Intelligent power distribution network feed automation control system and method based on multi-source information fusion

Abstract

The invention relates to the field of power system automation, and discloses an intelligent power distribution network feed automation control system based on multi-source information fusion, which comprises a multi-source sensor module used for collecting current data in a power distribution network in real time, and an information fusion module used for receiving the preprocessed current data and sending the preprocessed current data to the power distribution network. The multi-source fusion module is used for carrying out multi-source fusion processing on the current data by adopting a Bayesian reasoning algorithm, the fault positioning module is used for receiving a fusion result, and the optimal control module is used for receiving a grounding fault occurrence position and adopting an optimal control theory and a dynamic planning algorithm; the invention further discloses an intelligent power distribution network feed automatic control method based on multi-source information fusion. The method comprises the following steps of an acquisition stage, a fusion stage, a positioning and decision-making stage and an execution and self-healing stage. According to the invention, by constructing the Bayesian fusion and dynamic programming control mechanism, the purposes of more accurate ground fault positioning, more efficient strategy removal and more adaptive response process are achieved.

Description

Technical Field

[0001] The present invention relates to the field of power system automation, and in particular to an intelligent distribution network feed automation control system and method based on multi-source information fusion. Background Art

[0002] In actual distribution network operations, ground faults are one of the most common, complex, and hidden fault types. If misjudged and unresolved, they can not only cause voltage fluctuations and degrade power quality, but can also lead to widespread power outages, impacting residents' lives and the stable operation of critical infrastructure. Especially in densely populated urban areas and in scenarios involving new energy integration, the accuracy of ground fault response and the intelligence of the strategy directly determine the reliability and resilience of the distribution system.

[0003] Some existing distribution automation systems already possess certain fault detection and handling capabilities. For example, they employ fixed threshold and zero-sequence current change methods for fault identification, enabling rapid alerts in most obvious ground fault situations. Some systems also employ strategic response models based on expert rule bases to minimize outage areas and implement load transfer operations. These methods offer low implementation costs, mature algorithms, and the ability to cover most common operating conditions, making them suitable for distribution areas with simple structures and stable loads.

[0004] However, existing technologies also have a number of technical bottlenecks that are difficult to avoid. First, fault identification methods generally rely on static rules, which can easily lead to misjudgment in the face of atypical signals, especially when the noise background is strong or the load changes suddenly, the identification effect is significantly reduced. Secondly, the data fusion link mostly uses linear superposition or local maximum method, lacks systematic time series update capabilities, and the fusion results are prone to deviate from the actual state. Furthermore, most control strategies are based on preset processes and cannot be flexibly generated according to the fault location and operating status, which can easily lead to excessive removal or misoperation. Finally, traditional control methods are generally single-shot triggering, do not have a feedback mechanism, lack dynamic optimization and correction capabilities, and expose strong limitations in complex power distribution scenarios. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides an intelligent distribution network feed automation control system and method based on multi-source information fusion, which solves the problems in the existing technology that ground fault identification relies on static rules, data fusion lacks dynamic updates, control strategies are rigid and there is no feedback optimization mechanism.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent distribution network feed automation control system based on multi-source information fusion, comprising: A multi-source sensor module is used to collect current data in the distribution network in real time, the current data including voltage data, current data, load data and environmental data, and normalize and denoise the current data to obtain pre-processed current data; an information fusion module, configured to receive the preprocessed current data, perform multi-source fusion processing on the current data using a Bayesian inference algorithm to generate a fusion result, and perform dynamic Bayesian updating of the posterior probability of the ground fault location based on the fusion result; a fault location module, configured to receive the fusion result and determine the occurrence location of the ground fault based on the posterior probability; The optimal control module is used to receive the ground fault location, and calculate and generate a corresponding optimal fault line removal strategy using optimal control theory and dynamic programming algorithm.

[0007] Preferably, the multi-source sensor module includes: Voltage sensors are used to collect voltage data at each node of the distribution network; Current sensors are used to collect current data at each node of the distribution network; Load sensor, used to collect operating status data of distribution network load; Environmental sensors are used to collect environmental parameter data, including temperature, humidity and meteorological data.

[0008] Preferably, the multi-source sensor module uses the following normalization formula when performing normalization and denoising processing: ; in, For the original collected data, and are the minimum and maximum values ​​of the data in the historical records, is the normalized data value.

[0009] Preferably, the information fusion module includes: a data receiving unit, configured to receive the pre-processed current data; A priori knowledge base, used to store historical operating characteristic data of each feeder or node; A Bayesian fusion unit, which constructs a Bayesian reasoning model based on the prior knowledge base and the preprocessed current data, and updates the posterior probability of the ground fault; The information entropy calculation unit is used to calculate the information entropy change during the Bayesian fusion process to evaluate the confidence and reliability of data fusion.

[0010] Preferably, the multi-source fusion processing of the current data using the Bayesian inference algorithm includes: Construct a Bayesian network based on the observation data of each sensor to calculate different fault locations The posterior probability ,in It is multi-source observation data; The formula for calculating the posterior probability is: ; in, is the likelihood function, is the prior probability, is the normalization coefficient, is the likelihood function, which means that under the assumption events Under the conditions that occur.

[0011] Preferably, the fault location module includes: a posterior probability calculation unit, configured to receive the posterior probability calculated by the Bayesian fusion unit; A fault judgment unit is used to judge whether the posterior probability meets the preset fault judgment criteria; The fault location determination unit is used to select the location corresponding to the maximum a posteriori probability that meets the conditions as the final ground fault location.

[0012] Preferably, the posterior probability includes: Updated single-node posterior probability after multi-source data fusion; The multi-node joint posterior probability uses the following update model: ; in, is the maximum posterior probability value, represents the overall set of fault locations, is the observation data of each node, is the index variable, the conditional posterior probability, Indicates that the node Observational data Under known conditions, the fault occurs at the location The probability of is the window length parameter.

[0013] Preferably, the optimal control module includes: State modeling unit, used to establish a dynamic mathematical model of the power grid operation state; An objective function construction unit, used to construct an optimization objective function based on power loss, equipment operation cost and load transfer cost; The optimization algorithm unit is used to solve the optimization objective function using a dynamic programming algorithm to generate an optimal fault line removal strategy.

[0014] Preferably, the method of calculating the location of the ground fault using optimal control theory and dynamic programming algorithm includes: Define the state space, including the current load, voltage and operating status of each distribution node; Define the action space, including the set of removable lines and their corresponding operations; Construction phase cost function: ; in, represents the load imbalance cost, represents the voltage offset penalty, Indicates the recovery time, , , is the weight coefficient, is the total cost function, which means that in state Take control action The comprehensive cost value brought about is used as an evaluation indicator for control optimization or strategy selection.

[0015] The present invention also provides a smart distribution network feed automation control method based on multi-source information fusion, comprising the following steps: Acquisition phase: Collect multi-source data from the distribution network, including voltage, current, load and environmental information, and transmit it to the data processing center via wireless communication; Fusion stage: After normalizing and denoising the received data, the Bayesian inference algorithm is used to fuse multi-source information and dynamically update the posterior probability distribution of the ground fault location; Positioning and decision-making stage: Based on the fusion results, the fault location is determined, and the optimal fault removal plan is generated by combining the optimal control strategy and dynamic programming algorithm; Execution and self-healing phase: Control decisions are transmitted to each control terminal through an information synchronization mechanism, driving the distributed self-healing module to complete fault isolation and power supply reconstruction.

[0016] The present invention provides an intelligent distribution network feed automation control system and method based on multi-source information fusion. It has the following beneficial effects: 1. This invention implements a fusion analysis of multi-source preprocessed current data by constructing an information fusion module based on a Bayesian inference algorithm, thereby dynamically updating the posterior probability of the fault location. Compared to traditional fault diagnosis systems based on threshold or fixed template recognition, this solution dynamically adjusts judgments based on data real-timeness and prior knowledge, addressing the shortcomings of existing models, such as susceptibility to noise and lack of contextual adaptability.

[0017] 2. This invention incorporates a dynamic programming algorithm to construct an optimal fault removal strategy, minimizing operational costs while ensuring power continuity. Compared to existing methods that use fixed rules or manual experience to formulate isolation strategies, this method achieves cascaded strategy optimization without human intervention, avoiding issues such as response delays and policy redundancy, making it more suitable for the automated deployment of self-healing distribution network systems.

[0018] 3. This invention utilizes a rolling horizon control mechanism and a heuristic pruning path search strategy to effectively improve computational efficiency in large-scale distribution networks while ensuring timely control response. Traditional algorithms are prone to path explosion problems under multiple node state combinations. This solution significantly reduces system load and improves deployment flexibility through control horizon pruning and priority queue construction.

[0019] 4. This invention establishes a closed-loop information feedback mechanism for the entire positioning-control link, linking the confidence level of fault location results with control strategy judgment. This is more flexible than the existing decoupled "static judgment-static execution" architecture. Especially in fuzzy scenarios with non-centralized probability distributions, it can achieve policy delay and fault-tolerant execution, effectively avoiding mis-cutting issues and improving the system's anti-interference capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of the system architecture of the present invention; Figure 2 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] Please see the attached Figure 1 The embodiment of the present invention provides an intelligent distribution network feed automation control system based on multi-source information fusion, including: A multi-source sensor module is used to collect current data in the distribution network in real time, the current data including voltage data, current data, load data and environmental data, and normalize and denoise the current data to obtain pre-processed current data; In the intelligent distribution network feeder automation control system based on multi-source information fusion, the multi-source sensor module serves as the perception foundation for system operation. The quality and integrity of the data it collects directly determine the accuracy and robustness of subsequent information fusion, fault location, and control strategy generation. To achieve comprehensive understanding of the distribution network state and provide high-confidence input data for Bayesian information fusion, this embodiment incorporates refined modeling and functional integration of this module's design and preprocessing mechanisms.

[0023] In this embodiment, multi-source sensor modules are configured to be deployed on important feeders and key nodes in the distribution network. This module can collect multiple types of data related to electrical operating status in real time. The types of data collected include, but are not limited to, voltage data, current data, load operation data, and environmental perception data. In this system, the above data is collectively referred to as "current data," where "current" does not refer to a narrow current value, but rather a broad set of dynamic parameters that describe the distribution status.

[0024] Typically, voltage data is collected through voltage transformers, current data is obtained through current transformers, load data is captured by smart energy meters installed at the terminal loads, and environmental data is collected by temperature and humidity sensors and meteorological data access modules installed around the distribution station. Alternatively, these sensors can use wireless communication modules such as ZigBee and NB-IoT to establish communication with the data processing center, enabling rapid data transmission and synchronization.

[0025] In one possible implementation, the sampling period for sensor data is set to 500ms to 1s, depending on the importance of the line and the frequency of power load fluctuations. To improve the uniformity and robustness of the data fusion phase, the system performs unified normalization and denoising on all raw sensor data before uploading it to the information fusion module.

[0026] Specifically, in the preprocessing stage, the system performs normalization on each data channel separately. The normalization formula is: ; in, For the original collected data, and are the minimum and maximum values ​​of the data in the historical records, is the normalized data value.

[0027] This normalization operation can not only effectively avoid the inconsistency problem of numerical scales of different types of data, but also provide consistent input for prior probability and likelihood estimation in the Bayesian fusion algorithm.

[0028] In addition to normalization, the system also integrates a denoising algorithm module to filter out sudden changes in values ​​caused by electromagnetic interference, sampling anomalies, etc. In this embodiment, the system uses a combination of median filtering and sliding average filtering to perform data denoising.

[0029] Median filtering is used to remove spike mutation data. Its basic algorithm is: Take the middle value as the current sampling value; The sliding average filter is used to smooth the fluctuation noise, and its calculation formula is: ; in, For time point The smoothed value of is the value of each sampling point in the window, is the window length parameter, is the index variable, the conditional posterior probability.

[0030] In some embodiments, the denoising algorithm also introduces an outlier detection mechanism. When the difference between multiple consecutive sampling points is greater than a set threshold (for example, the current change rate exceeds 20% of the rated value), the system marks the data segment as invalid and fills it with the most recent valid value to prevent erroneous data from participating in subsequent decision-making.

[0031] Generally, to ensure the credibility and timeliness of preprocessed data, the system prioritizes normalization and denoising during each data update cycle before passing the results to the information fusion module. While this process increases the data processing burden, it effectively improves the accuracy of posterior probability calculations and reduces fault location errors caused by data distortion.

[0032] As an extension, in actual deployments, the sensor module supports remote parameter configuration. For example, background commands can be used to adjust the normalized history window length, filtering algorithm type, and threshold parameters to adapt to seasonal, environmental, or load characteristics. This enables adaptive adjustment of the data perception layer.

[0033] Furthermore, to accommodate distributed deployment requirements, the multi-source sensor module is designed with edge computing capabilities. This means that some normalization and preliminary filtering can be performed locally, reducing the burden of data transmission and improving system response speed. This feature is particularly suitable for power distribution areas with remote locations or limited communication conditions.

[0034] an information fusion module, configured to receive the preprocessed current data, perform multi-source fusion processing on the current data using a Bayesian inference algorithm to generate a fusion result, and perform dynamic Bayesian updating of the posterior probability of the ground fault location based on the fusion result; The information fusion module integrates preprocessed current data from multiple sensor modules and uses a Bayesian inference algorithm to generate a posterior probability of the fault location, thereby accurately locating the ground fault. As a key component of the system, the information fusion module's output provides crucial decision support for subsequent fault location modules. Specifically, in this embodiment, the information fusion module uses a Bayesian inference algorithm to perform multi-source fusion processing on the collected data and dynamically updates the posterior probability of the ground fault location based on the fusion results.

[0035] Typically, preprocessed current data is transmitted to the information fusion module via the data receiving unit. This module first formats the received current data and then fuses the multi-source data using a Bayesian inference algorithm. Through this process, the system can derive the probability distribution of the fault location based on real-time and historical data, providing high-confidence input information for the fault location module.

[0036] Specifically, in this embodiment, the information fusion module includes the following key parts: Data receiving unit: This unit is responsible for receiving pre-processed current data from multi-source sensor modules. This data contains information such as voltage, current, load, and environment, and provides input for subsequent fusion processing.

[0037] Prior knowledge base: This module stores historical operational characteristic data, including information such as the long-term operating behavior of each feeder or node in the distribution network and load variation characteristics. Historical data provides prior probabilistic support during the fusion process, enabling Bayesian reasoning to make reasonable inferences based on past experience.

[0038] Bayesian Fusion Unit: Based on prior data obtained from the prior knowledge base and real-time current data, this unit uses a Bayesian inference algorithm to update the posterior probability and calculate the probability of the fault location. In one possible implementation, this algorithm constructs a Bayesian network, using observation data from different sensors as conditional variables to perform joint probability calculations.

[0039] The core of the Bayesian inference algorithm is the update process of the posterior probability, which is as follows: ; in, is the likelihood function, is the prior probability, is the normalization coefficient, is the likelihood function, which means that under the assumption events Under the conditions that occur.

[0040] Using this formula, the system dynamically calculates the posterior probability of the fault location based on real-time data and historical prior knowledge. The system updates this posterior probability each time new current data arrives, forming a fault location model that becomes increasingly accurate over time.

[0041] Alternatively, the Bayesian fusion unit can be combined with the information entropy calculation unit to assess the credibility of the fusion results. Specifically, by calculating the change in information entropy, the system can quantify the uncertainty in the data fusion process. Low entropy values ​​indicate a relatively certain fusion result, and the system can directly output the final fault location result. High entropy values, however, require more data to improve the accuracy of the fusion result.

[0042] For example, when the entropy value of the posterior probability exceeds a preset threshold, the system will require re-collection of data or adjustment of sensor parameters to reduce the impact of uncertainty on subsequent fault location. The specific entropy calculation formula is: ; in, represents the entropy of the posterior probability distribution, Expressing assumptions The posterior probability of is the logarithmic operation, is the window length parameter, is the likelihood function, which means that under the assumption events Under the conditions that occur, is the index variable, the conditional posterior probability.

[0043] In some embodiments, entropy calculations can be performed simultaneously with probability calculations, with dynamically adjusted thresholds ensuring stable system operation even with incomplete or poor data quality. To improve processing efficiency, entropy calculations and posterior probability updates can be performed in parallel to ensure efficient system response even with high data throughput.

[0044] In one possible implementation, the results of the Bayesian inference calculations are dynamically transmitted to the fault location module, which determines the specific location of the fault based on these posterior probabilities. By using Bayesian inference algorithms and information entropy evaluation, the system can effectively integrate data from different sensors and quickly and accurately locate the fault when a ground fault occurs in the distribution network, providing a reliable basis for subsequent fault handling and recovery strategies.

[0045] Furthermore, to improve the system's real-time performance and adaptability, the information fusion module in this embodiment also has certain adaptive capabilities. Specifically, the system can automatically adjust the prior probabilities and likelihood functions in the Bayesian inference model based on the operating status of different distribution networks to account for factors such as environmental changes and load fluctuations.

[0046] a fault location module, configured to receive the fusion result and determine the occurrence location of the ground fault based on the posterior probability; After the aforementioned information fusion module completes the fusion of multi-source current data and dynamically updates the posterior probability distribution of each possible ground fault location using a Bayesian inference algorithm, the fault location module, acting as the system's key execution unit, further determines the actual location of the ground fault based on the fusion results. To ensure the logical and coherent flow of data between modules, the fault location module's input is directly derived from the fusion results and posterior probability distribution output by the information fusion module. The core function of this module is to transform uncertainty in the probability space into clear location criteria and, based on this, output the specific node or feeder location where the fault occurred.

[0047] The fault location module receives the fusion results from the information fusion module, which include the posterior probability distribution of each candidate fault location. By performing quantitative analysis and discriminant operations on this probability distribution, the system can determine the most likely fault location based on the maximum a posteriori criterion.

[0048] Specifically, the posterior probability obtained in the Bayesian inference stage uses the following update model: ; in, is the maximum posterior probability value, represents the overall set of fault locations, is the observation data of each node, is the index variable, the conditional posterior probability, Indicates that the node Observational data Under known conditions, the fault occurs at the location The probability of is the window length parameter.

[0049] The fault location module will perform the following discriminant operations on the above probability set: ; in, It represents the fault location decision result output by the fault location module. is the set of all possible candidate fault locations, Select the function for the maximum value, which is used to find the maximum probability value , Expressing assumptions The posterior probability of .

[0050] In general, when the maximum posterior probability value When the value is higher than the preset judgment threshold (such as 0.85), the system considers that the current fault judgment has sufficient confidence and can The final fault location is output. This threshold can be configured according to the actual situation of different distribution networks to balance the sensitivity and false alarm rate of positioning.

[0051] As an option, to further improve the accuracy and robustness of fault determination, this embodiment introduces a confidence interval verification mechanism, which verifies the judgment result based on the concentration of the posterior probability distribution.

[0052] Specifically, by calculating the relative difference between the maximum a posteriori probability and the second largest a posteriori probability: ; in, Indicates the position corresponding to the second largest posterior probability, The larger it is, the more exclusive and certain the judgment is. represents the load imbalance cost, is the posterior probability of the most likely fault location.

[0053] when When the value is lower than a certain threshold (such as 0.1), it indicates that the probability of multiple fault locations is similar. The system will delay the output of the final result and obtain more criteria by requesting additional sampling or expanding the fusion window to avoid incorrect positioning due to insufficient data.

[0054] In one possible implementation, the fault location module also integrates spatial constraint factors and topological weight information. Even if the posterior probabilities of two nodes are close, the system also considers their relative distance and connectivity characteristics in the distribution network topology.

[0055] In some embodiments, to avoid misjudgment, the fault location module will also integrate auxiliary features such as current change rate, voltage drop rate, and load mutation amplitude before finally outputting the location result, and incorporate them into the decision criteria in the form of a rule set. For example: If a node experiences a sudden drop in current amplitude within a short period of time (the rate of drop is greater than 40% of the rated value), accompanied by a sudden drop in load, and has the highest posterior probability, it is directly determined to be a fault point; If there are two or more nodes with similar posterior probability values, but one of the nodes meets the complex conditions of voltage drop exceeding the threshold, abnormal temperature and humidity changes, etc., then this node will be selected first.

[0056] In the case of inconsistent data or insufficient quality, the system also has a feedback re-sampling mechanism, allowing the fusion module to readjust the sampling frequency or expand the perception range, and re-perform Bayesian reasoning and positioning analysis to ensure the stability of the results.

[0057] In addition, the output of the fault location module can be directly connected to the human-computer interaction interface or remote monitoring platform to achieve visual display and remote alarm linkage.

[0058] An optimal control module is configured to receive the ground fault location, calculate and generate a corresponding optimal fault line removal strategy using optimal control theory and dynamic programming algorithm; After the fault location module identifies the ground fault location, the system automatically enters the fault response phase. The optimal control module, a key control decision-making unit, uses the identified ground fault location information and the current operating status of the distribution network to systematically evaluate possible control strategies using optimal control theory and dynamic programming algorithms, ultimately generating the optimal fault line removal solution.

[0059] To achieve seamless connection from positioning results to control actions, the optimal control module and the fault location module directly transmit fault location information through a data interface, and synchronously access system status parameters such as topology structure, node load, and power distribution to ensure that the generated control strategy is highly targeted and operationally feasible.

[0060] In this embodiment, the optimal control module receives the ground fault location output by the fault location module and constructs a control objective function based on the system's current operating status. This objective function comprehensively considers factors such as line power flow, voltage stability, load importance, and feeder redundancy to achieve a multi-objective balance between system safety, power supply continuity, and control cost in the disconnection strategy.

[0061] In general, the construction phase cost function can be expressed as follows: ; in, represents the load imbalance cost, represents the voltage offset penalty, Indicates the recovery time, , , is the weight coefficient, is the total cost function, which means that in state Take control action The comprehensive cost value brought about is used as an evaluation indicator for control optimization or strategy selection.

[0062] In the dynamic programming modeling process, the state variables represent the current topological configuration and node voltage status of the distribution network, and the control variables are the switch action sequences. The recursive relationship is defined as follows: ; in, Indicates that the status The minimum cumulative control cost under The control actions that can be executed in this state (such as opening a branch switch) Status The set of all possible actions. From the state Execute an action The immediate control cost generated, The next state after executing the action.

[0063] To ensure that the generated strategy has engineering feasibility and satisfies time constraints, this embodiment introduces a control action execution time function: ; in, For a specified action The total execution delay time, Indicates the basic response delay time, It is the action delay time of a specific switch device under current conditions, determined by combining the actual switch type and remote control signal quality.

[0064] As an option, the optimal control module also introduces a "load priority factor" in the candidate shedding strategy , used to preserve power supply to core users when the load cannot be fully restored. Its weight is set as follows: ; in, For the The weight coefficient of a load node indicates the importance of the node in the entire network and is used to optimize the scheduling strategy or fault recovery priority allocation. Indicates the The power demand of each load node, is the total load, The business importance evaluation coefficient of the node (for example, a higher value should be set for hospitals and data centers).

[0065] In one possible implementation, in order to balance rapid response and the global optimal control path, this embodiment introduces a "rolling time domain control mechanism," which uses the current moment as the starting point to optimize and predict the control path within a fixed time domain window in the future, and executes the most advanced control action as planned. The state is then rolled over and the calculation is repeated to form a dynamic feedback closed-loop control.

[0066] To cope with the complexity of high-order systems and large-scale cross-section combination problems, the optimal control module adopts a branch and bound technology based on heuristic pruning to eliminate path nodes with obviously suboptimal solutions or violation constraint solutions in the strategy space, thereby greatly improving the efficiency of dynamic programming search.

[0067] In the model input part, the optimal control module also combines the confidence information from the information fusion module (such as the maximum a posteriori probability value, information entropy index, etc.). When the probability is highly dispersed, it can avoid misoperation caused by positioning errors through strategy delay, upsampling or secondary confirmation mechanism.

[0068] The intelligent distribution network feeding automation control method based on multi-source information fusion described below and the intelligent distribution network feeding automation control system based on multi-source information fusion described above can refer to each other.

[0069] Please see the attached Figure 2 The present invention also provides a smart distribution network feed automation control method based on multi-source information fusion, comprising the following steps: S1. Collection phase: Collect multi-source data from the distribution network, including voltage, current, load and environmental information, and transmit it to the data processing center via wireless communication; S2, fusion stage: After normalizing and denoising the received data, the Bayesian inference algorithm is used to fuse multi-source information and dynamically update the posterior probability distribution of the ground fault location; S3, Positioning and Decision-making Stage: Based on the fusion results, the fault location is determined, and the optimal fault removal plan is generated by combining the optimal control strategy and dynamic programming algorithm; S4, execution and self-healing stage: The control decision is transmitted to each control terminal through the information synchronization mechanism, driving the distributed self-healing module to complete fault isolation and power supply reconstruction.

[0070] The method of this embodiment can be used to execute the above system embodiment. Its principles and technical effects are similar and will not be described in detail here.

[0071] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. Intelligent distribution network feed automation control system based on multi-source information fusion, characterized by: include: A multi-source sensor module is used to collect current data in the distribution network in real time, the current data including voltage data, current data, load data and environmental data, and normalize and denoise the current data to obtain pre-processed current data; an information fusion module, configured to receive the preprocessed current data, perform multi-source fusion processing on the current data using a Bayesian inference algorithm to generate a fusion result, and perform dynamic Bayesian updating of the posterior probability of the ground fault location based on the fusion result; a fault location module, configured to receive the fusion result and determine the occurrence location of the ground fault based on the posterior probability; The optimal control module is used to receive the ground fault location, and calculate and generate a corresponding optimal fault line removal strategy using optimal control theory and dynamic programming algorithm.

2. The intelligent distribution network feed automation control system based on multi-source information fusion according to claim 1 is characterized in that: The multi-source sensor module comprises: Voltage sensors are used to collect voltage data at each node of the distribution network; Current sensors are used to collect current data at each node of the distribution network; Load sensor, used to collect operating status data of distribution network load; Environmental sensors are used to collect environmental parameter data, including temperature, humidity and meteorological data.

3. The intelligent distribution network feed automation control system based on multi-source information fusion according to claim 1 is characterized in that: When the multi-source sensor module performs normalization and denoising, the following normalization formula is used: ； in, For the original collected data, and are the minimum and maximum values ​​of the data in the historical records, is the normalized data value.

4. The intelligent distribution network feed automation control system based on multi-source information fusion according to claim 1 is characterized in that: The information fusion module includes: a data receiving unit, configured to receive the pre-processed current data; A priori knowledge base, used to store historical operating characteristic data of each feeder or node; A Bayesian fusion unit, which constructs a Bayesian reasoning model based on the prior knowledge base and the preprocessed current data, and updates the posterior probability of the ground fault; The information entropy calculation unit is used to calculate the information entropy change during the Bayesian fusion process to evaluate the confidence and reliability of data fusion.

5. The intelligent distribution network feed automation control system based on multi-source information fusion according to claim 1 is characterized in that: The multi-source fusion processing of the current data using the Bayesian inference algorithm includes: Construct a Bayesian network based on the observation data of each sensor to calculate different fault locations The posterior probability ,in It is multi-source observation data; The formula for calculating the posterior probability is: ； in, is the likelihood function, is the prior probability, is the normalization coefficient, is the likelihood function, which means that under the assumption events Under the conditions that occur.

6. The intelligent distribution network feed automation control system based on multi-source information fusion according to claim 1 is characterized in that: The fault location module includes: a posterior probability calculation unit, configured to receive the posterior probability calculated by the Bayesian fusion unit; A fault judgment unit is used to judge whether the posterior probability meets the preset fault judgment criteria; The fault location determination unit is used to select the location corresponding to the maximum a posteriori probability that meets the conditions as the final ground fault location.

7. The intelligent distribution network feed automation control system based on multi-source information fusion according to claim 1 is characterized in that: The posterior probability includes: Updated single-node posterior probability after multi-source data fusion; The multi-node joint posterior probability uses the following update model: ; in, is the maximum posterior probability value, represents the overall set of fault locations, is the observation data of each node, is the index variable, the conditional posterior probability, Indicates that the node Observational data Under known conditions, the fault occurs at the location The probability of is the window length parameter.

8. The intelligent distribution network feed automation control system based on multi-source information fusion according to claim 1 is characterized in that: The optimal control module includes: State modeling unit, used to establish a dynamic mathematical model of the power grid operation state; An objective function construction unit, used to construct an optimization objective function based on power loss, equipment operation cost and load transfer cost; The optimization algorithm unit is used to solve the optimization objective function using a dynamic programming algorithm to generate an optimal fault line removal strategy.

9. The intelligent distribution network feed automation control system based on multi-source information fusion according to claim 1, characterized in that: The method of calculating the location of the ground fault by using the optimal control theory and the dynamic programming algorithm includes: Define the state space, including the current load, voltage and operating status of each distribution node; Define the action space, including the set of removable lines and their corresponding operations; Construction phase cost function: ； in, represents the load imbalance cost, represents the voltage offset penalty, Indicates the recovery time, , , is the weight coefficient, is the total cost function, which means that in state Take control action The comprehensive cost value brought about is used as an evaluation indicator for control optimization or strategy selection.

10. A method for controlling the feeding automation of an intelligent distribution network based on multi-source information fusion, according to the intelligent distribution network feeding automation control system based on multi-source information fusion according to any one of claims 1 to 9, characterized in that: The following steps are involved: Acquisition phase: Collect multi-source data from the distribution network, including voltage, current, load and environmental information, and transmit it to the data processing center via wireless communication; Fusion stage: After normalizing and denoising the received data, the Bayesian inference algorithm is used to fuse multi-source information and dynamically update the posterior probability distribution of the ground fault location; Positioning and decision-making stage: Based on the fusion results, the fault location is determined, and the optimal fault removal plan is generated by combining the optimal control strategy and dynamic programming algorithm; Execution and self-healing phase: Control decisions are transmitted to each control terminal through an information synchronization mechanism, driving the distributed self-healing module to complete fault isolation and power supply reconstruction.

Images