Generation method, device and equipment of electric power system operation mode scene

Abstract

According to the technical scheme provided by the embodiment of the invention, the dynamic deduction model is constructed by integrating multi-dimensional uncertainty factors such as new energy output and load change, the limitation of a traditional method is broken through, the capability of coping with the uncertainty of a complex operation environment of a power system is improved, and a support is provided for coping with system fluctuation under a high proportion of new energy; a mode scene is efficiently generated in combination with fast load flow calculation and an optimization scheduling module, then classification is carried out through an FCM algorithm, typical operation modes are accurately extracted, redundancy analysis is reduced, the efficiency and precision of scene generation are optimized, and a focusing basis is provided for decision making; based on a time-sharing zoning safety risk index system, a scene is evaluated from multiple dimensions, the weak link identification capability is enhanced, problems can be accurately positioned, making of a prevention and control strategy is assisted, and the safety and stability level of the system is improved; meanwhile, due to standardized processing of multi-source data and application of related technologies, the quality of input data and the fitting degree of the model to actual working conditions are guaranteed, and a foundation is laid for reliability of deduction results.

Description

Technical Field

[0001] This application provides embodiments in the field of power system technology, and particularly relates to a method, apparatus, and equipment for generating power system operation mode scenarios. Background Technology

[0002] As the proportion of new energy sources in the power system continues to increase, load characteristics are becoming increasingly complex. At the same time, situations such as hydropower dispatch and interaction with external power sources are also increasing, leading to a significant increase in uncertainties faced by the system.

[0003] Traditional power system analysis methods exhibit limitations in addressing fluctuations in renewable energy output, load changes, grid equipment status adjustments (such as N-1 faults and temporary outages), and uncertainties in external power supply. They struggle to fully adapt to the demands of rapidly changing operational scenarios and pose challenges to the derivation of vulnerable modes. These limitations place higher demands on power system security risk assessments in both planned and operational states, and also create potential risks to the safe and stable operation of the power system. Summary of the Invention

[0004] The embodiments of this application provide a method, apparatus, and equipment for generating power system operation mode scenarios, which solves the technical problem that existing power system analysis methods have limitations in generating power operation mode scenarios when dealing with uncertain factors, and are difficult to adapt to the generation requirements of operation mode scenarios with rapidly changing boundaries.

[0005] In a first aspect, embodiments of the present invention provide a method for generating power system operation mode scenarios, the method comprising:

[0006] Acquire multi-source operational data of the target power system, wherein the multi-source operational data refers to a heterogeneous data set collected from different equipment areas of the target power system;

[0007] The multi-source operational data is standardized to obtain a standardized input dataset;

[0008] The method for extracting typical high-risk moments based on the tSNE-FCM algorithm uses the standardized input dataset to construct a power system extrapolation model that considers multiple uncertainties, including at least the fluctuations in new energy output, load changes, grid equipment status adjustments, and uncertainties in external power supply.

[0009] Based on the power system simulation model, a sequence of scenario sequences is generated using a simulation algorithm. Based on the sequence of scenario sequences, the IMLE method is used to generate scenario sequences under rapidly changing boundaries.

[0010] The generated scenario is evaluated based on a pre-defined time-sharing and partitioned security risk indicator system.

[0011] Secondly, embodiments of the present invention also provide an apparatus for generating power system operation mode scenarios, the apparatus comprising:

[0012] The data acquisition unit is used to acquire multi-source operating data of the target power system, wherein the multi-source operating data refers to a heterogeneous data set collected from different equipment areas of the target power system;

[0013] The data processing unit is used to standardize the multi-source operating data to obtain a standardized input dataset;

[0014] The model extrapolation unit is used for the typical high-risk moment extraction method based on the tSNE-FCM algorithm. It uses the standardized input dataset to construct a power system extrapolation model that considers multiple uncertainties, including at least the fluctuation of new energy output, load changes, grid equipment status adjustment, and uncertainty of external power supply.

[0015] The scenario generation unit is used to generate a sequence of scenario modes based on the power system simulation model using a simulation algorithm, and to generate scenario modes under rapidly changing boundaries using the IMLE method based on the sequence of scenario modes.

[0016] The scenario evaluation unit is used to evaluate the generated scenario based on a preset time-division and partition-based security risk indicator system.

[0017] Thirdly, embodiments of the present invention also provide a device for generating power system operation mode scenarios, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the power system operation mode scenario generation method described in the first aspect.

[0018] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program thereon, which, when executed in a computer, causes the computer to execute the method for generating power system operation mode scenarios described in the first aspect.

[0019] The technical solution provided in this application integrates multiple dimensions of uncertainties, such as renewable energy output and load changes, to construct a dynamic simulation model. This breaks through the limitations of traditional methods, enhances the ability to cope with uncertainties in the complex operating environment of the power system, and provides support for dealing with system fluctuations under a high proportion of renewable energy. By combining the efficient generation of scenarios through fast power flow calculation and optimized scheduling modules, and then classifying them using the FCM algorithm, typical operating modes are accurately extracted, reducing redundant analysis and optimizing the efficiency and accuracy of scenario generation, providing a focused basis for decision-making. Based on a time-sharing and zone-based safety risk index system, scenarios are evaluated from multiple dimensions, strengthening the ability to identify weak links, accurately locating problems, and assisting in the formulation of prevention and control strategies, thereby improving the level of system safety and stability. Simultaneously, the standardized processing of multi-source data and the application of related technologies ensure the quality of input data and the model's fit to actual operating conditions, laying the foundation for the reliability of the simulation results. Attached Figure Description

[0020] Figure 1 This is a flowchart of a method for generating a power system operation mode scenario provided in an embodiment of this application;

[0021] Figure 2 This is a distribution feature diagram of the target power system operating time in the t-SNE nested space according to an embodiment of the present invention;

[0022] Figure 3 This is a scene result diagram showing the method of generating boundaries under rapid changes provided in the embodiments of the present invention;

[0023] Figure 4 This is a comparison chart of transformer overload time and longest continuous overload time provided in an embodiment of the present invention;

[0024] Figure 5 This is a structural diagram of a device for generating power system operation mode scenarios provided in an embodiment of the present invention;

[0025] Figure 6 This is a structural diagram of a device for generating power system operation mode scenarios provided in an embodiment of the present invention. Detailed Implementation

[0026] The present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0027] Figure 1 This is a flowchart of a method for generating a power system operation mode scenario provided in an embodiment of this application. The method can be executed by a device for generating a power system operation mode scenario. The device can be implemented by software and / or hardware and can be configured in an electronic device such as a computer.

[0028] like Figure 1As shown, the technical solution provided in this application includes the following steps:

[0029] S101, acquire multi-source operation data of the target power system. Multi-source operation data refers to a heterogeneous data set collected from different equipment areas of the target power system.

[0030] Optionally, S101 specifically includes:

[0031] Obtain historical power output data for wind farms and photovoltaic power plants;

[0032] Obtain curve data of power system load, including actual and predicted loads of load centers in each region of the target power system;

[0033] Acquire the recorded power supply operating parameters of the target power system, including at least the start / stop status, minimum and maximum output limits, ramp rate, and start / stop cost.

[0034] Acquire operational data of the energy storage system, which includes at least the energy storage capacity, charge and discharge power limits, and state of charge curve;

[0035] Obtain status information of power grid equipment, transaction contracts for electricity from outside the region, operating status of power grids in other regions, water inflow forecast data and dispatch rules for hydropower stations.

[0036] Specifically, historical output data from wind farms and photovoltaic power plants should be collected, with a data resolution at the hourly level, covering the complete operating conditions within a typical meteorological year. Recorded power system operating parameters should be obtained, including those for conventional power sources such as coal-fired units. Status information for grid equipment should include N-1 fault probability and temporary outage triggering conditions.

[0037] S102, standardize the multi-source running data to obtain a standardized input dataset.

[0038] Optionally, S102 specifically includes: performing time alignment on multi-source running data from different sources and sampling frequencies, uniformly converting it to an hourly time scale, and obtaining a standardized input dataset.

[0039] Specifically, multi-source operational data from different sources and sampling frequencies are time-aligned and uniformly converted to an hourly time scale to ensure consistency and comparability of various time-series data on the time axis; for missing data segments, a weighted moving average method is used to complete them; for outliers that deviate from historical statistical patterns, an anomaly detection method based on local density is used for identification.

[0040] Optionally, time alignment is performed on multi-source runtime data from different sources and sampling frequencies, and the data is uniformly converted to an hourly time scale to obtain a standardized input dataset, including:

[0041] For multi-source operational data with a sampling frequency higher than the target frequency, based on the formula Downsampling is performed using the sliding window averaging method, where x t This represents downsampled observations over a specified time period. The original sampling point observations are defined within a specified time period, where n is the number of sampling points within the specified time period, t is the sampling time, and i = 1, 2, 3, ...;

[0042] For multi-source operational data with a sampling frequency lower than the target frequency, based on the formula Upsampling is performed using linear interpolation, where y t The observed values ​​obtained through interpolation. and These are the observation values ​​at adjacent known time points;

[0043] For missing data segments, based on the formula The weighted moving average method is used for completion, where, These are estimates for the missing time points. These are the weighting coefficients. These are the valid observation values ​​located within the time window before and after the missing point;

[0044] For outliers that deviate from historical statistical patterns, based on the formula An anomaly detection method based on local density is used for identification and processing, where D(p) t () indicates a point in time. The local density, p t For the current observation value, is the time point observation within the neighborhood, and m is the size of the neighborhood.

[0045] It should be noted that if D(p) t If the value exceeds the set threshold, it is considered an outlier. For isolated outliers, the mean of the nearest neighbor values ​​is used as a substitute; for continuous outlier segments, they are directly removed and excluded from subsequent modeling.

[0046] S103 is a method for extracting typical high-risk moments based on the tSNE-FCM algorithm. It uses a standardized input dataset to construct a power system simulation model that considers multiple uncertainties, including at least the fluctuations in new energy output, load changes, grid equipment status adjustments, and uncertainties in external power supply.

[0047] Optionally, constructing a power system simulation model that considers multiple uncertainties includes:

[0048] Based on historical data and predictive models, a probability distribution sample of new energy output is generated. A Monte Carlo simulation method combined with meteorological data and load characteristic analysis is used to generate a continuous variation sequence of new energy output. For load characteristics, a Long Short-Term Memory (LSTM) Network (LSTM) model is used to fit the load change trend, taking into account seasonality and time-period differences. For hydropower output, a stochastic output model is established by combining inflow forecast data and scheduling rules. For power from outside the region, its output range and variation patterns are determined based on power trading contracts and the operating status of the external power grid. For the status of power grid equipment, a discrete event simulation method is used to set the triggering conditions and probabilities of N-1 faults and temporary outages. By recording system operation data at each moment, a sequence containing multiple scenarios is finally generated.

[0049] Considering the various operational bottlenecks in new energy systems, such as peak-shaving capacity constraints and cross-sectional transmission limitations, which often affect the system state in combination with different indicators, this invention introduces a method combining tSNE (t-DistributedStochastic Neighbor Embedding, a nonlinear technique for dimensionality reduction and visualization of high-dimensional data) and FCM (Fuzzy C-Means) clustering to identify operating patterns and compress structures for high-risk candidate moments.

[0050] Optionally, typical high-risk moment extraction methods based on the tSNE-FCM algorithm include:

[0051] Based on the time-series simulation data of the target power system, key operational indicators of the target power system are extracted. These key operational indicators include at least: wind and solar curtailment, main section power flow load factor, reserve capacity, and thermal power peak-shaving margin. A high-dimensional state vector characterizing the operational risk features of the target power system is constructed based on the key operational indicators. The high-dimensional state vector is mapped to a low-dimensional nested space using tSNE technology to preserve the nonlinear coupling structure of various bottleneck mechanisms. At the same time, the FCM clustering method is combined to identify time clusters where the similarity of risk structures meets the set similarity, and the data within the corresponding time clusters are extracted as operational samples.

[0052] Specifically, this method first extracts several key operational indicators, such as wind and solar curtailment, main section power flow load factor, reserve capacity, and thermal power peak-shaving margin, based on time-series simulation data of the target power system, and constructs a high-dimensional state vector characterizing the system's operational risk features. Then, tSNE technology is used to map this vector to a low-dimensional nested space to preserve the nonlinear coupling structure of various bottleneck mechanisms. On this basis, FCM clustering is combined to identify time clusters with similar risk structures, thereby extracting representative operational samples. This method balances the multidimensional complexity of risk indicators with the abstract representation of typical scenarios, providing a clearly structured and controllable set of representative time moments for addressing the problem of limited renewable energy consumption.

[0053] After generating the scene sequences, FCM clustering analysis is used to classify and analyze the scenes: First, the generated scene data is dimensionality-reduced by using the t-SNE algorithm to map the high-dimensional data to a two-dimensional space for easier visualization and analysis; then, the FCM clustering algorithm is applied to classify the dimensionality-reduced data, and the formula is as follows:

[0054] ;

[0055] in Indicates the first The sample belongs to the first Membership degree of a class For sample data, As cluster center, The fuzzy index is used. Through multiple iterations, the cluster centers and membership matrix are updated to obtain the final scene classification result. This process helps identify similar and typical scenes, providing a basis for subsequent analysis.

[0056] Among them, the FCM clustering algorithm is an unsupervised learning algorithm developed based on the traditional K-means clustering algorithm. As an extension of K-means, it differs from the hard clustering method of K-means by introducing the concept of fuzzy membership degree to achieve flexible fuzzy partitioning. Therefore, it has greater flexibility and robustness, and is widely used among many clustering algorithms with excellent clustering results. Specifically, the core idea of ​​FCM is to divide all samples into several fuzzy subsets. Each sample has a membership degree with each fuzzy subset, which is used to characterize the similarity between the sample and the subset. By calculating the membership degree value of a sample to each subset, the category to which the sample belongs is determined, thereby realizing the clustering processing of the data. Its advantage lies in its ability to handle the fuzzy classification problem of sample points and more accurately characterize the fuzzy classification relationship between samples.

[0057] The main implementation steps of FCM are as follows: First, initialize the cluster centers and membership matrix. The initial values ​​can be randomly generated or set based on experience, and the fuzziness parameters, iteration termination conditions, and number of clusters are given. Second, calculate the cluster centers of each fuzzy subset based on the membership matrix, that is, the representative samples of each subset. Third, based on the currently calculated cluster centers, recalculate the membership degree of each sample to each fuzzy subset to complete the update of the membership matrix. Finally, judge the difference between the current membership matrix and the result of the previous iteration. If the preset iteration termination condition is met, stop the iteration; otherwise, return to the second step and repeat the iteration process.

[0058] The calculation of cluster centers is as follows:

[0059] ;

[0060] ;

[0061] Where, m ab x represents the membership degree of the a-th sample to class b; i d is the sample point; m is the membership factor; n is the total number of samples; c is the number of clusters; t is the current iteration round; d ab Let be the Euclidean distance from the a-th sample to the b-th cluster center.

[0062] Using multi-dimensional operational indicators corresponding to high-risk moments of the target power system as clustering input variables, a clustering structure characterizing typical operational features is constructed. To comprehensively identify typical high-risk scenarios where renewable energy output is limited, this embodiment of the invention further extracts operational characteristics covering both transmission network bottlenecks and power source-side peak-shaving limitations as supplementary features, based on two key physical constraints of renewable energy curtailment.

[0063] To characterize the impact of transmission capacity limitations on renewable energy output, a "renewable energy consumption per unit" indicator is introduced. This indicator measures the incremental renewable energy output that the system can release after a specific line or main transformer increases its transmission capacity by one unit. Its expression is as follows:

[0064] ;

[0065] Among them, K res，i K represents the installed capacity (wind power or solar power) of the i-th renewable energy power station; total，i This represents the total installed capacity of new energy sources at the station; △Q overlimit，i This refers to the amount of electricity wasted at the station under transmission-constrained conditions; PTDF -1This is the inverse matrix of the power flow distribution factor, reflecting the contribution of the injected power from the power station to the power flow of the target line. The larger this value, the stronger the physical constraint of the current line or main transformer on the absorption of new energy, and the higher the system benefit of marginal capacity expansion. It is an important structural indicator for identifying network bottlenecks and the moment when new energy is restricted.

[0066] Secondly, to reflect the constraints of power supply-side regulation capacity, a system peak-shaving margin index R is introduced to measure whether the operating state of thermal power units is close to the minimum output boundary, defined as follows:

[0067] ;

[0068] in, This represents the minimum technical output of the i-th thermal power unit. For real-time power output, N represents the number of thermal power units participating in regulation. A larger R indicates a smaller system peak-shaving capacity, making it more difficult for renewable energy output to be fully utilized during peak periods. By constructing a clustering input space for high-risk candidate moments using the above supplementary features, it is possible to effectively distinguish renewable energy constraint modes constrained by different mechanisms.

[0069] To visually represent the structural distribution characteristics of 8760 operational samples throughout the year in a multidimensional risk indicator space, this embodiment of the invention employs the tSNE nonlinear dimensionality reduction algorithm, based on the normalized time-level state vector, to uniformly map high-dimensional data to a two-dimensional nested space. This algorithm preserves the local correlations of samples in the high-dimensional space, enabling the low-dimensional distribution to more accurately reflect the risk characteristic correlations at different operational times. Figure 2 This is a distribution feature map of the target power system operating time in the t-SNE nested space according to an embodiment of the present invention. Green, yellow and red respectively mark the samples of mild, moderate and extreme risk levels, clearly showing the clustering pattern of samples of each level in the low-dimensional space: extreme risk samples form dense clusters, moderate risk samples are distributed around them in a gradient, and mild risk samples are relatively scattered. This structural difference intuitively confirms the pattern separability of the multidimensional index state vector, providing a solid structural foundation for subsequent cluster analysis and typical sample extraction.

[0070] It should be noted that in practical engineering applications, tSNE dimensionality reduction is usually only performed on the selected high-risk candidate time periods, without needing to cover the entire year's samples. This is because the regular operating periods account for over 90% and mostly exhibit low-risk, normal characteristics. Including them in dimensionality reduction would significantly dilute the structural density of high-risk samples, leading to a decrease in the clustering algorithm's sensitivity to subtle risk differences. Meanwhile, the mild, moderate, and extreme risk samples fully cover core bottleneck mechanisms such as limited transmission sections and insufficient peak-shaving capacity, and their representativeness is sufficient to support the compression of typical time periods and the extraction of operating modes. Therefore, nested modeling of a subset of high-risk samples can reduce the computational load by more than 80% while ensuring analytical focus, making it the optimal strategy that balances efficiency and effectiveness.

[0071] After obtaining the separable distribution of high-risk moments in the nested space, the FCM clustering algorithm is further used to conduct cluster analysis on the dimensionality-reduced high-risk sample set, aiming to extract typical operating moments with strong representativeness and significant structural differences. Compared with hard clustering algorithms, the core advantage of FCM lies in constructing a soft membership relationship between samples and clusters—that is, each high-risk moment has a membership value between 0 and 1 for multiple clusters, rather than a simple "either / or" division. This characteristic allows it to not only clearly identify typical clusters dominated by various risks, such as the transmission section overload-dominated cluster and the peak-shaving capacity-dominated cluster, but also accurately capture the fuzzy state of the transition zone between different risk mechanisms: for example, when the system faces both renewable energy fluctuations and local line load limits at the same time, the membership degree of the sample to the "renewable energy absorption limited" cluster and the "network congestion" cluster may both be at a medium level, thus intuitively revealing the complex operating characteristics under the coupling of multiple mechanisms.

[0072] By employing this flexible partitioning method, FCM can effectively avoid the arbitrary classification of boundary samples by hard clustering, and better aligns with the actual characteristics of the power system's "gradual change in risk mechanisms." During the clustering process, by iteratively optimizing the objective function (minimizing the weighted distance from the sample to the cluster center), several clearly structured risk clusters can be formed. The central sample of each cluster serves as a typical representative of that type of risk characteristic, providing precise scenario support for subsequent targeted regulation strategies.

[0073] S104, Based on the power system simulation model, a sequence of scenario sequences is generated through simulation algorithms. Based on the scenario sequences, the IMLE method is used to generate scenario sequences under rapidly changing boundaries.

[0074] Specifically, by constructing a power system simulation model that considers multiple uncertainties, a sequence of scenario sequences is generated. The core of this method lies in the coordinated operation of fast power flow calculation and optimal scheduling. The simulation algorithms include the following: Forward-Backward Sweep Method, Newton-Raphson Method, and Improved Euler Method. In this embodiment, the fast power flow calculation uses the improved Newton-Raphson method to simplify the Jacobian matrix calculation, thereby improving efficiency. Optimal scheduling aims at system safety constraints and economic efficiency, setting objective functions and constraints—the objective function encompasses generation costs, network loss costs, and curtailment penalty costs, while the constraints include equipment capacity limitations, voltage stability requirements, and power flow balance conditions.

[0075] Implicit Maximum Likelihood Estimation (IMLE) is a widely used deep learning method in generative modeling in recent years. Its excellent convergence and high-quality generated samples demonstrate significant applicability in complex distribution learning tasks. Compared to common generative models such as VAE, GMMN, and GAN, IMLE's core feature lies in its ability to construct a stable and efficient non-adversarial learning mechanism during training by minimizing the minimum Euclidean distance between actual and generated samples. This characteristic makes it particularly suitable for high-fidelity simulation tasks in multidimensional uncertain scenarios.

[0076] 1. Optimization loss function for IMLE

[0077] The core idea of ​​IMLE-based scene generation is to train a scene generator model using typical high-risk state vectors extracted from historical operational data. This model can map random noise following a prior distribution into multi-dimensional high-risk state scenes that conform to the actual operational distribution characteristics. Essentially, the generator is an implicit, complex mapping function between random disturbances and system operational risk states, used to generate representative risk disturbance samples. It should be noted that the prior distribution used in this embodiment is not derived from the actual error distribution of operational risk indicators, but rather from an easily sampled distribution (such as the common Gaussian distribution) to ensure generation efficiency and model generalization ability.

[0078] Based on this, the constructed loss function uses the minimum distance between the actual high-risk state vector and the generated scene as the similarity criterion to guide the parameter update and optimization of the scene generator, so that the generated samples are as close as possible to the real distribution structure of historical high-risk operating states in the index space.

[0079] To ensure sampling efficiency and model generalization ability, this study chooses the simple and easy-to-sample Gaussian distribution as the prior distribution without loss of generality. Its sampling and mapping process can be expressed as follows:

[0080] ;

[0081] In the formula: K is Gaussian noise; Y is the newly generated system operating state sample; g(Z) is the trained state vector generator.

[0082] The relationship between the Gaussian noise K and the actual operating state distribution X in the generative model is characterized using a distributional approach:

[0083] ;

[0084] Here, q(X) represents the distribution of high-risk operating state scenarios generated by the generator; q(K) is the prior Gaussian distribution used as input; and q(X|K) is any conditional Gaussian distribution. Clearly, the integral form of this equation can characterize various complex distributions and theoretically fit any probability distribution, naturally encompassing the probability distribution structure of system operating risk indicators under different scenarios. After determining this distribution form, the unknown parameters can be solved through maximum likelihood estimation, enabling the automatic generation of typical high-risk scenarios and uncertainty extension modeling of the system. This process leverages the ability of statistical learning to approximate complex distributions, accurately capturing the coupling characteristics of risk indicators in historical data (such as the nonlinear correlation between abandoned power and transmission section load rate), and achieving reasonable extrapolation of unobserved boundary scenarios through parameter optimization, providing a quantitative modeling method to cover potential extreme operating states of the system.

[0085] Let the actual data distribution be p(X), then the objective of maximum likelihood estimation can be expressed as:

[0086] ;

[0087] In the formula: Let lgq(X) be the expectation of p(X).

[0088] The core optimization objective of IMLE is to maximize the matching probability between the generated scene and the real data, which essentially falls under the category of maximum likelihood estimation. However, since neural network parameter optimization typically relies on gradient descent (suitable for minimizing objectives), the original maximization objective needs to be transformed into an equivalent minimization form. Specifically, IMLE transforms the maximum likelihood estimation objective into a loss function basis that can be directly used for gradient descent by taking the negative logarithm of the maximum likelihood estimation objective—this is because the monotonicity of the logarithmic function makes "maximizing likelihood" and "minimizing the negative log-likelihood" completely equivalent.

[0089] Building upon this foundation, to enhance the structural matching between generated samples and real data, IMLE further introduces a minimum distance matching mechanism between Gaussian noise perturbation and actual samples: after the prior Gaussian noise is mapped by the generator to obtain the generated scene, the minimum Euclidean distance between the generated scene and the real state vector is calculated. Then, by taking the expectation of all samples, this distance metric is incorporated into the construction of the loss function. Ultimately, the loss function expression derived above retains the requirement of maximum likelihood estimation for distribution matching while strengthening the local structural consistency between generated samples and real data through distance constraints, providing a clear and effective objective guide for gradient descent optimization. Its specific form is as follows:

[0090] ;

[0091] During the training of IMLE, each iteration inputs m actual system operating state vectors (such as combinations of indicators from historical high-risk moments), denoted as... and n Gaussian noise samples These noise samples, after being mapped by the generative network, form candidate system state scenarios.

[0092] 2. The composition structure of a scene generator

[0093] Convolutional neural networks (CNNs), with their strong feature extraction capabilities, are used to build IMLE scene generators. The core of this generator is mapping low-dimensional Gaussian noise into a high-dimensional high-risk operating state vector of the system. The generator is based on deconvolutional layers and fully connected layers: after inputting low-dimensional Gaussian noise, it is progressively upsampled through 3-5 deconvolutional layers to increase the dimensionality while preserving the correlation between indicators; then, it is precisely mapped to the target dimension through 2-3 fully connected layers. The deconvolutional layers are activated with LeakyReLU, the fully connected layers use ReLU in the middle, and the last layer has no activation to cover the actual range. During training, the generated vector and real samples are substituted into the IMLE loss function, and the Adam optimizer is used to iteratively update the parameters, ultimately generating a high-fidelity, high-risk scene. It mainly consists of deconvolutional layers and fully connected layers; the specific steps are as follows:

[0094] First, the input features and weight matrix Perform a deconvolution operation to obtain a mixed feature matrix. Then, combine this matrix with the bias vector. Perform summation and apply the activation function. The output vector of this layer is obtained after processing. As shown in the following formula: In the formula, " " indicates deconvolution operation.

[0095] The operation process of a fully connected layer is similar to that of a deconvolutional layer, that is, the weight matrix... and input features Multiply them, then add the product to the bias vector. Perform summation and apply the activation function. Obtain the output vector by performing a nonlinear transformation. As shown in the following formula.

[0096] ;

[0097] 3. Modeling process for scenarios with rapidly changing boundaries

[0098] The numerous high-risk operational scenarios generated by the IMLE model are essentially unstructured, disordered sets unrelated to actual scheduling logic. They only reflect overall distribution characteristics but lack a clear correlation with the system's actual boundary operating conditions (such as transmission section limits and peak-shaving capacity thresholds) or predicted patterns (such as load peak-valley trends and renewable energy fluctuation patterns). Therefore, an additional screening mechanism is needed: on the one hand, calculating the feature matching degree between the generated scenarios and the typical high-risk moments (central samples with clear physical mechanisms) obtained from previous clustering, retaining samples that meet the similarity criteria; on the other hand, introducing scheduling boundary constraint verification (such as power flow limits and reserve capacity thresholds) to eliminate abnormal scenarios that do not conform to actual operational logic. Through this dual screening, the disordered states are constrained to the vicinity of the expected typical structure, ensuring both the rationality of scenario distribution and enhancing its physical correlation with actual scheduling, thereby improving its credibility and application value.

[0099] Specifically, let t be the operating time of the target power system, and define... express to The historical operating state vector sequence of the system within the time period, This indicates the future obtained based on existing operating trends or predictive models. to Predicted state sequence for a given time period.

[0100] The IMLE generator takes Gaussian noise as input and produces... Group of candidate high-risk state scenarios Each set of scenes Divided into two sections:

[0101] ;

[0102] ;

[0103] In the formula, , These represent the portion of the generated state aligned with the historical segment and the extended portion representing future risks, respectively.

[0104] To ensure the credibility of the generated scenario in terms of mechanism structure, the following two types of constraints must be met simultaneously: (1) the historical segment of the generated state Compared with the actual historical state They are highly similar; (2) Risk scenarios in the future to State prediction segment of the time period It should be located where some predictions have been made. Within a certain tolerance range, ensure that its variation range is reasonable.

[0105] Therefore, the following constrained minimization problem is constructed to filter the generated scenarios that satisfy the above conditions:

[0106] ;

[0107] The upper and lower limits are given by the following expressions:

[0108] ;

[0109] The parameter α controls the width of the interval formed by the upper and lower limits.

[0110] Considering the high difficulty of solving constrained optimization problems, based on the characteristics of the constraints, a penalty function is constructed using the Big M method commonly used in optimization theory, and the constraints are added to the objective function, thus transforming the problem into a nonlinear unconstrained optimization problem.

[0111] ;

[0112] in, Let M be a unit step function, and M be a penalty factor that is much larger than the upper bound of the principal term error.

[0113] Through the above optimization strategies, the scenario samples generated by IMLE under rapidly changing generation boundaries are structurally closer to the historical evolution path and reasonably fluctuate around the predicted value in terms of trend, thereby constructing a set of random high-risk scenarios with a credible physical mechanism foundation and representative fluctuation characteristics. Figure 3 This diagram illustrates the results of a scenario where the generation boundary changes rapidly, as provided in this embodiment of the invention. Different colors represent different clusters, and the black crosses mark the center points of each cluster. It can be observed that the clustering results exhibit good distribution coverage and aggregation in the dimensionality-reduced space. Each cluster has a compact structure and clear spacing, verifying the effectiveness of the previous feature construction and screening strategies in pattern recognition.

[0114] S105 evaluates the generated scenarios based on a preset time-sharing and partitioned security risk indicator system.

[0115] Optionally, S105 specifically includes:

[0116] Construct a safety risk indicator system that includes the risk of power fluctuations in new energy sources; evaluate the mode and scenario based on the safety risk indicator system, and analyze the relationship between transformer overload time and the longest continuous overload time.

[0117] Specifically, in the security risk indicator evaluation module, the generated scenarios are evaluated based on a time-division and zone-based security risk indicator system. This indicator system covers dimensions such as power flow distribution, voltage stability, frequency stability, and equipment load rate. For each scenario, relevant indicators are extracted and their values ​​are calculated.

[0118] 1. New Energy Power Indicators

[0119] This invention, based on output data processed with a unified time scale, points out that current risk assessment indicators for renewable energy power generation systems, such as wind and solar, largely follow traditional reliability assessment indicator systems. These indicators describe the system risk level before and after renewable energy integration through load shedding rate, expected power shortage, and expected power shortage amount. However, these indicators fail to reflect the individual contribution of renewable energy power plant output to system risk. Unlike traditional approaches, this example correlates risk with wind and solar grid-connected electricity prices, project costs, and wind and solar power utilization hours, measuring risk from the perspective of future investment policies and renewable energy development, rather than conducting risk assessments based on operational modes.

[0120] The risk function used in conventional power system risk assessment is the expected value of power system losses, i.e., the unconditional expectation, which is defined as follows:

[0121] ;

[0122] ;

[0123] Where: X is a random variable representing a certain power system risk index; x is a value of the random variable X, i.e., a specific numerical value; p(t) is the probability density function of the risk index; Q(x) is the cumulative distribution function of the risk index; and V(X) is the risk value of the risk index.

[0124] Based on conventional risk indicators such as system-level power shortage risk and component-level overload risk and load shedding risk, this paper proposes a risk indicator for renewable energy power fluctuation to quantify the contribution of renewable energy access to system risk, thus expanding the risk indicator system.

[0125] Wind Power Variability Risk (WPVR) is the risk of system load loss and wind / solar curtailment caused by insufficient ramp-up capability of conventional generating units due to deviations in renewable power output from planned values, under system conditions that consider random outages of generating units and transmission lines. The risk indicators are defined as follows:

[0126] ;

[0127] ;

[0128] ;

[0129] Where n is the total number of system states; o i Let be the probability of the i-th state of the system; o Inet The system net load is the system load power. I With new energy power w The difference; o ac R represents the actual output of the conventional unit after being limited by the ramp-up capability and output boundary of the renewable energy source; R is the maximum ramp-up rate of the conventional unit; Δt is the ramp-up time; o g max and o g min These represent the maximum and minimum output of a conventional unit.

[0130] If WPVR is positive, it indicates that the system is at risk of insufficient power supply (i.e., there is a demand for load shedding); if WPVR is negative, it indicates that the system is at risk of curtailment of wind and solar power.

[0131] The calculation steps for WPVR are as follows:

[0132] The first step is to calculate the net load based on the current power output plan curve Pg and the predicted value of renewable energy power;

[0133] The second step is to determine the time and predicted wind speed corresponding to each state, obtain the conditional fluctuation value of the new energy power under each state by sampling the joint conditions of k, and then update the net load curve according to the magnitude and sign of the fluctuation value.

[0134] The third step is to consider the random outage of system components and refresh the power generation base value of conventional units at the initial time of each period according to the outage status of the components.

[0135] The fourth step is to calculate the unit output boundary value during the study period, based on the formula, since the ramp rate of the conventional unit is within the maximum ramp rate range and is adjustable, and the unit output can be rescheduled and adjusted at any time.

[0136] The fifth step is to calculate the WPVR value according to the formula and determine the system risk type based on its positive or negative sign.

[0137] The renewable energy power fluctuation risk index is used to assess the coordination between renewable energy power generation and the ramp-up rate of conventional generating units, and to study the system risks caused by the untimely or insufficient ramp-up capacity of conventional generating units in a short period of time. To address the risks of load shedding or wind / solar curtailment, a new power generation plan needs to be formulated, and wind curtailment operations can be implemented by activating standby generating units or coordinating with appropriate market mechanisms. This index can reflect the risk level of renewable energy power fluctuations at the system level, providing a reference for dispatchers to balance grid risks and benefits and to formulate and modify wind and solar grid connection plans.

[0138] 2. System load gap percentage

[0139] Whether the system has sufficient power to meet overall electricity demand during peak load periods is a core indicator for measuring power supply security. The system load gap ratio effectively reflects the robustness of the power grid in terms of power supply configuration and reserve capacity. To quantitatively assess the supply-demand matching situation, this indicator is introduced to describe the relative degree to which the system cannot meet load demand during a specific period, and its definition is as follows:

[0140] ;

[0141] In the above formula, Q DEFICIENY P represents the proportion of the load demand gap. DEMAND For system load requirements, P ACTUAL This represents the actual load that the system needs to meet.

[0142] 3. System reserve gap ratio

[0143] Spinning reserve capacity serves as the first line of defense for ensuring power system frequency stability, and its supply-demand balance directly impacts the system's ability to respond to sudden disturbances such as generator trips and load spikes. This indicator, by precisely quantifying the supply-demand gap of spinning reserve resources, can not only assess the short-term dynamic stability of the system but also reflect the rationality of grid frequency regulation resource allocation. Its definition is as follows:

[0144] ;

[0145] In the above formula, O DEFICIENY P represents the proportion of the spinning reserve demand gap. DEMAND For system spin-off backup requirements, P ACTUAL This is a backup for rotation that the system actually needs to meet.

[0146] 4. System standby gap percentage

[0147] As a crucial support for the long-term regulation capacity of a power system, the supply-demand balance of outage reserves directly affects the system's regulatory margin in responding to issues such as sustained load fluctuations, planned unit maintenance, and fuel supply interruptions. This indicator, by quantifying the actual supply-demand gap in outage reserves, can effectively assess the system's power supply reliability over medium- to long-term timescales. Its definition is as follows:

[0148] ;

[0149] In the above formula, O SHUTDOWN_DEFICIENY The proportion of the demand gap for standby power outages, P SHUTDOWN_DEMAND For system downtime backup requirements, P SHUTDOWN_ACTUAL This is for the actual shutdown backup required by the system.

[0150] 5. System main transformer over-limit rate

[0151] As the core hub for energy conversion between different voltage levels in a power system, the main transformer's overload operation can lead to a significant increase in equipment losses and may also cause faults such as insulation aging and localized overheating, seriously threatening power supply reliability. As a critical link connecting the main grid and the load side, the main transformer's over-limit status directly reflects the capacity bottleneck of the transmission and transformation system. Its over-limit rate is defined as follows: ;

[0152] In the above formula, O OUTOFLIMIT Main variable over-limit rate, P OUTOFLIMIT Main transformer over-limit power, P CAPACITY Main variable capacity.

[0153] 6. System line over-limit rate

[0154] As the core channel for power transmission, the operating status of transmission lines directly affects the safety and stability of the power grid. Power flow distribution on lines is influenced by multiple factors, including system power balance, network topology, and control strategies. Once power flow exceeds limits, it will force a reconfiguration of the power flow, potentially leading to systemic risks such as protection malfunctions and cascading trips. To quantitatively assess line overload conditions, a line overload rate index is introduced, defined as follows:

[0155] ;

[0156] In the above formula, O OUTOFLIMIT P represents the line over-limit rate. OUTOFLIMIT For line over-limit power, P MAX This represents the upper limit of the line's current-carrying capacity.

[0157] 7. System critical channel over-limit rate

[0158] Critical transmission channels, as the core carriers of inter-regional power exchange, typically have their spatial distribution highly overlapping with the system's power flow center. Channel overruns often reflect deeper structural problems such as imbalances in inter-regional power supply distribution, insufficient transmission capacity, or limited dispatch flexibility. To accurately assess inter-regional transmission bottlenecks, the critical channel overrun rate is defined as follows:

[0159] ;

[0160] In the above formula, R INTERFACE_OUTOFLIMIT P represents the critical channel over-limit rate. INTERFACE_OUTOFLIMIT For critical channel over-limit power, P INTERFACE_MAX This represents the maximum power limit for critical channels.

[0161] 8. Failure probability

[0162] Failure probability refers to the probability that a power system will fail or exceed its safe operating range when faced with various uncertainties. For example, the uncertainty of new energy power generation may lead to grid instability, while improper load and hydropower dispatch may cause power outages.

[0163] ;

[0164] Where: O(Wind) is the probability of system failure caused by wind energy fluctuations; O(Hydro) is the probability of failure caused by improper hydropower dispatch; and O(Grid Failure) is the probability of failure caused by grid equipment failure.

[0165] Failure probability helps power systems understand the likelihood of system failure under various uncertainties. For example, when grid equipment is old or poorly maintained, the probability of equipment failure is higher; in regions heavily reliant on wind and solar power, if the output of new energy sources fluctuates significantly, the probability of system failure will also increase accordingly. By calculating failure probabilities, power companies can strengthen the monitoring and optimization of potential failure points, thereby reducing the incidence of failure events.

[0166] 9. Evaluate transformer overload time and longest continuous overload time based on safety evaluation indicators.

[0167] Figure 4 This is a comparison chart of transformer overload time and longest continuous overload time provided in an embodiment of the present invention. Figure 4 As shown, 10 sets of transformer data are displayed. Blue represents "transformer overload time," and orange represents "longest continuous overload time." Overall, "transformer overload time" is generally higher than "longest continuous overload time." The degree of difference varies among different data sets (sets 1 to 10), reflecting the complex operating conditions of transformer overload in power system operation.

[0168] Power system load is significantly affected by user behavior and industry production patterns. For example, in Groups 1-3, summer air conditioning load and the operation of high-power industrial equipment on weekdays cause a sudden increase in load in a short period of time, leading to continuous transformer overload. The longest continuous overload time is a threshold under ideal operating conditions (stable load and environmental adaptation) based on theoretical calculations of transformer thermal capacity and insulation life. However, actual load fluctuates randomly and experiences peak impacts, which can easily exceed this threshold, resulting in a longer overload time.

[0169] If the power grid dispatch strategy lags behind and fails to balance the load in a timely manner by transferring power to other areas or switching on and off standby equipment, or if the power grid plan does not match the regional load growth, transformer capacity will be "supply short of demand." For example, in groups 4-6, transformers may be overloaded for a long time due to insufficient support from the surrounding power grid, with the actual overload duration far exceeding the theoretical maximum continuous overload time.

[0170] The transformer's own condition affects its overload tolerance. Poor heat dissipation accelerates heat accumulation, shortens insulation withstand time, and impacts the theoretical maximum continuous overload time. Simultaneously, the combined effect of high-temperature environment and load further compresses the transformer's safe overload duration, exacerbating the difference between actual and theoretical overload times.

[0171] Prolonged overload threatens transformer lifespan and affects reliable power supply. Recommendations: Optimize load forecasting and dynamic dispatching; improve power grid planning, increase capacity and construct new facilities as needed; strengthen equipment inspections, address defects proactively, and ensure stable power system operation.

[0172] Throughout the process, the various units within the power system operation mode scenario generation device are tightly interconnected through data flow and logical relationships. The data acquisition and processing units provide standardized input datasets to the mode scenario generation module. The model inference and scenario generation units generate mode scenario sequences through the fast power flow calculation and optimized scheduling modules. The fuzzy C-means clustering module classifies and analyzes the scenarios. The scenario evaluation unit evaluates the scenarios based on the classification results, identifies weaknesses, and outputs the final results. The collaboration between these units ensures the efficiency and accuracy of the entire method, thereby achieving comprehensive generation of power system operation mode scenarios and precise identification of weaknesses.

[0173] Figure 5 This is a structural diagram of a device for generating power system operation mode scenarios provided in an embodiment of the present invention.

[0174] like Figure 5 As shown, the device includes:

[0175] The data acquisition unit 51 is used to acquire multi-source operation data of the target power system. Multi-source operation data refers to a heterogeneous data set collected from different equipment areas of the target power system.

[0176] Data processing unit 52 is used to standardize multi-source running data to obtain a standardized input dataset;

[0177] Model extrapolation unit 53 is used for the extraction of typical high-risk moments based on the tSNE-FCM algorithm. It uses a standardized input dataset to construct a power system extrapolation model that considers multiple uncertainties, including at least the fluctuation of new energy output, load changes, grid equipment status adjustment, and uncertainty of external power supply.

[0178] The scenario generation unit 54 is used to generate a sequence of scenario patterns based on a power system simulation model and a simulation algorithm, and to generate scenario patterns under rapidly changing boundaries using the IMLE method based on the sequence of scenario patterns.

[0179] The scenario evaluation unit 55 is used to evaluate the generated scenario based on a preset time-sharing and partitioned security risk indicator system.

[0180] Optionally, the data acquisition unit 51 is specifically used for:

[0181] Obtain historical power output data for wind farms and photovoltaic power plants;

[0182] Obtain curve data of power system load, including actual and predicted loads of load centers in each region of the target power system;

[0183] Acquire the recorded power supply operating parameters of the target power system, including at least the start / stop status, minimum and maximum output limits, ramp rate, and start / stop cost.

[0184] Acquire operational data of the energy storage system, which includes at least the energy storage capacity, charge and discharge power limits, and state of charge curve;

[0185] Obtain status information of power grid equipment, transaction contracts for electricity from outside the region, operating status of power grids in other regions, water inflow forecast data and dispatch rules for hydropower stations.

[0186] Optionally, the data processing unit 52 is specifically used for:

[0187] Time alignment is performed on multi-source running data from different sources and sampling frequencies, and the data is uniformly converted to an hourly time scale to obtain a standardized input dataset.

[0188] Optionally, the data processing unit 52 is further specifically used for:

[0189] For multi-source operational data with a sampling frequency higher than the target frequency, based on the formula Downsampling is performed using the sliding window averaging method, where x t This represents downsampled observations over a specified time period. The original sampling point observations are defined within a specified time period, where n is the number of sampling points within the specified time period, t is the sampling time, and i = 1, 2, 3, ...;

[0190] For multi-source operational data with a sampling frequency lower than the target frequency, based on the formula Upsampling is performed using linear interpolation, where y t The observed values ​​obtained through interpolation. and These are the observation values ​​at adjacent known time points;

[0191] For missing data segments, based on the formula The weighted moving average method is used for completion, where, These are estimates for the missing time points. These are the weighting coefficients. These are the valid observation values ​​located within the time window before and after the missing point;

[0192] For outliers that deviate from historical statistical patterns, based on the formula An anomaly detection method based on local density is used for identification and processing, where D(p) t () indicates a point in time. The local density, p t For the current observation value, is the time point observation within the neighborhood, and m is the size of the neighborhood.

[0193] Optionally, the model derivation unit 53 is specifically used for:

[0194] Based on the time-series simulation data of the target power system, the key operating indicators of the target power system are extracted. Among them, the key operating indicators include at least: wind and solar curtailment, main section power flow load rate, reserve capacity, and thermal power peak-shaving margin.

[0195] Construct a high-dimensional state vector characterizing the operational risk features of the target power system based on key operational indicators;

[0196] Based on tSNE technology, high-dimensional state vectors are mapped to low-dimensional nested spaces to preserve the nonlinear coupling structure of various bottleneck mechanisms. At the same time, the FCM clustering method is combined to identify time clusters where the similarity of risk structures meets the set similarity, and the data in the corresponding time clusters are extracted as running samples.

[0197] Optionally, the model derivation unit 53 is also specifically used for:

[0198] Based on historical data and prediction models, a probability distribution sample of new energy output is generated.

[0199] The Monte Carlo simulation method, combined with meteorological data and load characteristic analysis, is used to generate a continuous variation sequence of new energy output.

[0200] For load characteristics, a long short-term memory network model is used to fit the load change trend;

[0201] For hydropower output, a stochastic output model is established by combining water inflow forecast data and scheduling rules;

[0202] For electricity supplied from outside the region, its output range and variation pattern are determined based on the electricity trading contract and the operating status of the power grid in the other region.

[0203] For the status of power grid equipment, the triggering conditions and probabilities of N-1 faults and temporary outages are set using a discrete event simulation method.

[0204] Optionally, the scene evaluation unit 55 is specifically used for:

[0205] Construct a safety risk indicator system that includes the risk of fluctuations in new energy power.

[0206] The method and scenario are evaluated based on the safety risk index system, and the relationship between transformer overload time and longest continuous overload time is analyzed.

[0207] The device for generating power system operation mode scenarios provided in this embodiment of the invention has the same technical features as the method for generating power system operation mode scenarios provided in the above embodiments, so it can also solve the same technical problems and achieve the same technical effects.

[0208] like Figure 6 As shown in the figure, this application embodiment also provides a power system operation mode scenario generation device, including a processor 111, a communication interface 112, a memory 113, and a communication bus 114, wherein the processor 111, the communication interface 112, and the memory 113 communicate with each other through the communication bus 114.

[0209] Memory 113 is used to store computer programs;

[0210] In one embodiment of this application, when the processor 111 executes a program stored in the memory 113, it implements the method provided in any of the foregoing method embodiments, including:

[0211] Acquire multi-source operational data of the target power system, wherein the multi-source operational data refers to a heterogeneous data set collected from wind farms, photovoltaic power plants, power system loads, conventional power sources, energy storage systems and grid equipment;

[0212] The multi-source operational data is standardized to obtain a standardized input dataset;

[0213] The method for extracting typical high-risk moments based on the tSNE-FCM algorithm uses the standardized input dataset to construct a power system extrapolation model that considers multiple uncertainties, including at least the fluctuations in new energy output, load changes, grid equipment status adjustments, and uncertainties in external power supply.

[0214] Based on the power system simulation model, a sequence of scenario sequences is generated using a simulation algorithm, and the IMLE method is used to generate scenario sequences under rapidly changing boundaries.

[0215] The generated scenario is evaluated based on a pre-defined time-sharing and partitioned security risk indicator system.

[0216] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method provided in any of the foregoing method embodiments.

[0217] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0218] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0219] The above embodiments are merely illustrative examples and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this application.

Claims

1. A method for generating power system operation mode scenarios, characterized in that, The method includes: Acquire multi-source operational data of the target power system, wherein the multi-source operational data refers to a heterogeneous data set collected from different equipment areas of the target power system; The multi-source operational data is standardized to obtain a standardized input dataset; The method for extracting typical high-risk moments based on the tSNE-FCM algorithm uses the standardized input dataset to construct a power system extrapolation model that considers multiple uncertainties, including at least the fluctuations in new energy output, load changes, grid equipment status adjustments, and uncertainties in external power supply. Based on the power system simulation model, a sequence of scenario sequences is generated using a simulation algorithm. Based on the sequence of scenario sequences, the IMLE method is used to generate scenario sequences under rapidly changing boundaries. The generated scenario is evaluated based on a pre-defined time-sharing and partitioned security risk indicator system.

2. The method for generating power system operation mode scenarios according to claim 1, characterized in that, Acquiring multi-source operational data of the target power system includes: Obtain historical power output data for wind farms and photovoltaic power plants; Obtain curve data of power system load, wherein the curve data includes the actual load and predicted load of each regional load center of the target power system; The recorded power supply operating parameters of the target power system are obtained, wherein the power supply operating parameters include at least the start / stop status, minimum and maximum output limits, ramp rate, and start / stop cost; Acquire operational data of the energy storage system, wherein the operational data of the energy storage system includes at least energy storage capacity, charge and discharge power limits, and state of charge curve; Obtain status information of power grid equipment, transaction contracts for electricity from outside the region, operating status of power grids in other regions, water inflow forecast data and dispatch rules for hydropower stations.

3. The method for generating power system operation mode scenarios according to claim 1, characterized in that, The multi-source operational data is standardized to obtain a standardized input dataset, which includes: The multi-source running data from different sources and sampling frequencies are time-aligned and uniformly converted into an hourly time scale to obtain the standardized input dataset.

4. The method for generating power system operation mode scenarios according to claim 3, characterized in that, The multi-source runtime data from different sources and sampling frequencies are time-aligned and uniformly converted to an hourly time scale to obtain the standardized input dataset, which includes: For the multi-source operational data with a sampling frequency higher than the target frequency, based on the formula Downsampling is performed using the sliding window averaging method, where x t This represents downsampled observations over a specified time period. The original sampling point observations are defined within a specified time period, where n is the number of sampling points within the specified time period, t is the sampling time, and i = 1, 2, 3, ...; For the multi-source operational data with a sampling frequency lower than the target frequency, based on the formula Upsampling is performed using linear interpolation, where y t The observed values ​​obtained through interpolation. and These are the observation values ​​at adjacent known time points; For missing data segments, based on the formula The weighted moving average method is used for completion, where, These are estimates for the missing time points. These are the weighting coefficients. These are the valid observation values ​​located within the time window before and after the missing point; For outliers that deviate from historical statistical patterns, based on the formula An anomaly detection method based on local density is used for identification and processing, where D(p) t () indicates a point in time. The local density, p t For the current observation value, is the time point observation within the neighborhood, and m is the size of the neighborhood.

5. The method for generating power system operation mode scenarios according to claim 2, characterized in that, Typical high-risk moment extraction methods based on the tSNE-FCM algorithm include: Based on the time-series simulation data of the target power system, the key operating indicators of the target power system are extracted. The key operating indicators include at least: wind and solar curtailment, main section power flow load factor, reserve capacity, and thermal power peak-shaving margin. Construct a high-dimensional state vector characterizing the operational risk features of the target power system based on the aforementioned key operational indicators; Based on tSNE technology, the high-dimensional state vector is mapped to a low-dimensional nested space to preserve the nonlinear coupling structure of various bottleneck mechanisms. At the same time, the FCM clustering method is combined to identify time clusters where the risk structure similarity meets the set similarity, and the data in the corresponding time clusters are extracted as running samples.

6. The method for generating power system operation mode scenarios according to claim 5, characterized in that, Constructing a power system extrapolation model that considers multiple uncertainties includes: Based on historical data and prediction models, a probability distribution sample of new energy output is generated. The Monte Carlo simulation method, combined with meteorological data and load characteristic analysis, is used to generate a continuous variation sequence of new energy output. For load characteristics, a long short-term memory network model is used to fit the load change trend; For hydropower output, a stochastic output model is established by combining water inflow forecast data and scheduling rules; For electricity supplied from outside the region, its output range and variation pattern are determined based on the electricity trading contract and the operating status of the power grid in the other region. For the status of power grid equipment, the triggering conditions and probabilities of N-1 faults and temporary outages are set using a discrete event simulation method.

7. The method for generating power system operation mode scenarios according to claim 1, characterized in that, The evaluation of the generated scenario based on the preset time-sharing and partitioned security risk indicator system includes: Construct a safety risk indicator system that includes the risk of fluctuations in new energy power. The scenario is evaluated based on the aforementioned safety risk index system, and the relationship between transformer overload time and longest continuous overload time is analyzed.

8. A device for generating power system operation mode scenarios, characterized in that, The device includes: The data acquisition unit is used to acquire multi-source operating data of the target power system, wherein the multi-source operating data refers to a heterogeneous data set collected from different equipment areas of the target power system; The data processing unit is used to standardize the multi-source operating data to obtain a standardized input dataset; The model extrapolation unit is used for the typical high-risk moment extraction method based on the tSNE-FCM algorithm. It uses the standardized input dataset to construct a power system extrapolation model that considers multiple uncertainties, including at least the fluctuation of new energy output, load changes, grid equipment status adjustment, and uncertainty of external power supply. The scenario generation unit is used to generate a sequence of scenario modes based on the power system simulation model using a simulation algorithm, and to generate scenario modes under rapidly changing boundaries using the IMLE method based on the sequence of scenario modes. The scenario evaluation unit is used to evaluate the generated scenario based on a preset time-division and partition-based security risk indicator system.

9. A device for generating power system operation mode scenarios, characterized in that, The system includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the method for generating power system operation mode scenarios as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed in a computer, causes the computer to perform the method for generating a power system operation mode scenario as described in any one of claims 1-7.

Images