Feeder terminal automation processing method, device and electronic equipment

Abstract

The application provides a kind of feeder terminal automation processing method, device and electronic equipment, belong to power distribution network technical field.The method includes: respectively to electrical quantity data, equipment state data and environmental data are extracted features, obtain the first feature corresponding to electrical quantity data, the second feature corresponding to equipment state data and the third feature corresponding to environmental data;First feature, second feature and third feature are fused, and a fusion feature vector is obtained;Control strategy is generated based on electrical quantity data, equipment state data and fusion feature vector using a hybrid driving model;The hybrid driving model is based on a physical mechanism model and a data-driven model to construct.The application is used to solve the problem that the existing feeder terminal is difficult to achieve accurate fault detection and rapid recovery when facing complex power distribution network operation scenarios, such as multiple fault types and variable environmental conditions, which affects the reliability and stability of the power distribution network.

Description

Technical Field

[0001] This invention relates to the field of power distribution network technology, and more specifically to a feeder terminal automated processing method, a feeder terminal automated processing device, an electronic device, a machine-readable storage medium, and a computer program product. Background Technology

[0002] In power distribution networks, feeder terminals, as crucial equipment, undertake key tasks of data acquisition, processing, and control. Specifically, existing feeder terminals generally consist of a data acquisition module, a processing module, and a communication module. The data acquisition module collects limited electrical parameters, the processing module performs simple logical judgments and generates control commands based on a pre-set physical mechanism model, and the communication module uploads information to the master station or receives commands from the master station. This technical solution struggles to achieve accurate fault detection and rapid recovery in complex distribution network operation scenarios, such as multiple fault types and changing environmental conditions, thus impacting the reliability and stability of the distribution network. Summary of the Invention

[0003] The purpose of this invention is to provide an automated processing method, device, and electronic device for feeder terminals, in order to solve the shortcomings of existing feeder terminals in complex power distribution network operation scenarios, such as multiple fault types and changing environmental conditions, which make it difficult to achieve accurate fault detection and rapid recovery, thus affecting the reliability and stability of the power distribution network.

[0004] To achieve the above objectives, embodiments of the present invention provide an automated processing method for feeder terminals, comprising:

[0005] Acquire electrical quantity data, equipment status data, and environmental data of the power distribution network within a set time period;

[0006] Feature extraction is performed on the electrical quantity data, equipment status data, and environmental data respectively to obtain a first feature corresponding to the electrical quantity data, a second feature corresponding to the equipment status data, and a third feature corresponding to the environmental data;

[0007] The first feature, the second feature, and the third feature are fused to obtain a fused feature vector.

[0008] A control strategy is generated using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector, so that the execution equipment in the distribution network can achieve automated processing based on the control strategy;

[0009] The hybrid driving model is constructed based on a physical mechanism model and a data-driven model.

[0010] Optionally, the step of generating a control strategy using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector includes:

[0011] The first fault type prediction result is determined based on the electrical quantity data and the equipment status data using the physical mechanism model.

[0012] The data-driven model is used to determine the prediction result of the second fault type based on the fused feature vector;

[0013] Using a weighted voting mechanism, the final prediction result of the fault type is determined based on the prediction results of the first fault type and the prediction results of the second fault type;

[0014] Using the physical mechanism model and the data-driven model, a control strategy is generated based on the final prediction result of the fault type.

[0015] Optionally, the step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes:

[0016] If the final prediction result of the fault type indicates that no fault has occurred in the distribution network, a basic operation strategy for the distribution network is generated based on the physical mechanism model.

[0017] Based on the data-driven model, load forecasting of the distribution network is performed, and load forecasting data of the distribution network is output.

[0018] The basic operation strategy is dynamically adjusted based on the load forecast data to generate a dynamic adjustment strategy; the dynamic adjustment strategy serves as the control strategy.

[0019] Optionally, the step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes:

[0020] When the final prediction result of the fault type indicates that a fault has occurred in the distribution network, the data-driven model is used to determine the fault location prediction result based on the fused feature vector;

[0021] Using the physical mechanism model, a self-healing control strategy is generated based on the fault location prediction results; the self-healing control strategy serves as the control strategy.

[0022] Optionally, after generating a control strategy using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector, so that the execution equipment of the distribution network can achieve automated processing based on the control strategy, the method further includes:

[0023] Determine the fault isolation time achievement rate and / or load recovery rate of the distribution network after implementing the control strategy; the fault isolation time achievement rate is directly proportional to the actual fault isolation time of the distribution network after implementing the control strategy; the load recovery rate is directly proportional to the number of users whose power supply has been restored.

[0024] If the fault isolation time achievement rate and / or load recovery rate do not meet the set conditions, the process jumps to the step of generating a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid drive model, so that the execution equipment of the distribution network can realize automated processing based on the control strategy.

[0025] Optionally, the step of fusing the first feature, the second feature, and the third feature to obtain a fused feature vector includes:

[0026] The first feature, the second feature, and the third feature are fused using a weighted fusion method and / or a neural network fusion method to obtain a fused feature vector.

[0027] On the other hand, embodiments of the present invention also provide an automated processing device for feeder terminals, comprising:

[0028] The data acquisition module is used to acquire electrical quantity data, equipment status data, and environmental data of the power distribution network within a set time period;

[0029] The edge computing module is used to extract features from the electrical quantity data, equipment status data, and environmental data respectively, to obtain a first feature corresponding to the electrical quantity data, a second feature corresponding to the equipment status data, and a third feature corresponding to the environmental data.

[0030] The data fusion module is used to fuse the first feature, the second feature, and the third feature to obtain a fused feature vector;

[0031] An automation control module is used to generate a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid drive model, so that the execution equipment of the distribution network can realize automated processing based on the control strategy;

[0032] The hybrid driving model is constructed based on a physical mechanism model and a data-driven model.

[0033] Optionally, the step of generating a control strategy using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector includes:

[0034] The first fault type prediction result is determined based on the electrical quantity data and the equipment status data using the physical mechanism model.

[0035] The data-driven model is used to determine the prediction result of the second fault type based on the fused feature vector;

[0036] Using a weighted voting mechanism, the final prediction result of the fault type is determined based on the prediction results of the first fault type and the prediction results of the second fault type;

[0037] Using the physical mechanism model and the data-driven model, a control strategy is generated based on the final prediction result of the fault type.

[0038] Optionally, the step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes:

[0039] If the final prediction result of the fault type indicates that no fault has occurred in the distribution network, a basic operation strategy for the distribution network is generated based on the physical mechanism model.

[0040] Based on the data-driven model, load forecasting of the distribution network is performed, and load forecasting data of the distribution network is output.

[0041] The basic operation strategy is dynamically adjusted based on the load forecast data to generate a dynamic adjustment strategy; the dynamic adjustment strategy serves as the control strategy.

[0042] Optionally, the step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes:

[0043] When the final prediction result of the fault type indicates that a fault has occurred in the distribution network, the data-driven model is used to determine the fault location prediction result based on the fused feature vector;

[0044] Using the physical mechanism model, a self-healing control strategy is generated based on the fault location prediction results; the self-healing control strategy serves as the control strategy.

[0045] Optionally, after generating a control strategy using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector, so that the execution equipment of the distribution network can achieve automated processing based on the control strategy, the method further includes:

[0046] Determine the fault isolation time achievement rate and / or load recovery rate of the distribution network after implementing the control strategy; the fault isolation time achievement rate is directly proportional to the actual fault isolation time of the distribution network after implementing the control strategy; the load recovery rate is directly proportional to the number of users whose power supply has been restored.

[0047] If the fault isolation time achievement rate and / or load recovery rate do not meet the set conditions, the process jumps to the step of generating a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid drive model, so that the execution equipment of the distribution network can realize automated processing based on the control strategy.

[0048] Optionally, the step of fusing the first feature, the second feature, and the third feature to obtain a fused feature vector includes:

[0049] The first feature, the second feature, and the third feature are fused using a weighted fusion method and / or a neural network fusion method to obtain a fused feature vector.

[0050] On the other hand, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described feeder terminal automation processing method.

[0051] On the other hand, the present invention also provides a machine-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described feeder terminal automation processing method.

[0052] On the other hand, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described feeder terminal automated processing method.

[0053] Through the above technical solution, this embodiment of the invention utilizes a hybrid driving model to generate a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector, enabling the execution equipment in the distribution network to achieve automated processing based on the control strategy. Thus, this embodiment of the invention, through multimodal sensing, collects various types of data and combines a hybrid driving approach of physical mechanism models and data-driven models to achieve self-healing control from data acquisition. In the face of complex distribution network operation scenarios, it achieves accurate fault detection and rapid recovery, improving the reliability and stability of the distribution network.

[0054] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0055] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0056] Figure 1 This is one of the flowcharts illustrating the automated processing method for feeder terminals provided by the present invention;

[0057] Figure 2 This is the second flowchart of the automated processing method for feeder terminals provided by the present invention;

[0058] Figure 3 This is the third flowchart of the automated processing method for feeder terminals provided by the present invention;

[0059] Figure 4 This is the fourth flowchart of the automated processing method for feeder terminals provided by the present invention;

[0060] Figure 5 This is one of the structural schematic diagrams of the automated processing device for feeder terminals provided by the present invention;

[0061] Figure 6 This is the second schematic diagram of the structure of the automated processing device for feeder terminals provided by the present invention;

[0062] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0063] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0064] Method Implementation Examples

[0065] Please refer to Figure 1 This invention provides an automated processing method for feeder terminals, comprising:

[0066] Step 100: Obtain electrical quantity data, equipment status data, and environmental data of the power distribution network within a set time period.

[0067] The automated processing method for feeder terminals in this invention is applied to electronic equipment. In one embodiment, the electronic equipment can be a feeder terminal of a power distribution network. The feeder terminal acquires electrical quantity data, equipment status data, and environmental data of the power distribution network within a set time period. The set time period can be various time periods as needed, such as 6:00 to 22:00. The electrical quantity data can include at least one of the voltage, current, power, and frequency of the power distribution network. To consider as many electrical quantity factors as possible and form a multi-type data perception, the electrical quantity data can include the voltage, current, power, and frequency of the power distribution network. This electrical quantity data can be collected by sensors such as voltage transformers and current transformers. The equipment status data can include at least one of the temperature, vibration amplitude, and insulation resistance of the equipment (e.g., switchgear, power parameter detection equipment, etc.). To consider as many equipment status factors as possible and form a multi-type data perception, the equipment status data can include the temperature, vibration amplitude, and insulation resistance of the equipment (e.g., switchgear, power parameter detection equipment, etc.). This equipment status data can be acquired by temperature sensors, vibration sensors, and insulation monitoring devices. The environmental data can include at least one of the temperature, humidity, wind speed, and rainfall of the location of the power distribution network. To consider as many environmental factors as possible and generate multi-type data, environmental data can include temperature, humidity, wind speed, and rainfall at the location of the power distribution network. This environmental data can be collected through environmental sensors.

[0068] It should be noted that all the aforementioned sensors can synchronously collect data at a preset frequency, thereby forming a time-aligned multimodal raw dataset. This multimodal raw dataset includes time-series data of electrical quantities, equipment status, and environmental conditions.

[0069] It should be noted that after obtaining electrical quantity data, equipment status data, and environmental data, please refer to... Figure 2 The feeder terminal performs data preprocessing on the collected raw data. Data preprocessing can include data standardization and data cleaning. Specifically, the feeder terminal performs format standardization on the collected raw data, unifying the data units and timestamp format. Then, it removes noise interference through filtering algorithms (such as Kalman filtering and wavelet denoising), and uses normalization methods (min-max normalization or Z-score normalization) to map feature data from different ranges to a unified interval, forming a preprocessed multimodal standard dataset.

[0070] Step 200: Extract features from the electrical quantity data, equipment status data, and environmental data respectively to obtain the first feature corresponding to the electrical quantity data, the second feature corresponding to the equipment status data, and the third feature corresponding to the environmental data.

[0071] The feeder terminal can utilize machine learning algorithms (such as convolutional neural networks and long short-term memory networks) to extract features from the preprocessed time series data, obtaining a first feature corresponding to the electrical quantity data, a second feature corresponding to the equipment status data, and a third feature corresponding to the environmental data, thereby capturing the temporal correlation and nonlinear characteristics of the data. For example, in embodiments of the present invention, frequency domain features (dominant frequency, harmonic components) are extracted from vibration amplitude, and transient process features (voltage sag amplitude, recovery time) are extracted from electrical quantity data.

[0072] In another embodiment, the feeder terminal can also calculate the theoretical operating parameters of the equipment under ideal operating conditions (such as theoretical voltage distribution and equipment temperature rise threshold) based on the distribution network circuit theory and the equipment operation physical mechanism model. It then compares the measured data (electrical quantity data and equipment status data) with the theoretical operating parameters to identify abnormal data points that significantly deviate from the theoretical values, generating a preliminary abnormal state signal. This facilitates the formation of a composite feature vector containing both physical and data-driven features based on the aforementioned feature extraction results and abnormal data points.

[0073] Step 300: Perform feature fusion on the first feature, the second feature and the third feature to obtain a fused feature vector.

[0074] In some embodiments, the feeder terminal can use a weighted fusion method and / or a neural network fusion method to fuse the first feature, the second feature, and the third feature to obtain a fused feature vector. That is, the feeder terminal can use a weighted fusion method, a neural network fusion method, or a combination of weighted fusion methods and neural network fusion methods to fuse the first feature, the second feature, and the third feature to obtain a fused feature vector.

[0075] For example, in one embodiment, the feeder terminal can assign different weights to the first feature, the second feature, and the third feature, such as a weight of 0.5 for the first feature, a weight of 0.3 for the second feature, and a weight of 0.2 for the third feature. The first feature, the second feature, and the third feature are then weighted and fused using the following formula to obtain the fused feature vector:

[0076] The fused feature vector = first feature * 0.5 + second feature * 0.3 + third feature * 0.2.

[0077] In another embodiment, the feeder terminal first inputs the first feature, the second feature, and the third feature into a temporal convolutional network for temporal alignment; then, it inputs the temporally aligned first feature, second feature, and third feature into a Tranformer encoder, and achieves feature fusion through a cross-modal attention mechanism to obtain a fused feature vector.

[0078] In another embodiment, the feeder terminal can use a weighted fusion method to initially fuse feature vectors from different modes, dynamically adjusting the weights based on the reliability of the data source and scenario requirements (e.g., increasing the weight of electrical quantity data in fault scenarios). Further, a cross-modal fusion model is constructed using a deep neural network. This model takes multi-modal feature vectors as input and outputs a fused feature vector that comprehensively reflects the operating status of the distribution network, achieving a multi-dimensional assessment of equipment health status and grid operational stability. For example, integrating the anomaly degree (weight 0.6), health degree (weight 0.3), and environmental risk value (weight 0.1) output from the feature extraction in step 200 generates a comprehensive status assessment result ("normal," "warning," "fault"), outputting a probability distribution matrix. The output of the comprehensive status assessment result and its probability distribution provides a decision-making basis for subsequent control strategy generation.

[0079] Step 400: Using the hybrid drive model, a control strategy is generated based on the electrical quantity data, the equipment status data, and the fused feature vector, so that the execution equipment of the distribution network can achieve automated processing based on the control strategy.

[0080] The hybrid driving model is constructed based on a physical mechanism model and a data-driven model. The physical mechanism model refers to mathematical equations such as circuit models and equipment operation models built upon prior physical principles. The data-driven model refers to models that rely on historical or real-time data and use machine learning algorithms (such as neural networks and fuzzy logic) to uncover patterns. This embodiment of the invention, based on the electrical quantity data, the equipment status data, and the fused feature vector, combines the physical-data hybrid driving model to generate corresponding control strategies, enabling the power distribution network's execution equipment to achieve automated processing based on these control strategies.

[0081] In one embodiment, step 400, generating a control strategy using a hybrid driving model based on the electrical quantity data, the equipment status data, and the fused feature vector, includes: using the physical mechanism model based on the electrical quantity data and the equipment status data to determine a first fault type prediction result; using the data-driven model based on the fused feature vector to determine a second fault type prediction result; using a weighted voting mechanism to determine a final fault type prediction result based on the first fault type prediction result and the second fault type prediction result; and using the physical mechanism model and the data-driven model to generate a control strategy based on the final fault type prediction result.

[0082] The physical mechanism model of this invention can establish mathematical equations for the transient process of power distribution network faults based on Kirchhoff's laws and the equivalent circuit model of equipment, clarifying the mapping relationship between fault characteristics and physical parameters (e.g., the relationship between short-circuit fault current and fault distance). Using this physical mechanism model, the fault type can be determined. For example, the phase current and zero-sequence current are obtained through electrical quantity data and equipment status data. By calculating and comparing the phase current and zero-sequence current, if the voltage of a certain phase is close to zero and the current surges, while the zero-sequence current increases significantly, the physical mechanism model will directly indicate that this is a "single-phase ground fault" (first fault type prediction result). The feeder terminal then uses the data-driven model based on the fused feature vector to determine the second fault type prediction result. The data-driven model can utilize patterns and associations learned from historical data to identify and predict faults. That is, the data-driven model uses historical fault data to train a fault classification model (such as support vector machine, random forest) to identify fault types such as single-phase ground fault and two-phase short circuit. For example, the data-driven model is trained based on historical fused feature vectors and corresponding fault type labels obtained from electrical quantity data, equipment status data, and environmental data. The calculation process for the historical fused feature vectors and the fused feature vectors is the same and will not be repeated here. Thus, the feeder terminal uses the data-driven model to determine the second fault type prediction result based on the fused feature vectors. Finally, using a weighted voting mechanism, the final fault type prediction result is determined based on the first and second fault type prediction results. For example, the first fault type prediction result of the physical mechanism model is a single-phase ground fault with a probability of 0.6 and a two-phase short-circuit fault with a probability of 0.4. The second fault type prediction result of the data-driven model is a single-phase ground fault with a probability of 0.3 and a two-phase short-circuit fault with a probability of 0.7. The weight of the physical mechanism model is 2, and the weight of the data-driven model is 3. The weighted probability of a single-phase ground fault is: (0.6*2+0.3*3) / (2+3)=0.42;

[0083] The weighted probability of the fault being a two-phase short-circuit fault is: (0.4*2+0.7*3) / (2+3)=0.58. Since 0.58 is greater than 0.42, the final predicted fault type is a two-phase short-circuit fault.

[0084] The feeder terminal utilizes the physical mechanism model and the data-driven model to generate control strategies based on the final prediction results of the fault type. For example, if the final prediction result of the fault type indicates no fault has occurred, an economical operation control strategy is generated based on the physical mechanism model, and dynamically adjusted in conjunction with the load fluctuation trend predicted by the data-driven model. If the final prediction result indicates a fault has occurred, the data-driven model is used to quickly identify the fault location, and a reasonable control strategy, such as fault isolation, load transfer, and power switching, is formulated in conjunction with the physical mechanism model. The feeder terminal converts the control strategy into executable control commands (such as circuit breaker opening and closing commands, distributed power supply start and stop signals) through a standardized interface and sends them to the relevant execution equipment in the distribution network. It should be noted that a secure encryption module is used for two-way authentication (device identity verification + command integrity verification) during command transmission to ensure the security and reliability of the control commands.

[0085] This invention utilizes a hybrid driving model to generate a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector, enabling the distribution network's execution equipment to achieve automated processing based on this control strategy. Thus, this invention achieves self-healing control from data acquisition through multimodal sensing, collecting various types of data, and combining a physical mechanism model with a data-driven model. In complex distribution network operation scenarios, it enables accurate fault detection and rapid recovery, improving the reliability and stability of the distribution network.

[0086] In other aspects of the embodiments of the present invention, the step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes: generating a basic operation strategy for the distribution network based on the physical mechanism model when the final prediction result of the fault type indicates that no fault has occurred in the distribution network; performing load forecasting for the distribution network based on the data-driven model and outputting load forecast data for the distribution network; dynamically adjusting the basic operation strategy based on the load forecast data to generate a dynamic adjustment strategy; and using the dynamic adjustment strategy as the control strategy.

[0087] In normal operation scenarios without faults, the feeder terminal can generate multi-objective optimization functions (e.g., minimizing line network losses and maximizing photovoltaic absorption rate) based on the physical mechanism model, and solve these functions to obtain the basic operation strategy. Then, a data-driven model is used to perform load forecasting for the distribution network, obtaining load forecast data. Finally, based on the load forecast data from the data-driven model, the basic operation strategy of the physical mechanism model is dynamically adjusted.

[0088] For example, taking a 10kV urban distribution network (including 3 residential transformer substations, 1 commercial transformer substation, and 1 1MW distributed photovoltaic power station) as an example under normal operating conditions, the implementation process of the control strategy is illustrated:

[0089] First, a physical mechanism model generates a basic economic operation strategy. Based on distribution network line parameters (resistance R, reactance X), transformer efficiency curves, and photovoltaic inverter characteristics, the physical mechanism model establishes the core optimization objective: minimizing line losses and maximizing photovoltaic absorption rate. According to Kirchhoff's laws, the formula for line active power loss is: P Loss = 3 I 2 R *10 -3 (I is the line current, in A; R is the line resistance, in Ω), and the current I =3 U cos φS (S represents apparent power, U represents line voltage, and cosφ represents power factor). Next, the basic operation strategy is output: The physical mechanism model calculates that, under the current average load (approximately 800kW), economical operation can be achieved through the following operations: The transformer tap is set to "Level 3" to stabilize the output voltage at 10.5kV (+5% of the rated voltage of the 10kV line, reducing current to minimize line loss); two sets of parallel capacitors (total capacity 300kvar) are put into operation to maintain the line power factor at 0.95 (avoiding additional losses due to reactive current); the photovoltaic inverter operates in "local consumption priority" mode, with excess power only fed back to the grid at a rate of ≤100kW (avoiding voltage exceedances due to excessive reverse power flow).

[0090] Secondly, load fluctuations are predicted using a data-driven model. A Long Short-Term Memory (LSTM) network model is trained using 12 months of historical load data (sampling point every 15 minutes), weather data (temperature, humidity), and date type (weekday / weekend / holiday) to predict the load fluctuation trend of the distribution network for the next 24 hours. The output load prediction data is as follows: 10:00-15:00 the next day: Due to the weekday and temperature rising to 35℃, the air conditioning load in commercial areas surges, and the total load is predicted to increase from 800kW to 1100kW (an increase of 37.5%); 00:00-06:00 the next day: Residential and commercial loads are at their lowest point, and the total load is predicted to drop to 450kW (only 56% of the average load); Photovoltaic output prediction: The next day is sunny, and the photovoltaic output reaches a peak of 800kW from 12:00-14:00, overlapping with the peak load period.

[0091] Finally, based on the load fluctuation prediction of the data-driven model, the basic operating strategy of the physical mechanism model is dynamically adjusted, as follows:

[0092] To cope with the load peak from 10:00 to 15:00: The basic operating strategy of the physical mechanism model will lead to an increase in line loss due to increased current during load surges (calculated based on a base voltage of 10.5kV, network loss will increase by 22%). Adjustment measures: 1 hour in advance (9:00), adjust the transformer tap to "level 4" to raise the voltage to 10.7kV (using higher voltage to reduce current, theoretically reducing line loss by 15%); put an additional 150kvar capacitor into operation to maintain a power factor ≥0.94 (to offset the reactive power demand brought about by the increased load); unlock the "full grid connection" restriction of the photovoltaic inverter (because the peak load can absorb all photovoltaic output), avoiding curtailment losses.

[0093] Addressing the 00:00-06:00 load trough: The basic operating strategy of the physical mechanism model is that high voltage may lead to increased transformer iron losses during low load periods (iron losses are proportional to the square of the voltage when the voltage is too high). Adjustment measures: 30 minutes in advance (23:30), adjust the transformer tap to "Level 2" to reduce the voltage to 10.3kV (reducing iron losses; actual measured iron losses decreased by 18%); disconnect one group of 150kvar capacitors (to avoid voltage rise caused by overcompensation under low load); photovoltaic power output is zero at night, so no inverter strategy adjustment is needed.

[0094] Thus, by combining the theoretical optimal solution of the physical mechanism model (basic operation strategy) with the load forecast of the data-driven model (dynamic adjustment), the daily grid loss of the line is reduced by 12.3% compared with the simple physical mechanism model strategy, and the photovoltaic absorption rate is increased from 85% to 98% of the basic operation strategy, achieving the dual goals of "economic operation + adaptive adjustment".

[0095] In other aspects of the embodiments of the present invention, the step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes: when the final prediction result of the fault type indicates that a fault has occurred in the distribution network, determining the fault location prediction result based on the fused feature vector using the data-driven model; generating a self-healing control strategy based on the fault location prediction result using the physical mechanism model; and using the self-healing control strategy as the control strategy.

[0096] Data-driven models can also utilize historical fault data to train fault location identification models (such as support vector machines and random forests). For example, a data-driven model can be trained based on historical fusion feature vectors obtained from electrical quantity data, equipment status data, and environmental data, along with the corresponding fault location labels. When the final prediction result of the fault type indicates a fault in the distribution network, the feeder terminal prioritizes using a data-driven model to quickly locate the fault location (response time ≤ 50ms), calculates a self-healing control strategy based on a physical mechanism model (such as disconnecting the nearest circuit breaker combination), and generates a control strategy set containing switch operation commands and load transfer paths (such as "trip the circuit breaker in the fault section → close the tie switch to transfer the load → report to the master station")

[0097] For example, the following example of a single-phase grounding fault in a 10kV distribution network illustrates how to calculate a self-healing control strategy based on a physical mechanism model and how to work in conjunction with a data-driven model to complete fault handling.

[0098] Please refer to Figure 3 The physical model calculation process for the self-healing control strategy is as follows: First, input data: power grid topology (including switch and circuit breaker locations), line parameters (impedance per unit length: R=0.27Ω / km, X=0.35Ω / km), fault section information (SW1-SW2 section), and load current distribution (SW1: 200A, SW2: 150A). Next, the physical model calculation process is performed, including fault current path analysis and minimum outage range calculation. The fault current path analysis is as follows: Upstream current path from the fault point: Substation A → CB1 → SW1 → Fault point; Downstream current path from the fault point: Fault point → SW2 → Load (150A). The minimum outage range calculation is as follows: Option 1: Only trip SW1. Disadvantage: The downstream area (SW1-SW2 section) remains energized, and the ground fault continues → potentially causing equipment damage. Option 2: Trip SW1+SW2. Advantage: Completely isolates the fault section, but causes a power outage for all loads (including healthy parts) in the SW1-SW2 section. Option 3: Trip CB1 (substation outlet circuit breaker). Disadvantage: The entire F1 feeder line is de-energized, resulting in an excessively large area. Based on Options 1 to 3, the self-healing control strategy is: Option 2 (trip SW1+SW2) is selected because: it conforms to the principle of "minimum fault isolation range" (only the SW1-SW2 section is de-energized). There are no important loads downstream of the fault point, and a short-term power outage is acceptable. The capacity of the tie switch SW3 is 400A, which is greater than the current remaining load of F1 (150A). Voltage drop calculation: After transfer, the terminal voltage = 0.97pu > the lower limit of 0.95pu. Operation sequence: Trip SW1+SW2 (isolate the fault), close SW3 (transfer power from substation B to the downstream load of SW2 via feeder F2). Please refer to... Figure 4The final physical mechanism model output is: self-healing control strategy = {skip SW1, skip SW2, close SW3}; verification result = {voltage qualified, no overload}.

[0099] The embodiments of the present invention combine the reliability of the physical mechanism model and the adaptability of the data-driven model with a hybrid driving model, which enables more accurate fault diagnosis, prediction and control strategy generation under complex working conditions, overcoming the limitations of a single model.

[0100] In other aspects of the embodiments of the present invention, after generating a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid drive model to enable the execution equipment of the distribution network to automate processing based on the control strategy, the method further includes: determining the fault isolation time achievement rate and / or load recovery rate of the distribution network after executing the control strategy; the fault isolation time achievement rate is directly proportional to the actual fault isolation time of the distribution network after executing the control strategy; the load recovery rate is directly proportional to the number of users whose power supply has been restored; if the fault isolation time achievement rate and / or load recovery rate do not meet the set conditions, the method jumps to the step of generating a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid drive model to enable the execution equipment of the distribution network to automate processing based on the control strategy.

[0101] After the power distribution network's execution equipment operates, the feeder terminal collects post-control status data in real time (such as switch position signals and post-fault voltage and current recovery values). The feeder terminal calculates the control target achievement rate (fault isolation time achievement rate and / or load recovery rate). If the fault isolation time achievement rate and / or load recovery rate do not meet the set conditions, the control strategy iteration mechanism is triggered, returning to step 400 for execution, forming a closed-loop processing flow of "collection-analysis-control-feedback". It should be noted that, in this embodiment of the invention, the control target achievement rate can be calculated based on the fault isolation time achievement rate, or the load recovery rate, or both.

[0102] The fault isolation time achievement rate is calculated based on the ratio of the actual fault isolation time to the target fault isolation time. Expected target: Fault isolation time ≤ 100ms (the time from strategy execution to the disappearance of the fault current). For example, the actual fault isolation time means that after SW1 and SW2 are tripped, the zero-sequence current in area B drops from 300A to 5A in 70ms. The target fault isolation time is 100ms. Fault isolation time achievement rate = Actual fault isolation time / Target fault isolation time * 100% = 70ms / 100ms * 100% = 70% (not met; ≥ 90% is required for acceptance).

[0103] The load recovery rate is calculated based on the ratio of users with restored power to the total number of users. Expected target: Load recovery rate for non-faulty transformer areas (A, C) ≥ 95% (number of users with restored power / total number of users). For example, if all 300 users in transformer area A have their power restored, and 190 out of 200 users in transformer area C have their power restored (10 users were not restored due to a voltage dip), then the load recovery rate = (300 + 190) / (300 + 200) * 100% = 490 / 500 * 100% = 98% (meets the target).

[0104] This invention uses the fault isolation time achievement rate as the evaluation metric. This invention found that a fault isolation time achievement rate of 70% < 90% indicates a significant bias in the assessment. Further analysis of the reasons for this bias is as follows:

[0105] Data shows that the K1 switch tripping time is 20ms slower than expected (due to mechanical jamming), resulting in a delay in fault current cutoff. This triggers the control strategy iteration mechanism, returning to step 400 to re-optimize the hybrid drive model. The physical model is corrected as follows: based on the actual SW1 tripping time (50ms), the timing of "tripping SW2 first, then SW1" is recalculated (SW2 has no fault current, so tripping is faster), theoretically shortening the isolation time to 50ms. The data-driven model is corrected as follows: the "emergency strategy for switch jamming" from historical data is called, adding redundant control of "repeatedly sending the SW1 tripping command 3 times"; the new strategy output is: "Send SW2 tripping command → send SW1 tripping command (repeated 3 times) after 30ms → close SW3."

[0106] After implementing the iterative strategy, the secondary calculation fault isolation time achievement rate was shortened to 45ms, achieving 90% (meeting the target); the load recovery rate remained at 98%. Ultimately, a closed loop of "control-feedback-optimization" was formed, ensuring that the control effect met expectations.

[0107] Existing feeder terminal technology solutions typically employ a single data acquisition method, such as collecting only electrical quantity data (e.g., voltage, current), or simply combining it with a small amount of environmental data. Their processing often relies on fixed physical models or empirical rules, resulting in limited data fusion and analysis capabilities. In terms of control, they are mostly open-loop controls, lacking real-time dynamic response and self-healing control capabilities to the distribution network's operating status. This invention, through multimodal sensing technology, collects various types of data and combines a hybrid driving approach of physical mechanism models and data-driven models to achieve closed-loop processing from data acquisition to self-healing control. This improves the distribution network's fault diagnosis, self-healing control, and operational optimization capabilities, enabling accurate fault detection and rapid recovery while ensuring system stability and reliability.

[0108] In summary, the automated processing method for feeder terminals according to embodiments of the present invention has the following advantages:

[0109] 1. Multimodal data acquisition: It can acquire more comprehensive power distribution network operation data, including electrical quantity data, equipment status data and environmental data, providing rich information support for accurately assessing the operation status of the power distribution network and solving the problem of single data in existing technologies.

[0110] 2. Advantages of the hybrid driving model: The physical-data hybrid driving model combines the reliability of physical mechanisms with the adaptability of data-driven approaches. Under complex operating conditions, it can more accurately perform fault diagnosis, prediction, and control strategy generation, overcoming the limitations of a single model.

[0111] 3. Closed-loop processing and self-healing control: A complete closed-loop processing flow has been formed, which can dynamically adjust the control strategy according to real-time data, realize rapid fault isolation and recovery, significantly improve the reliability and self-healing capability of the distribution network, and shorten the fault recovery time.

[0112] Device Examples

[0113] Please refer to Figure 5 On the other hand, embodiments of the present invention also provide an automated processing device for feeder terminals, comprising:

[0114] The data acquisition module 501 is used to acquire electrical quantity data, equipment status data, and environmental data of the distribution network within a set time period. Specifically, the data acquisition module 501 is used to collect multimodal data during the operation of the distribution network, including electrical quantity data (such as voltage, current, power, frequency, etc.), equipment status data (such as temperature, vibration, insulation resistance, etc.), and environmental data (such as temperature, humidity, wind speed, rainfall, etc.). Specifically, it acquires electrical quantity data through voltage transformers and current transformers; acquires equipment status data using temperature sensors, vibration sensors, and insulation monitoring devices; and acquires environmental data using environmental sensors. The acquired data is output in a standardized format and transmitted to the edge computing module 502 through a standardized interface.

[0115] Edge computing module 502 is used to extract features from the electrical quantity data, equipment status data, and environmental data respectively, obtaining a first feature corresponding to the electrical quantity data, a second feature corresponding to the equipment status data, and a third feature corresponding to the environmental data. Edge computing module 502 performs preliminary processing on the collected raw data, including data cleaning and normalization, to reduce data transmission volume. Edge computing module 502 receives data transmitted from data acquisition module 501 and performs local real-time processing and analysis. On the one hand, it preprocesses the raw data, including filtering, noise reduction, and data normalization, to improve data quality; on the other hand, based on a preset physical mechanism model and preliminary data-driven algorithms, it performs feature extraction and preliminary fault detection on the data. For example, it calculates the theoretical operating parameters of the equipment through the physical mechanism model and compares them with the actual collected data to preliminarily determine whether there are any anomalies; it uses machine learning algorithms to learn the features of the data and identify potential fault modes. The processed intermediate data and feature information are transmitted to data fusion module 503.

[0116] The data fusion module 503 is used to fuse the first feature, the second feature, and the third feature to obtain a fused feature vector. The data fusion module 503 performs deep fusion of multi-source data from the edge computing module 502. It employs multi-dimensional data fusion algorithms, such as weighted fusion and / or neural network fusion, combining the temporal, spatial, and feature dimensions of the data to establish a comprehensive power distribution network operation status model. By fusing data from different sources and of different types, it mines the potential correlations and patterns between the data, achieving a more accurate description and evaluation of the power distribution network operation status. The fused data is stored in a local database and transmitted to the automation control module 504 and the multi-mode communication module 506.

[0117] The automation control module 504 is used to generate a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid drive model, so that the execution equipment of the distribution network can achieve automated processing based on the control strategy. The automation control module 504 realizes automatic control of feeder equipment, such as the closing and opening operations of switches, based on the comprehensive information transmitted from the data fusion module 503 and the preset control strategy. According to the distribution network operating status information provided by the data fusion module 503, and combined with the physical-data hybrid drive model, a corresponding control strategy is generated. The physical-data hybrid drive model combines the physical mechanism model of the distribution network (such as circuit model, equipment operation model, etc.) and the data-driven model (such as neural network, fuzzy logic, etc.). Under normal operating conditions, the physical mechanism model is used as the basis for optimized control combined with the data-driven model; under fault or abnormal conditions, the data-driven model is used to quickly identify the fault type and location, and a reasonable self-healing control strategy is formulated in combination with the physical mechanism model, such as fault isolation, load transfer, and power switching. Control commands are transmitted to the execution equipment through a standardized interface, and the control results are fed back to the data fusion module 503, forming a closed-loop control. The hybrid driving model is constructed based on a physical mechanism model and a data-driven model.

[0118] Optionally, the step of generating a control strategy using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector includes:

[0119] The first fault type prediction result is determined based on the electrical quantity data and the equipment status data using the physical mechanism model.

[0120] The data-driven model is used to determine the prediction result of the second fault type based on the fused feature vector;

[0121] Using a weighted voting mechanism, the final prediction result of the fault type is determined based on the prediction results of the first fault type and the prediction results of the second fault type;

[0122] Using the physical mechanism model and the data-driven model, a control strategy is generated based on the final prediction result of the fault type.

[0123] Optionally, the step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes:

[0124] If the final prediction result of the fault type indicates that no fault has occurred in the distribution network, a basic operation strategy for the distribution network is generated based on the physical mechanism model.

[0125] Based on the data-driven model, load forecasting of the distribution network is performed, and load forecasting data of the distribution network is output.

[0126] The basic operation strategy is dynamically adjusted based on the load forecast data to generate a dynamic adjustment strategy; the dynamic adjustment strategy serves as the control strategy.

[0127] Optionally, the step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes:

[0128] When the final prediction result of the fault type indicates that a fault has occurred in the distribution network, the data-driven model is used to determine the fault location prediction result based on the fused feature vector;

[0129] Using the physical mechanism model, a self-healing control strategy is generated based on the fault location prediction results; the self-healing control strategy serves as the control strategy.

[0130] Optionally, after generating a control strategy using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector, so that the execution equipment of the distribution network can achieve automated processing based on the control strategy, the method further includes:

[0131] Determine the fault isolation time achievement rate and / or load recovery rate of the distribution network after implementing the control strategy; the fault isolation time achievement rate is directly proportional to the actual fault isolation time of the distribution network after implementing the control strategy; the load recovery rate is directly proportional to the number of users whose power supply has been restored.

[0132] If the fault isolation time achievement rate and / or load recovery rate do not meet the set conditions, the process jumps to the step of generating a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid drive model, so that the execution equipment of the distribution network can realize automated processing based on the control strategy.

[0133] Optionally, the step of fusing the first feature, the second feature, and the third feature to obtain a fused feature vector includes:

[0134] The first feature, the second feature, and the third feature are fused using a weighted fusion method and / or a neural network fusion method to obtain a fused feature vector.

[0135] Please refer to Figure 6 The feeder terminal automation processing device also includes a security encryption module 505 and a multi-mode communication module 506.

[0136] The security encryption module 505 is used to protect the data transmission and storage of the entire device. During data acquisition, transmission, processing, and storage, encryption algorithms (such as AES encryption, RSA encryption, etc.) are used to encrypt the data to prevent it from being illegally stolen or tampered with. Simultaneously, authentication and access control are implemented for communication between modules to ensure the security and reliability of the device.

[0137] The multi-mode communication module 506 supports various communication methods, such as Ethernet, wireless communication (4G, 5G, WiFi, Bluetooth, etc.), and power line carrier communication, enabling communication between the feeder terminal and the upstream master station and other intelligent devices. It automatically selects the appropriate communication method based on different communication scenarios and requirements, ensuring real-time data transmission and interaction.

[0138] The feeder terminal automation processing device proposed in this invention consists of a data acquisition module 501, an edge computing module 502, a data fusion module 503, an automation control module 504, a security encryption module 505, and a multi-mode communication module 506, forming an integrated architecture of "sensing-processing-decision-execution-security-communication". This device collects various types of data through multi-modal sensing technology and combines a hybrid driving method of physical mechanism model and data-driven model to achieve closed-loop processing from data acquisition to self-healing control, improving the fault diagnosis, self-healing control, and operation optimization capabilities of the distribution network, while ensuring the safety and reliability of the system.

[0139] The automated processing device for feeder terminals has the following advantages:

[0140] 1. This invention provides a multimodal sensing architecture, which constructs a data acquisition system that includes electrical quantity data, equipment status data, environmental data, and other multimodal data, to achieve comprehensive perception of the operating status of the power distribution network. The feeder terminal system architecture based on multimodal sensing includes a data acquisition module 501, an edge computing module 502, a data fusion module 503, an automation control module 504, a security encryption module 505, a multimodal communication module 506, and their interconnection through standardized interfaces.

[0141] 2. This invention proposes a physical-data hybrid driving model, which combines the physical mechanism model of the power distribution network with the data-driven model, giving full play to the advantages of both and improving the accuracy and adaptability of fault diagnosis, prediction and control. The construction method and application of the physical-data hybrid driving model creatively combine the interpretability of physical mechanisms with the adaptability of data driving, achieving complementary advantages of "rapid location (data driving) + reliable execution (physical verification)" in fault handling.

[0142] 3. Through standardized interfaces and collaborative work between modules, a closed-loop processing flow is formed from data acquisition, processing, fusion to control execution and feedback, realizing automated self-healing control of the feeder terminal. The closed-loop processing flow from data acquisition to self-healing control includes the data interaction between modules and the generation and feedback mechanism of control commands.

[0143] The feeder terminal automated processing device includes a processor and a memory. The aforementioned data acquisition module 501, edge computing module 502, data fusion module 503, automated control module 504, security encryption module 505, and multimode communication module 506 are all stored in the memory as program units. The processor executes the aforementioned program units stored in the memory to realize the corresponding functions.

[0144] A processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured.

[0145] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0146] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other through the communication bus 740. The processor 710 can call logical instructions in the memory 730 to execute a feeder terminal automation processing method, which includes: acquiring electrical quantity data, equipment status data, and environmental data of the distribution network within a set time period; extracting features from the electrical quantity data, equipment status data, and environmental data respectively to obtain a first feature corresponding to the electrical quantity data, a second feature corresponding to the equipment status data, and a third feature corresponding to the environmental data; fusing the first feature, the second feature, and the third feature to obtain a fused feature vector; and generating a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid driving model, so that the execution equipment of the distribution network can achieve automated processing based on the control strategy; wherein the hybrid driving model is constructed based on a physical mechanism model and a data-driven model.

[0147] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0148] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a machine-readable storage medium. When the computer program is executed by a processor, the computer can execute a feeder terminal automation processing method. The method includes: acquiring electrical quantity data, equipment status data, and environmental data of the distribution network within a set time period; extracting features from the electrical quantity data, equipment status data, and environmental data respectively to obtain a first feature corresponding to the electrical quantity data, a second feature corresponding to the equipment status data, and a third feature corresponding to the environmental data; fusing the first feature, the second feature, and the third feature to obtain a fused feature vector; and generating a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid driving model, so that the execution equipment of the distribution network can achieve automated processing based on the control strategy; wherein, the hybrid driving model is constructed based on a physical mechanism model and a data-driven model.

[0149] In another aspect, the present invention also provides a machine-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a feeder terminal automation processing method. The method includes: acquiring electrical quantity data, equipment status data, and environmental data of a distribution network within a set time period; extracting features from the electrical quantity data, equipment status data, and environmental data respectively to obtain a first feature corresponding to the electrical quantity data, a second feature corresponding to the equipment status data, and a third feature corresponding to the environmental data; fusing the first feature, the second feature, and the third feature to obtain a fused feature vector; and generating a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid driving model, so that the execution equipment of the distribution network can achieve automated processing based on the control strategy; wherein the hybrid driving model is constructed based on a physical mechanism model and a data-driven model.

[0150] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0151] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An automated processing method for feeder terminals, characterized in that, include: Acquire electrical quantity data, equipment status data, and environmental data of the power distribution network within a set time period; Feature extraction is performed on the electrical quantity data, equipment status data, and environmental data respectively to obtain a first feature corresponding to the electrical quantity data, a second feature corresponding to the equipment status data, and a third feature corresponding to the environmental data; The first feature, the second feature, and the third feature are fused to obtain a fused feature vector. A control strategy is generated using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector, so that the execution equipment in the distribution network can achieve automated processing based on the control strategy; The hybrid driving model is constructed based on a physical mechanism model and a data-driven model. The process of generating a control strategy using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector includes: The first fault type prediction result is determined based on the electrical quantity data and the equipment status data using the physical mechanism model. The data-driven model is used to determine the prediction result of the second fault type based on the fused feature vector; Using a weighted voting mechanism, the final prediction result of the fault type is determined based on the prediction results of the first fault type and the prediction results of the second fault type; Using the physical mechanism model and the data-driven model, a control strategy is generated based on the final prediction result of the fault type.

2. The automated processing method for feeder terminals according to claim 1, characterized in that, The step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes: If the final prediction result of the fault type indicates that no fault has occurred in the distribution network, a basic operation strategy for the distribution network is generated based on the physical mechanism model. Based on the data-driven model, load forecasting of the distribution network is performed, and load forecasting data of the distribution network is output. The basic operation strategy is dynamically adjusted based on the load forecast data to generate a dynamic adjustment strategy; the dynamic adjustment strategy serves as the control strategy.

3. The automated processing method for feeder terminals according to claim 1, characterized in that, The step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes: When the final prediction result of the fault type indicates that a fault has occurred in the distribution network, the data-driven model is used to determine the fault location prediction result based on the fused feature vector; Using the physical mechanism model, a self-healing control strategy is generated based on the fault location prediction results; the self-healing control strategy serves as the control strategy.

4. The automated processing method for feeder terminals according to claim 1, characterized in that, After generating a control strategy using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector, so that the execution equipment in the distribution network can achieve automated processing based on the control strategy, the method further includes: Determine the fault isolation time achievement rate and / or load recovery rate of the distribution network after implementing the control strategy; the fault isolation time achievement rate is directly proportional to the actual fault isolation time of the distribution network after implementing the control strategy; the load recovery rate is directly proportional to the number of users whose power supply has been restored. If the fault isolation time achievement rate and / or load recovery rate do not meet the set conditions, the process jumps to the step of generating a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid drive model, so that the execution equipment of the distribution network can realize automated processing based on the control strategy.

5. The automated processing method for feeder terminals according to claim 1, characterized in that, The step of fusing the first feature, the second feature, and the third feature to obtain a fused feature vector includes: The first feature, the second feature, and the third feature are fused using a weighted fusion method and / or a neural network fusion method to obtain a fused feature vector.

6. An automated processing device for feeder terminals, characterized in that, include: The data acquisition module is used to acquire electrical quantity data, equipment status data, and environmental data of the power distribution network within a set time period; The edge computing module is used to extract features from the electrical quantity data, equipment status data, and environmental data respectively, to obtain a first feature corresponding to the electrical quantity data, a second feature corresponding to the equipment status data, and a third feature corresponding to the environmental data. The data fusion module is used to fuse the first feature, the second feature, and the third feature to obtain a fused feature vector; An automation control module is used to generate a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid drive model, so that the execution equipment of the distribution network can realize automated processing based on the control strategy; The hybrid driving model is constructed based on a physical mechanism model and a data-driven model. The process of generating a control strategy using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector includes: The first fault type prediction result is determined based on the electrical quantity data and the equipment status data using the physical mechanism model. The data-driven model is used to determine the prediction result of the second fault type based on the fused feature vector; Using a weighted voting mechanism, the final prediction result of the fault type is determined based on the prediction results of the first fault type and the prediction results of the second fault type; Using the physical mechanism model and the data-driven model, a control strategy is generated based on the final prediction result of the fault type.

7. The automated processing device for feeder terminals according to claim 6, characterized in that, The step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes: If the final prediction result of the fault type indicates that no fault has occurred in the distribution network, a basic operation strategy for the distribution network is generated based on the physical mechanism model. Based on the data-driven model, load forecasting of the distribution network is performed, and load forecasting data of the distribution network is output. The basic operation strategy is dynamically adjusted based on the load forecast data to generate a dynamic adjustment strategy; the dynamic adjustment strategy serves as the control strategy.

8. The automated processing device for feeder terminals according to claim 6, characterized in that, The step of generating a control strategy based on the final prediction result of the fault type using the physical mechanism model and the data-driven model includes: When the final prediction result of the fault type indicates that a fault has occurred in the distribution network, the data-driven model is used to determine the fault location prediction result based on the fused feature vector; Using the physical mechanism model, a self-healing control strategy is generated based on the fault location prediction results; the self-healing control strategy serves as the control strategy.

9. The automated processing device for feeder terminals according to claim 6, characterized in that, After generating a control strategy using a hybrid drive model based on the electrical quantity data, the equipment status data, and the fused feature vector, so that the execution equipment in the distribution network can achieve automated processing based on the control strategy, the method further includes: Determine the fault isolation time achievement rate and / or load recovery rate of the distribution network after implementing the control strategy; the fault isolation time achievement rate is directly proportional to the actual fault isolation time of the distribution network after implementing the control strategy; the load recovery rate is directly proportional to the number of users whose power supply has been restored. If the fault isolation time achievement rate and / or load recovery rate do not meet the set conditions, the process jumps to the step of generating a control strategy based on the electrical quantity data, the equipment status data, and the fused feature vector using a hybrid drive model, so that the execution equipment of the distribution network can realize automated processing based on the control strategy.

10. The automated processing device for feeder terminals according to claim 6, characterized in that, The step of fusing the first feature, the second feature, and the third feature to obtain a fused feature vector includes: The first feature, the second feature, and the third feature are fused using a weighted fusion method and / or a neural network fusion method to obtain a fused feature vector.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the feeder terminal automated processing method according to any one of claims 1 to 5.

12. A machine-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the feeder terminal automated processing method according to any one of claims 1 to 5.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the feeder terminal automated processing method according to any one of claims 1 to 5.

Images