Power distribution network reliability fast evaluation method and system based on uncertainty factor
By combining multi-scale uncertainty spectral decomposition and causal decoupling mechanisms with topological elastic mapping and adaptive reduction framework, this method solves the complex coupling problem of multiple uncertainty factors that traditional distribution network reliability assessment methods fail to handle, and achieves fast and accurate distribution network reliability assessment, applicable to distribution networks of different sizes and complexities.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JILIN ELECTRIC POWER RES INST LTD
- Filing Date
- 2026-05-06
- Publication Date
- 2026-06-05
AI Technical Summary
Traditional distribution network reliability assessment methods fail to fully consider the complex coupling relationships between various uncertainties, resulting in high computational complexity. This makes it difficult to meet the needs of modern power systems for rapid and efficient assessment, especially in large-scale and complex distribution networks where real-time or near-real-time reliability assessment is difficult to achieve.
A multi-scale uncertainty spectrum decomposition-causal decoupling fast equivalent evaluation mechanism is adopted, combined with a hierarchical fast iterative framework of topological elastic mapping and adaptive reduction of uncertainty coupling strength. The correlation strength between uncertainty factors is dynamically identified, and the reliability of the distribution network is rapidly evaluated through equivalent reliability perturbation parameters and fast reconfiguration algorithm.
It significantly improves the accuracy and efficiency of assessment results, enables real-time monitoring and evaluation of the reliability recovery capability of distribution networks under abnormal conditions, provides support for emergency management and smart grid optimization decision-making, and is applicable to distribution networks of different sizes and complexities.
Smart Images
Figure CN122155540A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power systems, and more specifically, to a method and system for rapid reliability assessment of distribution networks based on uncertainty factors. Background Technology
[0002] In power systems, the distribution network, as a crucial link connecting the transmission network and end users, directly impacts users' electricity experience and the stable operation of the power system. With the large-scale integration of distributed power sources, such as solar and wind power, and the increasing diversification of electricity demand, the distribution network faces numerous uncertainties, which significantly affect its reliability.
[0003] Traditional methods for assessing the reliability of distribution networks have several limitations in handling these uncertainties. Firstly, they often fail to adequately consider the complex coupling relationships between various uncertainties. For example, factors such as weather fluctuations, distributed generation output fluctuations, load randomness, equipment aging and degradation, and the probability of protection malfunctions all interact and jointly affect distribution network reliability, and traditional methods struggle to accurately characterize this combined impact. Secondly, traditional assessment methods are computationally complex, making it difficult to meet the demands of modern power systems for rapid and efficient assessments. For instance, while the traditional Monte Carlo method can assess distribution network reliability relatively accurately, it involves enormous computational demands and is time-consuming, making it difficult to achieve real-time or near-real-time reliability assessments, especially in large-scale, complex distribution networks.
[0004] Furthermore, with the continuous expansion of distribution network scale and the increasing complexity of its structure, traditional assessment methods based on deterministic models cannot effectively cope with the challenges posed by various uncertainties. This leads to significant deviations between assessment results and actual conditions, failing to provide accurate and reliable decision-making basis for the planning, operation, and maintenance of distribution networks. Therefore, developing a fast and accurate distribution network reliability assessment method that can comprehensively consider multiple uncertainty factors is of significant practical importance. Summary of the Invention
[0005] The purpose of this invention is to provide a method and system for rapid reliability assessment of distribution networks based on uncertainty factors. It solves the problem that traditional distribution network reliability assessment methods are usually based on static models and ignore the influence of many dynamic uncertainty factors, resulting in low accuracy and high computational complexity of the assessment results. In particular, when the distribution network involves multiple uncertainty factors, such as weather changes and distributed power source fluctuations, existing assessment methods often cannot handle them efficiently and cannot provide accurate reliability indicators in real time, thus failing to meet the application requirements.
[0006] This invention achieves the above objective through the following technical solution: a rapid reliability assessment method for distribution networks based on uncertainty factors, comprising the following steps:
[0007] S1. Obtain basic operating data and raw data of multiple uncertainty factors of the power distribution network, and preprocess the basic operating data and raw data of uncertainty factors to obtain a standardized dataset;
[0008] S2. Construct a fast equivalent evaluation mechanism for multi-scale uncertainty spectrum decomposition and causal decoupling, which maps multiple types of uncertainty factors to a unified reliability influence spectrum domain and transforms them into equivalent reliability perturbation parameters.
[0009] S3. Establish a fast reliability reconfiguration algorithm based on topology elastic mapping to characterize the structural recoverability of the distribution network under fault reconfiguration, load transfer, and microgrid islanding operation, and quickly solve the equivalent reliability index of the system.
[0010] S4. Build a hierarchical rapid iterative framework for adaptive reduction of uncertainty coupling strength, dynamically identify the correlation strength between uncertainty factors and automatically reduce the participation of weak coupling factors, and realize rapid iterative evaluation of distribution network reliability in the main coupling factor subspace.
[0011] Furthermore, in step S1, the basic operational data includes:
[0012] The topology parameters of the distribution network, the rated technical parameters of the equipment, the basic load data of the nodes, and the basic parameters of the installed capacity and output of distributed power sources;
[0013] The original data for the various uncertainty factors include meteorological element fluctuation monitoring data, distributed power generation actual output fluctuation time series data, distribution network node load random change data, power equipment full life cycle aging and degradation status monitoring data, and relay protection device operation and malfunction probability statistics, corresponding to five types of uncertainty factors: meteorological fluctuations, distributed power generation output fluctuations, load randomness, equipment aging and degradation, and protection malfunction probability.
[0014] Furthermore, in step S1, the preprocessing includes:
[0015] The basic operational data and the original data of uncertainty factors are sequentially processed by outlier removal, missing value completion, and dimension normalization. The dimension normalization adopts the min-max standardization method.
[0016] The standardized dataset is divided into training, validation and test sets according to a preset ratio, which are used for iterative optimization of model parameters, hyperparameter tuning and overfitting monitoring, and final performance evaluation of the model, respectively.
[0017] Furthermore, in step S2, the construction of a fast equivalent evaluation mechanism for multi-scale uncertainty spectral decomposition-causal decoupling includes:
[0018] A time-frequency joint spectral representation model for multi-source uncertainty factors is established, and time-series and static uncertainty factors are transformed into corresponding spectral features and integrated to form a multi-source uncertainty spectral matrix.
[0019] Construct an influence transmission path matrix based on causal strength, and calculate the causal influence strength of various uncertainty factors on the core reliability indicators of the distribution network by combining the electrical topology and fault propagation law of the distribution network.
[0020] The multi-source uncertainty spectrum matrix and the influence transmission path matrix are operated on, and the results are subjected to spectral domain dimensionality reduction to extract the dominant risk modes;
[0021] The quantitative impact coefficients of each dominant risk mode on the reliability of the distribution network are calculated by regression model, and used as equivalent reliability disturbance parameters to achieve dimensionality reduction equivalence of high-dimensional uncertainty factors. The performance of the regression model is then verified.
[0022] Furthermore, the spectral domain dimensionality reduction process employs principal component analysis, which sets a preset threshold for variance contribution rate based on the accuracy requirements of the power distribution network assessment, and extracts principal components that meet the threshold requirements as the dominant risk modes.
[0023] The regression model is a ridge regression model. The optimal regularization coefficient is determined by cross-validation and the model training is completed. The model is required to have a determination coefficient on the test set that is not lower than a preset value. If this is not met, the principal component extraction threshold is readjusted and the operation is repeated.
[0024] Furthermore, in step S3, establishing a fast reliability reconstruction algorithm based on topological elastic mapping includes:
[0025] A topological elasticity tensor model of the distribution network is established. The topological elasticity tensor is constructed with the connection relationship of distribution network nodes and the transmission capacity limit of branches as tensor dimensions. The model is trained and validated based on the historical fault reconstruction case dataset of the distribution network.
[0026] Network structure disturbances are mapped to elastic response coefficients, and the topology disturbance values of the distribution network are calculated by combining equivalent reliability disturbance parameters. The elastic response coefficients of each node and branch are obtained by topology elastic tensor operation.
[0027] A reliability recovery function is constructed, with the elastic response coefficient as the core input, combined with the load importance coefficient and the power supply recovery time during power failure. The function is established through nonlinear fitting and the fitting effect is verified.
[0028] Tensor product operation is performed on the topological elasticity tensor and the reliability recovery function. Combined with the original reliability benchmark value of the distribution network, the system-level equivalent reliability index is quickly solved by tensor contraction and index aggregation.
[0029] Furthermore, the load importance coefficient is weighted according to the node load type, and different value ranges are divided according to important load nodes, general load nodes, and non-important load nodes;
[0030] The system-level equivalent reliability index includes at least one of the following: average outage time at load points, average outage frequency of the system, and system power supply reliability rate.
[0031] Furthermore, in step S4, the hierarchical rapid iterative framework for adaptive reduction of uncertainty coupling strength includes:
[0032] Construct an uncertainty coupling strength matrix, and calculate the coupling strength between any two types of uncertainty factors based on a weighted fusion of linear and nonlinear correlation analysis methods;
[0033] Train a coupling strength determination model to classify and identify the coupling types between uncertainty factors;
[0034] Calculate the coupling sensitivity gradient to obtain the sensitivity of the impact of changes in coupling strength on the core reliability indicators of the distribution network;
[0035] An adaptive reduction threshold is set, and weak coupling factors are identified and eliminated based on the threshold to form a main coupling factor subspace.
[0036] Within the main coupling factor subspace, a fast reliability reconstruction algorithm is substituted, and a hierarchical iterative calculation method is adopted to realize the fast iterative evaluation of the reliability of the distribution network. The generalization of the iterative evaluation model is then verified.
[0037] Furthermore, the adaptive reduction threshold includes a coupling strength threshold and a coupling sensitivity gradient threshold, which are set comprehensively based on the accuracy and computational efficiency requirements of rapid assessment of distribution network reliability.
[0038] In the hierarchical iterative calculation process, only the relevant parameters of the main coupling factor are updated, while the weak coupling factor remains at a fixed baseline value. An iterative convergence threshold is set, and the iteration is terminated when the difference between the core reliability index of the system in two adjacent iterations is less than the convergence threshold, thus obtaining the reliability assessment result.
[0039] The generalization verification requires selecting test cases of distribution networks with different topologies, and the relative error between the evaluation results and the traditional Monte Carlo method should not exceed a preset value.
[0040] A rapid reliability assessment system for distribution networks based on uncertainty factors is used to execute the aforementioned rapid reliability assessment method for distribution networks based on uncertainty factors. The system includes:
[0041] Data acquisition and preprocessing module, uncertainty factor equivalent transformation module, topological elastic mapping and reliability solution module, hierarchical iterative evaluation module;
[0042] The data acquisition and preprocessing module is used to acquire basic operation data of the distribution network and raw data of various uncertainty factors, and to complete data cleaning, normalization and dataset division.
[0043] The uncertainty factor equivalent transformation module is used to construct a multi-scale uncertainty spectrum decomposition-causal decoupling mechanism to transform multiple types of uncertainty factors into equivalent reliability perturbation parameters.
[0044] The topology elastic mapping and reliability solution module is used to establish a topology elastic tensor model and a reliability recovery function, characterize the recoverability of the distribution network structure, and quickly solve the equivalent reliability index of the system.
[0045] The hierarchical iterative evaluation module is used to construct an uncertainty coupling strength matrix, dynamically identify and reduce weak coupling factors, form a main coupling factor subspace, and realize rapid evaluation of the reliability of the distribution network through hierarchical iterative calculation.
[0046] The beneficial effects of this invention are as follows:
[0047] 1. By introducing multi-scale uncertainty spectrum decomposition and causal decoupling mechanism, various uncertainty factors in the distribution network are transformed into equivalent reliability disturbance parameters. Combined with topology elastic mapping and fast reconfiguration algorithm, the calculation time is greatly shortened, the lengthy calculation steps in traditional methods are effectively reduced, the evaluation efficiency is improved, and it can quickly respond to real-time needs, which is especially suitable for the dynamic evaluation of smart distribution networks.
[0048] 2. By using a precise regression model to quantify and analyze various uncertainty factors in the distribution network, the impact of these factors is incorporated into the reliability assessment, which significantly improves the accuracy of the assessment results. Through a unified reliability impact spectrum, the synergistic effect of multiple factors can be comprehensively considered, making the assessment results more scientific and practical.
[0049] 3. By establishing an adaptive reduction mechanism for uncertainty coupling strength, this method can dynamically identify the influence strength of each uncertainty factor in the distribution network and eliminate factors with weaker influence, thereby optimizing the evaluation process. This adaptive capability ensures that the evaluation process remains efficient under different environments and operating conditions, while avoiding unnecessary calculations of unimportant factors, thus further improving computational efficiency and evaluation accuracy.
[0050] 4. This invention is not only applicable to conventional distribution network reliability assessment, but also capable of handling complex operating states such as fault reconfiguration, load transfer, and microgrid islanding. Through rapid reconfiguration algorithms and dynamic assessment mechanisms, it can monitor and assess the reliability recovery capability of the distribution network under abnormal conditions in real time, providing strong support for emergency management and smart grid optimization decisions.
[0051] 5. By adopting a hierarchical and rapid iterative evaluation framework, this method can handle distribution networks of different sizes and perform real-time evaluation of multiple reliability indicators. This high scalability makes the method suitable for rapid reliability analysis of small local distribution networks and large complex power grids, ensuring its wide applicability under different environments and needs, and possessing strong future development potential. Attached Figure Description
[0052] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:
[0053] Figure 1 This is a flowchart illustrating the overall method of the present invention;
[0054] Figure 2 This is a flowchart of the data preprocessing process of the present invention;
[0055] Figure 3 This is a flowchart of the uncertainty spectral decomposition of the present invention;
[0056] Figure 4 This is a flowchart illustrating the adaptive reduction of coupling strength according to the present invention.
[0057] Figure 5 This is a system block diagram of the present invention. Detailed Implementation
[0058] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.
[0059] Example 1:
[0060] Please see Figures 1-4 This invention provides a technical solution: a rapid reliability assessment method for distribution networks based on uncertainty factors, the method comprising:
[0061] S1. Obtain basic operation data of the distribution network and raw data of various uncertainty factors, and preprocess the data;
[0062] The basic operational data of the distribution network refers to various data reflecting the normal operating status of the distribution network, which may include line current, voltage, power, equipment operating time, fault records, and user electricity consumption information. This data is the basis for assessing the reliability of the distribution network. The raw data includes various uncertainties. Uncertainties refer to data related to various factors that affect the reliability assessment results of the distribution network but are uncertain. For example, the output of renewable energy generation, such as wind power and solar power, is random and fluctuating; its power generation data is the raw data of uncertainties. Uncertainty in load demand, such as fluctuations in user electricity consumption over different time periods, also falls into this category. Data preprocessing involves cleaning, transforming, and standardizing the acquired basic operational data and raw data of various uncertainties. Data cleaning may include removing noisy data and handling missing values; data transformation may involve unifying data from different formats into a format suitable for subsequent analysis; and data standardization ensures that the data has the same scale and range, facilitating subsequent processing and comparison.
[0063] S2. Construct a fast equivalent evaluation mechanism for multi-scale uncertainty spectrum decomposition and causal decoupling, which maps multiple types of uncertainty factors to a unified reliability influence spectrum domain and transforms them into equivalent reliability perturbation parameters.
[0064] Among them, multi-scale uncertainty spectral decomposition is a method to decompose multiple types of uncertainty factors according to different scales. Multi-scale means analyzing uncertainty factors from multiple different levels or ranges, such as from different perspectives such as time scale and spatial scale. Spectral decomposition decomposes uncertainty factors in a specific spectral domain, similar to decomposing a signal in the frequency domain, in order to better understand and handle the uncertainty of different components. Causal decoupling aims to analyze the causal relationship between uncertainty factors and decouple the coupling relationship between them. By decoupling causality, it is possible to clarify which uncertainties are causes, which are effects, and how they influence each other, thus decomposing complex uncertainties into relatively independent causal chains, facilitating subsequent assessment and processing. The rapid equivalent assessment mechanism is a mechanism capable of quickly assessing the reliability of distribution networks. It maps multiple types of uncertainties, after multi-scale uncertainty spectrum decomposition and causal decoupling, into a unified reliability influence spectrum domain, transforming them into equivalent reliability disturbance parameters. These disturbance parameters can be directly used to assess the reliability of the distribution network, avoiding complex direct calculations and improving assessment efficiency. The reliability influence spectrum domain is a specific conceptual space used to describe the impact of uncertainties on the reliability of the distribution network. In this domain, different types of deterministic factors can be represented and analyzed according to their influence methods and degrees, facilitating unified processing and comparison of their impact on reliability. The equivalent reliability disturbance parameters are parameters obtained after processing multiple types of uncertainties in the reliability influence spectrum domain. These parameters can equivalently reflect the disturbance effect of uncertainties on the reliability of the distribution network, allowing for rapid assessment of the degree of impact of uncertainties on the reliability of the distribution network.
[0065] S3. Establish a fast reliability reconfiguration algorithm based on topology elastic mapping to characterize the structural recoverability of the distribution network under fault reconfiguration, load transfer, and microgrid islanding operation, and quickly solve the equivalent reliability index of the system.
[0066] Topology elastic mapping is a method that elastically processes the topology of a distribution network and maps it to a mathematical space or model. Elasticity means considering the changes and adaptability of the distribution network's topology under fault conditions. This mapping facilitates the analysis and calculation of the distribution network's reliability under different topological states. The fast reliability reconfiguration algorithm is specifically designed for quickly calculating the reliability of a distribution network after experiencing fault reconfiguration, load shifting, and microgrid islanding. Based on topology elastic mapping, this algorithm characterizes the structural recoverability of the distribution network under these special conditions—that is, the ability of the distribution network to recover to normal operation or maintain a certain power supply capacity after being disturbed—and quickly solves for the system's equivalent reliability index. Fault reconfiguration involves adjusting the state of switches in the distribution network when a fault occurs, changing the network's topology to isolate the faulty part while restoring power to the non-faulty areas, thereby reducing the impact of faults. The scope and duration of power outages are defined to improve power supply reliability. Load transfer involves transferring power from faulty lines or equipment to other operating lines or equipment when a distribution network fails or requires maintenance, ensuring continuous power supply and minimizing the impact of outages on users. Microgrid islanding refers to a small-scale power generation and distribution system composed of distributed power sources, energy storage devices, energy conversion devices, loads, monitoring and protection devices. When the distribution network fails or disconnects from the main grid, the microgrid can operate independently, forming an island to continue supplying power to local loads, improving the reliability of power supply and energy utilization efficiency in local areas. The system equivalent reliability index is used to measure the overall reliability of the distribution network. It is calculated using a rapid reliability reconstruction algorithm and comprehensively considers the reliability performance of the distribution network under various operating conditions, such as outage frequency, outage duration, and power availability, providing a quantitative basis for assessing the reliability level of the distribution network.
[0067] S4. Build a hierarchical fast iterative framework for adaptive reduction of uncertainty coupling strength, dynamically identify the correlation strength between uncertainty factors and automatically reduce the participation of weak coupling factors, and realize fast iterative evaluation of distribution network reliability in the main coupling factor subspace.
[0068] Among these, the adaptive reduction of uncertainty coupling strength addresses the issue that multiple uncertainty factors in distribution network reliability assessment may be interconnected and influential, exhibiting coupling relationships. Uncertainty coupling strength measures the degree of this interconnection. Adaptive reduction refers to automatically adjusting the processing method for these factors based on their coupling strength, reducing the participation of weakly coupled factors in reliability assessment. This reduces computational load while focusing on the main coupling factors that have a significant impact on reliability, improving assessment efficiency and accuracy. The hierarchical rapid iteration framework divides the distribution network reliability assessment process into multiple levels and performs rapid iteration at each level. Through hierarchical processing, complex assessment problems can be decomposed into several relatively simple sub-problems, facilitating management and computation. Rapid iteration refers to the gradual approximation of accurate reliability assessment results through repeated calculations and updates at each level, improving the speed and accuracy of the assessment. The main coupling factor subspace, when considering the coupling relationships between uncertainty factors, defines the space formed by those uncertain factors with strong coupling strength and significant impact on distribution network reliability as the main coupling factor subspace. Rapid iterative assessment of distribution network reliability within this subspace allows for focused processing of factors that play a key role in reliability, improving the relevance and efficiency of the assessment.
[0069] It should be noted that, during use, S1 acquires and preprocesses data to provide an accurate and comprehensive foundation for subsequent evaluation, ensuring that the analysis is based on reliable information. S2 constructs a rapid equivalent evaluation mechanism that uniformly processes and transforms multiple types of uncertainty factors, efficiently grasping their impact on reliability and simplifying complex problems. S3 establishes a rapid reconstruction algorithm that can accurately characterize the structural recovery capability of the distribution network under various special operating conditions, quickly solve reliability indicators, and improve the evaluation capability of actual operating conditions. S4 builds a hierarchical rapid iteration framework that dynamically identifies the coupling strength of uncertainty factors, automatically reduces the participation of weakly coupled factors, focuses on the main coupled factors, reduces the amount of computation, and improves the evaluation speed and relevance. The overall design, from the data foundation to the evaluation mechanism, algorithm, and framework, is optimized layer by layer, enabling rapid, accurate, and comprehensive evaluation of the reliability of the distribution network.
[0070] In one embodiment, basic operational data of the distribution network and raw data of various uncertainty factors are acquired, and the data is preprocessed, including:
[0071] The system acquires the topology parameters of the distribution network, the rated technical parameters of the equipment, the basic load data of the nodes, and the basic parameters of the installed capacity and output of distributed power sources as basic operating data. At the same time, it acquires meteorological element fluctuation monitoring data, actual output fluctuation time series data of distributed power sources, random load change data of distribution network nodes, aging and degradation status monitoring data of power equipment throughout its entire life cycle, and statistical data on the operation and maloperation probability of relay protection devices as raw data for multiple types of uncertainty factors. These multiple types of uncertainty factors specifically include five categories: meteorological fluctuations, output fluctuations of distributed power sources, load randomness, equipment aging and degradation, and protection maloperation probability.
[0072] The basic operational data and the original data of uncertainty factors are sequentially processed by outlier removal, missing value completion, and dimension normalization to obtain a standardized dataset. The normalization adopts the min-max normalization method, and the expression is:
[0073]
[0074] in, For the first in the dataset One set of raw data, This is the global minimum value of the corresponding dataset. This represents the global maximum value of the corresponding dataset. The normalized dimensionless data has a range of values. ;
[0075] The standardized dataset is divided into training, validation, and test sets in a 7:2:1 ratio. The training set is used for iterative optimization of model parameters, the validation set is used for hyperparameter tuning and overfitting monitoring, and the test set is used for final performance evaluation of the model.
[0076] This design ensures that obtaining comprehensive and accurate data is a prerequisite for reliable evaluation. It covers basic data from multiple aspects such as topology and equipment parameters, as well as raw data from five types of uncertainty factors. In preprocessing, outlier removal and missing value completion ensure data integrity, dimension normalization eliminates differences in data scales, and the dataset is divided proportionally. The training set is used for model optimization, the validation set is used to tune hyperparameters and monitor overfitting, and the test set is used to evaluate the final performance. Each part performs its specific function, providing high-quality, standardized, and reasonably divided data for subsequent evaluations. This ensures that model training and evaluation are based on reliable data, thereby improving the accuracy and reliability of the evaluation results.
[0077] In one embodiment, a multi-scale uncertainty spectral decomposition-causal decoupling fast equivalent evaluation mechanism is constructed, which maps multiple types of uncertainty factors to a unified reliability impact spectral domain and transforms them into equivalent reliability perturbation parameters, including:
[0078] A joint time-frequency spectral representation model for multi-source uncertainty factors is established. For time-series uncertainty factors such as meteorological fluctuations, distributed power generation output fluctuations, and load randomness, a short-time Fourier transform is used to convert them from the time domain to time-frequency domain spectral features. The Hanning window is selected as the window function for the short-time Fourier transform, with a window length of 256 and a step size of 64. For static uncertainty factors such as equipment aging and degradation and the probability of protection malfunction, they are converted into static spectral vectors through equal-interval discretization and spectral feature mapping. The spectral features of all factors are integrated to form a multi-source uncertainty spectral matrix.
[0079]
[0080] in, For the number of uncertainty factors, The number of sampling points for the time spectrum;
[0081] An impact propagation path matrix based on causal strength is constructed. The causal impact strength of each type of uncertainty factor on the core reliability indicators of the distribution network is calculated using the Granger causality test algorithm. The training parameters of the Granger causality test model are as follows:
[0082] The lag order was determined using the AIC criterion, with a value ranging from 1 to 8. The significance level was set to 0.05, and the number of iterations was set to 1000.
[0083] Based on the electrical topology relationships, equipment connection relationships, and fault propagation patterns of the distribution network, a system is constructed. Influence propagation path matrix of order Elements in the matrix Indicates the first Uncertainty factors are determined by the first The strength of the causal impact of a fault propagation path on the reliability of the distribution network; the larger the value, the more significant the impact of the path.
[0084] Multi-source uncertainty spectrum matrix With influence transmission path matrix Matrix multiplication is performed to obtain the reliability impact spectrum matrix of the fused impact paths. Principal component analysis is then used to perform spectral domain dimensionality reduction on the reliability impact spectrum matrix. Principal components with variance contribution rates greater than a preset threshold are extracted as the dominant risk modes of distribution network reliability. The preset threshold is determined by the following rules:
[0085] The threshold is set according to the classification of the accuracy requirements for distribution network assessment. If the assessment requires high accuracy, the threshold is set at 90% to 95%; if the assessment requires conventional accuracy, the threshold is set at 85% to 90%.
[0086] The high-dimensional stochastic process is transformed into equivalent reliability perturbation parameters of a finite number of dominant risk modes. The quantitative impact coefficients of each dominant risk mode on the reliability of the distribution network are calculated using a ridge regression model. The training steps of the ridge regression model are as follows:
[0087] ① Initialize regularization coefficients The search scope is ;
[0088] ② Use 5-fold cross-validation to traverse the search range on the validation set. Choose the value that minimizes the mean square error. As the optimal regularization coefficient;
[0089] ③ Use the optimal Train the ridge regression model on the training set to obtain the influence coefficients of each dominant risk mode;
[0090] Use this influence coefficient as the equivalent reliability disturbance parameter , ,in To extract the number of dominant risk modes, This achieves dimensionality reduction equivalence for high-dimensional uncertainty factors;
[0091] To validate the ridge regression model, the coefficient of determination on the test set is required. If the conditions are not met, readjust the principal component extraction threshold and repeat the above steps.
[0092] This design constructs a multi-scale uncertainty spectral decomposition-causal decoupling rapid equivalent assessment mechanism. First, a time-frequency joint spectral representation model is established to handle different types of uncertainty factors and integrate them to form a spectral matrix. Then, an influence transmission path matrix is constructed to clarify the intensity of causal influence. Through matrix operations and principal component analysis, dimensionality reduction is achieved, the dominant risk modes are extracted, and they are transformed into equivalent reliability disturbance parameters. This unified processing and dimensionality reduction of complex high-dimensional uncertainty factors focuses on the dominant factors that have a significant impact on reliability, reduces the amount of computation, and considers causal relationships, making the assessment more scientific and reasonable, and accurately reflecting the impact of uncertainty factors on the reliability of the distribution network.
[0093] In one embodiment, a fast reliability reconfiguration algorithm based on topology elastic mapping is established to characterize the structural recoverability of the distribution network under fault reconfiguration, load transfer, and microgrid islanding operation conditions, and to quickly solve for the system's equivalent reliability index, including:
[0094] A topological elasticity tensor model of the distribution network is established, using the node connection relationship, branch transmission capacity limit, equipment redundancy configuration degree, and microgrid islanding capability of the distribution network as tensor dimensions to construct a third-order topological elasticity tensor:
[0095]
[0096] in, This represents the total number of nodes in the distribution network. Let be the number of performance characteristic dimensions of the power equipment, and each element in the tensor represents the elastic recovery characteristics of the distribution network topology under the corresponding dimension.
[0097] The topological resilience tensor model is trained based on a dataset of historical fault reconfiguration cases in the distribution network. The training steps are as follows:
[0098] ① Collect at least 1,000 sets of distribution network fault reconfiguration cases, including labeled information such as fault type, topology disturbance value, and actual recovery capability value;
[0099] ② Divide the case data into a tensor model training set and a validation set in a 7:3 ratio;
[0100] ③ Use gradient descent to optimize tensor element values. Set the loss function to the mean square error between the predicted and actual recovery values, the learning rate to 0.001, the number of iterations to 5000, and the batch size to 32. Stop training when the validation set loss stops decreasing after 50 consecutive iterations.
[0101] ④ Validate the model on the test set, requiring the mean absolute error (MAE) to be ≤0.05; otherwise, adjust the tensor dimension or increase the amount of training data and retrain.
[0102] Network structural perturbations are mapped to elastic response coefficients, combined with equivalent reliability perturbation parameters. Based on the distribution network fault propagation model, the topology disturbance values under three typical operating scenarios—fault reconfiguration, load transfer, and microgrid islanding—are calculated. Through the reduction operation of the topology elastic tensor, the topology disturbance values are quantitatively mapped to the elastic response coefficients of each node and branch of the distribution network. , Indicates the first Node to the The topological elastic recovery capability of the branch under structural disturbances has a numerical range of [value missing]. The closer the value is to 1, the stronger the recovery ability;
[0103] To replace the traditional path-by-path enumeration analysis method, a reliability recovery function is constructed using the elastic response coefficient. Using the load importance coefficients of each node in the distribution network and the power supply recovery time after a fault as the core input, a reliability recovery function is established through nonlinear fitting:
[0104]
[0105] in, These are correction coefficients obtained by fitting historical operating data of the distribution network, and they satisfy... Normalization constraints, The function output value represents the reliability recovery time of the corresponding node or branch after a power failure. The load importance coefficient is determined by weighting the values according to the load type of the node: 0.7 to 1.0 for important load nodes, 0.3 to 0.7 for general load nodes, and 0 to 0.3 for non-important load nodes.
[0106] The training steps for nonlinear fitting are as follows:
[0107] ① Use the Levenberg-Marquardt algorithm to fit the correction coefficients The initial value is set to ;
[0108] ② Set the fitting termination condition to the sum of squared residuals. Or the number of iterations is ≥100;
[0109] ③ Validate the fit on the validation set, requiring a goodness of fit. ;
[0110] The equivalent reliability index of the system is obtained by fast tensor solution. The tensor product operation is performed on the topological elastic tensor and the reliability recovery function. Combined with the original reliability benchmark value of the distribution network under the disturbance-free state, the system-level equivalent reliability index is obtained by fast solution through tensor contraction and index aggregation. The reliability index includes three core indicators: average outage time at load point, average outage frequency of the system, and system power supply reliability rate.
[0111] This design establishes a fast reliability reconstruction algorithm based on topological elastic mapping, constructs a topological elastic tensor model to describe the topological elastic recovery characteristics of the distribution network, trains the model based on historical cases, maps network structural disturbances to elastic response coefficients, calculates disturbance values by combining equivalent reliability disturbance parameters, constructs a reliability recovery function to replace traditional methods, and determines correction coefficients through nonlinear fitting. This quickly characterizes the structural recoverability of the distribution network under different operating scenarios, avoids tedious calculations of path-by-path enumeration, improves computational efficiency, and considers multiple factors to make the solution of the system's equivalent reliability index more accurate and comprehensive, providing strong support for reliability assessment.
[0112] In one embodiment, a hierarchical fast iterative framework for adaptive reduction of uncertainty coupling strength is established. This framework dynamically identifies the correlation strength between uncertainty factors and automatically reduces the participation of weakly coupled factors. It enables rapid iterative evaluation of distribution network reliability within the main coupling factor subspace, including:
[0113] Construct an uncertainty coupling strength matrix, and calculate the linear and nonlinear coupling strength between any two types of uncertainty factors using a weighted fusion method based on Pearson correlation coefficient and mutual information values. Coupling strength matrix of order Elements in the matrix:
[0114] ,
[0115] in, The Pearson correlation coefficient is used. For the normalized result of mutual information values, The weights of the Pearson correlation coefficient, The weights are the mutual information values, and satisfy the following conditions: The rules for determining the weights are as follows:
[0116] If the linear correlation among the uncertainty factors is significant, the significance of the linear correlation is determined by the F-test. At that time, take , ;
[0117] If the nonlinear correlation is significant, take , ;
[0118] If the linear and nonlinear correlations are roughly equal, then take... ;
[0119] Matrix elements The numerical range is The closer the value is to 1, the higher the degree of coupling between the two types of factors;
[0120] The steps for training the coupling strength determination model are as follows:
[0121] ① Construct a labeled dataset containing pairs of coupling strength factors, with the label categories being strong coupling and weak coupling;
[0122] ② A random forest classifier is used as the coupling strength determination model. The model parameters are: the number of decision trees is set to 100, the maximum depth is set to 10, and the minimum number of sample splits is set to 5.
[0123] ③ Train the model on the training set, optimize the hyperparameters using grid search, and achieve a classification accuracy of ≥0.9 on the validation set;
[0124] Calculate the coupling sensitivity gradient and the coupling strength matrix. Taking the first-order partial derivative of the system's equivalent reliability index, we obtain the coupling sensitivity gradient matrix. Elements in the matrix Indicates the first Class and the The sensitivity of the influence of the coupling strength change of the class factor on the core reliability index of the distribution network is such that the larger the absolute value, the more significant the influence of the coupling strength change on the reliability assessment result.
[0125] An adaptive reduction threshold is set, and the coupling strength threshold is comprehensively set based on the accuracy requirements and computational efficiency requirements of rapid distribution network reliability assessment. and coupling sensitivity gradient threshold ;
[0126] The rule for determining the threshold is:
[0127] If the evaluation focuses on computational efficiency, then take , ;
[0128] If the evaluation focuses on accuracy, then take , ;
[0129] If a balance between accuracy and efficiency is required, then take... , ;
[0130] When the coupling strength of two types of factors And the absolute value of the coupling sensitivity gradient When this occurs, the two types of factors are determined to be weakly coupled factors;
[0131] A main coupling factor subspace is formed, and all weak coupling factors and their corresponding coupling strength and influence path parameters are eliminated. The remaining strong coupling factors are taken as the main coupling factors. The equivalent reliability perturbation parameters of the main coupling factors, the simplified coupling strength matrix, and the simplified influence transmission path matrix are used as the core elements to construct a low-dimensional main coupling factor subspace, thereby achieving a second dimensionality reduction of the evaluation dimension.
[0132] A fast iterative reliability assessment is performed within the main coupling factor subspace. Various low-dimensional parameters within the subspace are used as inputs, and a fast reliability reconstruction algorithm based on topological elastic mapping is applied. A hierarchical iterative calculation method is employed to evaluate the distribution network reliability. During the iteration process, only the relevant parameters of the main coupling factor are updated, while the weak coupling factor maintains a fixed baseline value. An iterative convergence threshold is set. The convergence threshold is determined by the following rule:
[0133] Set according to the accuracy requirements of the reliability index. If the index requires retaining 4 decimal places, take... If you require three decimal places, take... If you require two decimal places, take... When the difference between the core reliability indices of the system obtained from two consecutive iterations is less than the preset convergence threshold... When the iteration ends, the final reliability assessment result of the distribution network is obtained;
[0134] To verify the generalization of the iterative evaluation model, at least three distribution network test cases with different topologies were selected, and the relative error between the evaluation results and the traditional Monte Carlo method was required to be ≤5%.
[0135] This design establishes a hierarchical, rapid iterative framework for adaptive reduction of uncertainty coupling strength. First, a coupling strength matrix is constructed to measure the degree of coupling between uncertainty factors. A judgment model is trained to identify strong and weak coupling factors, the coupling sensitivity gradient is calculated, and an adaptive reduction threshold is set to identify weak coupling factors, forming a main coupling factor subspace. Rapid iterative evaluation is performed within this space, dynamically identifying and reducing the participation of weak coupling factors, achieving a second dimensionality reduction in the evaluation dimension, reducing computational load, improving evaluation efficiency, and ensuring evaluation accuracy. Generalization verification ensures that the model is applicable to distribution networks with different topologies, making reliability evaluation more efficient, accurate, and universal.
[0136] In one embodiment, the various uncertainty factors include weather fluctuations, distributed power output fluctuations, load randomness, equipment aging and degradation, and protection malfunction probability. Weather fluctuations include random fluctuations in wind speed, sunlight, temperature, and rainfall; equipment aging and degradation include performance degradation of transformers, circuit breakers, and transmission lines. The optimization algorithm used during model training also includes the Adam algorithm, with a learning rate decay strategy of 10% every 100 iterations, and a weight decay coefficient set to... To prevent overfitting.
[0137] This design clearly defines the specific content of multiple uncertainty factors and employs various optimization algorithms and learning rate decay and weight decay strategies during model training. The detailed definition of uncertainty factors makes the assessment more targeted. Considering the impact of multiple factors on reliability, optimization algorithms such as the Adam algorithm accelerate model convergence and improve training efficiency. The learning rate decay strategy adjusts the learning rate according to the training progress to avoid getting trapped in local optima. The weight decay coefficient prevents overfitting and ensures the model's generalization ability. By comprehensively utilizing multiple methods to optimize the model training process, the model's performance is improved, making the distribution network reliability assessment based on this model more accurate and reliable, and adaptable to various complex situations.
[0138] Example 2:
[0139] Please see Figure 5 A rapid reliability assessment system for distribution networks based on uncertainty factors is used to perform the aforementioned rapid reliability assessment method for distribution networks based on uncertainty factors. The system includes:
[0140] Data acquisition and preprocessing module, uncertainty factor equivalent transformation module, topological elastic mapping and reliability solution module, hierarchical iterative evaluation module;
[0141] The data acquisition and preprocessing module is used to acquire basic operational data of the distribution network and raw data of various uncertainty factors, and to complete data cleaning, normalization and dataset partitioning.
[0142] The uncertainty factor equivalent transformation module is used to construct a multi-scale uncertainty spectrum decomposition-causal decoupling mechanism to transform multiple types of uncertainty factors into equivalent reliability perturbation parameters.
[0143] The topology elasticity mapping and reliability solution module is used to establish the topology elasticity tensor model and reliability recovery function, characterize the recoverability of the distribution network structure, and quickly solve the equivalent reliability index of the system.
[0144] The hierarchical iterative evaluation module is used to construct an uncertainty coupling strength matrix, dynamically identify and reduce weak coupling factors, form a main coupling factor subspace, and realize rapid evaluation of distribution network reliability through hierarchical iterative calculation.
[0145] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented 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.
[0146] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.
Claims
1. A rapid reliability assessment method for distribution networks based on uncertainty factors, characterized in that, Includes the following steps: S1. Obtain basic operating data and raw data of multiple uncertainty factors of the power distribution network, and preprocess the basic operating data and raw data of uncertainty factors to obtain a standardized dataset; S2. Construct a fast equivalent evaluation mechanism for multi-scale uncertainty spectrum decomposition and causal decoupling, which maps multiple types of uncertainty factors to a unified reliability influence spectrum domain and transforms them into equivalent reliability perturbation parameters. S3. Establish a fast reliability reconfiguration algorithm based on topology elastic mapping to characterize the structural recoverability of the distribution network under fault reconfiguration, load transfer, and microgrid islanding operation, and quickly solve the equivalent reliability index of the system. S4. Build a hierarchical rapid iterative framework for adaptive reduction of uncertainty coupling strength, dynamically identify the correlation strength between uncertainty factors and automatically reduce the participation of weak coupling factors, and realize rapid iterative evaluation of distribution network reliability in the main coupling factor subspace.
2. The method for rapid reliability assessment of distribution networks based on uncertainty factors according to claim 1, characterized in that, In step S1, the basic operational data includes: The topology parameters of the distribution network, the rated technical parameters of the equipment, the basic load data of the nodes, and the basic parameters of the installed capacity and output of distributed power sources; The original data for the various uncertainty factors include meteorological element fluctuation monitoring data, distributed power generation actual output fluctuation time series data, distribution network node load random change data, power equipment full life cycle aging and degradation status monitoring data, and relay protection device operation and malfunction probability statistics, corresponding to five types of uncertainty factors: meteorological fluctuations, distributed power generation output fluctuations, load randomness, equipment aging and degradation, and protection malfunction probability.
3. The method for rapid reliability assessment of distribution networks based on uncertainty factors according to claim 1, characterized in that, In step S1, the preprocessing includes: The basic operational data and the original data of uncertainty factors are sequentially processed by outlier removal, missing value completion, and dimension normalization. The dimension normalization adopts the min-max standardization method. The standardized dataset is divided into training, validation and test sets according to a preset ratio, which are used for iterative optimization of model parameters, hyperparameter tuning and overfitting monitoring, and final performance evaluation of the model, respectively.
4. The method for rapid reliability assessment of distribution networks based on uncertainty factors according to claim 1, characterized in that, In step S2, the construction of a fast equivalent evaluation mechanism for multi-scale uncertainty spectral decomposition and causal decoupling includes: A time-frequency joint spectral representation model for multi-source uncertainty factors is established, and time-series and static uncertainty factors are transformed into corresponding spectral features and integrated to form a multi-source uncertainty spectral matrix. Construct an influence transmission path matrix based on causal strength, and calculate the causal influence strength of various uncertainty factors on the core reliability indicators of the distribution network by combining the electrical topology and fault propagation law of the distribution network. The multi-source uncertainty spectrum matrix and the influence transmission path matrix are operated on, and the results are subjected to spectral domain dimensionality reduction to extract the dominant risk modes; The quantitative impact coefficients of each dominant risk mode on the reliability of the distribution network are calculated by regression model, and used as equivalent reliability disturbance parameters to achieve dimensionality reduction equivalence of high-dimensional uncertainty factors. The performance of the regression model is then verified.
5. The method for rapid reliability assessment of distribution networks based on uncertainty factors according to claim 4, characterized in that: The spectral domain dimensionality reduction process adopts the principal component analysis method. According to the accuracy requirements of the power distribution network assessment, a preset threshold for the variance contribution rate is set in stages, and the principal components that meet the threshold requirements are extracted as the dominant risk modes. The regression model is a ridge regression model. The optimal regularization coefficient is determined by cross-validation and the model training is completed. The model is required to have a determination coefficient on the test set that is not lower than a preset value. If this is not met, the principal component extraction threshold is readjusted and the operation is repeated.
6. The method for rapid reliability assessment of distribution networks based on uncertainty factors according to claim 1, characterized in that, In step S3, establishing a fast reliability reconstruction algorithm based on topological elastic mapping includes: A topological elasticity tensor model of the distribution network is established. The topological elasticity tensor is constructed with the connection relationship of distribution network nodes and the transmission capacity limit of branches as tensor dimensions. The model is trained and validated based on the historical fault reconstruction case dataset of the distribution network. Network structure disturbances are mapped to elastic response coefficients, and the topology disturbance values of the distribution network are calculated by combining equivalent reliability disturbance parameters. The elastic response coefficients of each node and branch are obtained by topology elastic tensor operation. A reliability recovery function is constructed, with the elastic response coefficient as the core input, combined with the load importance coefficient and the power supply recovery time during the fault. The function is established through nonlinear fitting and the fitting effect is verified. Tensor product operation is performed on the topological elasticity tensor and the reliability recovery function. Combined with the original reliability benchmark value of the distribution network, the system-level equivalent reliability index is quickly solved by tensor contraction and index aggregation.
7. The method for rapid reliability assessment of distribution networks based on uncertainty factors according to claim 6, characterized in that: The load importance coefficient is assigned a weighted value based on the node load type, and different value ranges are divided according to important load nodes, general load nodes, and non-important load nodes. The system-level equivalent reliability index includes at least one of the following: average outage time at load points, average outage frequency of the system, and system power supply reliability rate.
8. The method for rapid reliability assessment of distribution networks based on uncertainty factors according to claim 1, characterized in that, In step S4, the hierarchical fast iterative framework for adaptive reduction of uncertainty coupling strength includes: Construct an uncertainty coupling strength matrix, and calculate the coupling strength between any two types of uncertainty factors based on a weighted fusion of linear and nonlinear correlation analysis methods; Train a coupling strength determination model to classify and identify the coupling types between uncertainty factors; Calculate the coupling sensitivity gradient to obtain the sensitivity of the impact of changes in coupling strength on the core reliability indicators of the distribution network; An adaptive reduction threshold is set, and weak coupling factors are identified and eliminated based on the threshold to form a main coupling factor subspace. Within the main coupling factor subspace, a fast reliability reconstruction algorithm is substituted, and a hierarchical iterative calculation method is adopted to realize the fast iterative evaluation of the reliability of the distribution network. The generalization of the iterative evaluation model is then verified.
9. The method for rapid reliability assessment of distribution networks based on uncertainty factors according to claim 8, characterized in that: The adaptive reduction threshold includes a coupling strength threshold and a coupling sensitivity gradient threshold, which are set comprehensively based on the accuracy and computational efficiency requirements of rapid assessment of distribution network reliability. In the hierarchical iterative calculation process, only the relevant parameters of the main coupling factor are updated, while the weak coupling factor remains at a fixed baseline value. An iterative convergence threshold is set, and the iteration is terminated when the difference between the core reliability index of the system in two adjacent iterations is less than the convergence threshold, thus obtaining the reliability assessment result. The generalization verification requires selecting test cases of distribution networks with different topologies, and the relative error between the evaluation results and the traditional Monte Carlo method should not exceed a preset value.
10. A rapid reliability assessment system for distribution networks based on uncertainty factors, characterized in that, The system, applied to the rapid reliability assessment method for distribution networks based on uncertainty factors according to any one of claims 1-9, comprises: Data acquisition and preprocessing module, uncertainty factor equivalent transformation module, topological elastic mapping and reliability solution module, hierarchical iterative evaluation module; The data acquisition and preprocessing module is used to acquire basic operation data of the distribution network and raw data of various uncertainty factors, and to complete data cleaning, normalization and dataset division. The uncertainty factor equivalent transformation module is used to construct a multi-scale uncertainty spectrum decomposition-causal decoupling mechanism to transform multiple types of uncertainty factors into equivalent reliability perturbation parameters. The topology elastic mapping and reliability solution module is used to establish a topology elastic tensor model and a reliability recovery function, characterize the recoverability of the distribution network structure, and quickly solve the equivalent reliability index of the system. The hierarchical iterative evaluation module is used to construct an uncertainty coupling strength matrix, dynamically identify and reduce weak coupling factors, form a main coupling factor subspace, and realize rapid evaluation of the reliability of the distribution network through hierarchical iterative calculation.