rigid loads, and calculate their proportion with the total rigid power demand. Based on the user's power supply guarantee level weight, perform weighted summation to obtain the risk adjustment capability index corresponding to the supply reliability dimension. Extract the number of electricity markets for which the user has trading rights and the scale of working capital reserves used for electricity trading from the market trading permission parameters. Calculate the ratio of the number of markets to the total number of tradable markets in the region and the ratio of the capital reserves to the expected total electricity cost. Apply the weights of the trading system to perform weighted merging to obtain the risk adjustment capability index corresponding to the contract flexibility dimension. Extract the total self-generated and self-consumed electricity of distributed renewable energy and the maximum electricity saving that can be achieved through energy-saving renovation from the new energy consumption parameters. Calculate the ratio of the two in the total electricity demand and weight them according to the configuration weight to obtain the risk adjustment capability index corresponding to the green premium dimension. Finally, aggregate the indicators calculated from each dimension to generate the multi-dimensional risk adjustment capability index set.

4. The method for modeling the stochastic utility of multidimensional risk preferences of electricity users as described in claim 1, characterized in that, The constructed multi-dimensional risk utility function includes: For each risk dimension, the product of the heterogeneous risk aversion coefficient to be estimated, the standardized risk index of the corresponding dimension, and the result of subtracting the risk adjustment capacity index of the corresponding dimension from 1 is multiplied and the opposite number is taken as the exponent of the natural constant e to obtain the exponent term. Then, the exponent term is subtracted from the base value 1 to obtain the single-dimensional risk utility loss of the dimension. Extract the risk dimension attention weights that are intrinsically determined by the user's own attributes, and then sum the calculated single-dimensional risk utility losses of all single dimensions according to the risk dimension attention weights to obtain the total risk utility loss of multiple dimensions. The basic electricity utility is obtained by multiplying the marginal utility of a unit of electricity consumption by the scale of electricity demand. The basic electricity utility is obtained by multiplying the attention weight of each risk dimension with the corresponding risk adjustment capability index, summing the results, and then multiplying by the marginal utility gain brought by the unit comprehensive adjustment capability. The deterministic utility term is obtained by subtracting the multidimensional total risk utility loss from the basic electricity utility and adding the adjustment capability utility gain. The deterministic utility term is then added to the random disturbance utility term, which follows an independent and identically distributed Gumbel extreme value distribution, to generate the multidimensional risk utility function.

5. The method for modeling the stochastic utility of multidimensional risk preferences of electricity users as described in claim 1, characterized in that, The optimal estimate of the heterogeneity risk aversion coefficient to be estimated is output, including: After obtaining the multidimensional risk utility function, the deterministic utility term of each decision option is extracted, a multivariate Logit probability model is constructed to which the probability of a user choosing a specific decision option follows, and the selection probability output by the multivariate Logit probability model is multiplied and accumulated with the logarithm of the actual selection indicator variable to generate the log-likelihood function of the overall sample. The Bayesian hierarchical model framework is applied to deconstruct all the parameters to be estimated contained in the log-likelihood function, and to divide them into a population mean parameter layer that reflects the average level of all users and follows a normal prior distribution, and an individual heterogeneity variance parameter layer that reflects the differences between individuals and follows an inverse gamma prior distribution. When maximizing the log-likelihood function, Newton iteration is used to calculate the gradient and Hessian matrix, and the step size coefficient of Newton iteration is adaptively adjusted according to the change in the likelihood function of the previous and next iterations. During the iterative optimization calculation, a Gibbs sampling mechanism is embedded to alternately update the population mean parameter layer and the individual heterogeneity variance parameter layer. If the potential scaling factor of the Markov chain formed by sampling is less than the set discrimination threshold, it indicates that the chain has converged, and the mean of the posterior distribution result is extracted to derive the optimal estimate.

6. The method for modeling the stochastic utility of multidimensional risk preferences of electricity users as described in claim 5, characterized in that, Also includes: The individual risk aversion coefficient in the derived optimal estimate is compared with the risk-neutral benchmark coefficients of four dimensions based on the average conditions of the regional electricity market to classify users’ single-dimensional risk preference types under the dimensions of price, supply, contract and green. Based on the risk dimension attention weight of each dimension, the individual risk aversion coefficients of the four dimensions are weighted and accumulated to output a comprehensive risk aversion coefficient. The comprehensive risk aversion coefficient is then compared with the comprehensive benchmark coefficient generated by weighting all single-dimensional benchmark coefficients to determine the user's final comprehensive risk preference type. A stochastic differential evolution equation with mean regression velocity coefficient and long-term equilibrium value parameters is constructed, and a standard Wiener process increment to characterize the impact of market random disturbances is superimposed on the stochastic differential evolution equation. The optimal estimated value is substituted into the stochastic differential evolution equation as the initial evolution state for time-series recursive calculation, and the time-varying dynamic risk aversion coefficient trajectory curve representing the evolution of user's future characteristics is output.

7. A device for modeling the stochastic utility of multidimensional risk preferences of electricity users, characterized in that, include: The acquisition module is used to acquire electricity market transaction data, power grid operation data, and user-side characteristic data. It calculates the price fluctuation conditional variance, power shortage probability, expected power shortage amount, contract rigidity constraint degree, unit contract adjustment cost, green premium rate conditional variance, and absorption responsibility gap cost through the electricity market transaction data and the power grid operation data, respectively. It then performs extreme value standardization processing on the above calculation results to generate a multi-dimensional standardized risk indicator set. The generation module extracts load response parameters, emergency reserve parameters, market transaction permission parameters, and new energy consumption parameters from the user-side feature data based on the risk dimensions included in the multi-dimensional standardized risk indicator set. It then calculates the multi-dimensional risk adjustment capability indicator set corresponding to each risk dimension through the composite ratio and weight allocation of the corresponding dimensions. The construction module takes the generated set of multi-dimensional standardized risk indicators, the set of multi-dimensional risk adjustment capability indicators, and the introduced heterogeneous risk aversion coefficient to be estimated, and substitutes them into the absolute risk aversion model to perform endogenous coupling calculations to obtain the utility loss and utility gain of each dimension. Then, it combines them with the basic electricity utility to generate a deterministic utility term, and then superimposes a random disturbance utility term that follows an extreme value distribution to construct a multi-dimensional risk utility function. The output module, based on the constructed multidimensional risk utility function, derives and generates the multivariate discrete choice probability of the user when facing different decision options. It establishes a corresponding log-likelihood function with the goal of maximizing the multivariate discrete choice probability, and iteratively optimizes the log-likelihood function using an extreme value estimation algorithm containing adaptive step size adjustment and Monte Carlo sampling. Finally, it calculates and outputs the optimal estimate of the heterogeneous risk aversion coefficient to be estimated.

8. The stochastic utility modeling device for multidimensional risk preferences of electricity users as described in claim 7, characterized in that: The acquisition module, the generation module, the construction module, and the output module are controlled to execute the stochastic utility modeling method for multidimensional risk preferences of power users as described in any one of claims 2 to 6.

9. An electronic device, characterized in that, include: Communication interface, processor, memory; The memory is used to store program instructions, which, when executed by the processor connected to the memory via the communication interface, cause the electronic device to implement the stochastic utility modeling method for multidimensional risk preferences of power users as described in any one of claims 1 to 6.

10. A computer-readable storage medium having program instructions stored thereon, characterized in that, When the program instructions are executed by a computer, the computer implements the stochastic utility modeling method for multidimensional risk preferences of power users as described in any one of claims 1 to 6.

Images