Service output control method and device for power grid regulation, equipment and medium

Abstract

This invention belongs to the field of service output control technology for power grid regulation, specifically relating to a service output control method, device, equipment, and medium for power grid regulation. The method includes: time alignment and windowing of multi-source measurements; anomaly detection and construction of a mask matrix; construction of topology conditions based on switch state vectors and operating mode encoding; inputting measurements, masks, and topology conditions into a conditional diffusion model under topology constraints, iteratively denoising and generating data for missing or abnormal locations, and performing hold-up or weak update on reliable observation locations; calculating uncertainty through multiple samplings and mapping it to a reliability matrix; gating or weighted fusion of service outputs based on reliability, and outputting analytical support and risk warnings. This solution is applicable to scenarios involving communication jitter, bad pixels, pseudo data, and mode switching, reducing the risk of misjudgment.

Description

Technical Field

[0001] This invention belongs to the field of business output control technology for power grid regulation, and specifically relates to a business output control method, device, equipment and medium for power grid regulation. Background Technology

[0002] With the development of distribution network automation, dispatch automation, and edge sensing devices, the amount of power grid measurement information available to the control side is constantly increasing. The measurement types have expanded from traditional analog quantities such as voltage, current, and power to multi-source data such as switch status, alarm events, and operating mode information. Measurement data from different sources differ in sampling frequency, timestamp accuracy, data triggering methods, and transmission links. In actual operation, they are also susceptible to factors such as communication jitter, abnormal acquisition devices, clock synchronization deviations, and sensor drift, leading to data quality problems such as missing data, jams, out-of-bounds errors, sudden jumps, drift, and time stamp misalignments.

[0003] Existing methods for handling missing or abnormal power grid measurements typically include interpolation, filtering, rule correction, and predictive completion based on historical data. These methods can be effective in cases of short-term missing measurements, single measurement anomalies, or relatively stable operating conditions, but they usually rely on the smoothness of data from adjacent time points or statistical regularities under fixed operating conditions. When there are long-term missing measurements, multiple channels experiencing simultaneous anomalies, coexistence of anomalies and missing measurements, or switching operations due to switch changes, traditional completion methods are prone to producing overly smoothed, discontinuous, or inconsistent completion results with the current power grid operating state.

[0004] Meanwhile, during the operation of the distribution network, there are adjustments to the operating mode, such as maintenance, switching operations, load transfer, and changes in the status of bus tie or sectionalizing switches. These changes in operating mode will cause changes in the power grid topology, power flow distribution, and the correlations between various measurement channels. If a data completion model based on a fixed topology or fixed operating condition assumption is still used, the model will find it difficult to distinguish whether the measurement changes are caused by the switching of normal operating modes or by data loss, bad points, or abnormal disturbances, thus affecting the accuracy and stability of the completion results.

[0005] In recent years, denoising diffusion probability models have been increasingly applied to tasks such as time series generation, data recovery, and missing data completion. These models can generate data through a process of progressive noise addition and reverse denoising, and possess a certain degree of uncertainty modeling capability. However, existing diffusion model methods are mostly geared towards general time series scenarios, often failing to fully incorporate power system-specific information such as grid switching states, operating mode coding, and topology changes. Existing power system-related diffusion model solutions also tend to focus on anomaly detection, attack identification, or single measurement recovery, making them difficult to directly apply to control and regulation scenarios involving multiple missing or anomaly sources and simultaneous operating mode switching.

[0006] Furthermore, existing measurement completion methods typically output only a deterministic completion value, rarely along with the reliability of that value. In dispatching operations, the completed grid measurement data is often used for grounding assessment, mode adjustment risk alerts, alarm analysis, or decision support. If the completion result itself is unstable or has low reliability, but the system still directly outputs automatic conclusions based on that result, it may lead to misjudgments, omissions, or invalid actions. Therefore, the dispatching side needs a mechanism that can quantify the reliability of the completion result to determine whether the relevant output can be automatically accepted or whether it needs to enter a manual review process. Summary of the Invention

[0007] The purpose of this invention is to provide a service output control method, device, equipment and medium for power grid regulation, and to solve the problem of inaccurate service output control in the prior art for power grid regulation.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a service output control method for power grid regulation, characterized in that it includes: Acquire multi-source power grid measurement data, perform time alignment and windowing processing on the multi-source power grid measurement data, and obtain windowed multi-source power grid measurement data; Anomaly detection is performed on the windowed multi-source power grid measurement data, anomaly type identifiers are generated, and a mask matrix corresponding to the windowed multi-source power grid measurement data is constructed based on the anomaly type identifiers. Constructing topology conditions based on power grid operation mode information; The windowed multi-source power grid measurement data, the mask matrix, and the topology conditions are used as inputs. The conditional diffusion model is repeatedly called to perform multiple generation samplings to obtain multiple sets of completion results. The generation sampling includes: iteratively denoising the inputs and generating them. After each iteration, the reliable observation locations are maintained or weakly updated, and the missing or abnormal locations are generated and updated. The uncertainty is calculated based on the multiple sets of completion results; The uncertainty is mapped to a confidence score, and the location-level confidence matrix corresponding to the completed multi-source measurement data is output. The channel-level confidence and / or time-level confidence are obtained by summing the location-level confidence matrix. Based on the location-level credibility matrix, the channel-level credibility, and / or the time-level credibility, a credibility assessment result is generated. Based on the credibility assessment result, the service outputs used for power grid regulation are gated or weighted and fused. When the credibility of a key channel related to power grid regulation is lower than a preset threshold, a manual review prompt is output and automatic conclusion output is suppressed. When the credibility of the key channel is not lower than the preset threshold, automatic service prompt information is output, and the credibility assessment result is used as a weight to weight and fuse downstream judgment indicators. The service outputs include grounding judgment support prompts and / or mode adjustment risk prompts.

[0009] Preferably, the time alignment includes: resampling, interpolation alignment, or event alignment of data with different sampling rates using a unified timestamp or a unified time grid; the windowing process includes segmenting the multi-source measurement data according to a preset window length and sliding step size.

[0010] Preferably, the anomaly detection includes at least one or more of the following: missing, stuck, out of bounds, sudden jump, drift, and timescale misalignment, and the anomaly type is encoded as an anomaly type identifier. ,when At this time, the corresponding position of the mask matrix is ​​set to ,when At this time, the corresponding position of the mask matrix is ​​set to ;in, For time step index, For channel index, For position Anomaly type identifier, For the mask matrix at position The anomaly detection further includes using a learning-based anomaly localization model to output anomaly probabilities for each time step and channel position, and generating or updating the mask matrix according to a preset threshold. The learning-based anomaly localization model includes a graph neural network model, a temporal neural network model, or a combination thereof.

[0011] Preferably, in the step of constructing topology conditions based on power grid operation mode information, the conditions include at least a switch state vector and an operation mode code; wherein the switch state vector is obtained by sorting and encoding one or more of the open / close states of circuit breakers, disconnectors, tie switches, bus tie switches, sectionalizing switches, and feeder switches according to a fixed rule; the operation mode code is used to characterize the operation mode category or operation mode number within a preset time period, and is generated using either one-hot encoding or embedded encoding; wherein the topology conditions also include node graph embedding and / or line graph embedding based on the power grid topology map; wherein the node graph is constructed with buses as nodes and lines or transformers as edges, and the line graph uses the edges of the original topology map as line graph nodes and establishes connections when sharing endpoint buses; the node graph embedding and / or line graph embedding are extracted using a message passing network, and the node graph embedding and / or line graph embedding are concatenated with the switch state vector and the operation mode code as topology conditions input into the condition diffusion model.

[0012] Preferably, the step of obtaining the completed multi-source measurement data includes: For positions indicated by the mask matrix as reliable observations, a maintain-update or weak-update strategy is adopted; for positions indicated by the mask matrix as missing or abnormal, a generate-update strategy is adopted to obtain the completed multi-source measurement data. The weak-update strategy involves weighting and fusing reliable observations with temporary results generated by iterative denoising according to a preset weak-update coefficient.

[0013] Preferably, in the iterative denoising generation process, a consistency constraint term is introduced, which includes at least one of topological consistency constraint and change boundary constraint. The topological consistency constraint is used to constrain the completion result to satisfy the power flow relationship and / or Kirchhoff's laws. The change boundary constraint is used to constrain the change amplitude of the completion result at adjacent times to not exceed a preset boundary. The conditional diffusion model adopts a multi-resolution diffusion architecture, performs seasonal-trend decomposition or multi-scale downsampling on windowed multi-source measurements to obtain a multi-scale sequence from coarse to fine, and performs diffusion completion in a cascade manner from coarse to fine. When performing fine-scale completion, the coarse-scale completion result is used as the conditional input.

[0014] Preferably, in the step of mapping the uncertainty to a confidence score, the following condition is met:

[0015] in, For position The credibility score; The uncertainty of the position; For channel The uncertainty-to-confidence mapping coefficient; For the time step index within the window, For channel indexing; It is an exponential function; it also includes local control over the alarm triggering window, i.e., within the alarm triggering window. Internal calculation of local critical channel reliability:

[0016] in, The average confidence level of the key channel location within the alarm trigger window; Select an indicator for the location, when the location belong The value is 1 if the condition is met, otherwise it is 0. This is the set of locations corresponding to the alarm trigger window; A set of key channel locations; and based on The comparison result with the preset threshold determines whether to suppress the automatic conclusion output in the alarm trigger window.

[0017] In a second aspect, the present invention provides a service output control device for power grid regulation, comprising: The data access and alignment module is used to acquire multi-source measurement data of the power grid, perform time alignment and windowing processing on the multi-source measurement data of the power grid, and obtain windowed multi-source measurement data of the power grid. An anomaly detection and mask generation module is used to perform anomaly detection on windowed multi-source power grid measurement data, generate anomaly type identifiers, and construct a mask matrix corresponding to the windowed multi-source power grid measurement data based on the anomaly type identifiers. The topology condition construction module is used to construct topology conditions based on power grid operation mode information; The conditional diffusion completion module is used to take the windowed multi-source power grid measurement data, the mask matrix and the topology conditions as input, and repeatedly call the conditional diffusion model to perform multiple generation samplings to obtain multiple sets of completion results; the generation sampling includes: iteratively denoising the input to generate, and after each iteration, performing a hold-up update or weak update on the reliable observation position and a generation update on the missing or abnormal position. The uncertainty assessment module is used to calculate the uncertainty based on the multiple sets of completion results; The confidence calculation module is used to map the uncertainty into a confidence score, output a location-level confidence matrix corresponding to the completed multi-source measurement data, and obtain channel-level confidence and / or time-level confidence by summing the location-level confidence matrix. The business gating and fusion output module is used to generate a credibility assessment result based on the location-level credibility matrix, the channel-level credibility, and / or the time-level credibility. Based on the credibility assessment result, it performs gating or weighted fusion on the business outputs used for power grid regulation. When the credibility of a key channel related to power grid regulation is lower than a preset threshold, it outputs a manual review prompt and suppresses automatic conclusion output. When the credibility of the key channel is not lower than the preset threshold, it outputs automatic business prompt information and uses the credibility assessment result as a weight to perform weighted fusion on downstream judgment indicators. The business outputs include grounding judgment support prompts and / or mode adjustment risk prompts.

[0018] In a third aspect, the present invention provides an electronic device including a processor and a memory, the processor being configured to execute a computer program stored in the memory to implement the aforementioned business output control method for power grid regulation.

[0019] In a fourth aspect, the present invention provides a computer-readable storage medium storing at least one instruction that, when executed by a processor, implements the aforementioned service output control method for power grid regulation.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: By performing time alignment and windowing processing on the multi-source measurement data of the power grid, windowed multi-source measurement data of the power grid is obtained, realizing the organization of measurement data with different sampling frequencies, timestamps and triggering methods under a unified time scale, providing consistent data input for subsequent anomaly detection, diffusion completion and confidence calculation.

[0021] By performing anomaly detection on multi-source measurement data of the power grid, anomaly type identifiers are generated. Based on the anomaly type identifiers, a mask matrix corresponding to the windowed multi-source measurement data of the power grid is constructed to distinguish between reliable observation locations, missing locations, and anomaly locations, providing a location basis for the differentiated update of the subsequent conditional diffusion model.

[0022] By constructing topological conditions based on grid operation mode information, including at least switch state vectors and operation mode codes, the current grid switch state and operation mode are incorporated into the completion process, enabling the completion model to adapt to topological changes caused by switching operations, maintenance switching, and operation mode adjustments.

[0023] By inputting windowed multi-source power grid measurement data, mask matrix, and topology conditions into the conditional diffusion model, and performing iterative denoising generation under topology constraints, generative completion of missing or abnormal locations under the current topology conditions can be achieved.

[0024] By performing a preserve update or weak update on the reliable observation position based on the mask matrix after each iteration, and performing a generate update on the missing or abnormal position, the reliable observation data can be preserved or flexibly corrected. At the same time, the missing or abnormal data is generated in a targeted manner to avoid unnecessary disturbance to the reliable measurement during the completion process.

[0025] By performing multiple iterations of denoising generation using windowed multi-source power grid measurement data, mask matrix, and topology condition-based conditional diffusion model, multiple sets of completion results are obtained. Based on these multiple completion results, the completed multi-source measurement data and its uncertainty are determined, thereby achieving a quantitative characterization of the stability of the completion results.

[0026] By mapping uncertainty to confidence scores, the system outputs a location-level confidence matrix corresponding to the completed multi-source measurement data. The channel-level confidence and / or time-level confidence are obtained by summarizing the location-level confidence matrices, thus achieving confidence evaluation of the completion results from the dimensions of location, channel, and time. Attached Figure Description

[0027] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of a service output control method for power grid regulation according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the topology condition construction for a service output control method for power grid regulation according to Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the reliability gating and weighted fusion output of a business output control method for power grid regulation according to Embodiment 1 of the present invention; Figure 4 This is a bar chart comparing the accuracy of business judgment under automatic and random review versus credibility-guided review in a business output control method for power grid regulation according to Embodiment 1 of the present invention, with a 10% manual review budget. Figure 5 This is a comparison curve of the average business judgment accuracy across all scenarios as a function of manual budget review in a business output control method for power grid regulation according to Embodiment 1 of the present invention. Figure 6 This is a structural block diagram of a business output control device for power grid regulation according to an embodiment of the present invention; Figure 7 This is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0028] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0029] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0030] Example 1 In this embodiment, for ease of explanation, the meanings of the symbols are agreed upon as follows: Location index: Time index within the window Channel Index The position is recorded as ; Windowed multi-source measurement: ; Mask matrix: ,in Indicates a reliable observation. Indicates missing or abnormal; Topological conditions: At least by the switch state vector Coding of operating mode constitute; Diffusion variables: For the first Intermediate variables in each diffusion / denoising iteration step ; Uncertainty and Confidence: Position-level uncertainty is The location-level confidence level is The credibility matrix is ​​.

[0031] like Figure 1 As shown, a business output control method for power grid regulation includes the following specific steps: S1, Multi-source measurement access and time alignment, windowing Acquire multi-source measurement data from the power grid, which includes one or more of the following: voltage measurements, current measurements, power measurements, switch status measurements, and alarm event data. Each type of measurement data carries a timestamp and channel identifier.

[0032] A unified time grid is established according to a preset time granularity, and multi-source power grid measurement data with different sampling frequencies or different triggering methods are mapped to the unified time grid. For continuous measurement data such as voltage, current and power measurements, resampling and / or interpolation alignment methods are used to generate measurement values ​​for the corresponding time grid points; for switch status measurements and alarm event data, event alignment methods are used to generate status values ​​or event values ​​for the corresponding time grid points.

[0033] Data from each channel at the same time grid point are combined according to a preset channel order to obtain a time-aligned multi-source measurement sequence of the power grid. Then, a preset window length is used... With step size Perform windowed slicing to obtain windowed input. .

[0034] S2. Anomaly Detection and Mask Matrix Construction Anomaly detection is performed on the multi-source measurement data of the power grid, and anomaly type identifiers are generated. The anomaly detection includes one or more of the following: missing data detection, stuck data detection, out-of-bounds data detection, sudden jump data detection, drift data detection, and timescale misalignment data detection.

[0035] Based on the anomaly detection results, an anomaly type identifier is generated for each time step and each channel within the window. For locations where no anomalies are detected and valid observations exist, the location is marked as a reliable observation location; for locations with missing, stuck, out-of-bounds, abrupt jumps, drifts, or time-scale misalignments, the location is marked as a missing or anomaly location.

[0036] Based on the anomaly type identifier, a mask matrix corresponding to the windowed power grid multi-source measurement data is constructed. Confidential observation location Missing or abnormal locations .

[0037] S3, Construction of Topological Conditions like Figure 2 As shown, topology conditions are constructed based on power grid operation mode information. The power grid operation information includes switch status information and operation mode information.

[0038] Based on the open / closed states of one or more of the following: circuit breaker, disconnector, tie switch, bus tie switch, sectionalizing switch, and feeder switch, the states are encoded in a preset order to obtain the switch state vector. The operating mode code is obtained by encoding the current window according to its operating mode category or operating mode number. .

[0039] The topology conditions are obtained by concatenating the switch state vector and the operating mode code. It can be represented as:

[0040] in, This represents the vector concatenation operation; The corresponding switch-on / off state encoding vector for the window is used to characterize the network connectivity state; An encoding vector for the operating mode category or mode number, used to characterize the operating condition category; It is a topological condition vector, and serves as the conditional input to the conditional diffusion model.

[0041] In another embodiment, node graph embeddings and / or line graph embeddings can be generated based on the power grid topology map, and the node graph embeddings and / or line graph embeddings can be concatenated with the switch state vector and the operation mode code as the topology conditions of the input condition diffusion model.

[0042] S4, Conditional Diffusion Completion The windowed multi-source power grid measurement data, the mask matrix, and the topology conditions are input into a pre-constructed conditional diffusion model, and iterative denoising is performed under the constraints of the topology conditions.

[0043] During each iteration of denoising, the conditional diffusion model generates a temporary denoising result based on the current iteration variable, diffusion step number, mask matrix, and topological conditions. After each iteration, different update methods are performed on different positions according to the mask matrix.

[0044] Specifically, including: Forward diffusion process:

[0045] in, Forward diffusion from the first Step to the first The transition probability distribution of the step; For the first An intermediate variable for each diffusion / noise-adding step; Number the diffusion step. , This represents the total number of diffusion steps. The mean is Covariance is Gaussian distribution; For noise scheduling parameters (usually vary) Incremental (used to control the intensity of noise added at each step). An identity matrix consistent with the data dimensions.

[0046] definition , Then, noise data from any diffusion step can be directly sampled:

[0047] in, For the first Step signal retention coefficient; The cumulative signal retention coefficient; To window multi-source measurement data; It is a standard Gaussian noise random variable; This represents a Gaussian distribution with zero mean and unit covariance.

[0048] Reverse denoising process:

[0049] in, For the first The first step of denoising obtained Step-by-step provisional results; This is a noise prediction network (denoising network). Its network parameters; Indicates the condition The predicted value for noise; To sample the noise intensity parameter, this embodiment takes... ; It is Gaussian random noise, satisfying .

[0050] Mask constraint update:

[0051] in, For position In the The updated value of the step; For time step index, For channel indexing; These are the original observations; For mask identifier, Indicates the location of a reliable observation. Indicates a missing or abnormal location; when Maintain observation constraints while, when Update using the generated results. As needed, it can also be done in... Position is updated using weak methods: ,in It is a weak update coefficient.

[0052] After the iterative denoising reaches the preset number of iterations, a set of completion results is obtained. In this completion result, the reliable observation locations retain the original observation values ​​or are corrected by weak updates, while missing or anomalous locations are completed by the conditional diffusion model.

[0053] S5. Multiple Sampling and Uncertainty Calculation Based on the windowed multi-source power grid measurement data, the mask matrix, and the topology conditions, a conditional diffusion model is invoked to perform multiple iterative denoising generation. Each iteration of denoising generation uses different random noise initializations or different random sampling paths to obtain a set of completion results.

[0054] Specifically: For the same input Repeated completion Next, get Location-level uncertainty, taking standard deviation as an example:

[0055] in, For position Uncertainty; Number of samples; For sampling index, ; For the first The completion result obtained from the second sampling is located at... The value; for The sample mean of the sampled items; This is the correction term for the degrees of freedom of the unbiased standard deviation.

[0056] The summaries of channel-level and time-level uncertainties are as follows:

[0057] in, For channel The average uncertainty (statistics only for the completed positions); For a moment The average uncertainty is only statistically analyzed for the complete channel; The length of the time window; Number of channels; To complete the location indicator, when If the value is 1, then the value is 0; otherwise, the value is 0.

[0058] S6. Map uncertainty to confidence level and output the confidence matrix. Mapping uncertainty to a confidence score:

[0059] in, For position The credibility score, with a value range of 100%. A larger value indicates more reliable completion; This represents the uncertainty at the corresponding location; For channel The uncertainty-to-confidence mapping coefficient is used to adjust the sensitivity of different channels; It is an exponential function. For reliable observation locations ( (Can be directly ordered) .

[0060] Output credibility matrix .

[0061] Furthermore, channel-level confidence and / or time-level confidence are obtained by summing the location-level confidence matrices. Channel-level confidence can be obtained by averaging, weighted averaging, or quantile statistics of the confidence of the same channel at each time step within the window; time-level confidence can be obtained by averaging, weighted averaging, or quantile statistics of the confidence of each channel at the same time step.

[0062] S7. Trust-based business gating and weighted fusion output like Figure 3 As shown, based on the location-level credibility matrix, channel-level credibility, and / or time-level credibility, the control service output generated based on the completed multi-source measurement data is gated or weighted and fused to generate a credibility assessment result.

[0063] For the current control operation, a set of key channels corresponding to this operation is pre-determined. The credibility of the key channels is determined based on the location-level credibility matrix and / or channel-level credibility.

[0064] When the credibility of the key channel is lower than a preset threshold, the automatic conclusion output is suppressed and a manual review prompt is output; when the credibility of the key channel is not lower than the preset threshold, an automatic business prompt message is output.

[0065] When weighted fusion is required, the location-level credibility matrix, channel-level credibility, and / or time-level credibility are used as weights to weight and fuse downstream judgment indicators to obtain a comprehensive judgment indicator. The control service output includes grounding judgment support prompts and / or mode adjustment risk prompts.

[0066] Specifically: Set the threshold When the average confidence level of the critical channel Output review prompts and suppress automatic conclusion output; when The system can automatically output business prompts and weight and integrate downstream assessment indicators based on their credibility.

[0067] in, This is the gate threshold; Average confidence level of key channels; For participating fusion indicator (1 represents position) (Participation in integration, with 0 indicating non-participation). For normalized fusion weights; For position Corresponding downstream analysis and derivative indicators; This refers to the integrated indicators after fusion. For summation index. The denominator is the normalized term for the positions involved in the merging, used to ensure that there is a summation index when the denominator is non-zero. .

[0068] In a specific application scenario, when communication jitter causes intermittent missing key voltage or current channels, and a switch change occurs within the same window, this embodiment constructs topological conditions based on the switch state vector and operating mode encoding, and performs conditional diffusion to complete the missing or abnormal positions under these topological conditions. Subsequently, multiple sets of completion results are generated through multiple iterations of noise reduction, and the corresponding uncertainty and confidence matrices are calculated. When the confidence of the key channel near the switching moment is lower than a preset threshold, the automatic conclusion output of grounding assessment support prompts or mode adjustment risk prompts is suppressed, and the process is transferred to manual review.

[0069] like Figures 4-5 As shown, by adopting the business output control method in this embodiment, the accuracy of business judgment is improved by using only automatic, random and credibility-guided review under a 10% manual review budget; the average business judgment accuracy across all scenarios increases with the increase of the manual review budget.

[0070] Example 2 like Figure 6 As shown, based on the same inventive concept as the above embodiments, the present invention also provides a service output control device for power grid regulation, comprising: The data access and alignment module is used to acquire multi-source measurement data of the power grid, perform time alignment and windowing processing on the multi-source measurement data of the power grid, and obtain windowed multi-source measurement data of the power grid. An anomaly detection and mask generation module is used to perform anomaly detection on windowed multi-source power grid measurement data, generate anomaly type identifiers, and construct a mask matrix corresponding to the windowed data based on the anomaly type identifiers. The topology condition construction module is used to construct topology conditions based on power grid operation mode information; The conditional diffusion completion module is used to take the windowed multi-source measurement data, the mask matrix and the topological conditions as the same input, call the conditional diffusion model to perform iterative denoising and generation, and perform maintain or weak update on the reliable observation positions and generate update on the missing or abnormal positions after each iteration to obtain the completed multi-source measurement data. The uncertainty assessment module is used to perform multiple generation samplings on the same windowed input to obtain multiple sets of completion results, and calculate the uncertainty based on the multiple sets of completion results. The confidence calculation module is used to map the uncertainty into a confidence score, output a location-level confidence matrix corresponding to the completed multi-source measurement data, and obtain channel-level confidence and / or time-level confidence by summing the location-level confidence matrix. The business gating and fusion output module is used to generate a credibility assessment result based on the location-level credibility matrix, the channel-level credibility, and / or the time-level credibility. Based on the credibility assessment result, it performs gating or weighted fusion on the business outputs used for power grid regulation. When the credibility of a key channel related to power grid regulation is lower than a preset threshold, it outputs a manual review prompt and suppresses automatic conclusion output. When the credibility of the key channel is not lower than the preset threshold, it outputs automatic business prompt information and uses the credibility assessment result as a weight to perform weighted fusion on downstream judgment indicators. The business outputs include grounding judgment support prompts and / or mode adjustment risk prompts.

[0071] In one embodiment, a storage and logging module is also included.

[0072] Example 3 like Figure 7 As shown, the present invention also provides an electronic device 100 for implementing a service output control method for power grid regulation; The electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on at least one processor 102, and at least one communication bus 104.

[0073] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the business output control method for power grid regulation in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0074] The memory 101 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0075] At least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 102 may be a microprocessor or any conventional processor. Processor 102 is the control center of electronic device 100, connecting various parts of electronic device 100 via various interfaces and lines.

[0076] The memory 101 in the electronic device 100 stores multiple instructions to implement a service output control method for power grid regulation, and the processor 102 can execute multiple instructions to achieve the following: Acquire multi-source power grid measurement data, perform time alignment and windowing processing on the multi-source power grid measurement data, and obtain windowed multi-source power grid measurement data; Anomaly detection is performed on windowed multi-source power grid measurement data, anomaly type identifiers are generated, and a mask matrix corresponding to the windowed data is constructed based on the anomaly type identifiers. Constructing topology conditions based on power grid operation mode information; The windowed multi-source measurement data, the mask matrix, and the topological conditions are used as the same input. The conditional diffusion model is called to perform iterative denoising and generation. After each iteration, the reliable observation positions are maintained or weakly updated, and the missing or abnormal positions are generated and updated to obtain the completed multi-source measurement data. Multiple generation samplings are performed on the same windowed input to obtain multiple sets of completion results, and the uncertainty is calculated based on the multiple sets of completion results; The uncertainty is mapped to a confidence score, and the location-level confidence matrix corresponding to the completed multi-source measurement data is output. The channel-level confidence and / or time-level confidence are obtained by summing the location-level confidence matrix. Based on the location-level credibility matrix, the channel-level credibility, and / or the time-level credibility, a credibility assessment result is generated. Based on the credibility assessment result, the service outputs used for power grid regulation are gated or weighted and fused. When the credibility of a key channel related to power grid regulation is lower than a preset threshold, a manual review prompt is output and automatic conclusion output is suppressed. When the credibility of the key channel is not lower than the preset threshold, automatic service prompt information is output, and the credibility assessment result is used as a weight to weight and fuse downstream judgment indicators. The service outputs include grounding judgment support prompts and / or mode adjustment risk prompts.

[0077] in, Figure 6 The external inputs shown represent the data sources received by the system from external dispatch automation systems, measurement acquisition devices, or alarm event systems. They include at least one or more of SCADA measurements, PMU measurements, power distribution measurements, switch status, and alarm events. After being received by the data access and alignment module, the data enters the subsequent processing flow. Figure 6 The external output shown represents the processing results provided by the system to the outside world, including at least one or more of the following: completion results, credibility matrix, business prompts, and automatic / manual review markers.

[0078] Example 4 If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, and read-only memory (ROM).

[0079] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0080] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0081] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0082] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0083] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A business output control method for power grid regulation, characterized in that, include: Acquire multi-source power grid measurement data, perform time alignment and windowing processing on the multi-source power grid measurement data, and obtain windowed multi-source power grid measurement data; Anomaly detection is performed on the windowed multi-source power grid measurement data, anomaly type identifiers are generated, and a mask matrix corresponding to the windowed multi-source power grid measurement data is constructed based on the anomaly type identifiers. Constructing topology conditions based on power grid operation mode information; The windowed multi-source power grid measurement data, the mask matrix, and the topology conditions are used as inputs. The conditional diffusion model is repeatedly called to perform multiple generation samplings to obtain multiple sets of completion results. The generation sampling includes: iteratively denoising the inputs and generating them. After each iteration, the reliable observation locations are maintained or weakly updated, and the missing or abnormal locations are generated and updated. The uncertainty is calculated based on the multiple sets of completion results; The uncertainty is mapped to a confidence score, and the location-level confidence matrix corresponding to the completed multi-source measurement data is output. The channel-level confidence and / or time-level confidence are obtained by summing the location-level confidence matrix. Based on the location-level credibility matrix, the channel-level credibility, and / or the time-level credibility, a credibility assessment result is generated. Based on the credibility assessment result, the business outputs used for power grid regulation are gated or weighted and fused. Specifically, when the credibility of a key channel related to power grid regulation is lower than a preset threshold, a manual review prompt is output and automatic conclusion output is suppressed. When the credibility of the key channel is not lower than the preset threshold, automatic business prompt information is output, and the credibility assessment result is used as a weight to weight and fuse downstream judgment indicators. The business outputs include grounding judgment support prompts and / or mode adjustment risk prompts.

2. The service output control method for power grid regulation as described in claim 1, characterized in that, The time alignment includes: resampling, interpolation alignment, or event alignment of data with different sampling rates using a unified timestamp or a unified time grid; the windowing process includes segmenting multi-source measurement data according to a preset window length and sliding step size.

3. The service output control method for power grid regulation as described in claim 1, characterized in that, The anomaly detection includes at least one or more of the following: missing, stuck, out of bounds, sudden jump, drift, and timescale misalignment; Anomaly detection is performed on windowed multi-source power grid measurement data, anomaly type identifiers are generated, and a mask matrix corresponding to the windowed multi-source power grid measurement data is constructed based on the anomaly type identifiers, including: Encode the exception type as an exception type identifier. ; when At this time, the corresponding position of the mask matrix is ​​set to ,when At this time, the corresponding position of the mask matrix is ​​set to ;in, For time step index, For channel index, For position Anomaly type identifier, For the mask matrix at position The anomaly detection further includes using a learning-based anomaly localization model to output anomaly probabilities for each time step and channel position, and generating or updating the mask matrix according to a preset threshold. The learning-based anomaly localization model includes a graph neural network model, a temporal neural network model, or a combination thereof.

4. The service output control method for power grid regulation as described in claim 1, characterized in that, In the step of constructing topology conditions based on power grid operation mode information, the conditions include at least a switch state vector and an operation mode code. The switch state vector is obtained by sorting and encoding one or more of the open / closed states of circuit breakers, disconnectors, tie switches, bus tie switches, sectionalizing switches, and feeder switches according to a fixed rule; the operation mode code is used to characterize the operation mode category or operation mode number within a preset time period, and is generated using either hot-swappable encoding or embedded encoding. The topology conditions also include node graph embedding and / or line graph embedding based on the power grid topology graph; The node graph is constructed with buses as nodes and lines or transformers as edges, while the line graph uses the edges of the original topology graph as line graph nodes and establishes connections when sharing endpoint buses. The node graph embedding and / or line graph embedding are extracted using a message passing network, and the node graph embedding and / or line graph embedding are concatenated with the switch state vector and the operation mode encoding to form a topology condition input condition diffusion model.

5. The service output control method for power grid regulation as described in claim 1, characterized in that, The step of obtaining the completed multi-source measurement data includes: For positions indicated by the mask matrix as reliable observations, a maintain-update or weak-update strategy is adopted; for positions indicated by the mask matrix as missing or abnormal, a generate-update strategy is adopted to obtain the completed multi-source measurement data. The weak-update strategy involves weighting and fusing reliable observations with temporary results generated by iterative denoising according to a preset weak-update coefficient.

6. The service output control method for power grid regulation as described in claim 1, characterized in that, In the iterative denoising process, a consistency constraint term is introduced, which includes at least one of topological consistency constraint and change boundary constraint. Topology consistency constraints are used to ensure that the completion result satisfies the power flow relationship and / or Kirchhoff's laws; variation boundary constraints are used to ensure that the variation of the completion result at adjacent time points does not exceed the preset boundary. The conditional diffusion model adopts a multi-resolution diffusion architecture, which performs seasonal-trend decomposition or multi-scale downsampling on windowed multi-source measurements to obtain multi-scale sequences from coarse to fine, and performs diffusion completion in a cascade manner from coarse to fine. When performing fine-scale completion, the coarse-scale completion result is used as the conditional input.

7. The service output control method for power grid regulation as described in claim 1, characterized in that, In the step of mapping the uncertainty to a confidence score, the following condition is met: in, For position The credibility score; The uncertainty of the position; For channel The uncertainty-to-confidence mapping coefficient; For the time step index within the window, For channel indexing; It is an exponential function; it also includes the alarm trigger window. Internal calculation of local critical channel reliability: in, The average confidence level of the key channel location within the alarm trigger window; Select an indicator for the location, when the location belong The value is 1 if the condition is met, otherwise it is 0. This is the set of locations corresponding to the alarm trigger window; A set of key channel locations; and based on The comparison result with the preset threshold determines whether to suppress the automatic conclusion output in the alarm trigger window.

8. A business output control device for power grid regulation, characterized in that, include: The data access and alignment module is used to acquire multi-source measurement data of the power grid, perform time alignment and windowing processing on the multi-source measurement data of the power grid, and obtain windowed multi-source measurement data of the power grid. An anomaly detection and mask generation module is used to perform anomaly detection on windowed multi-source power grid measurement data, generate anomaly type identifiers, and construct a mask matrix corresponding to the windowed multi-source power grid measurement data based on the anomaly type identifiers. The topology condition construction module is used to construct topology conditions based on power grid operation mode information; The conditional diffusion completion module is used to take the windowed multi-source power grid measurement data, the mask matrix and the topology conditions as input, and repeatedly call the conditional diffusion model to perform multiple generation samplings to obtain multiple sets of completion results; the generation sampling includes: iteratively denoising the input to generate, and after each iteration, performing a hold-up update or weak update on the reliable observation position and a generation update on the missing or abnormal position. The uncertainty assessment module is used to calculate the uncertainty based on the multiple sets of completion results; The confidence calculation module is used to map the uncertainty into a confidence score, output a location-level confidence matrix corresponding to the completed multi-source measurement data, and obtain channel-level confidence and / or time-level confidence by summing the location-level confidence matrix. The business gating and fusion output module is used to generate a credibility assessment result based on the location-level credibility matrix, the channel-level credibility, and / or the time-level credibility. Based on the credibility assessment result, it performs gating or weighted fusion on the business outputs used for power grid regulation. When the credibility of a key channel related to power grid regulation is lower than a preset threshold, it outputs a manual review prompt and suppresses automatic conclusion output. When the credibility of the key channel is not lower than the preset threshold, it outputs automatic business prompt information and uses the credibility assessment result as a weight to perform weighted fusion on downstream judgment indicators. The business outputs include grounding judgment support prompts and / or mode adjustment risk prompts.

9. An electronic device, characterized in that, It includes a processor and a memory, the processor being used to execute a computer program stored in the memory to implement the service output control method for power grid regulation as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, which, when executed by a processor, implements the service output control method for power grid regulation as described in any one of claims 1 to 7.

Images