Source-load bilateral uncertainty characterization and traceability method and system for electric power system

Abstract

The invention relates to the technical field of electric power system analysis, in particular to an electric power system source load bilateral uncertainty characterization and traceability method and system.The method comprises the steps that a first-level index, a second-level index and a third-level index are divided; the method comprises the following steps: acquiring data of a first-level index on a plurality of time sections, setting a plurality of data judgment thresholds, classifying a current system operation state based on the judgment thresholds, and carrying out step-by-step analysis and calculation through a feature importance score of each index about a certain index output in a machine learning model, through the layered and progressive analysis framework and in combination with a collinearity robust machine learning method, the source-load bilateral uncertainty in the high-proportion new energy power system can be systematically quantified and traced, and a brand-new, more accurate and scientific analysis tool is provided for risk early warning, planning decision and operation optimization of a power grid.

Description

Technical Field

[0001] This invention relates to the field of power system analysis technology, and more specifically, to a method and system for characterizing and tracing the source-load dual-sided uncertainty of a power system. Background Technology

[0002] With the deepening of the "dual-carbon" strategy, a new power system based on new energy sources is being rapidly constructed. The strong volatility of new energy output and the increasingly complex electricity consumption behavior on the load side are intertwined, resulting in high uncertainty on both the power source and load sides, as well as in system regulation resources. The traditional power system's balancing mechanism of "deterministic generation tracking uncertain load" has evolved into a new model of "bidirectional random matching of uncertainties on both the power source and load sides." This strong uncertainty on both sides poses significant challenges to the system's reliable power supply (load loss risk) and efficient absorption of new energy (curtailment risk).

[0003] Existing technologies have the following shortcomings in dealing with uncertainty: (1) Single analytical dimension: Most studies only start from a few factors on the source side or the load side, which makes it difficult to cope with the increasing number of uncertain factors and their complex coupling effects, and face the problem of "combinatorial explosion".

[0004] (2) Lack of a systematic source tracing framework: Existing research focuses on the optimization scheduling under uncertain environment, but lacks a systematic analytical framework for the quantitative characterization of uncertainty itself, especially the transmission path tracing from system-level operational risks (phenomena) to basic technical indicators (root causes).

[0005] (3) Limitations of analytical methods: Traditional correlation analysis cannot handle nonlinear relationships, while methods such as global sensitivity analysis (GSA) make strong assumptions about the independence of input variables, which limits their application in real-world scenarios where indicators are collinear.

[0006] Therefore, there is an urgent need for a new method that can systematically quantify the strong uncertainty on both the source and load sides, accurately identify key influencing factors, and clearly reveal the transmission path of uncertainty, so as to improve the power grid risk perception and decision support capabilities. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for characterizing and tracing the uncertainty of both the source and load sides of a power system, in order to solve the above-mentioned problems in the prior art.

[0008] This invention is achieved through the following technical solution: In a first aspect, the present invention provides a method for characterizing and tracing the source-load dual-sided uncertainty of a power system, including: Data from several types of power systems is acquired and classified into several indicators, including primary, secondary, and tertiary indicators. Acquire data for primary indicators across multiple time segments, set several judgment thresholds for the data, and classify the current system operating status based on the judgment thresholds. If the system is classified as a definite state, no action is taken. If the classification is uncertain, then several first machine learning models are established. Each first machine learning model takes a certain first-level indicator as its unique output and all second-level indicators as its input. The feature importance score array of each second-level indicator output by each first machine learning model with respect to a certain first-level indicator is obtained. The second-level indicator corresponding to the feature importance score with the largest score in the feature importance score array is selected as the key second-level indicator. Then, several second machine learning models are established. Each second machine learning model takes a certain key secondary indicator as its unique output and all tertiary indicators as its input. The feature importance score array of each tertiary indicator output by each second machine learning model with respect to a certain key secondary indicator is obtained. The secondary indicator corresponding to the feature importance score with the largest score in the feature importance score array is selected as the key tertiary indicator. Output key tertiary indicators as sources of uncertainty in the power system.

[0009] Preferably, the primary indicators include load shedding rate, expected load shedding, wind and solar curtailment rate, and expected renewable energy curtailment. The secondary indicators include the daily maximum peak-valley difference rate of net load, the peak-shaving characteristics of new energy sources, the matching degree of source-load fluctuations, the daily average new energy penetration rate, and the new energy penetration rate during peak load periods. The three-level indicators include new energy-side indicators and load-side indicators. The new energy-side indicators include the proportion of new energy installed capacity, the accuracy of wind and solar power daily output forecast, the number of wind and solar power daily utilization hours, and the wind and solar power hourly output change rate. The load-side indicators include the daily total load power forecast deviation rate, the hourly load change rate, the daily maximum peak-valley difference rate, and the daily peak duration.

[0010] Preferably, the step of setting a judgment threshold and classifying the current system operating state based on the judgment threshold includes: When both the load shedding rate and the wind and solar curtailment rate are greater than their respective preset thresholds, the condition is classified as a state of strong uncertainty. When only the load loss rate is greater than its preset threshold, it is classified as a load-side uncertainty state; When only the wind curtailment rate exceeds its preset threshold, it is classified as a source-side uncertainty state. When neither of them is greater than their respective preset thresholds, it is classified as a deterministic state.

[0011] Preferably, the machine learning model is a random forest model. Based on random forest, a score reflecting the importance of features is obtained by perturbing the out-of-bag data that did not participate in the training of the decision tree and then calculating the change in classification accuracy.

[0012] Preferably, the load loss rate includes:

[0013]

[0014]

[0015]

[0016] In the formula, For the first The amount of load loss at each point in time. For the first The actual net load at each point in time This represents the operating capacity of the thermal power units on that day. This represents the operating capacity of other generating units for the day. This represents the maximum charge and discharge power of the energy storage. For the first The actual load at each point in time For the first The actual output power of wind power at a given time point For the first The actual output power of photovoltaics at each point in time. The load shedding rate, 0-1 auxiliary variables for judgment, This corresponds to the maximum time.

[0017] Preferably, the expected amount of renewable energy curtailment includes:

[0018]

[0019]

[0020] In the formula, For the first The amount of wind and solar power curtailed at each point in time. This represents the minimum output power of the thermal power unit. This is the minimum output power for other units. This is in response to expectations regarding the amount of electricity that has been wasted from renewable energy sources.

[0021] Preferably, the average daily new energy penetration rate includes:

[0022] The renewable energy penetration rate during peak load periods includes:

[0023]

[0024] In the formula, This represents the average daily penetration rate of new energy sources. This represents the penetration rate of new energy sources during peak load periods. For the corresponding time, This corresponds to the year.

[0025] Preferably, the daily maximum peak-to-valley difference rate of net load includes:

[0026] In the formula, The maximum peak-to-valley difference rate of net load on a given day. To correspond to a certain day.

[0027] Preferably, the daily utilization hours of the wind power photovoltaic system include:

[0028]

[0029] In the formula, This refers to the number of hours of wind power generation. For the first Predicted wind turbine output at each point in time. The number of hours of photovoltaic power generation utilization. For the first Predicted photovoltaic output values ​​at specific time points.

[0030] Secondly, the present invention also provides a power system source-load two-sided uncertainty characterization and tracing system for performing the above-described power system source-load two-sided uncertainty characterization and tracing method, comprising: The status judgment module is configured to acquire data from several types of power systems, classify the data into several indicators, including primary, secondary, and tertiary indicators; acquire data of primary indicators at multiple time points, set judgment thresholds for several data points, and classify the current system operating status based on the judgment thresholds. If the system is classified as a definite state, no further action is taken. The source tracing module is configured to, if classified as an uncertain state, establish several first-level machine learning models. Each first-level machine learning model takes a specific primary indicator as its unique output and all secondary indicators as its input. It then obtains a feature importance score array for each secondary indicator output by the first-level machine learning model with respect to a specific primary indicator. The secondary indicator corresponding to the highest feature importance score in the feature importance score array is selected as the key secondary indicator. Next, it establishes several second-level machine learning models. Each second-level machine learning model takes a specific key secondary indicator as its unique output and all tertiary indicators as its input. It then obtains a feature importance score array for each tertiary indicator output by the second-level machine learning model with respect to a specific key secondary indicator. The secondary indicator corresponding to the highest feature importance score in the feature importance score array is selected as the key tertiary indicator. The key tertiary indicator is output as the source of uncertainty in the power system.

[0031] The technical solution of the present invention has at least the following advantages and beneficial effects: The method provided by this invention includes primary, secondary, and tertiary indicators; it acquires data on primary indicators across multiple time segments, sets several data judgment thresholds, classifies the current system operating status based on these thresholds, and performs hierarchical analysis and calculation using the feature importance score of each indicator output by the machine learning model. Through this hierarchical and progressive analysis framework, combined with machine learning methods robust to collinearity, it can systematically quantify and trace the source-load dual-side uncertainties in high-proportion renewable energy power systems, providing a new, more accurate, and scientific analytical tool for power grid risk warning, planning decisions, and operation optimization. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 This is a schematic diagram of the source-load two-sided strong uncertainty index system of the present invention; Figure 3 This is a heatmap showing the correlation matrix between the secondary indicators of this invention. Figure 4 The importance of the secondary metric of random forest for load shedding rate in this invention; Figure 5This invention uses the secondary index of random forest to assess the importance of wind and solar power abandonment rate. Figure 6 This is a schematic diagram illustrating the global sensitivity analysis of "load loss rate" according to the present invention. Figure 7 This invention provides a global sensitivity analysis of the "wind curtailment and solar curtailment rate"; Figure 8 This demonstrates the importance of the three-level indicators of the present invention for the peak modulation characteristics of new energy sources; Figure 9 This invention highlights the importance of the three-level indicators for the net load peak-valley ratio. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0035] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. The naming or numbering of steps in this application does not imply that the steps in the method flow must be executed in the chronological / logical order indicated by the naming or numbering. The execution order of named or numbered process steps can be changed according to the desired technical objective, as long as the same or similar technical effect is achieved.

[0036] The independently described modules or sub-modules may or may not be physically separated; they may be implemented in software or hardware, and some modules or sub-modules may be implemented in software, with the processor calling the software to implement the function of these modules or sub-modules, while other modules or sub-modules may be implemented in hardware, such as through hardware circuits. Furthermore, some or all of the modules can be selected to achieve the purpose of this application's solution according to actual needs.

[0037] Please refer to Figures 1-9 The present invention provides a method for characterizing and tracing the source-load dual-sided uncertainty of a power system, comprising: S101: Obtain data from several types of power systems, and classify the data from these power systems into several indicators, including primary indicators, secondary indicators, and tertiary indicators. S102: Obtain data of primary indicators at multiple time points, set judgment thresholds for several data points, classify the current system operating status based on the judgment thresholds, and do not process if the classification is a definite state. This step aims to perform a macro-level qualitative classification of the uncertainty level at each time point based on the actual consequences of system operation, providing clear scenario labels for subsequent source tracing analysis. Classification basis: The load shedding rate and wind / solar curtailment rate from the primary indicators are used as two-dimensional evaluation coordinates.

[0038] S103: If the classification is uncertain, then establish several first machine learning models. Each first machine learning model takes a certain first-level indicator as its unique output and all second-level indicators as its input. The feature importance score array of each second-level indicator output by each first machine learning model with respect to a certain first-level indicator is obtained. The second-level indicator corresponding to the feature importance score with the largest score in the feature importance score array is selected as the key second-level indicator. S104: Establish several second machine learning models. Each second machine learning model takes a certain key secondary indicator as its unique output and all tertiary indicators as its input. The feature importance score array of each tertiary indicator output by each second machine learning model with respect to a certain key secondary indicator is obtained. The secondary indicator corresponding to the feature importance score with the largest score in the feature importance score array is selected as the key tertiary indicator. S104: Output key tertiary indicators as sources of uncertainty in the power system.

[0039] The method provided by this invention includes primary, secondary, and tertiary indicators; it acquires data on primary indicators across multiple time segments, sets several data judgment thresholds, classifies the current system operating status based on these thresholds, and performs hierarchical analysis and calculation using the feature importance score of each indicator output by the machine learning model. Through this hierarchical and progressive analysis framework, combined with machine learning methods robust to collinearity, it can systematically quantify and trace the source-load dual-side uncertainties in high-proportion renewable energy power systems, providing a new, more accurate, and scientific analytical tool for power grid risk warning, planning decisions, and operation optimization.

[0040] In one exemplary embodiment of the present invention, the primary indicators include load shedding rate, expected load shedding, wind and solar curtailment rate, and expected renewable energy curtailment. These primary indicators represent the ultimate adverse outcome resulting from the system's failure to fully mitigate risks under the impact of uncertainties on both the power source and load sides. Load shedding rate and expected load shedding are core indicators for measuring the overall power supply capacity of the system, directly quantifying the ultimate negative impact on socio-economic activities when the system copes with uncertainties. Wind and solar curtailment rate and expected renewable energy curtailment are core indicators for measuring the renewable energy absorption capacity, directly quantifying the negative impact on renewable energy development goals and energy transition strategies when the system copes with uncertainties.

[0041] Loss of Load Rate (LOLR):

[0042] Net load

[0043]

[0044]

[0045] Expected Energy Not Supplied (EENS):

[0046]

[0047] Wind Solar Curtailment Rate (WSCR):

[0048]

[0049]

[0050] Expected Curtailed Renewable Energy (ECRE):

[0051]

[0052] In the formula, This represents the load loss at time point t; a positive value indicates that a load loss has occurred. This represents the actual net load at time point t; This indicates the operating capacity of the thermal power unit on that day; This indicates the operating capacity of other generating units (hydropower, nuclear power, etc.) on that day; Indicates the maximum charging and discharging power of the energy storage; This represents the actual load at time point t; This represents the actual output power of the wind power at time point t; This represents the actual output power of the photovoltaic system at time point t. This represents the amount of wind and solar power curtailment at time point t, with a positive value indicating that wind and solar power curtailment has occurred. This indicates the minimum output power of the thermal power unit; This indicates the minimum output power of other generating units (hydropower, nuclear power, etc.); It is a 0-1 auxiliary variable used for judgment. This corresponds to the maximum time.

[0053] The secondary indicators include the daily maximum peak-to-valley difference rate of net load, renewable energy peak-shaving characteristics, source-load fluctuation matching degree, daily average renewable energy penetration rate, and renewable energy penetration rate during peak load periods. These secondary indicators quantitatively reflect the system's uncertainty characteristics resulting from the interaction and mutual influence between the source and load sides, and do not belong solely to the source or grid side. The daily average renewable energy penetration rate measures the proportion of renewable energy in the daily power balance; the higher the penetration rate, the greater the impact of the inherent uncertainty of renewable energy on the entire system. The instantaneous renewable energy penetration rate during peak load periods is representative, reflecting the support capacity of renewable energy for power supply during peak load periods. Net load is the final object that conventional power sources need to balance; the daily maximum peak-to-valley difference rate of net load integrates the peak-to-valley characteristics of the load and the output characteristics of renewable energy, making it the most core and direct comprehensive indicator for assessing the system's demand for flexibility resources. Renewable energy peak-shaving characteristics measure the degree of change between the net load peak-to-valley difference and the load peak-to-valley difference after renewable energy integration, reflecting the matching relationship between renewable energy output and peak load. The source-load fluctuation matching degree quantitatively reflects the temporal synchronicity of the fluctuation trends of renewable energy output and load.

[0054] Specifically, the Daily Average Renewable Energy Penetration (DAREP):

[0055] Renewable Energy Penetration during Peak Load (REP-PL):

[0056]

[0057] Net Load Peak-to-Valley Rate (NLPVR):

[0058] In the formula, To correspond to a certain moment, To correspond to a certain year, To correspond to a certain day.

[0059] Renewable Energy Peak Regulation Characteristic (RE-PRC): Original load peak-to-valley difference:

[0060] Net load peak-to-valley difference:

[0061] Peak-shaving characteristics, i.e., the change in peak-to-valley difference after the integration of new energy sources:

[0062] A value greater than 0 indicates that it has a positive peak-shaving effect, reducing the peak-to-valley difference.

[0063] Source-Load Fluctuation Matching Degree (SL-FMD) is the correlation coefficient between renewable energy output and load.

[0064]

[0065] New energy output:

[0066] In the formula, This indicates the peak-to-valley difference in the daily load. This represents the peak-to-valley difference in net load for the day; This represents the actual output of new energy sources (wind power and photovoltaics) at time point t; This represents the average value of the actual output of new energy sources on that day. This represents the average actual load for that day.

[0067] The three-level indicators include new energy-side indicators and load-side indicators. The new energy-side indicators include the proportion of new energy installed capacity, the accuracy of wind and solar power daily output forecast, the number of wind and solar power daily utilization hours, and the 1-hour output change rate of wind and solar power. The load-side indicators include the daily total power forecast deviation rate, the 1-hour load change rate, the daily maximum peak-valley difference rate of load, and the daily peak duration of load.

[0068] The three-level indicators quantify the inherent, independent statistical characteristics and prediction accuracy of new energy loads from both sides, and are the basic indicators for measuring the uncertainty of the power system. Similar to penetration rate, the proportion of installed capacity of new energy sources reflects the impact of uncertainty on the entire system. The daily power output forecast accuracy of wind and solar power quantifies the forecast deviation caused by insufficient understanding (fuzziness) or meteorological randomness; improving forecast accuracy is one of the keys to reducing uncertainty. Daily utilization hours of wind and solar power reflect the daily endowment of new energy resources and are a direct manifestation of "intermittency." Low utilization hours mean long periods of low output, increasing the demand for system reserve capacity. The 1-hour output change rate of wind and solar power directly quantifies the "fluctuation" of new energy, reflecting the rate of increase or decrease in new energy output within a short period, determining the system's flexibility requirements. The daily total power output forecast deviation rate quantifies the uncertainty of load-side forecasts. The 1-hour load change rate reflects the fluctuation characteristics of the load itself and is also a major indicator of system flexibility requirements. The daily maximum peak-valley difference rate reflects the basic demand of the load on the system's peak-shaving capacity. The daily peak duration reflects the increasingly significant load "peaking" characteristics. It should be noted that the longer the peak duration, the smaller the overall load change rate and the less significant the "peaking characteristic."

[0069] New energy sector indicators: Renewable Energy Installed Capacity Ratio (RE-ICR):

[0070] Daily Wind / Solar Power Forecast Accuracy (WPFA / SPFA):

[0071]

[0072] Daily Wind / Solar Power Utilization Hours (WPUH / SPUH):

[0073]

[0074] Hourly Wind / Solar Power Change Rate (WPCR / SPCR):

[0075]

[0076] Load-side indicators: Daily Load Forecast Accuracy (LFA):

[0077] Hourly Load Change Rate (LCR):

[0078] Daily Load Peak-to-Valley Rate (DLPVR):

[0079] Daily Peak Load Duration (DPLD): The time when the load is above 98% of the daily maximum load is defined as the load peak time.

[0080]

[0081] In the formula, This indicates the operating capacity of the wind turbines on that day; This indicates the installed capacity of photovoltaic systems on that day. This represents the predicted value of the wind turbine output at time point t; This represents the predicted photovoltaic output at time point t. This represents the wind power output at time point t+4. The reason it is... This is because the time interval in the embodiment is 15 minutes, and 4 time intervals make up 1 hour. Similarly; This represents the predicted load at time point t.

[0082] Data acquisition: Based on the actual operation data or production simulation data of the power system, calculate the values ​​of each indicator in the three-level indicator system at multiple consecutive time sections (such as daily) to form a multi-dimensional time series dataset.

[0083] In one exemplary embodiment of the present invention, the step of setting a judgment threshold and classifying the current system operating state based on the judgment threshold includes: When both the load shedding rate and the wind and solar curtailment rate are greater than their respective preset thresholds, the condition is classified as a state of strong uncertainty. When only the load loss rate is greater than its preset threshold, it is classified as a load-side uncertainty state; When only the wind curtailment rate exceeds its preset threshold, it is classified as a source-side uncertainty state. When neither of them is greater than their respective preset thresholds, it is classified as a deterministic state.

[0084] It should be noted that the "load-side uncertainty system" and "source-side uncertainty system" here do not mean that the uncertainty originates from the load side or the new energy side, but rather that the uncertainty at the system level is ultimately reflected in the supply guarantee on the load side or the absorption of new energy on the new energy side.

[0085] In one exemplary embodiment of the present invention, the machine learning model is a random forest model. Based on random forest, a score reflecting the importance of features is obtained by perturbing the out-of-bag data that did not participate in the training of the decision tree and then calculating the change in classification accuracy.

[0086] Specifically, this involves a hierarchical, progressive identification and tracing of key factors based on random forests. This step is the core of the invention, aiming to overcome the limitations of traditional analysis methods in handling multi-factor, nonlinear, and collinear problems, and to identify key driving factors of uncertainty layer by layer and construct transmission paths.

[0087] (1) Method selection: Random Forest algorithm, which is robust to collinearity and does not require strict linearity assumptions, is used as the main analysis tool. Its built-in feature importance evaluation function based on permutation prediction error is used to obtain feature importance by perturbing the out-of-bag (OOB) data that did not participate in the training of decision trees and then calculating the change in classification accuracy.

[0088] Suppose a sample has M features, and a random forest randomly selects K datasets, resulting in K out-of-bounds (OOB) datasets. The steps for ranking feature importance are as follows: 1) Initialization .

[0089] 2) Using the first Training decision trees on a dataset. Calculate the first Classification accuracy of OOB datasets .

[0090] 3) Features in the OOB dataset Apply a perturbation and recalculate the accuracy. .

[0091] 4) For Repeat steps 2) and 3).

[0092] 5) Calculate the characteristics using the following formula. Importance, that is:

[0093] 6) To Sort the features in descending order to obtain a ranking of feature importance; the higher the ranking, the more important the feature.

[0094] In the formula, Let be the importance score of the i-th feature.

[0095] (2) Identification of key coupling indicators. Two independent random forest regression surrogate models were trained with load loss rate (LOLR) and wind / solar curtailment rate (RECR) as target outputs and all secondary indicators as inputs. Using the feature importance assessment function built into random forest based on permutation prediction error, the importance scores of each secondary indicator were quantitatively calculated and ranked to identify the 1-2 key secondary indicators that contribute the most to LOLR and RECR.

[0096] (3) In-depth source tracing of uncertainty. Using the key secondary indicators identified in the previous step as the target output and all tertiary indicators (including source and load sides) as input, a new random forest surrogate model is trained. Similarly, using feature importance assessment, the key tertiary indicators with the greatest impact on the key secondary indicators are identified by calculating and ranking the importance scores of each tertiary indicator.

[0097] (4) Transmission path construction: Combining the analysis results of the second and third layers, a data-driven uncertainty transmission path is constructed from "key third-level indicators" to "key second-level indicators" and then to "first-level risk indicators".

[0098] 4. Uncertainty Source Identification and Cross-Validation. This step aims to comprehensively assess and verify the reliability of the analysis results.

[0099] (1) Source judgment: Statistically analyze the importance weight ratio of all the key third-level indicators identified, based on their predefined attributes (source side or load side). The larger the corresponding importance weight ratio, the higher the ranking of the third-level indicator and the greater the influence of the indicator, so as to judge the main source of uncertainty leading to the overall risk of the system.

[0100] (2) Cross-validation: To ensure the reliability of the random forest analysis conclusions, cross-validation was performed using global sensitivity analysis based on variance (GSA-Sobol' method). Specifically, to address the instability of the surrogate model in GSA on small sample data, this invention employs ensemble neural network technology to construct the surrogate model. This technology ensures reproducibility by fixing the random seed and integrating the prediction results of multiple (e.g., 10) independently trained neural networks, thereby obtaining a stable and highly accurate surrogate model. The consistency of the key factor identification results was verified by comparing the sensitivity index (especially the total sensitivity index TSI) ranking calculated by GSA with the importance ranking of the random forest.

[0101] This invention also provides a power system source-load two-sided uncertainty characterization and tracing system for performing the above-described power system source-load two-sided uncertainty characterization and tracing method, comprising: The status judgment module is configured to acquire data from several types of power systems, classify the data into several indicators, including primary, secondary, and tertiary indicators; acquire data of primary indicators at multiple time points, set judgment thresholds for several data points, and classify the current system operating status based on the judgment thresholds. If the system is classified as a definite state, no further action is taken. The source tracing module is configured to, if classified as an uncertain state, establish several first-level machine learning models. Each first-level machine learning model takes a specific primary indicator as its unique output and all secondary indicators as its input. It then obtains a feature importance score array for each secondary indicator output by the first-level machine learning model with respect to a specific primary indicator. The secondary indicator corresponding to the highest feature importance score in the feature importance score array is selected as the key secondary indicator. Next, it establishes several second-level machine learning models. Each second-level machine learning model takes a specific key secondary indicator as its unique output and all tertiary indicators as its input. It then obtains a feature importance score array for each tertiary indicator output by the second-level machine learning model with respect to a specific key secondary indicator. The secondary indicator corresponding to the highest feature importance score in the feature importance score array is selected as the key tertiary indicator. The key tertiary indicator is output as the source of uncertainty in the power system.

[0102] This embodiment provides a specific example to further explain the above content.

[0103] Indicator system construction, data acquisition, and system status classification.

[0104] Data preparation: Based on the 2024 power grid operation data of Shandong Province. The data includes: daily meteorological data (wind speed, solar intensity), daily peak and valley load values, and the installed capacity of various types of generating units (thermal power, wind power, photovoltaic, and energy storage). Based on the above basic data, load curves, wind power and photovoltaic output curves with a time resolution of 15 minutes and a duration of 365 days were generated through production simulation.

[0105] Indicator Calculation: Based on the indicator system defined in Section 1 of the "Invention Content" section of this invention, the values ​​of 20 primary, secondary, and tertiary indicators for each day of the 365 days are calculated to form a 365x20 data matrix.

[0106] System state classification: The risk thresholds for both LOLR and RECR were set at 1%. Classification was performed over 365 days. The results showed that the "deterministic" state had the most days (288 days); the "strong uncertainty" state had the fewest days (2 days); and the "source-side uncertainty" and "load-side uncertainty" states had 33 and 42 days respectively. This classification provides sample sets of different risk levels for subsequent analysis.

[0107] 2. Identification of key coupling indicators.

[0108] Collinearity Diagnosis: The correlation matrix heatmap of all indicators shows strong correlations among some indicators, with clusters of indicators pointing to the same scenario, such as a surge in new energy generation or heavy load on a given day. This can be considered redundant, thus necessitating the identification of key uncertainty indicators. The correlation matrix heatmap and variance inflation factor (VIF) of the five secondary indicators were calculated. The results show strong correlations between SL-FMD, REP-PL, and NLPVR, with VIF values ​​all greater than 20, indicating severe collinearity. This diagnostic result confirms the limitations of directly using GSA for the entire analysis, highlighting the necessity of using random forest in this invention.

[0109] Random Forest Analysis: ① Model configuration: Use the fitrensemble function in MATLAB, set 'Method' to 'Bag' and 'NumLearningCycles' to 100.

[0110] ② Analysis of LOLR: The model was trained using five secondary indicators as input and LOLR as output. The predictorImportance function was used to calculate feature importance. The results show that RE-PRC (Renewable Energy Peak Shaving Characteristics) has the highest importance score.

[0111] ③ Analysis of RECR: Using five secondary indicators as input and RECR as output, the same analysis was performed. The results show that NLPVR (Net Load Daily Maximum Peak-to-Valley Ratio) has the highest importance score, far exceeding other indicators.

[0112] ④ Identification Results: Therefore, this step identifies RE-PRC and NLPVR as key secondary coupling indicators that lead to supply risk and absorption risk, respectively.

[0113] GSA cross-validation: ① Input selection: To conduct effective GSA, three secondary indicators with weak collinearity were selected as input variables: DAREP (daily average new energy penetration rate), RE-PRC, and NLPVR.

[0114] ② Proxy Model Construction: Construct an ensemble neural network proxy model. Set the ensemble size to 10, the random seed rng to 123, and the network structure to [10, 5].

[0115] ③GSA execution: Perform Sobol' analysis with a basic sample size N of 1024. Calculate the first-order (FSI) and total-order (TSI) sensitivity indices of each input to LOLR and RECR.

[0116] ④ Verification conclusion: FSI reflects the input variables The effect of TSI on system output reflects the input variables. as well as The combined effect of the interaction with all other input random variables on the system output can be used to reflect the non-superposition characteristic of the input's influence on the output. GSA results show that the peak-shaving characteristics of new energy sources have the largest FSI for LOLR, and the daily maximum peak-valley difference rate of net load has the largest FSI for RECR, indicating that they have the largest individual impact on the primary indicators, verifying that they are the key indicators in the source-load coupling index. However, the TSI of other indicators is not 0 and is even greater than that of the key indicators, indicating that other indicators do have an impact on the primary indicators; their impact is reflected in the primary indicators through their interaction with the key indicators.

[0117] 3. In-depth tracing and source determination of uncertainty.

[0118] (1) In-depth source tracing: ① Model configuration: The fitrensemble function in MATLAB is still used, with 'Method' set to 'Bag' and 'NumLearningCycles' set to 100.

[0119] ② Source analysis for RE-PRC: Train a random forest model using all tertiary indicators as input and RE-PRC as output. Evaluate the importance of each tertiary indicator.

[0120] ③ Source analysis for NLPVR: The same analysis is performed with all three levels of indicators as input and NLPVR as output.

[0121] (2) Results analysis and transmission path construction: Based on the importance ranking results of the two models above, the example identified RE-ICR (renewable energy installed capacity share), SPUH (daily photovoltaic power generation utilization hours), and DPLD (daily peak load duration) as the top three key tertiary indicators in terms of overall contribution. This result indicates that: ①The proportion of new energy installed capacity is the most important indicator among the three levels of indicators. The large-scale integration of new energy is the primary factor that increases uncertainty.

[0122] ② The number of hours of photovoltaic power generation utilization is the second most important indicator among the three levels of indicators, indicating that the uncertainty of photovoltaic power generation has a significant impact. This is consistent with the current situation of large-scale distributed photovoltaic grid connection in Shandong Province.

[0123] Based on this, a clear path of uncertainty transmission can be constructed, for example: "The increase in RE-ICR (a tertiary indicator) is the primary factor leading to the increase in NLPVR (a secondary indicator), while the increase in NLPVR is the key driving force leading to the deterioration of RECR (a primary indicator). (3) Determination of sources of uncertainty: Among the three key tertiary indicators identified, RE-ICR and SPUH are indicators on the new energy side, while DPLD is an indicator on the load side. Based on the quantity (2:1) and cumulative importance weight, it can be concluded that in this example, the strong uncertainty faced by the system originates more from the basic characteristics of the source side.

[0124] Through the above steps, the embodiments of the present invention achieve a comprehensive characterization of uncertainties on both the source and load sides of the power system, accurate identification of key factors, and clear tracing of transmission paths, providing strong technical support for the power grid to cope with the challenges of uncertainty.

[0125] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0126] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This computer software product, stored in a storage medium, includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0127] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for characterizing and tracing the source-load dual-sided uncertainty in a power system, characterized in that, include; Data from several types of power systems is acquired and then classified into several indicators, including primary, secondary, and tertiary indicators. Acquire data for primary indicators across multiple time segments, set several judgment thresholds for the data, and classify the current system operating status based on the judgment thresholds. If the system is classified as a definite state, no action is taken. If the classification is uncertain, then several first machine learning models are established. Each first machine learning model takes a certain first-level indicator as its unique output and all second-level indicators as its input. The feature importance score array of each second-level indicator output by each first machine learning model with respect to a certain first-level indicator is obtained. The second-level indicator corresponding to the feature importance score with the largest score in the feature importance score array is selected as the key second-level indicator. Then, several second machine learning models are established. Each second machine learning model takes a certain key secondary indicator as its unique output and all tertiary indicators as its input. The feature importance score array of each tertiary indicator output by each second machine learning model with respect to a certain key secondary indicator is obtained. The secondary indicator corresponding to the feature importance score with the largest score in the feature importance score array is selected as the key tertiary indicator. Output key tertiary indicators as sources of uncertainty in the power system.

2. The method for characterizing and tracing the source-load dual-sided uncertainty of a power system according to claim 1, characterized in that, The primary indicators include load shedding rate, expected load shedding, wind and solar curtailment rate, and expected renewable energy curtailment. The secondary indicators include the daily maximum peak-valley difference rate of net load, the peak-shaving characteristics of new energy, the matching degree of source-load fluctuation, the daily average new energy penetration rate, and the new energy penetration rate during the peak load period; The three-level indicators include new energy-side indicators and load-side indicators. The new energy-side indicators include the proportion of new energy installed capacity, the accuracy of wind and solar power daily output forecast, the number of wind and solar power daily utilization hours, and the wind and solar power hourly output change rate. The load-side indicators include the daily total load power forecast deviation rate, the hourly load change rate, the daily maximum peak-valley difference rate, and the daily peak duration.

3. The method for characterizing and tracing the source-load dual-sided uncertainty of a power system according to claim 2, characterized in that, The step of setting a judgment threshold and classifying the current system operating state based on the judgment threshold includes: When both the load shedding rate and the wind and solar curtailment rate are greater than their respective preset thresholds, the condition is classified as a state of strong uncertainty. When only the load loss rate is greater than its preset threshold, it is classified as a load-side uncertainty state; When only the wind curtailment rate exceeds its preset threshold, it is classified as a source-side uncertainty state. When neither of them is greater than their respective preset thresholds, it is classified as a deterministic state.

4. The method for characterizing and tracing the source-load dual-sided uncertainty of a power system according to claim 3, characterized in that, The machine learning model is a random forest model. Based on random forest, a score reflecting the importance of features is obtained by perturbing the out-of-bag data that did not participate in the training of the decision tree and then calculating the change in classification accuracy.

5. The method for characterizing and tracing the source-load dual-sided uncertainty of a power system according to claim 4, characterized in that, The load failure rate includes: In the formula, For the first The amount of load loss at each point in time. For the first The actual net load at each point in time This represents the operating capacity of the thermal power units on that day. This represents the operating capacity of other generating units for the day. This represents the maximum charge and discharge power of the energy storage. For the first The actual load at each point in time For the first The actual output power of wind power at a given time point For the first The actual output power of photovoltaics at each point in time. The load shedding rate, 0-1 auxiliary variables for judgment, This corresponds to the maximum time.

6. The method for characterizing and tracing the source-load dual-sided uncertainty of a power system according to claim 5, characterized in that, The expected amount of abandoned renewable energy includes: In the formula, For the first The amount of wind and solar power curtailed at each point in time. This represents the minimum output power of the thermal power unit. This is the minimum output power for other units. This is in response to expectations regarding the amount of electricity that has been wasted from renewable energy sources.

7. The method for characterizing and tracing the source-load dual-sided uncertainty of a power system according to claim 6, characterized in that, The average daily new energy penetration rate includes: The renewable energy penetration rate during peak load periods includes: In the formula, This represents the average daily penetration rate of new energy sources. For the penetration of renewable energy, For the corresponding time, For the corresponding year.

8. The method for characterizing and tracing the source-load dual-sided uncertainty of a power system according to claim 7, characterized in that, The maximum daily peak-to-valley difference rate of net load includes: In the formula, The maximum peak-to-valley difference rate of net load on a given day. To correspond to a certain day.

9. The method for characterizing and tracing the source-load dual-sided uncertainty of a power system according to claim 8, characterized in that, The daily utilization hours of wind and solar power include: In the formula, This refers to the number of hours of wind power generation. For the first Predicted wind turbine output at each point in time. The number of hours of photovoltaic power generation utilization. For the first Predicted photovoltaic output values ​​at specific time points.

10. A power system source-load two-sided uncertainty characterization and tracing system, used to execute the power system source-load two-sided uncertainty characterization and tracing method according to any one of claims 1-9, characterized in that, include: The status judgment module is configured to acquire data from several types of power systems, classify the data into several indicators, including primary, secondary, and tertiary indicators; acquire data of primary indicators at multiple time points, set judgment thresholds for several data points, and classify the current system operating status based on the judgment thresholds. If the system is classified as a definite state, no further action is taken. The source tracing module is configured to, if classified as an uncertain state, establish several first-level machine learning models. Each first-level machine learning model takes a specific primary indicator as its unique output and all secondary indicators as its input. It then obtains a feature importance score array for each secondary indicator output by the first-level machine learning model with respect to a specific primary indicator. The secondary indicator corresponding to the highest feature importance score in the feature importance score array is selected as the key secondary indicator. Next, it establishes several second-level machine learning models. Each second-level machine learning model takes a specific key secondary indicator as its unique output and all tertiary indicators as its input. It then obtains a feature importance score array for each tertiary indicator output by the second-level machine learning model with respect to a specific key secondary indicator. The secondary indicator corresponding to the highest feature importance score in the feature importance score array is selected as the key tertiary indicator. The key tertiary indicator is output as the source of uncertainty in the power system.

Images