Power distribution network reliability index evaluation method and system based on binary neural network

Abstract

The invention discloses a power distribution network reliability index evaluation method and system based on a binary neural network, and belongs to the technical field of power system reliability evaluation. The method comprises the following steps: acquiring configuration data of the circuit breaker of the power distribution network and preprocessing; constructing a multi-task binary neural network model comprising a shared encoder and a plurality of task specific headers; calculating conservative double-boundary constraints of the key node indexes based on the training data; constructing a multi-task total loss function fusing constraint penalty terms through an augmented Lagrange method; and taking the loss function as a target training model, and performing evaluation and prediction by using the trained model. According to the method, the calculation complexity is reduced through the binary neural network, the engineering constraint is fused into the training through the augmented Lagrange method, the lightweight, constraint-aware and multi-task cooperative high-efficiency evaluation of the reliability of the power distribution network is realized, and the evaluation speed, precision and practicability are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of power system reliability assessment technology, specifically relating to a method and system for assessing the reliability index of distribution networks based on binary neural networks. Background Technology

[0002] As a crucial component of the power system, the reliability of the distribution network directly impacts the power quality for users and the economic benefits for power companies. Traditional methods for assessing distribution network reliability primarily include analytical methods and Monte Carlo simulation. Analytical methods calculate system reliability indicators by establishing mathematical models, but they struggle to handle complex network topologies and diverse fault scenarios. Monte Carlo simulation, on the other hand, obtains statistical results through extensive random sampling, offering high accuracy but requiring lengthy computation times and failing to meet real-time requirements.

[0003] In recent years, with the development of artificial intelligence technology, neural network methods have been introduced into the field of power distribution network reliability assessment. While traditional full-precision neural networks can establish a nonlinear mapping relationship between circuit breaker configuration and reliability indicators, they suffer from the following problems: (1) The model has a large number of parameters and high computational complexity. The reliability assessment of the distribution network needs to consider system-level indicators (such as SAIFI, SAIDI, EENS) and local indicators of key nodes (such as CIF, CID, eens). Traditional neural networks are difficult to achieve lightweight deployment while ensuring accuracy.

[0004] (2) Lack of constraints and guarantees for key nodes. In distribution network planning, the reliability indicators of key nodes need to meet specific constraints (such as the principle of conservative estimation). Traditional methods have failed to effectively integrate these constraints into the model training process.

[0005] Therefore, there is an urgent need to study a computationally efficient and constraint-aware method for assessing the reliability of distribution networks to meet the needs of real-time planning and decision-making. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method and system for evaluating the reliability index of distribution networks based on binary neural networks, which addresses the shortcomings of the prior art. This method solves the technical problems of high computational complexity, difficulty in task balancing in multi-task learning, and lack of constraint guarantees for key nodes in traditional distribution network reliability evaluation methods.

[0007] The present invention adopts the following technical solution: The method for evaluating the reliability index of distribution networks based on binary neural networks includes the following steps: The circuit breaker configuration data of the distribution network is acquired and preprocessed to obtain a dataset containing input features and output labels. The output labels include system-level reliability indicators and local reliability indicators of key nodes. A multi-task binary neural network model is constructed, the multi-task binary neural network model including a shared encoder for extracting shared features from input features, and multiple task-specific heads corresponding to the system-level reliability index and the local reliability index of the key node, respectively, wherein the shared encoder contains at least one binary layer; Conservative double-boundary constraints are used to calculate the local reliability index of the key node based on training data. Based on the aforementioned dual-boundary constraints, a multi-task total loss function containing constraint penalty terms is constructed using the augmented Lagrange method. Using the preprocessed dataset, the multi-task total loss function is used as the optimization objective to train the multi-task binary neural network model. The trained multi-task binary neural network model is then used to evaluate and predict the reliability indicators of the distribution network.

[0008] Preferably, the preprocessing includes: normalizing the input data as a whole, and normalizing the output labels by task grouping.

[0009] Preferably, the shared encoder comprises, in sequence, an input layer, at least one full-precision dense layer, and at least one binary dense layer; the task-specific header is a full-precision dense layer.

[0010] Preferably, the conservative double-boundary constraint for calculating the local reliability index of the key node based on training data specifically includes: Based on the data distribution of local reliability indicators of each key node in the training data, the lower bound quantile and upper bound quantile are calculated respectively, and the lower boundary and upper boundary of each indicator are determined based on the quantile and the preset margin.

[0011] Preferably, based on the dual-boundary constraints, constructing a multi-task total loss function including a constraint penalty term using the augmented Lagrangian method specifically includes: The model predicts the amount of violation of the dual boundary constraints; the constraint penalty term is constructed based on the amount of violation, the Lagrange multiplier, and the penalty parameter; the constraint penalty term is added to the multi-task uncertainty weighted loss to obtain the multi-task total loss function.

[0012] Preferably, when constructing the constraint penalty term, different weights are assigned to the lower boundary constraint and the upper boundary constraint, wherein the weight of the lower boundary constraint is greater than the weight of the upper boundary constraint.

[0013] Preferably, during training, the Lagrange multipliers are periodically updated according to the constraint violation situation, wherein the update rate of the multipliers corresponding to the lower boundary constraint is greater than the update rate of the multipliers corresponding to the upper boundary constraint.

[0014] Preferably, the penalty parameter is increased when the constraint violation persists.

[0015] Preferably, the training process employs an early stopping strategy. When the comprehensive verification index, calculated based on the verification mean square error and constraint violation penalty for each task, fails to improve within a preset period, training is stopped and the optimal model parameters are restored.

[0016] Secondly, embodiments of the present invention provide a power distribution network reliability index evaluation system based on a binary neural network, comprising: The data module is used to acquire circuit breaker configuration data of the distribution network and preprocess it to obtain a dataset containing input features and output labels. The output labels include system-level reliability indicators and local reliability indicators of key nodes. A building module is used to build a multi-task binary neural network model, the multi-task binary neural network model including a shared encoder for extracting shared features from input features, and multiple task-specific heads corresponding to the system-level reliability index and the local reliability index of the key node, respectively, wherein the shared encoder contains at least one binary layer; The constraint module is used to calculate conservative double-boundary constraints for the local reliability index of the key node based on the training data. The function module is used to construct a multi-task total loss function containing constraint penalty terms based on the aforementioned dual-boundary constraints using the augmented Lagrange method; The prediction module is used to train the multi-task binary neural network model using the preprocessed dataset and the multi-task total loss function as the optimization objective, and to evaluate and predict the reliability indicators of the distribution network using the trained multi-task binary neural network model.

[0017] Thirdly, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for evaluating the reliability index of a power distribution network based on a binary neural network.

[0018] Fourthly, embodiments of the present invention provide a computer-readable storage medium including a computer program, which, when executed by a processor, implements the steps of the above-described method for evaluating the reliability index of a power distribution network based on a binary neural network.

[0019] Fifthly, a chip includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for evaluating the reliability index of a power distribution network based on a binary neural network.

[0020] In a sixth aspect, embodiments of the present invention provide an electronic device, including a computer program, which, when executed by the electronic device, implements the steps of the above-described method for evaluating the reliability index of a power distribution network based on a binary neural network.

[0021] Compared with the prior art, the present invention has at least the following beneficial effects: A method for evaluating distribution network reliability indicators based on binary neural networks is proposed. By introducing a binary layer, the method significantly reduces the number of model parameters and computational complexity, improving inference efficiency. A multi-task framework is employed to simultaneously output system-level and node-level indicators, enhancing the comprehensiveness of the evaluation. Conservative double-boundary constraints are incorporated into the training using an augmented Lagrangian method to ensure that predicted values ​​for key nodes meet engineering constraints, thereby enhancing model reliability. This method achieves lightweight, constraint-aware, and multi-task collaborative distribution network reliability evaluation, effectively addressing the problems of high computational cost, missing constraints, and task imbalance inherent in traditional methods.

[0022] Furthermore, overall normalization eliminates the dimensional differences between input features, improving the model's training stability and convergence speed; group normalization, on the other hand, processes the output metrics for different tasks (system-level, node-level) separately, avoiding training bias caused by different metric magnitudes, and further optimizing the balance of multi-task learning.

[0023] Furthermore, the introduction of binary layers significantly reduces model size and improves computational efficiency; the full-precision layer, acting as the task head, ensures the prediction accuracy of the output metrics. This hybrid structure of binary feature extraction and full-precision task output achieves a good balance between efficiency and accuracy.

[0024] Furthermore, using a data-driven quantile method to determine the boundary can adapt to the data distribution and enhance the rationality and robustness of the constraints; by introducing margins to form conservative boundaries, a safe buffer is provided for the prediction results, which is more in line with the conservative estimation principle in engineering practice.

[0025] Furthermore, the constraints are transformed into differentiable penalty terms, enabling the model to explicitly perceive and satisfy the constraints during training; the introduction of the augmented Lagrangian method makes the constraint optimization process more stable and efficient.

[0026] Furthermore, the reliability indicators of key nodes in the distribution network must prioritize meeting the minimum requirements, while the upper boundary can be relatively lenient. Greater weight is given to the lower boundary constraints, making the model more stringent in penalizing violations of the lower boundary during training. This prioritizes ensuring that the indicators of key nodes do not fall below the safety threshold, addressing the irrationality of traditional constraint methods that treat different boundaries equally. This differentiated design precisely matches the actual needs of distribution network planning, further reducing the power supply risk of key nodes and improving the engineering practicality of the evaluation method.

[0027] Furthermore, the lower boundary constraint multiplier updates faster, enabling rapid response to constraint violations and timely increases in penalty intensity to ensure that key node indicators meet minimum requirements as quickly as possible. The upper boundary multiplier updates slower, avoiding over-constraint that could lead to a decrease in overall model accuracy. This addresses the mismatch between the constraint update mechanism and actual needs, allowing multiplier updates to dynamically adapt to the priorities of different boundaries, further improving the effectiveness of constraint satisfaction and ensuring that the lower boundary violation rate on the test set is kept at a low level.

[0028] Furthermore, the penalty parameter is increased stepwise from an initial value of 0.01 to 0.9. As constraint violations persist, the penalty intensity is gradually increased, forcing the model to adjust its parameters to satisfy the constraints. Simultaneously, a maximum penalty parameter is set to prevent excessive penalty that could cause model divergence. This makes the constraint penalty adaptive, ensuring convergence speed in the early stages of model training and strengthening constraint enforcement in the later stages, ultimately ensuring that the constraint violation rate on the test set is below 5%, thus achieving a balance between model accuracy and constraint satisfaction.

[0029] Furthermore, the comprehensive validation metric combines the validation mean squared error and constraint violation penalty for each task, comprehensively reflecting the model's accuracy and constraint satisfaction. Training is stopped and optimal parameters are restored when the metric fails to improve after 100 consecutive epochs, preventing overfitting of the training data in the later stages of training, which could lead to decreased accuracy on the test set or increased constraint violation rate. This design balances training accuracy and generalization ability, enabling the model to maintain high prediction accuracy (R²) on the test set. 2 The low constraint violation rate (≥0.957) improves the practicality and stability of the method.

[0030] It is understood that the beneficial effects of the second to sixth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0031] In summary, the method of this invention achieves lightweight model, automatic task balancing, and constraint awareness while ensuring prediction accuracy, providing an efficient and reliable technical means for real-time planning and decision-making in power distribution networks.

[0032] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0033] Figure 1 The training process monitoring diagram of the distribution network reliability index evaluation method based on binary neural network provided in an embodiment of the present invention is shown in which (a) is the training and verification mean square error curve of each task; (b) is the evolution curve of the learned task uncertainty parameter; (c) is the constraint violation penalty curve; and (d) is the evolution curve of the augmented Lagrange penalty parameter ρ. Figure 2The above is a comparison chart of the prediction results of the distribution network reliability index evaluation method based on binary neural network provided in an embodiment of the present invention. Among them, (a) to (c) are scatter plots comparing the predicted and actual values ​​of system-level indices SAIFI, SAIDI, and EENS, respectively; (d) to (f) are scatter plots comparing the predicted and actual values ​​of local indices CIF, CID, and eens of node 3, respectively; (g) to (i) are scatter plots comparing the predicted and actual values ​​of local indices CIF, CID, and eens of node 12, respectively. Figure 3 The task weight evolution diagram of a distribution network reliability index evaluation method based on binary neural network provided in one embodiment of the present invention; Figure 4 This is a schematic diagram of the process of the present invention; Figure 5 A schematic diagram of a computer device provided in an embodiment of the present invention; Figure 6 This is a block diagram of a chip provided according to an embodiment of the present invention.

[0034] Among them, 60. Computer equipment; 61. Processor; 62. Memory; 63. Computer program; 600. Electronic device; 610. Processing unit; 620. Storage unit; 6201. Random access memory unit; 6202. Cache memory unit; 6203. Read-only memory unit; 6204. Program / utility; 6205. Program module; 630. Bus; 640. Display unit; 650. Input / output interface; 660. Network adapter; 700. External device. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0037] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0038] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0039] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0040] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0041] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0042] This invention provides a method for evaluating the reliability indicators of a distribution network based on a binary neural network, including: data preparation and preprocessing, construction of a multi-task binary neural network model, conservative constraint calculation, design of a training framework incorporating constraints, model training and optimization, and final evaluation and prediction. This method reduces the computational burden through binaryization technology, simultaneously evaluates multiple indicators through a multi-task learning framework, and embeds key engineering constraints into the learning process using an augmented Lagrangian method. In terms of computational efficiency, the introduction of a binary neural network layer significantly reduces the number of model parameters and computational complexity, enabling the model to be quickly deployed to edge devices and achieve real-time evaluation. Regarding reliability assurance, it creatively adopts conservative double-boundary constraints based on data quantiles and seamlessly integrates them into the model training process using an augmented Lagrangian method, ensuring that the prediction results of key nodes strictly meet engineering safety requirements. In terms of multi-task processing, through reasonable network structure design and loss function construction, it achieves simultaneous, high-precision prediction of system-level and node-level indicators. Experimental results show that the method of this invention has high prediction accuracy (R0) on various reliability indicators. 2 It can achieve a score of over 0.985, and the violation rate of key node constraints is less than 5%. It is significantly better than traditional methods in terms of calculation speed, prediction accuracy and constraint satisfaction, and has important engineering application value.

[0043] Please see Figure 4 This invention discloses a method for evaluating the reliability index of a distribution network based on a binary neural network, comprising the following steps: S1. Obtain circuit breaker configuration data of the distribution network and perform preprocessing. S101, Data Acquisition To comprehensively reflect the uncertainties in distribution network operation, a physics-based simulation method is adopted. Using the IEEE 33-bus system or other actual distribution network topologies as prototypes, power system simulation software (such as OpenDSS, MATLAB / Simulink) is used to model and simulate the distribution network, simulating actual distribution network operation scenarios. Parameters such as circuit breaker switching states (0 / 1 representing open / closed), failure rate (times / year), mean time to repair (hours), and selectable installation location types are selected as input variables, while distribution network reliability indicators are used as output variables. Through Monte Carlo simulation or combined enumeration methods, a large number of different operation scenarios are generated.

[0044] For each scenario, the simulation calculates its corresponding reliability index as an output variable (label): System-level indicators such as System Average Outage Frequency Index (SAIFI), System Average Outage Duration Index (SAIDI), and Expected Power Shortage (EENS); and local indicators such as User Average Outage Frequency (CIF), User Average Outage Duration (CID), and Expected Power Shortage (EENS) for pre-selected key nodes (such as nodes 3 and 12 in the embodiment).

[0045] The final dataset consists of N samples, each sample consisting of M-dimensional input features (e.g., M=24) and K-dimensional output labels (e.g., K=9, containing 3 system indicators and 2 node × 3 local indicators).

[0046] S102, Data Preprocessing The collected data is preprocessed, including filtering and smoothing, outlier removal, and normalization.

[0047] Use statistical methods (such as the 3σ criterion) or model-based methods to detect and remove obvious outliers.

[0048] Calculate normalization parameters for the output metrics of different tasks, and then normalize the input data X as a whole.

[0049] in, The mean vector of the input data (24-dimensional). The standard deviation vector (24-dimensional) of the input data.

[0050] The output data Y is grouped and normalized into three groups: system-level indicators (dimensions 1-3), node 3 indicators (dimensions 4-6), and node 12 indicators (dimensions 7-9). The normalization parameters are calculated for each group.

[0051] in, Indicates a grouped index. and The first g The mean and standard deviation of the group output data.

[0052] The preprocessed complete dataset is randomly divided into a training set (for model parameter learning, accounting for 70%), a validation set (for hyperparameter tuning and early stopping detection, accounting for 15%), and a test set (for final performance evaluation, accounting for 15%).

[0053] S2. Construct a multi-task binary neural network model The multi-task binary neural network model of the present invention comprises two parts: a shared encoder and a task-specific head.

[0054] S201, Shared Encoder The goal of a shared encoder is to learn a general, high-level characteristic representation from circuit breaker configuration to grid operating status. Its architecture is designed to balance expressive power and efficiency. Input layer: 24-dimensional (circuit breaker configuration parameters), receiving M-dimensional normalized input features; First layer: Full-precision Dense layer, 24 → 512, ReLU activation, BatchNorm, Dropout(0.3); Second layer: BinaryDense, 512 → 256, ReLU activated, Dropout (0.3); Third layer: BinaryDense, 256 → 128, ReLU activated, Dropout (0.3); The introduction of the binary layer makes most of the model weights only 1 bit, which greatly reduces the model storage space (by about 32 times) and simplifies most multiply-accumulate operations into symbolic operations and accumulation, significantly improving the calculation speed and reducing energy consumption.

[0055] Output: 128-dimensional shared feature vector. The shared encoder ultimately outputs a fixed-dimensional shared feature vector for use by all downstream tasks.

[0056] S202, Task-Specific Header This invention designs three task-specific heads, corresponding to system-level metrics, node 3 metrics, and node 12 metrics, respectively. Each head has the same structure, forming a shallow, full-precision neural network. Input: 128-dimensional shared feature vector, receiving shared feature vectors from the shared encoder.

[0057] Hidden layer: Full-precision Dense layer, 128→128, ReLU activation, Dropout (0.3).

[0058] Output layer: Full-precision Dense layer, 128→3 (three reliability metrics).

[0059] Typically, it contains 1-2 fully connected layers, using activation functions such as ReLU, and finally a linear output layer to produce the predicted values ​​of all reliability metrics corresponding to the task (for example, a task head responsible for system-level metrics outputs 3 values: predictions of SAIFI, SAIDI, and EENS). This architecture of shared encoder + task-specific head realizes the decoupling of feature learning sharing and task specificity. It reduces the risk of overfitting and improves data utilization through sharing, while ensuring the flexibility of predicting different metrics through the task head.

[0060] S3. Calculate conservative double-boundary constraints based on training data. For the reliability metrics of key nodes (nodes 3 and 12), conservative double-boundary constraints are calculated based on the training data. The training and validation sets are merged, and the following steps are performed for each output metric: S301, Calculate the quantiles of the data Lower bound quantile In this embodiment, 0.20 (20th percentile) is used. Upper bound quantile In this embodiment, 0.85 (85th percentile) is used. Let the dataset be After sorting, it is ,but:

[0061]

[0062] S302, Calculation Data Range

[0063] S303, Calculate the constraint boundary

[0064]

[0065] in, This is the lower bound distance. This is the upper boundary distance.

[0066] S4. Incorporate constraints into model training using the augmented Lagrangian method. This is the core of constraint-aware evaluation, which transforms the boundary constraints obtained in step S3 into part of the training loss.

[0067] S401. Calculation of Constraint Violation Quantity For the reliability metrics of nodes 3 and 12, a double boundary constraint is defined:

[0068]

[0069] in, Let be the predicted value of the i-th indicator. and These are the corresponding lower and upper boundaries.

[0070] The amount of constraint violation is:

[0071]

[0072] when hour, This indicates a violation of the lower boundary constraint; when hour, If , it indicates a violation of the upper boundary constraint; otherwise, the constraint is satisfied, and the violation amount is 0.

[0073] S402, Conservative Augmented Lagrange Constraint Penalty To reflect the principle of conservative estimation, this invention uses different weights for the lower boundary constraints and the upper boundary constraints:

[0074] in: The weights of the lower boundary constraints indicate which constraints are more important. The weight of the upper boundary constraint indicates that the upper boundary constraint is relatively lenient; and These are the Lagrange multipliers for the lower and upper boundaries, respectively; This is the penalty parameter.

[0075] S403, Constraining the Total Loss of Multi-Task Tasks Add constraint penalties to the total multi-task loss:

[0076] in, This is the multi-task uncertainty weighted loss defined in step 3.

[0077] S404, Update of Lagrange Multipliers After several batches in each training cycle (updated every 10 epochs in this embodiment), the Lagrange multipliers are updated based on constraint violations:

[0078]

[0079] in: and This represents the average number of violations across all batches within that period. The upper bound of the lower boundary multiplier; This represents the upper bound of the multipliers; the update speed of the upper bound multipliers is halved, reflecting their lower priority.

[0080] S405, Update of penalty parameters Increase the penalty parameter when the constraint violation persists:

[0081] Among them, growth factors Maximum penalty parameter .

[0082] Please see Figure 1 (d), penalty parameter The value increases in a stepwise manner during training, from an initial value of 0.01 to a final value of 0.9.

[0083] S5, Model Training Using the prepared training and validation sets, perform end-to-end training on the model and loss function constructed above.

[0084] Training parameters: 1000 training epochs, 64 batch size, 0.001 learning rate, Adam optimizer, early stopping patience value 100, initial penalty parameter. The multiplier is updated every 5 epochs.

[0085] DataLoader is used to batch process the training and validation sets. In each training epoch, forward propagation, loss and constraint penalty calculation, backpropagation, and parameter updates are performed on each batch of the training set. The MSE for each task is calculated on the validation set. The Lagrange multipliers and penalty parameters are updated every 5 epochs.

[0086] Record training history, including training MSE and validation MSE for each task, and constraints and penalties. Penalty parameters. Calculate the comprehensive verification index:

[0087] If the overall validation metrics do not improve for 100 consecutive epochs, stop training and restore the optimal model parameters.

[0088] Please see Figure 1 (a) After training convergence, the validation MSE of each task stabilized at a low level: approximately 0.025 for system-level tasks, approximately 0.035 for node 3 tasks, and approximately 0.03 for node 12 tasks.

[0089] Please see Figure 1 (b) The learned task uncertainty parameters Maintain stability in the later stages of training.

[0090] Please see Figure 1 (c) The penalty for violating the constraint stabilized at around 0.002 in the later stages of training.

[0091] S6. Model Prediction and Evaluation The test set data is normalized, input into the trained model to obtain predicted values, and then inverse normalization is performed to restore the original dimensions.

[0092] Calculate the coefficient of determination for each reliability metric. Root mean square error (RMSE) and mean absolute error (MAE). Please refer to [link / reference]. Figure 2 The scatter plot of the prediction results for various reliability indicators shows: system-level indicators SAIFI, SAIDI, and EENS All are above 0.985; the R² values ​​for CIF, CID, and eens at node 3 are between 0.957 and 0.965; the R² values ​​for CIF, CID, and eens at node 12 are... The values ​​range from 0.979 to 0.988. The predicted values ​​closely match the actual values, with the scatter plots closely distributed near the diagonal.

[0093] The constraint satisfaction of reliability indices for nodes 3 and 12 was checked, and the constraint violation rates were calculated. Statistical results show that the lower boundary violation rate for each index of node 3 is between 2.1% and 3.2%, and the upper boundary violation rate is between 1.5% and 2.0%; for node 12, the lower boundary violation rate is between 1.5% and 1.8%, and the upper boundary violation rate is between 1.0% and 1.2%. The constraint violation rates are all below 5%, indicating that the model basically satisfies the constraints on the test set.

[0094] In another embodiment of the present invention, a distribution network reliability index evaluation system based on binary neural network is provided. This system can be used to implement the above-mentioned distribution network reliability index evaluation method based on binary neural network. Specifically, the distribution network reliability index evaluation system based on binary neural network includes a data module, a construction module, a constraint module, a function module, and a prediction module.

[0095] The data module is used to acquire circuit breaker configuration data of the distribution network and preprocess it to obtain a dataset containing input features and output labels. The output labels include system-level reliability indicators and local reliability indicators of key nodes. A building module is used to build a multi-task binary neural network model, the multi-task binary neural network model including a shared encoder for extracting shared features from input features, and multiple task-specific heads corresponding to the system-level reliability index and the local reliability index of the key node, respectively, wherein the shared encoder contains at least one binary layer; The constraint module is used to calculate conservative double-boundary constraints for the local reliability index of the key node based on the training data. The function module is used to construct a multi-task total loss function containing constraint penalty terms based on the aforementioned dual-boundary constraints using the augmented Lagrange method; The prediction module is used to train the multi-task binary neural network model using the preprocessed dataset and the multi-task total loss function as the optimization objective. The trained multi-task binary neural network model is then used to evaluate and predict the reliability indicators of the distribution network.

[0096] This invention provides a terminal device comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve corresponding method flows or corresponding functions. The processor described in this embodiment can be used in the operation of a distribution network reliability index evaluation method based on binary neural networks, including: The process involves acquiring and preprocessing circuit breaker configuration data from the distribution network to obtain a dataset containing input features and output labels. The output labels include system-level reliability indicators and local reliability indicators for critical nodes. A multi-task binary neural network model is constructed, comprising a shared encoder for extracting shared features from the input features, and multiple task-specific heads corresponding to the system-level reliability indicators and the local reliability indicators for critical nodes, respectively. The shared encoder contains at least one binary layer. Conservative double-boundary constraints for the local reliability indicators of the critical nodes are calculated based on training data. Based on these double-boundary constraints, a multi-task total loss function containing constraint penalty terms is constructed using the augmented Lagrangian method. The multi-task binary neural network model is then trained using the preprocessed dataset and the multi-task total loss function as the optimization objective. Finally, the trained multi-task binary neural network model is used to evaluate and predict the reliability indicators of the distribution network.

[0097] Please see Figure 5The terminal device is a computer device. In this embodiment, the computer device 60 includes a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When the processor 61 executes the computer program 63, it implements the distribution network reliability index evaluation method based on binary neural networks in this embodiment. To avoid repetition, these details are not elaborated here. Alternatively, when the processor 61 executes the computer program 63, it implements the functions of each model / unit in the distribution network reliability index evaluation system based on binary neural networks in this embodiment. To avoid repetition, these details are not elaborated here.

[0098] Computer device 60 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. Computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will understand that... Figure 5 This is merely an example of computer device 60 and does not constitute a limitation on computer device 60. It may include more or fewer components than shown, or combine certain components, or different components. For example, computer device may also include input / output devices, network access devices, buses, etc.

[0099] The processor 61 may be a Central Processing Unit (CPU), or other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0100] The memory 62 can be an internal storage unit of the computer device 60, such as a hard disk or RAM of the computer device 60. The memory 62 can also be an external storage device of the computer device 60, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., provided on the computer device 60.

[0101] Furthermore, the memory 62 may include both internal storage units of the computer device 60 and external storage devices. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 can also be used to temporarily store data that has been output or will be output.

[0102] Please see Figure 6 The terminal device is an electronic device 600, which is manifested in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including storage unit 620 and processing unit 610), a display unit 640, etc.

[0103] The storage unit stores program code, which can be executed by the processing unit 610 to perform the steps described in the method section of this specification according to various exemplary embodiments of the present invention. For example, the processing unit 610 can perform actions such as... Figure 4 The steps are shown in the figure.

[0104] Storage unit 620 may include a readable medium in the form of a volatile storage unit, such as random access memory (RAM) 6201 and / or cache memory 6202, and may further include a read-only memory (ROM) 6203.

[0105] Storage unit 620 may also include a program / utility 6204 having a set (at least one) program module 6205, such program module 6205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0106] Bus 630 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the multiple bus structures.

[0107] Electronic device 600 can also communicate with one or more external devices 700 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., router, modem). This communication can be performed via input / output interface 650. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network, wide area network, and / or public network, such as the Internet) via network adapter 660. Network adapter 660 can communicate with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0108] Example 4 This invention also provides a storage medium, specifically a computer-readable storage medium, which is a memory device in a terminal device for storing programs and data. It is understood that the computer-readable storage medium here can include both built-in storage media in the terminal device and extended storage media supported by the terminal device; it can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which can be one or more computer programs (including program code). More specific examples of the computer-readable storage medium include: an electrical connection with one or more wires, a portable disk, a hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical fiber, portable compact disk read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0109] Computer-readable storage media also include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium can also be any readable medium other than a readable storage medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, radio frequency, etc., or any suitable combination thereof.

[0110] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0111] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the distribution network reliability index evaluation method based on binary neural networks in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor in the following steps: The process involves acquiring and preprocessing circuit breaker configuration data from the distribution network to obtain a dataset containing input features and output labels. The output labels include system-level reliability indicators and local reliability indicators for critical nodes. A multi-task binary neural network model is constructed, comprising a shared encoder for extracting shared features from the input features, and multiple task-specific heads corresponding to the system-level reliability indicators and the local reliability indicators for critical nodes, respectively. The shared encoder contains at least one binary layer. Conservative double-boundary constraints for the local reliability indicators of the critical nodes are calculated based on training data. Based on these double-boundary constraints, a multi-task total loss function containing constraint penalty terms is constructed using the augmented Lagrangian method. The multi-task binary neural network model is then trained using the preprocessed dataset and the multi-task total loss function as the optimization objective. Finally, the trained multi-task binary neural network model is used to evaluate and predict the reliability indicators of the distribution network.

[0112] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0113] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0114] I. Simulation Experiment Design and Comparison Data To verify the superiority of this invention, a standard power distribution network test system (containing 33 nodes, 5 feeders, and 28 circuit breakers) was selected as the simulation object. A simulation platform was built based on PSCAD / EMTDC to simulate 10,000 fault scenarios (including single-point faults, multi-node cascading faults, and equipment aging faults), generating a dataset containing 100,000 samples (24-dimensional input features + 9-dimensional output labels). A comparative experiment was conducted with traditional evaluation methods. The experimental hardware environment consisted of an Intel Core i7-12700H processor and 32GB of memory, while the software environment was Python 3.8 + TensorFlow 2.8.

[0115] 1. Selection of comparison methods Traditional methods: analytical method (fault tree analysis), Monte Carlo simulation method (sampling times of 100,000 times) Existing neural network method: Full-precision multi-task neural network (structure is consistent with this invention, only the binary layer is replaced with a full-precision layer). 2. Core Evaluation Indicators Computational efficiency: time taken per evaluation (seconds / evaluation), number of model parameters (ten thousand); Prediction accuracy: Coefficient of determination (R²) 2 ), root mean square error (RMSE); Constraint satisfaction rate: Key node indicator constraint violation rate (%).

[0116] II. Simulation Experiment Results and Analysis 1. Comparison of computational efficiency data

[0117] This invention replaces the full-precision layer with a binary layer, reducing the number of parameters by 70.3% and shortening the time for a single evaluation by 73.3%, which is only 0.65% of the Monte Carlo simulation method and 4.2% of the analytical method. The total time for 1000 evaluations is only 120 seconds, meeting the real-time requirement for rapid evaluation (minute-level decision-making) after distribution network faults and solving the pain point of low computational efficiency of traditional methods.

[0118] 2. Comparison of prediction accuracy data

[0119] The various indicators R of this invention 2 Both outperform traditional methods and full-precision neural networks, with a system-level performance index R0. 2 The R-value reaches above 0.985, indicating a key local indicator at the critical node. 2Between 0.965 and 0.988, the average RMSE is reduced by 10.3% compared to the full-precision network. This is thanks to the multi-task architecture of shared encoder + task-specific head. Although the binary layer compresses parameters, the feature extraction accuracy is not lost through the complementarity of the full-precision input layer and the task head, thus resolving the contradiction between lightweight and high precision.

[0120] 3. Constraint Satisfaction Comparison Data

[0121] Because full-precision neural networks lack constraint mechanisms, the violation rate of key node indicators is as high as 9.7%, posing a risk to power supply security. This invention, through conservative double-boundary constraints and augmented Lagrangian methods, reduces the average constraint violation rate to 1.8%, and the lower boundary violation rate (minimum reliability guarantee for key nodes) is below 3%, strictly adhering to the conservative estimation principle of distribution network planning. This solves the problem of traditional methods lacking key node constraint guarantees, ensuring that the evaluation results meet the actual engineering requirements.

[0122] Simulation experiments and practical application data show that the present invention significantly outperforms traditional methods and existing neural network methods in terms of computational efficiency, prediction accuracy, and constraint satisfaction: the number of parameters is reduced by more than 70%, the computation speed is increased by 4-150 times, and the predicted R... 2 The accuracy is ≥0.957, and the constraint violation rate is ≤3.2%. This invention completely solves the core problems of slow calculation, low accuracy, and lack of constraints in traditional distribution network reliability assessment. It provides scientific and efficient technical support for real-time planning and fault recovery decision-making in distribution networks, and has extremely high engineering application value.

[0123] In summary, the present invention provides a method and system for evaluating the reliability index of a distribution network based on a binary neural network, which has the following advantages: (1) Lightweight model achieved through binary neural network. The shared encoder uses binary layer to binarize the weights to +1 or -1, which greatly reduces the number of model parameters and computational complexity. Compared with traditional full-precision neural network, the computational efficiency is significantly improved, making it easy to deploy on resource-constrained embedded devices.

[0124] (2) Feature sharing is achieved through multi-task learning. The shared encoder extracts the general features of the circuit breaker configuration, and multiple task-specific heads output system-level indicators and node-level indicators respectively. The correlation between tasks is used to improve the generalization ability of the model and avoid the redundancy of training the model separately for each task.

[0125] (3) Conservative estimation is achieved through data-driven constraint calculation. The constraint boundary is automatically calculated based on the quantiles and ranges of the training data, avoiding the subjectivity of manually setting thresholds; the asymmetric margin setting is adopted, with tighter lower boundary constraints to ensure conservative estimation and looser upper boundary constraints to prevent over-conservatism, which meets the actual needs of distribution network planning.

[0126] (4) Constraint fusion is achieved through augmented Lagrange method. The constraints of key nodes are integrated into the model training process. Dynamically adjusted Lagrange multipliers and penalty parameters are used. The model gradually satisfies the constraints while optimizing the prediction accuracy. Asymmetric weights and upper bounds of multipliers are set to ensure the priority of lower boundary constraints and reflect the conservative estimation principle.

[0127] (5) Improve numerical stability through grouping normalization. Calculate normalization parameters for the output indicators of different tasks to avoid mutual interference between indicators with different dimensions and numerical ranges, thereby improving the stability and convergence speed of model training.

[0128] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0129] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0130] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0131] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0132] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0133] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0134] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random-access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0135] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0136] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0137] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0138] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for evaluating the reliability index of distribution networks based on binary neural networks, characterized in that, Includes the following steps: The circuit breaker configuration data of the distribution network is acquired and preprocessed to obtain a dataset containing input features and output labels. The output labels include system-level reliability indicators and local reliability indicators of key nodes. A multi-task binary neural network model is constructed, the multi-task binary neural network model including a shared encoder for extracting shared features from input features, and multiple task-specific heads corresponding to the system-level reliability index and the local reliability index of the key node, respectively, wherein the shared encoder contains at least one binary layer; Conservative double-boundary constraints are used to calculate the local reliability index of the key node based on training data. Based on the aforementioned dual-boundary constraints, a multi-task total loss function containing constraint penalty terms is constructed using the augmented Lagrange method. Using the preprocessed dataset, the multi-task total loss function is used as the optimization objective to train the multi-task binary neural network model. The trained multi-task binary neural network model is then used to evaluate and predict the reliability indicators of the distribution network.

2. The method for evaluating the reliability index of a distribution network based on a binary neural network according to claim 1, characterized in that, The preprocessing includes: normalizing the input data as a whole, and normalizing the output labels by task grouping.

3. The method for evaluating the reliability index of a distribution network based on a binary neural network according to claim 1, characterized in that, The shared encoder comprises, in sequence, an input layer, at least one full-precision dense layer, and at least one binary dense layer; the task-specific header is a full-precision dense layer.

4. The method for evaluating the reliability index of a distribution network based on a binary neural network according to claim 1, characterized in that, The conservative double-boundary constraints for calculating the local reliability index of the key nodes based on training data specifically include: Based on the data distribution of local reliability indicators of each key node in the training data, the lower bound quantile and upper bound quantile are calculated respectively, and the lower boundary and upper boundary of each indicator are determined based on the quantile and the preset margin.

5. The method for evaluating the reliability index of a distribution network based on a binary neural network according to claim 1, characterized in that, Based on the aforementioned dual-boundary constraints, the multi-task total loss function, which includes a constraint penalty term, is constructed using the augmented Lagrange method, specifically including: The model predicts the amount of violation of the dual boundary constraints; the constraint penalty term is constructed based on the amount of violation, the Lagrange multiplier, and the penalty parameter; the constraint penalty term is added to the multi-task uncertainty weighted loss to obtain the multi-task total loss function.

6. The method for evaluating the reliability index of a distribution network based on a binary neural network according to claim 5, characterized in that, When constructing the constraint penalty term, different weights are assigned to the lower boundary constraint and the upper boundary constraint, with the weight of the lower boundary constraint being greater than that of the upper boundary constraint.

7. The method for evaluating the reliability index of a distribution network based on a binary neural network according to claim 5, characterized in that, During training, the Lagrange multipliers are updated periodically based on constraint violations, with the update rate of the multipliers corresponding to the lower boundary constraints being greater than that of the multipliers corresponding to the upper boundary constraints.

8. The method for evaluating the reliability index of a distribution network based on a binary neural network according to claim 7, characterized in that, The penalty parameter is increased when the constraint violation persists.

9. The method for evaluating the reliability index of a distribution network based on a binary neural network according to claim 1, characterized in that, The training process employs an early stopping strategy. When the comprehensive verification index, calculated based on the verification mean square error of each task and the constraint violation penalty, fails to improve within a preset period, training is stopped and the optimal model parameters are restored.

10. A distribution network reliability index evaluation system based on binary neural network, characterized in that, include: The data module is used to acquire circuit breaker configuration data of the distribution network and preprocess it to obtain a dataset containing input features and output labels. The output labels include system-level reliability indicators and local reliability indicators of key nodes. A building module is used to build a multi-task binary neural network model, the multi-task binary neural network model including a shared encoder for extracting shared features from input features, and multiple task-specific heads corresponding to the system-level reliability index and the local reliability index of the key node, respectively, wherein the shared encoder contains at least one binary layer; The constraint module is used to calculate conservative double-boundary constraints for the local reliability index of the key node based on the training data. The function module is used to construct a multi-task total loss function containing constraint penalty terms based on the aforementioned dual-boundary constraints using the augmented Lagrange method; The prediction module is used to train the multi-task binary neural network model using the preprocessed dataset and the multi-task total loss function as the optimization objective, and to evaluate and predict the reliability indicators of the distribution network using the trained multi-task binary neural network model.

Images