CN-122022503-A - Electric power metering data security risk assessment method and device

CN122022503ACN 122022503 ACN122022503 ACN 122022503ACN-122022503-A

Abstract

The invention discloses a power metering data security risk assessment method and device, and belongs to the technical field of power system data security. The method comprises the steps of collecting multi-source data of a metering core, scene dynamics and equipment health, constructing a high-dimensional feature matrix through standardized processing, constructing a dynamic risk assessment model fusing scene feature statistical distribution and equipment health attenuation rules, dynamically updating prior probability and likelihood functions, combining a GARCH-LSTM improved model and a Markov chain to realize risk value prediction and evolution deduction, outputting a hierarchical prevention and control strategy based on a risk value confidence interval and conduction probability, and optimizing model parameters through grid search. The device adopts a modularized design and is suitable for multi-scene deployment. The risk assessment method and the risk assessment system can improve the dynamic accuracy and prevention and control foresight of risk assessment, and reduce the loss of safety events and the operation and maintenance cost.

Inventors

LI LIN
CHEN JUAN
SUN YUANXIANG
YANG FENGXIA
XU HONGBO
FAN XIAOYU
LV WEIJIA
ZHAI SHURAN
CHEN XIN

Assignees

国网天津市电力公司营销服务中心
国网天津市电力公司
国家电网有限公司

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. The electric power metering data security risk assessment method is characterized by comprising the following steps of: Step 1, collecting power metering multi-source data of power distribution nodes or transformer areas in a target area, and constructing a high-dimensional feature matrix; Step 2, respectively quantifying distribution node or area scene feature statistical distribution rules and metering equipment health attenuation rules, constructing a dynamic risk assessment model integrating the distribution node or area scene feature statistical distribution rules and the metering equipment health attenuation rules based on quantified results, and dynamically updating prior probability and likelihood functions of the dynamic risk assessment model; Step 3, inputting the high-dimensional feature matrix into a dynamic risk assessment model, and carrying out probability distribution prediction of risk values by combining an improved LSTM time sequence prediction model and a Markov chain state transition algorithm to obtain a confidence interval of the risk values, and carrying out risk evolution trend deduction to obtain risk conduction probability; And 4, outputting a hierarchical prevention and control strategy and a model parameter iteration optimization instruction based on the confidence interval and the risk conduction probability of the risk value.
2. The method for evaluating the safety risk of electric power metering data according to claim 1, wherein, In the step 1, the electric power metering multisource data comprises metering core data, scene dynamic characteristic data and equipment health state data, wherein the metering core data comprises node three-phase current effective values, node three-phase voltage effective values, total active power, total reactive power, power factors and total harmonic distortion rate, the scene dynamic characteristic data comprises distributed photovoltaic access power, distributed wind power access power, new energy permeability, industrial/resident/commercial/public load ratio, time period load fluctuation rate and peak-to-valley load ratio, and the equipment health state data comprises metering equipment accumulated operation time, calibration period residual days, last calibration deviation values, real-time working temperature, equipment internal humidity, fault accumulated times and sensor drift amount.
3. The method for evaluating the safety risk of electric power metering data according to claim 2, wherein in step 1, the process of constructing the high-dimensional feature matrix specifically comprises the following steps: Step 1.1, performing standardization processing on the power metering multi-source data by adopting Z-score standardization to obtain first power metering multi-source data; step 1.2, aiming at the first electric power metering multi-source data, identifying and eliminating abnormal values by adopting a 3 sigma criterion, filling the missing values by a K neighbor interpolation method, and obtaining second electric power metering multi-source data; Step 1.3, arranging the second electric power metering multi-source data in turn according to the sorting sequence of metering core data, scene dynamic characteristic data and equipment health state data, wherein each data sample respectively forms characteristic vectors of a single sample, the characteristic vectors of all samples are stacked in rows and integrated into a high-dimensional characteristic matrix Wherein m is the number of collected effective samples, and n is the total number of features of the high-dimensional feature matrix.
4. The method for evaluating the safety risk of electric power metering data according to claim 3, wherein in the step 2, the step of quantifying the statistical distribution rule of the scene characteristics of the distribution node or the platform region is characterized by specifically comprising the steps of fitting the statistical distribution of the scene dynamic characteristic data by adopting a mixed Gaussian model, determining the number of model components by a Bayesian information rule, and obtaining the scene dynamic characteristic data under different risk levels R Probability density function of (2) ; The method for quantifying the health decay law of the metering equipment specifically comprises the following steps of calculating a health index H of the equipment, wherein the formula is as follows: ; Wherein, the A feature dimension number for the device health status data; Standardized statistics for class i device health status data; Counting a control threshold for the 95% confidence level of the i-th type health data, and determining based on the quantile of the historical health status data; fitting historical health state indexes and associated data of actual fault occurrence conditions to obtain a sigmoid function slope through a least square method, wherein the fitting target is the mean square error of the predicted health state and the actual state; fitting historical data based on equipment performance decay after failure for the cumulative decay coefficient; the relative duty ratio of the operation time length after the jth fault is a statistic; The number of times is accumulated for the failure.
5. The method for risk assessment of electrical power metering data according to claim 4, wherein, In step 2, the dynamic risk assessment model is a probability assessment model fusing multiple source features, and includes: an input layer for receiving a high-dimensional feature matrix; the feature fusion layer is used for integrating the high-dimensional feature matrix with the equipment health index H and probability density functions of the scene dynamic feature data S under different risk levels R Performing association integration to obtain comprehensive features fusing the multi-source scene features and the health state of the equipment; and the probability output layer is used for outputting posterior probability distribution of the risk level based on the comprehensive characteristics of the integrated multi-source scene characteristics and the equipment health state.
6. The method for risk assessment of electrical power metering data according to claim 5, wherein, In step 2, dynamically updating the prior probability of the dynamic risk assessment model, which specifically includes the following steps: Probability density function combining scene dynamic characteristic data S under different risk levels R And calculating posterior probability with the equipment health index H through a Bayesian posterior update formula, and updating the posterior probability into the prior probability of risk assessment, wherein the Bayesian posterior update formula is as follows: ; Wherein, the To combine the device health index And scene dynamic feature data Posterior probability of posterior risk class R; is a Beta distribution likelihood function of the device health index at risk level R, For different risk levels Dynamic feature data of lower scene Probability density functions of (2); is a risk grade Is a priori probability of the initial prior of (a); To integrate dummy variables, representing risk levels Is traversed and valued; the likelihood function of the dynamic risk assessment model is updated dynamically, and the method specifically comprises the following steps: Fitting the multidimensional normal distribution of the high-dimensional feature matrix under different risk levels R based on the historical high-dimensional feature matrix and the associated data of the corresponding risk levels to obtain likelihood functions For the probability density function of the multidimensional normal distribution, the formula is as follows: ; where n is the feature dimension of the high-dimensional feature matrix, Is the mean vector of the high-dimensional features at risk level R, As covariance matrix of high-dimensional features at risk level R, Is that T represents matrix inversion; probability output layer through fusion 、 And And outputting posterior probability distribution of the risk level.
7. The method for evaluating the safety risk of electric power metering data according to claim 6, wherein the step 3 specifically comprises the following steps: Adopting an improved LSTM time sequence prediction model fused with GARCH fluctuation rate modeling, integrating the fluctuation rate characteristics of the electric power metering data captured by the GARCH model into an LSTM model input layer, combining the output of a dynamic risk assessment model to complete the probability distribution prediction of a risk value, extracting the mathematical expectation and standard deviation statistical characteristics of the probability distribution prediction, combining a preset confidence level, and calculating a confidence interval of the risk value; Counting transition frequencies among risk levels based on historical transition records of the risk levels, constructing a state transition matrix through a Markov chain state transition algorithm after Laplace smoothing treatment, completing risk evolution trend deduction by combining the output of a dynamic risk assessment model, and obtaining risk conduction probability through quantitative calculation; Wherein the risk conduction probability The calculation formula of (2) is as follows: ; Wherein, the Is the number of time sequence windows A step state transition matrix is used to determine, For the current risk level, For risk levels higher than the current level, For k-step inner slave Transfer to Probability of (2); wherein the risk value Through the statistical feature fusion calculation of probability distribution, the calculation formula is as follows: ; Wherein, the For a mathematical expectation of the risk value, As a coefficient of variation of the risk profile, Is a weight coefficient.
8. The method for evaluating the security risk of electric power metering data according to claim 7, wherein in step 4, the output logic of the hierarchical prevention and control strategy is as follows: based on the 95% confidence interval upper limit of the risk value and the risk conduction probability, classifying the risk into three levels of low risk, medium risk and high risk; The upper limit of a confidence interval of a corresponding risk value of the low risk level is less than or equal to 0.3, the risk conduction probability is less than 10%, a conventional monitoring strategy is adopted, electric power metering multi-source data are collected according to an original preset collection period, and a dynamic risk assessment model is checked; The upper limit of the confidence interval of the risk value corresponding to the risk level in the stroke is 0.3-0.7, the risk conduction probability is more than or equal to 10% and less than 30%, the data acquisition period is shortened to 50% of the original preset acquisition period by adopting the enhanced monitoring strategy, and the special investigation of the health state data of the equipment is synchronously started; The upper limit of the confidence interval of the corresponding risk value of the high risk level is more than 0.7, the risk conduction probability is more than or equal to 30%, an emergency prevention and control strategy is adopted, the settlement use of the node metering data is suspended, and the off-line calibration or replacement flow of the equipment is triggered; And if the upper limit of the risk value confidence interval and the risk conduction probability do not meet the double-condition requirement of any level, executing the prevention and control strategy according to the higher level in the risk level respectively corresponding to the two indexes.
9. The method for evaluating the safety risk of electric power metering data according to claim 8, wherein in step 4, the process of outputting the model parameter iterative optimization instruction is as follows: triggering a parameter iteration if the newly acquired power metering multisource data reaches a preset number; Carrying out parameter optimization on covariance matrix weights of the dynamic risk assessment model and learning rate of the improved LSTM time sequence prediction model by adopting a grid search algorithm; The method comprises the steps of taking a KS value of a dynamic risk assessment model prediction result and a PSI value as core indexes, wherein the KS value is a distinguishing degree index of the model prediction result after the high risk level and the low risk level in three types of risk levels are singly divided and the low risk level and the medium risk level are combined into non-high risk levels, the KS value target threshold is more than or equal to 0.4, and the PSI value target threshold is less than 0.1; And feeding the optimized parameters back to the dynamic risk assessment model and the improved LSTM time sequence prediction model to realize the self-adaptive updating of the model.
10. A power metering data security risk assessment apparatus for performing the power metering data security risk assessment method according to any one of claims 1 to 9, characterized in that the apparatus comprises: The matrix construction module is used for collecting power metering multisource data of power distribution nodes or transformer areas in the target area and constructing a high-dimensional feature matrix; The model construction module is used for respectively quantifying distribution node or area scene characteristic statistical distribution rules and metering equipment health attenuation rules, constructing a dynamic risk assessment model integrating the distribution node or area scene characteristic statistical distribution rules and the metering equipment health attenuation rules based on quantification results, and dynamically updating prior probability and likelihood functions of the dynamic risk assessment model; The confidence interval and risk conduction probability calculation module is used for inputting the high-dimensional feature matrix into a dynamic risk assessment model, carrying out probability distribution prediction of a risk value by combining an improved LSTM time sequence prediction model and a Markov chain state transition algorithm to obtain a confidence interval of the risk value, and carrying out risk evolution trend deduction to obtain risk conduction probability; And the prevention and control strategy output and parameter optimization module is used for outputting a hierarchical prevention and control strategy and model parameter iteration optimization instruction based on the confidence interval and the risk conduction probability of the risk value.

Description

Electric power metering data security risk assessment method and device Technical Field The invention belongs to the technical field of data security of power systems, and particularly relates to a power metering data security risk assessment method and device. Background With the deep construction and power market reform of a novel power system, power metering data become core foundations of power grid safety scheduling, electric energy fair transaction and user electricity behavior analysis, and the safety and reliability of the power metering data are directly related to the stable operation of the power system. At present, the scale of the power system is continuously enlarged, a large amount of distributed new energy is accessed, diversified loads randomly fluctuate, so that the power metering data is characterized by massive growth and dynamic change, meanwhile, the security threat faced by the whole flow of data collection, transmission, storage and application is increasingly complex, such as metering deviation caused by data tampering, packet loss transmission and hardware aging, cross-node risk diffusion and the like, and higher requirements are provided for the power metering data security risk assessment technology. The current power metering data security risk assessment method mainly adopts a fixed threshold value or a static model, dynamic change characteristics of a power system, such as power fluctuation caused by access of distributed new energy and differentiated characteristics of power loads of different terminals, are not fully considered, so that assessment results are difficult to adapt to risk identification requirements of complex scenes and are easy to produce misjudgment or missed judgment, and the traditional scheme mainly responds to security risks passively, lacks active prediction capability and is difficult to meet the requirements of prospective security prevention and control of the power system. Disclosure of Invention Aiming at the technical problems pointed out in the background art, the invention aims to provide a power metering data security risk assessment method and device. In order to achieve the purpose of the invention, the technical scheme provided by the invention is as follows: First aspect The invention provides a power metering data security risk assessment method, which comprises the following steps: Step 1, collecting power metering multi-source data of power distribution nodes or transformer areas in a target area, and constructing a high-dimensional feature matrix; Step 2, respectively quantifying distribution node or area scene feature statistical distribution rules and metering equipment health attenuation rules, constructing a dynamic risk assessment model integrating the distribution node or area scene feature statistical distribution rules and the metering equipment health attenuation rules based on quantified results, and dynamically updating prior probability and likelihood functions of the dynamic risk assessment model; Step 3, inputting the high-dimensional feature matrix into a dynamic risk assessment model, and carrying out probability distribution prediction of risk values by combining an improved LSTM time sequence prediction model and a Markov chain state transition algorithm to obtain a confidence interval of the risk values, and carrying out risk evolution trend deduction to obtain risk conduction probability; And 4, outputting a hierarchical prevention and control strategy and a model parameter iteration optimization instruction based on the confidence interval and the risk conduction probability of the risk value. Further, in the step 1, the electric power metering multisource data comprises metering core data, scene dynamic characteristic data and equipment health state data, wherein the metering core data comprises node three-phase current effective values, node three-phase voltage effective values, total active power, total reactive power, power factors and total harmonic distortion rate, the scene dynamic characteristic data comprises distributed photovoltaic access power, distributed wind power access power, new energy permeability, industrial/resident/commercial/public load ratio, time period load fluctuation rate and peak-to-valley load ratio, and the equipment health state data comprises metering equipment accumulated operation time, calibration period residual days, last calibration deviation values, real-time working temperature, equipment internal humidity, fault accumulated times and sensor drift amount. Further, in step 1, the process of constructing the high-dimensional feature matrix specifically includes the following steps: Step 1.1, performing standardization processing on the power metering multi-source data by adopting Z-score standardization to obtain first power metering multi-source data; step 1.2, aiming at the first electric power metering multi-source data, identifying and eliminating abnormal values by adoptin