A power distribution network operation architecture construction method, system, device and storage medium

Abstract

This invention discloses a method, system, equipment, and storage medium for constructing a distribution network operation architecture, relating to the field of distribution network operation architecture. The method includes: extracting monitoring data from different equipment operations to establish a distribution network dataset with the same time-stamp; resampling all equipment data to the same timestamp sequence based on the kernel density formula; determining and performing alignment operations between the current equipment operation monitoring data and the same timestamp sequence based on the equipment's sampling frequency and timestamp; obtaining the quality assessment score and weight coefficient of the equipment operation monitoring data, and updating and optimizing it; constructing a distribution network operation architecture evaluation model and outputting the current operation status labels of the distribution network equipment; and dynamically adjusting the equipment operation parameters in the distribution network operation architecture based on fault tree analysis and expert experience rules. This achieves dynamic sampling of equipment operation data and applies the distribution network operation architecture evaluation model to solve the problem of synergistic operation of distribution network economy and reliability under multiple constraints.

Description

Technical Field

[0001] This invention relates to the field of power distribution network operation architecture, specifically to a method, system, device, and storage medium for constructing a power distribution network operation architecture. Background Technology

[0002] The distribution network operation architecture refers to the sum of organizational structures, control strategies, and operational procedures within a power system's distribution network to achieve efficient and reliable power distribution under normal operating and fault conditions. Its core objective is to ensure stable power transmission from the transmission network to end users, while simultaneously addressing faults, optimizing operational efficiency, and adapting to future needs. Multimodal data fusion, a key technology in artificial intelligence and big data, refers to the integration of data from different sources and modalities to provide a more comprehensive and accurate information representation. However, during multimodal data fusion, inconsistencies in the extension directions of self-defined protocols of different sampling devices lead to deviations in the frequency, accuracy, and time response of the sampled monitoring data, resulting in data matching imbalances during multimodal data fusion. Especially with the rapid development of intelligent algorithms, ensuring the consistency of multimodal data has become particularly important. Data consistency includes data volume, data accuracy, and data sequence. Ensuring that the input data are at the same time or in the same sequence is crucial for guaranteeing the accuracy of data prediction.

[0003] Traditional data preprocessing methods mainly include Kalman filtering, min-max normalization, Z-score standardization, and polynomial interpolation. These methods can improve data quality to a certain extent, but they cannot guarantee that the processing of each group of data will be optimal when fusing multimodal data. For example, conventional data standardization uses min-max normalization and Z-score standardization, but these two methods cannot handle high-precision floating-point data, especially data that is highly dependent on decimals. Simple data standardization is difficult to achieve the effect of data optimization, or it lacks a good comprehensive evaluation of data interpolation and downsampling. Under the combined effect of multimodal data, it may lead to data bias, resulting in inaccurate final output results. Summary of the Invention

[0004] To address the aforementioned technical problems, this paper provides a method, system, equipment, and storage medium for constructing a power distribution network operation architecture. This technical solution solves the problems mentioned in the background technology, such as the inability to guarantee optimal processing of each group of data during multimodal data fusion, the difficulty in achieving data optimization through simple data standardization for data with a high dependence on decimals, the lack of comprehensive evaluation for data interpolation and downsampling, and the potential for data deviation under the combined effect of multimodal data, leading to inaccurate final output results.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for constructing a power distribution network operation architecture includes: Based on historical data of power distribution network operation, monitoring data of different equipment operation are extracted, and a unified time-scaled dataset of power distribution network is established. Based on the same time-stamped dataset of the power distribution network, and using the kernel density formula, the data of all devices are resampled to the same timestamp sequence. Based on the device's sampling frequency and timestamp, determine and perform alignment operations between the current device operation monitoring data and the same timestamp sequence; Based on the historical error rate, signal-to-noise ratio, and data freshness of different equipment operation monitoring data, obtain the quality assessment score and weight coefficient of the equipment operation monitoring data, and update and optimize them accordingly; Based on the weighting coefficients of the equipment operation monitoring data, and using the long short-term memory network neural network model, a power distribution network operation architecture evaluation model is constructed, and the current operation status label of the power distribution network equipment is output. Based on the current operating status labels of the distribution network equipment, and using fault tree analysis and expert experience rules, the operating parameters of the equipment in the distribution network operation architecture are dynamically adjusted.

[0006] Preferably, the step of resampling the data of all devices to the same timestamp sequence based on the same time-stamped dataset of the distribution network and using the kernel density formula specifically includes: From the same time-scaled dataset of the power distribution network, the sampling frequency of all devices is extracted as a sample set, and then fitted into the kernel density formula to obtain the probability density curve of the sampling frequency. Based on the probability density curve of the sampling frequency, find the frequency range corresponding to the density peak. This range represents the main concentration range of equipment sampling frequencies, indicating the mainstream frequency level of distribution network data sampling, and can cover the original frequency characteristics of most equipment to the greatest extent. Within the frequency range corresponding to the density peak, three types of candidate frequencies are selected as preliminary candidate target frequencies. These three types of candidate frequencies include: high-frequency candidates, mid-frequency candidates, and low-frequency candidates. For each type of preliminary candidate target frequency, calculate the total information loss rate of all equipment operation monitoring data resampled to that frequency; The candidate frequency with the lowest total information loss rate is selected as the final target sampling frequency. If the total information loss rates of multiple candidate frequencies are similar, the candidate frequency with the lowest value is selected first. Using the target sampling frequency as the interval, generate an equally spaced sequence of the same timestamps that covers the complete time window of the data to be fused.

[0007] Preferably, the step of determining and performing the alignment operation between the current device operation monitoring data and the same timestamp sequence based on the device's sampling frequency and timestamp specifically includes: Based on the same time-scaled dataset of the power distribution network, obtain the sampling frequency and timestamp in the operation monitoring data sequence of each device; For each device's operational monitoring data sequence, compare the sampling frequencies. With the final target sampling frequency The size relationship is determined, and the entire sequence is optimized accordingly based on the comparison results; like Then, linear interpolation or cubic spline interpolation methods are used to resample the equipment operation monitoring data sequence onto the same timestamp sequence; like If so, it is determined that the equipment operation monitoring data has been aligned with the timestamp sequence and no further processing is required; like Then, the mean aggregation or maximum / minimum aggregation method is used to aggregate the equipment operation monitoring data sequence within a time window centered on each time point in the same timestamp sequence, and the result is used as the value of that time point, thereby resampling to the same timestamp sequence.

[0008] Preferably, the step of obtaining a quality assessment score and weighting coefficient for the equipment operation monitoring data based on the historical error rate, signal-to-noise ratio, and data freshness of different equipment operation monitoring data, and then updating and optimizing it, specifically includes: Based on the historical error rate, signal-to-noise ratio, and data freshness of different equipment operation monitoring data, a quality assessment score for the equipment operation monitoring data is obtained using a weighted summation formula. Based on the quality assessment score of the equipment operation monitoring data, the weight coefficients of the equipment operation monitoring data are obtained using the weighted average method. The latest quality assessment score and weighting coefficient are calculated using real-time monitoring data of equipment operation at fixed time intervals or event triggering methods. Based on the historical monitoring data of equipment operation, and using the Three Sigma criterion, the boundary thresholds for the weighting coefficients of the equipment operation monitoring data are obtained. For equipment operation monitoring data with a weighting coefficient less than the boundary threshold, the weighting coefficient will be set to zero and the data will be removed from this data fusion. If the sampling frequency of the real-time monitoring data of the equipment operation is obtained, and it is determined whether the frequency change value exceeds the tolerance or new monitoring parameters of the equipment operation are detected, the target sampling frequency is updated and a resampling and alignment operation is performed.

[0009] Furthermore, this embodiment proposes a distribution network operation architecture construction system for implementing the distribution network operation architecture construction method described above, including: The data optimization and processing module is used to extract monitoring data of different equipment operation based on historical data of distribution network operation and establish a common time-scaled dataset for the distribution network; based on the common time-scaled dataset of the distribution network and the kernel density formula, resample the data of all equipment to the same timestamp sequence; determine and perform alignment operation between the current equipment operation monitoring data and the same timestamp sequence by using the sampling frequency and timestamp of the equipment; and obtain the quality assessment score and weight coefficient of the equipment operation monitoring data based on the historical error rate, signal-to-noise ratio and data freshness of the different equipment operation monitoring data, and update and optimize them. The dynamic allocation module is used to construct a power distribution network operation architecture evaluation model based on the weight coefficients of equipment operation monitoring data and a long short-term memory network neural network model, and output the current operation status label of the power distribution network equipment; based on the current operation status label of the power distribution network equipment, and based on fault tree analysis and expert experience rules, dynamically adjust the equipment operation parameters in the power distribution network operation architecture. The data optimization processing module includes: The dataset unit is used to extract monitoring data of different equipment operation based on historical data of distribution network operation, and to establish a common time-scaled dataset of the distribution network. The timestamp alignment unit is used to resample the data of all devices to the same timestamp sequence based on the kernel density formula of the same time-scaled dataset of the power distribution network. A data timestamp alignment unit is used to determine and perform alignment operations between the current device operation monitoring data and the same timestamp sequence based on the device's sampling frequency and timestamp. The data optimization unit is used to obtain the quality assessment score and weight coefficient of the equipment operation monitoring data based on the historical error rate, signal-to-noise ratio and data freshness of different equipment operation monitoring data, and to update and optimize it. The dynamic allocation module includes: The model building unit is used to build a power distribution network operation architecture evaluation model based on the weight coefficients of the equipment operation monitoring data and the long short-term memory network neural network model, and output the current operation status label of the power distribution network equipment. The dynamic adjustment unit is used to dynamically adjust the operating parameters of the equipment in the distribution network operation architecture based on the current operating status labels of the distribution network equipment and fault tree analysis and expert experience rules.

[0010] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method for constructing a power distribution network operation architecture. It extracts monitoring data from different devices from historical power distribution network operation data, fits the device sampling frequency probability density curve using a kernel density formula, selects the target sampling frequency with the lowest total information loss rate, generates a sequence with the same timestamp, and, based on the relationship between the device sampling frequency and the target frequency, employs interpolation, aggregation, or direct alignment operations to achieve time-stamp synchronization of multi-device monitoring data. This lays a high-quality data foundation for subsequent fusion analysis. Simultaneously, it calculates the quality assessment score and weight coefficient of the device monitoring data using historical error rate, signal-to-noise ratio, and data freshness, and applies the three sigma criterion (3... The system sets weight thresholds to exclude low-quality data from the fusion process, and updates weights at fixed intervals or event triggers to ensure the high reliability of the data used in the analysis, avoiding a one-size-fits-all approach. Furthermore, it trains a Long Short-Term Memory (LSTM) neural network using weighted historical data with the same time stamp, using multi-dimensional operational status labels such as voltage deviation assessment values ​​and line overload risk probabilities as training targets, and optimizes the model through a weighted loss function. Compared to traditional assessment methods that rely on human experience, this model can output equipment operational status labels in real time and accurately, providing a quantitative basis for operational adjustments. Finally, based on the comparison between equipment operational status labels and preset safety thresholds, if an over-limit situation occurs, fault tree analysis is used to trace the root cause (directly monitorable equipment parameters), and combined with an expert experience rule base to generate targeted compensation instructions (such as adjusting distributed power output and switching reactive power compensation capacitors), realizing a shift from passively responding to faults to proactively predicting and adjusting, effectively balancing the safety, economy, and reliability of distribution network operation, and avoiding the problems of decision-making lag or blind measures in traditional dispatching. Attached Figure Description

[0011] Figure 1 This is a flowchart of a method for constructing a power distribution network operation architecture according to the present invention.

[0012] Figure 2 This is a flowchart illustrating the specific process of resampling data from all devices to the same timestamp sequence based on the kernel density formula of the present invention.

[0013] Figure 3 The flowchart for obtaining quality assessment scores and weighting coefficients for equipment operation monitoring data and updating and optimizing them is provided in this invention.

[0014] Figure 4 This invention provides a flowchart for outputting the current operating status labels of distribution network equipment in the construction of the distribution network operation architecture evaluation model.

[0015] Figure 5 This is a structural diagram of the electronic device proposed in this invention.

[0016] Figure 6 This is a schematic diagram of the structure of the computer-readable storage medium proposed in this invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1 As shown, a method for constructing a power distribution network operation architecture includes: Based on historical data of power distribution network operation, monitoring data of different equipment operation are extracted, and a unified time-scaled dataset of power distribution network is established. Based on the same time-stamped dataset of the power distribution network, and using the kernel density formula, the data of all devices are resampled to the same timestamp sequence. Based on the device's sampling frequency and timestamp, determine and perform alignment operations between the current device operation monitoring data and the same timestamp sequence; Based on the historical error rate, signal-to-noise ratio, and data freshness of different equipment operation monitoring data, obtain the quality assessment score and weight coefficient of the equipment operation monitoring data, and update and optimize them accordingly; Based on the weighting coefficients of the equipment operation monitoring data, an evaluation model for the power distribution network operation architecture is constructed using an LSTM neural network model, and the current operation status label of the power distribution network equipment is output. Based on the current operating status labels of the distribution network equipment, and using fault tree analysis and expert experience rules, the operating parameters of the equipment in the distribution network operation architecture are dynamically adjusted.

[0019] It can be explained that the prediction accuracy of existing neural network models depends on the quality of the training sample set. Among these factors, resampling equipment data to the same timestamp sequence and ensuring data quality are paramount. Inconsistent timestamp sequences can easily lead to significant deviations in the prediction results of neural network models. Therefore, this embodiment extracts monitoring data from different equipment from historical data of the power distribution network operation, fits the probability density curve of the equipment sampling frequency using the kernel density formula, selects the target sampling frequency with the lowest total information loss rate, generates the same timestamp sequence, and, based on the relationship between the equipment sampling frequency and the target frequency, employs interpolation, aggregation, or direct alignment operations to achieve time-stamp synchronization of multi-equipment monitoring data. This lays a high-quality data foundation for subsequent fusion analysis. Simultaneously, it calculates the quality assessment score and weight coefficient of the equipment monitoring data using historical error rate, signal-to-noise ratio, and data freshness, and based on 3... The criteria set weight thresholds to exclude low-quality data from the fusion process, and update the weights at fixed intervals or event triggers to ensure that the data involved in the analysis always has high credibility, avoiding the one-size-fits-all use of data. Furthermore, an LSTM neural network is trained using historical data with the same time scale and weighted, with multi-dimensional operating status labels such as voltage deviation assessment values ​​and line overload risk probability as training targets, and the model is optimized through a weighted loss function. Compared with traditional assessment methods that rely on human experience, this model can output equipment operating status labels in real time and accurately, providing a quantitative basis for operation adjustments. Finally, based on the comparison between equipment operating status labels and preset safety thresholds, if an over-limit situation occurs, fault tree analysis is used to trace the root cause (directly monitorable equipment parameters) to locate the root cause, and combined with an expert experience rule base to generate targeted compensation instructions (such as adjusting the output of distributed power sources and switching reactive power compensation capacitors), realizing the transformation from passively responding to faults to proactively predicting and adjusting, effectively balancing the safety, economy and reliability of distribution network operation, and avoiding the problems of decision lag or blind measures in traditional dispatching.

[0020] The step of extracting monitoring data of different equipment based on historical data of distribution network operation and establishing a common time-scaled dataset for the distribution network specifically includes: Historical data on the operation of the distribution network can be obtained through the historical database of the distribution network. Based on historical data of power distribution network operation, monitoring data of different equipment operation are extracted and grouped and numbered according to the type of monitoring data; Based on the monitoring data of different equipment operation, obtain the sampling frequency, accuracy and timestamp of each group of different equipment operation data; Establish a unified time-stamped dataset for the power distribution network, recording the type, monitoring data, sampling frequency, accuracy, and timestamp of the operating data of different equipment in each group.

[0021] It can be explained that data consistency is key to ensuring the accuracy and reliability of fusion results when multiple data are fused. The core challenge lies in the fact that differences in communication protocols, sampling mechanisms, and accuracy standards among different devices lead to significant inconsistencies in data timescale, sampling frequency, and data accuracy. Therefore, this embodiment extracts the operation monitoring data of multiple devices from the historical database of the distribution network, groups and numbers them according to data type, and structurally acquires and records key metadata such as sampling frequency, accuracy, and timestamps for each group of data. Based on this, a distribution network unified timescale dataset is established. While integrating the multiple monitoring data themselves, it more completely preserves their original sampling attributes, providing a solid and traceable data foundation for subsequent fusion optimization that achieves data consistency and accuracy coordination.

[0022] Reference Figure 2As shown, the specific steps of resampling data from all devices to the same timestamp sequence based on the kernel density formula include: From the same time-scaled dataset of the power distribution network, the sampling frequency of all devices is extracted as a sample set, and then fitted into the kernel density formula to obtain the probability density curve of the sampling frequency. Based on the probability density curve of the sampling frequency, find the frequency range corresponding to the density peak. This range represents the main concentration range of equipment sampling frequencies, indicating the mainstream frequency level of distribution network data sampling, and can cover the original frequency characteristics of most equipment to the greatest extent. Within the frequency range corresponding to the density peak, three types of candidate frequencies are selected as preliminary candidate target frequencies. These three types of candidate frequencies include: high-frequency candidates, mid-frequency candidates, and low-frequency candidates. Among them, the ranges of high-frequency candidates, mid-frequency candidates, and low-frequency candidates are obtained by judging the sampling frequency and timestamp of the device and performing an alignment operation between the current device operation monitoring data and the same timestamp sequence; For each type of preliminary candidate target frequency, calculate the total information loss rate of all equipment operation monitoring data resampled to that frequency; The candidate frequency with the lowest total information loss rate is selected as the final target sampling frequency. If the total information loss rates of multiple candidate frequencies are similar, the candidate frequency with the lowest value is selected first. Using the target sampling frequency as the interval, generate an equally spaced sequence of the same timestamps that covers the complete time window of the data to be fused.

[0023] It can be explained that when performing multi-data fusion, it is necessary to ensure that the input data has a consistent timestamp sequence. Since the data sampling frequencies of different devices are different, it is easy for multiple sets of data to be difficult to synchronize in time, resulting in data misalignment, which affects the accuracy of the fusion result. To this end, this embodiment extracts the sampling frequency from the historical data of each device, fits its probability density curve based on the kernel density estimation method, and determines the target sampling frequency that can retain the original frequency characteristics of most devices to the greatest extent by finding the frequency interval corresponding to the density peak. Furthermore, in order to balance the impact of data interpolation and downsampling operations on the overall quality, this embodiment calculates the total information loss rate generated by resampling all data to the frequency under different candidate target frequencies, selects the final target sampling frequency based on the principle of the lowest loss rate, and generates the same timestamp sequence at equal intervals accordingly, thereby maximizing the accuracy and reliability of data fusion while maintaining the same timestamp sequence. The expression for the nuclear density formula is as follows: In the formula, For the first Density estimate at point, For the total number of devices, To control the width of the kernel function, For the first The sampling frequency of each device Let be the independent variable of the probability density function, representing an arbitrary sampling frequency value. For kernel functions, the Gaussian kernel function is commonly used. ; The expression for the total information loss rate is: In the formula, The total information loss rate, For the total number of devices, , and are the weighting coefficients for interpolation and downsampling, respectively, representing the importance of interpolation error and downsampling loss in the total loss. These coefficients are set based on prior knowledge and expert experience. , For the first Normalization error caused by interpolation operation in individual device data For the first The loss of normalized information in device data due to downsampling operations; in, The main reason for this discrepancy is the difference between the interpolation function estimate and the actual operating state of the equipment. This discrepancy can be quantified using an error function, such as the maximum absolute error or mean square error between the interpolation points and the actual data points in linear or cubic spline interpolation, and then normalized to eliminate the influence of dimensions. This loss primarily stems from the loss of high-frequency detail information during data aggregation. This loss is quantified based on whether mean aggregation or maximum / minimum aggregation methods are used, such as by comparing the changes in variance, range, and spectral energy statistical characteristics of the data sequence before and after aggregation.

[0024] The step of determining and performing alignment operations between the current device operation monitoring data and the same timestamp sequence based on the device's sampling frequency and timestamp specifically includes: Based on the same time-scaled dataset of the power distribution network, obtain the sampling frequency and timestamp in the operation monitoring data sequence of each device; For each device's operational monitoring data sequence, compare its sampling frequency. With the final target sampling frequency The size relationship is determined, and the entire sequence is optimized accordingly based on the comparison results; like Then, linear interpolation or cubic spline interpolation methods are used to resample the equipment operation monitoring data sequence onto the same timestamp sequence; like If so, it is determined that the equipment operation monitoring data has been aligned with the timestamp sequence and no further processing is required; like Then, the mean aggregation or maximum / minimum aggregation method is used to aggregate the equipment operation monitoring data sequence within a time window centered on each time point in the same timestamp sequence, and use the result as the value of that time point, thereby resampling to the same timestamp sequence.

[0025] The purpose of determining the same resampling time reference is to: based on the scientifically selected final target sampling frequency. Each device's data sequence is individually processed to align it all to the same equally spaced timestamp sequence, laying the foundation for subsequent data fusion. Specifically: for sampling frequencies lower than […], For low-frequency devices, interpolation methods are needed to improve their time resolution in order to estimate their data values ​​at the target time point. For sampling frequencies equal to... The equipment's data is naturally synchronized with the target time series, therefore requiring no processing and preserving the originality of the data to the maximum extent. This is especially beneficial for sampling frequencies higher than [specific frequency range missing]. For high-frequency devices, the temporal resolution needs to be reduced through aggregation methods in order to suppress high-frequency noise and significantly reduce the amount of data while preserving the overall trend of the data, thereby improving computational efficiency. This step ensures the consistency of multi-source heterogeneous data in the temporal dimension and is a key technical prerequisite for ensuring the accuracy and reliability of the fusion results.

[0026] Reference Figure 3 As shown, the process of acquiring, updating, and optimizing the quality assessment scores and weighting coefficients of equipment operation monitoring data specifically includes: Based on the historical error rate, signal-to-noise ratio, and data freshness of different equipment operation monitoring data, a quality assessment score for the equipment operation monitoring data is obtained using a weighted summation formula. Based on the quality assessment score of the equipment operation monitoring data, the weight coefficients of the equipment operation monitoring data are obtained using the weighted average method. The latest quality assessment score and weighting coefficient are calculated using real-time monitoring data of equipment operation at fixed time intervals or event triggering methods. Based on historical equipment operation monitoring data, and using 3 Criteria for obtaining boundary thresholds for weighting coefficients of equipment operation monitoring data; For equipment operation monitoring data with a weighting coefficient less than the boundary threshold, set its weighting coefficient to zero and remove it from this data fusion. If the sampling frequency of the real-time monitoring data of the equipment operation is obtained, and it is determined whether the frequency change value exceeds the tolerance or new monitoring parameters of the equipment operation are detected, the target sampling frequency is updated and a resampling and alignment operation is performed.

[0027] It can be explained that the accuracy of multi-source data fusion in the power grid highly depends on the quality of the input data. If monitoring data from equipment with large errors, low signal-to-noise ratios, and poor timeliness is used directly, the fusion results will deviate from reality and may even lead to scheduling decision errors. Therefore, it is necessary to first assess the data quality, then assign high weights to high-quality data and low weights to low-quality data to ensure the reliability of the fusion. Among these, the historical error rate of monitoring data from different equipment reflects the accuracy of the data; the lower the error rate, the more reliable the data. The signal-to-noise ratio reflects the purity of the signal; the higher the signal-to-noise ratio, the higher the proportion of effective signals. Data freshness reflects timeliness; the newer the data, the more accurately it represents the current state of the power grid. This is the core dimension for measuring quality. Through a weighted summation formula, these three dimensions are quantified into a quality assessment score. The higher the score, the better the data quality. The weights of each dimension in the formula are determined by expert experience or historical verification based on the emphasis of the distribution network operation on accuracy, purity, and timeliness. The fixed time interval is set comprehensively based on the rate of change of equipment status, communication bandwidth limitations, and system computing resources. The quality assessment score expression for the equipment operation monitoring data is as follows: In the formula, For the first Quality assessment score of equipment operation monitoring data , , These are the weighting coefficients for historical error rate, signal-to-noise ratio, and data freshness, respectively. , , , The first Historical error rate, signal-to-noise ratio, and data freshness value for each device; The weighting coefficient expression for the equipment operation monitoring data is: In the formula, For the first Weighting coefficients for individual equipment operation monitoring data For the total number of devices, For the first The quality assessment score of the equipment operation monitoring data.

[0028] Reference Figure 4 As shown, the specific operating status labels of the current distribution network equipment output by the constructed distribution network operation architecture evaluation model include: Based on the historical monitoring data of equipment operation and the operation status labels after being aligned with the same timestamp sequence and weighted by weight coefficients, an LSTM neural network model training sample set is created, which includes a training set and a validation set. The operating status label includes one or more of the following: voltage deviation assessment value, line overload risk probability, and power supply reliability index. The equipment operation monitoring data, after being aligned with the same timestamp sequence and weighted by weight coefficients, is used as the input to the LSTM neural network model; The operating status labels of power distribution network equipment are used as the training target of the LSTM neural network model; With the goal of minimizing the error between the predicted and actual values ​​of the running status labels, a loss function for the LSTM neural network model is constructed based on the mean squared error formula. If there are multiple states for the operation status label, the total evaluation value of the operation status label is obtained by weighted summation. The loss function of the LSTM neural network model is constructed with the goal of minimizing the error between the predicted value and the actual value of the total evaluation value of the status label. The weight of the operation status label is determined based on the expert scoring method according to the degree of influence of the operation status label on the operation stability of the distribution network equipment. Based on the loss function of the LSTM neural network model, the gradient of the loss function with respect to the weights of each layer of the neural network model is calculated using the error backpropagation method. The Adam optimizer is used to iteratively update the weights of each layer of the neural network model based on the gradient of the weights of each layer. Based on validation set detection, when the error value no longer decreases or falls below a for N consecutive training epochs on the validation set, the training is completed and the final model weights are saved, where a is an error threshold preset according to the validation set performance. The trained LSTM neural network model is defined as the distribution network operation architecture evaluation model. By inputting equipment operation monitoring data that has been aligned with the same timestamp sequence and weighted by weight coefficients, the model outputs the current operation status label of the distribution network equipment.

[0029] This can be explained by the fact that the operating status of the power distribution network is a complex feature hidden behind massive amounts of multi-source monitoring data. Traditional methods struggle to directly and accurately extract status information from this data. This embodiment utilizes the powerful sequence modeling capabilities of the LSTM neural network to learn a mapping function from historical, aligned, and weighted data. This function can automatically output a quantitative indicator representing the current operating status of the power grid, i.e., an operating status label, based on the latest monitoring data. The LSTM neural network processes data step by step over time. It assumes that at each time step, the input feature vector contains the observation values ​​of all variables at the same moment. In the original data, the equipment frequencies are different, and at each time step... Some devices have data, while others have missing or misaligned data. This violates the basic input assumptions of LSTM neural networks. Therefore, this embodiment ensures that, based on the alignment of the same timestamp sequence, the model can receive a complete feature vector of all devices at the same time for each same timestamp sequence. This enables the LSTM neural network to correctly learn the correlation between data from different devices at the same time. Furthermore, by weighting the data with weight coefficients, the LSTM neural network can focus on high-quality device operation monitoring data, avoiding the impact of device monitoring data with large errors, low signal-to-noise ratios, and poor timeliness on network training, thereby improving the accuracy and precision of LSTM neural network inference.

[0030] The step of dynamically adjusting the equipment operating parameters in the distribution network operation architecture based on the current operating status labels of the distribution network equipment, fault tree analysis, and expert experience rules specifically includes: Based on the safety, economy, and reliability of power distribution network equipment operation, and using an expert scoring method, safety thresholds for equipment operating status labels are obtained. Based on the current operating status label of the distribution network equipment output by the distribution network operation architecture evaluation model, it is determined whether the current operating status label of the distribution network equipment exceeds the safety threshold. If so, the root cause of the operating status label exceeding the limit is traced according to the fault tree analysis, that is, one or more bottom events are located. Combined with the preset expert experience rule base, the corresponding equipment operation compensation instruction is generated. If not, it remains unchanged. The bottom events are the operating status parameters of the distribution network equipment that can be directly monitored. The equipment operation compensation command includes one or more of the following: adjusting the output of distributed power sources, switching reactive power compensation capacitors, and adjusting the tap changers of on-load tap-changing transformers.

[0031] This can be explained by using the distribution network operation architecture evaluation model to output the current operation status label of the distribution network equipment. By comparing it with the safety threshold of the equipment operation status label, it is possible to determine whether there is an anomaly in the operation status of the distribution network equipment before judgment. Based on fault tree analysis and expert experience rules, the root cause of the operation status label exceeding the limit can be traced, that is, one or more bottom events can be located. Combined with the preset expert experience rule library, corresponding equipment operation compensation instructions are generated. Thus, equipment operation compensation is completed according to the directly monitorable distribution network equipment operation status parameters, ensuring the stable and safe operation of the distribution network equipment. Among them, fault tree analysis is a top-down, deductive system reliability analysis method. It starts from an undesirable system failure event (called the "top event") and analyzes all the direct and indirect causes that led to its occurrence layer by layer until it is traced back to the most basic initial cause event that does not need to be further decomposed (called the "bottom event"). The entire process uses logic gates to connect the various events, and finally forms an inverted tree logic diagram, which clearly shows all possible paths and combinations of the top event and the bottom event.

[0032] Furthermore, based on the same inventive concept as the above-mentioned distribution network operation architecture construction method, this embodiment proposes a distribution network operation architecture construction system, including: The data optimization and processing module is used to extract monitoring data of different equipment operation based on historical data of distribution network operation and establish a common time-scaled dataset for the distribution network; based on the common time-scaled dataset of the distribution network and the kernel density formula, resample the data of all equipment to the same timestamp sequence; determine and perform alignment operation between the current equipment operation monitoring data and the same timestamp sequence by using the sampling frequency and timestamp of the equipment; and obtain the quality assessment score and weight coefficient of the equipment operation monitoring data based on the historical error rate, signal-to-noise ratio and data freshness of the different equipment operation monitoring data, and update and optimize them. The dynamic allocation module is used to construct a power distribution network operation architecture evaluation model based on the weight coefficients of equipment operation monitoring data and an LSTM neural network model, and output the current operation status label of the power distribution network equipment; based on the current operation status label of the power distribution network equipment, the module dynamically adjusts the equipment operation parameters in the power distribution network operation architecture based on fault tree analysis and expert experience rules. The data optimization processing module includes: The dataset unit is used to extract monitoring data of different equipment operation based on historical data of distribution network operation, and to establish a common time-scaled dataset of the distribution network. The timestamp alignment unit is used to resample the data of all devices to the same timestamp sequence based on the kernel density formula of the same time-scaled dataset of the power distribution network. A data timestamp alignment unit is used to determine and perform alignment operations between the current device operation monitoring data and the same timestamp sequence based on the device's sampling frequency and timestamp. The data optimization unit is used to obtain the quality assessment score and weight coefficient of the equipment operation monitoring data based on the historical error rate, signal-to-noise ratio and data freshness of different equipment operation monitoring data, and to update and optimize it. The dynamic allocation module includes: The model building unit is used to build a power distribution network operation architecture evaluation model based on the weight coefficients of the equipment operation monitoring data and the LSTM neural network model, and output the current operation status label of the power distribution network equipment. The dynamic adjustment unit is used to dynamically adjust the operating parameters of the equipment in the distribution network operation architecture based on the current operating status labels of the distribution network equipment and fault tree analysis and expert experience rules.

[0033] Furthermore, the method according to the embodiments of this application can also be achieved by means of... Figure 5 The architecture of the electronic device shown is used to implement this. For example... Figure 5 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store a method for constructing a power distribution network operation architecture provided in this application. The electronic device 500 may also include a user interface 508. Of course, Figure 5 The architecture shown is merely exemplary and can be omitted as needed when implementing different devices. Figure 5 One or more components in the illustrated electronic device.

[0034] Figure 6 This is a schematic diagram of a computer-readable storage medium structure provided in one embodiment of this application. Figure 6 The diagram illustrates a computer-readable storage medium 600 according to one embodiment of this application. The computer-readable storage medium 600 stores computer-readable instructions. When executed by a processor, the computer-readable instructions can perform a method for constructing a power distribution network operation architecture according to an embodiment of this application, as described above with reference to the accompanying drawings. The storage medium 600 includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0035] In summary, the advantages of this invention are: to achieve dynamic sampling of equipment operation data, to apply a power distribution network operation architecture evaluation model, and to solve the problem of synergistic operation of power distribution networks under multiple constraints in terms of economy and reliability.

[0036] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for constructing a power distribution network operation architecture, characterized in that, include: Based on historical data of power distribution network operation, monitoring data of different equipment operation are extracted, and a unified time-scaled dataset of power distribution network is established. Based on the same time-stamped dataset of the power distribution network, and using the kernel density formula, the data of all devices are resampled to the same timestamp sequence. Based on the device's sampling frequency and timestamp, determine and perform alignment operations between the current device operation monitoring data and the same timestamp sequence; Based on the historical error rate, signal-to-noise ratio, and data freshness of different equipment operation monitoring data, obtain the quality assessment score and weight coefficient of the equipment operation monitoring data, and update and optimize them accordingly; Based on the weighting coefficients of the equipment operation monitoring data, and using the long short-term memory network neural network model, a power distribution network operation architecture evaluation model is constructed, and the current operation status label of the power distribution network equipment is output. Based on the current operating status labels of the distribution network equipment, and using fault tree analysis and expert experience rules, the operating parameters of the equipment in the distribution network operation architecture are dynamically adjusted.

2. The method for constructing a power distribution network operation architecture according to claim 1, characterized in that, The step of extracting monitoring data of different equipment based on historical data of distribution network operation and establishing a common time-scaled dataset for the distribution network specifically includes: Historical data on the operation of the distribution network can be obtained through the historical database of the distribution network. Based on historical data of power distribution network operation, monitoring data of different equipment operation are extracted and grouped and numbered according to the type of monitoring data; Based on the monitoring data of different equipment operation, obtain the sampling frequency, accuracy and timestamp of each group of different equipment operation data; Establish a unified time-stamped dataset for the power distribution network, recording the type, monitoring data, sampling frequency, accuracy, and timestamp of the operating data of different equipment in each group.

3. The method for constructing a power distribution network operation architecture according to claim 2, characterized in that, The step of resampling all device data to the same timestamp sequence based on the same time-stamped dataset of the power distribution network and using the kernel density formula specifically includes: From the same time-scaled dataset of the power distribution network, the sampling frequency of all devices is extracted as a sample set, and then fitted into the kernel density formula to obtain the probability density curve of the sampling frequency. Based on the probability density curve of the sampling frequency, find the frequency range corresponding to the density peak. This range represents the main concentration range of equipment sampling frequencies, indicating the mainstream frequency level of distribution network data sampling, and can cover the original frequency characteristics of most equipment to the greatest extent. Within the frequency range corresponding to the density peak, three types of candidate frequencies are selected as preliminary candidate target frequencies. These three types of candidate frequencies include: high-frequency candidates, mid-frequency candidates, and low-frequency candidates. For each type of preliminary candidate target frequency, calculate the total information loss rate of all equipment operation monitoring data resampled to that frequency; The candidate frequency with the lowest total information loss rate is selected as the final target sampling frequency. If the total information loss rates of multiple candidate frequencies are similar, the candidate frequency with the lowest value is selected first. Using the target sampling frequency as the interval, generate an equally spaced sequence of the same timestamps that covers the complete time window of the data to be fused.

4. The method for constructing a power distribution network operation architecture according to claim 3, characterized in that, The step of determining and performing alignment operations between the current device operation monitoring data and the same timestamp sequence based on the device's sampling frequency and timestamp specifically includes: Based on the same time-scaled dataset of the power distribution network, obtain the sampling frequency and timestamp in the operation monitoring data sequence of each device; For each device's operational monitoring data sequence, compare the sampling frequencies. With the final target sampling frequency The size relationship is determined, and the entire sequence is optimized accordingly based on the comparison results; like Then, linear interpolation or cubic spline interpolation methods are used to resample the equipment operation monitoring data sequence onto the same timestamp sequence; like If so, it is determined that the equipment operation monitoring data has been aligned with the timestamp sequence and no further processing is required; like Then, the mean aggregation or maximum / minimum aggregation method is used to aggregate the equipment operation monitoring data sequence within a time window centered on each time point in the same timestamp sequence, and the result is used as the value of that time point, thereby resampling to the same timestamp sequence.

5. The method for constructing a power distribution network operation architecture according to claim 4, characterized in that, The process of obtaining quality assessment scores and weighting coefficients for equipment operation monitoring data based on historical error rates, signal-to-noise ratios, and data freshness of different equipment operation monitoring data, and then updating and optimizing them, specifically includes: Based on the historical error rate, signal-to-noise ratio, and data freshness of different equipment operation monitoring data, a quality assessment score for the equipment operation monitoring data is obtained using a weighted summation formula. Based on the quality assessment score of the equipment operation monitoring data, the weight coefficients of the equipment operation monitoring data are obtained using the weighted average method. The latest quality assessment score and weighting coefficient are calculated using real-time monitoring data of equipment operation at fixed time intervals or event triggering methods. Based on the historical monitoring data of equipment operation, and using the Three Sigma criterion, the boundary thresholds for the weighting coefficients of the equipment operation monitoring data are obtained. For equipment operation monitoring data with a weighting coefficient less than the boundary threshold, the weighting coefficient will be set to zero and the data will be removed from this data fusion. If the sampling frequency of the real-time monitoring data of the equipment operation is obtained, and it is determined whether the frequency change value exceeds the tolerance or a new monitoring parameter of the equipment operation is detected, the target sampling frequency is updated and a resampling and alignment operation is performed.

6. The method for constructing a power distribution network operation architecture according to claim 5, characterized in that, The step of constructing a power distribution network operation architecture evaluation model based on the weighting coefficients of equipment operation monitoring data and a long short-term memory network neural network model, and outputting the current operation status label of the power distribution network equipment, specifically includes: Based on the historical monitoring data of equipment operation and the operation status labels after being aligned with the same timestamp sequence and weighted by weight coefficients, a training sample set for the Long Short-Term Memory Network neural network model is created, which includes a training set and a validation set. The operating status label includes one or more of the following: voltage deviation assessment value, line overload risk probability, and power supply reliability index. The equipment operation monitoring data, after being aligned with the same timestamp sequence and weighted by weight coefficients, is used as the input to the long short-term memory network neural network model; The operating status labels of power distribution network equipment are used as the training target for the long short-term memory network neural network model; With the goal of minimizing the error between the predicted and actual values ​​of the running status labels, a loss function for the long short-term memory network neural network model is constructed based on the mean square error formula. If there are multiple states for the operation status label, the total evaluation value of the operation status label is obtained by weighted summation. The loss function of the long short-term memory network neural network model is constructed with the goal of minimizing the error between the predicted value and the actual value of the total evaluation value of the status label. The weight of the operation status label is determined based on the expert scoring method according to the degree of influence of the operation status label on the operation stability of the distribution network equipment. Based on the loss function of the Long Short-Term Memory (LSTM) network model, the gradient of the loss function with respect to the weights of each layer of the network model is calculated using backpropagation. The weights of each layer in the neural network model are iteratively updated based on the gradient of the weights of each layer. Based on validation set detection, when the error value no longer decreases or falls below a for N consecutive training rounds on the validation set, the training is completed and the final model weights are saved, where a is an error threshold preset according to the performance of the validation set. The trained Long Short-Term Memory (LSTM) neural network model is defined as the power distribution network operation architecture evaluation model. By inputting equipment operation monitoring data that has been aligned with the same timestamp sequence and weighted by weight coefficients, the model outputs the current operation status label of the power distribution network equipment.

7. The method for constructing a power distribution network operation architecture according to claim 6, characterized in that, The step of dynamically adjusting the operating parameters of equipment in the distribution network operation architecture based on the current operating status labels of the distribution network equipment, fault tree analysis, and expert experience rules specifically includes: Based on the safety, economy, and reliability of power distribution network equipment operation, and using an expert scoring method, safety thresholds for equipment operating status labels are obtained. Based on the current operating status label of the distribution network equipment output by the distribution network operation architecture evaluation model, it is determined whether the current operating status label of the distribution network equipment exceeds the safety threshold. If so, the root cause of the operating status label exceeding the limit is traced according to the fault tree analysis, that is, one or more bottom events are located. Combined with the preset expert experience rule base, the corresponding equipment operation compensation instruction is generated. If not, it remains unchanged. The bottom events are the operating status parameters of the distribution network equipment that can be directly monitored. The equipment operation compensation command includes one or more of the following: adjusting the output of distributed power sources, switching reactive power compensation capacitors, and adjusting the tap changers of on-load tap-changing transformers.

8. A power distribution network operation architecture construction system, characterized in that, The method for constructing a distribution network operation architecture as described in any one of claims 1-7 includes: The data optimization and processing module is used to extract monitoring data of different equipment operation based on historical data of distribution network operation and establish a common time-scaled dataset for the distribution network; based on the common time-scaled dataset of the distribution network and the kernel density formula, resample the data of all equipment to the same timestamp sequence; determine and perform alignment operation between the current equipment operation monitoring data and the same timestamp sequence by using the sampling frequency and timestamp of the equipment; and obtain the quality assessment score and weight coefficient of the equipment operation monitoring data based on the historical error rate, signal-to-noise ratio and data freshness of the different equipment operation monitoring data, and update and optimize them. The dynamic allocation module is used to construct a power distribution network operation architecture evaluation model based on the weight coefficients of equipment operation monitoring data and a long short-term memory network neural network model, and output the current operation status label of the power distribution network equipment; based on the current operation status label of the power distribution network equipment, and based on fault tree analysis and expert experience rules, dynamically adjust the equipment operation parameters in the power distribution network operation architecture. The data optimization processing module includes: The dataset unit is used to extract monitoring data of different equipment operation based on historical data of distribution network operation, and to establish a common time-scaled dataset of the distribution network. The timestamp alignment unit is used to resample the data of all devices to the same timestamp sequence based on the kernel density formula of the same time-scaled dataset of the power distribution network. A data timestamp alignment unit is used to determine and perform alignment operations between the current device operation monitoring data and the same timestamp sequence based on the device's sampling frequency and timestamp. The data optimization unit is used to obtain the quality assessment score and weight coefficient of the equipment operation monitoring data based on the historical error rate, signal-to-noise ratio and data freshness of different equipment operation monitoring data, and to update and optimize it. The dynamic allocation module includes: The model building unit is used to build a power distribution network operation architecture evaluation model based on the weight coefficients of the equipment operation monitoring data and the long short-term memory network neural network model, and output the current operation status label of the power distribution network equipment. The dynamic adjustment unit is used to dynamically adjust the operating parameters of the equipment in the distribution network operation architecture based on the current operating status labels of the distribution network equipment and fault tree analysis and expert experience rules.

9. An electronic device, characterized in that, include: At least one processor; And, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a method for constructing a power distribution network operation architecture as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements a method for constructing a power distribution network operation architecture according to any one of claims 1-7.

Images