Charging pile power network fault monitoring and positioning method based on digital twinning

Abstract

The application discloses a charging pile power network fault monitoring and positioning method based on digital twinning, and relates to the technical field of power system fault diagnosis. The method constructs a charging pile power network topology static model and a charging pile equipment dynamic model to simulate real-time operation state of the equipment; a charging pile fault detection data model is constructed, and fault features and data mapping are defined. A charging pile power network digital twin is established by fusing the above models to realize dynamic virtual mapping of the physical network. Based on the digital twin, a charging pile fault positioning data model is constructed to associate fault signals and topology positions. In the operation stage, the digital twin is used to monitor charging pile operation data in real time, and the fault detection model is used to identify abnormalities; when a fault occurs, the fault positioning data model is called to quickly and accurately locate the fault charging pile in combination with the network topology relationship. The application significantly improves the real-time performance of fault monitoring and the positioning accuracy of the charging pile network through the digital twinning technology of multiple models.

Description

Technical Field

[0001] This invention relates to the field of power system fault diagnosis technology, and in particular to a method for monitoring and locating faults in the power network of charging piles based on digital twins. Background Technology

[0002] Fault monitoring and location technology for charging pile power networks is crucial for ensuring the safety of electric vehicle charging. With the large-scale integration of high-power charging piles into urban power distribution networks, existing methods primarily rely on SCADA systems to collect basic electrical quantities and set fixed threshold alarms, combined with impedance methods or traveling wave methods to determine fault areas. While this technology is barely feasible in simple radial networks, it reveals serious shortcomings when facing complex scenarios with multiple branches of charging piles connected in parallel and dynamic load switching: fault detection real-time performance is limited by minute-level data refresh rates, making it difficult to capture millisecond-level transient anomalies; location accuracy is limited by static topology models, failing to distinguish fault signal crosstalk in parallel branches, resulting in a false positive rate exceeding 15%; and parameter fixing mechanisms cannot adapt to temperature drift and equipment aging, leading to high maintenance costs. Especially at sites where high-power fast charging piles account for more than 30%, traditional methods have resulted in an average of 12 missed fault reports and over one million yuan in losses due to false shutdowns annually.

[0003] Chinese patent CN117521498A, "A Guided Fault Diagnosis and Prediction Method and System for Charging Piles," reveals the specific manifestations of current technological bottlenecks. The solution includes: constructing a digital twin model of the charging pile and importing its operational data in real time; using this model to perform state simulation and predict its operational status; simultaneously, establishing a guided fault diagnosis engine, inputting the predicted status into the engine, and performing fault prediction through built-in analysis algorithms. This invention is based on dividing management boundaries using block grid nodes, with real-time data as the core, integrating digital twins, algorithm training, and diagnostic models. Through engine diagnosis, it provides feedback on fault results and outputs optimized maintenance solutions, achieving intelligent digital management, fault early warning, and efficient response for clustered charging piles, thereby significantly improving risk warning capabilities.

[0004] Based on the shortcomings of existing technologies, this field urgently needs to overcome three core challenges: First, how to integrate static network topology, device dynamic equations, and multi-dimensional fault characteristics to construct a high-fidelity digital twin capable of real-time mapping of the physical system state, thus resolving transient fault loss caused by minute-level response delays; second, how to establish a weighting function to quantify fault propagation intensity, accurately characterize the attenuation law of fault signals in charging pile clusters, and overcome the positioning ambiguity caused by crosstalk of parallel branch signals; third, how to achieve second-level fault location and online self-calibration of model parameters under the limited computing power of embedded devices, compressing the positioning error to within 5 meters to accurately identify specific faulty charging piles, while reducing the misjudgment rate to below 5%. Breakthroughs in these areas are crucial for promoting the intelligent operation and maintenance of electric vehicle charging networks. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] Therefore, this invention provides a method for monitoring and locating faults in the power network of charging piles based on digital twins, which solves the problems of transient fault loss caused by minute-level response delay and ambiguity in location caused by crosstalk of parallel branch signals.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, the present invention provides a method for monitoring and locating faults in the power network of charging piles based on digital twins, comprising the following steps:

[0009] Step S1: Construct a static model of the charging pile power network topology;

[0010] Step S2: Establish a dynamic model of the charging pile equipment Step S3: Construct a charging pile fault detection data model;

[0011] Step S4: Merge and construct a digital twin;

[0012] The static topology model, dynamic device model, and fault detection model are integrated through a unified data interface to establish a dynamic virtual mapping relationship:

[0013] Where G is a directed graph, DT(t) is the state of the digital twin at time t, Φ(·) is the fusion function driven by real-time data, and F k (t) represents the fault feature vector;

[0014] Step S5: Real-time monitoring and fault identification;

[0015] Collect real-time operating data from each node Input fault detection model calculation ||F k (t)||2, if Satisfy ||F k If (t)||2>||Θ||2, where Θ is the threshold vector, then a fault is determined to have occurred;

[0016] Step S6: Fault location;

[0017] Calling the fault location data model L(G,F) k Based on the topological adjacency matrix A and the fault signal propagation path, the fault probability distribution is calculated:

[0018]

[0019] Where d(v) k ,v src) represents the path from node k to the first abnormal node v src The shortest path hop count, where λ is the decay factor; argmax k P f (v k This was identified as the source of the fault.

[0020] As a preferred embodiment of the charging pile power network fault monitoring and location method based on digital twins described in this invention, step S1, which involves constructing a static topology model of the charging pile power network, specifically includes:

[0021] Based on graph theory, a directed graph G = (V, E) is constructed, where the vertex set V represents charging pile nodes and power distribution nodes, and the edge set E represents power connection lines.

[0022] Obtain the electrical wiring diagram of the power distribution network area where the charging pile is located through the power grid monitoring and data acquisition SCADA system or the power distribution management system DMS.

[0023] The physical devices in the wiring diagram are abstracted as nodes: charging piles are abstracted as load nodes, and transformers, switches, and busbars are abstracted as distribution nodes;

[0024] Based on the electrical connection relationship, the charging pile group is divided into a tree topology, and a tree topology structure with the power distribution node as the root node is constructed. The tree topology structure is a radial power distribution network topology, wherein: the root node is the outgoing switch or bus of the upper substation; the intermediate nodes are the power distribution branch box or ring network cabinet; and the leaf nodes are the charging pile equipment.

[0025] For each node v k Add attribute vector Q k = [ID, latitude and longitude, rated power];

[0026] Define the adjacency matrix A = [α] ij ] n×n Where: n is the total number of nodes, and α is the number of nodes when there is a direct power connection between node i and node j. ij =1, when there is no direct connection between node i and node j. ij =0, diagonal element α ii =0.

[0027] As a preferred embodiment of the charging pile power network fault monitoring and location method based on digital twins described in this invention, step S2, which establishes a dynamic model of the charging pile equipment, specifically includes the following operations:

[0028] For each charging pile node v k Establish the state-space equations in the dq rotating coordinate system;

[0029] Phase synchronization of grid voltage is achieved through a phase-locked loop (PLL):

[0030]

[0031] Where: θ(t) is the real-time phase angle, in rad;

[0032] w nom The nominal angular frequency is 314 rad / s for a 50 Hz system;

[0033] (The control objective is to bring the q-axis voltage to zero);

[0034] K p ,K i This refers to the proportional-integral gain coefficient.

[0035] Add a device noise model:

[0036]

[0037] Where: σ k The white noise intensity factor is taken as 0.5-2% of the rated value; h m f is the amplitude of the m-th harmonic, set according to the limits of IEC 61000-3-12 standard; f1 is the fundamental frequency of 50Hz or 60Hz; f s The sampling frequency is ≥10kHz.

[0038] As a preferred embodiment of the charging pile power network fault monitoring and location method based on digital twins described in this invention, wherein: in step S2, a dynamic model of the charging pile equipment is established. The state-space equations are:

[0039]

[0040] Where x k (t) is the state vector, including output voltage, current harmonic distortion rate, u k (t) is the input control vector, ξ k (t) represents equipment noise;

[0041] The term for each charging pile node v k The state-space equations in the dq rotating coordinate system are established as follows:

[0042]

[0043] Among them: I dq (t)=[I d (t),I q (t)] T ,I dq(t) is the current vector of the charging pile in the dq rotating coordinate system, where the d-axis is the active component and the q-axis is the reactive component.

[0044] V dq (t) represents the voltage vector on the grid side in the dq coordinate system;

[0045] R is the equivalent AC side resistance of the charging pile;

[0046] L is the equivalent AC side inductance of the charging pile;

[0047] w is the angular frequency of the power grid synchronization, w = 2π × 50Hz or 60Hz.

[0048] As a preferred embodiment of the charging pile power network fault monitoring and location method based on digital twins described in this invention, step S3, which involves constructing a charging pile fault detection data model, specifically includes the following operations:

[0049] Define a multidimensional fault feature vector F k (t);

[0050] Calculate the dynamic baseline value: Where X∈{I,THD}, I is the current, THD is the total harmonic distortion of the current; T w A 15-minute sliding time window;

[0051] Set the fault thresholds for different levels: θ = [θ1, θ2, θ3, θ4] T =[5.0,3.0,0.2,0.3] T ;

[0052] Establish fault determination rules: when ||F k (t)|| ∞ A warning is triggered when the value is >1.0; when Confirm the fault in time; where ||·|| ∞ Let ||·||1 be the L1 norm, and take the maximum value component. Let ||·||1 be the L1 norm, and take the sum of the absolute values ​​of the components.

[0053] As a preferred embodiment of the charging pile power network fault monitoring and location method based on digital twins described in this invention, the step S3 of constructing the charging pile fault detection data model specifically involves defining a fault feature vector F. k (t)=[ΔI k (t),ΔTHD k (t),Δcosφ k (t),ΔZ k (t)] T , where ΔI k ΔTHD is the rate of change of current. k Δcosφ is the change in harmonic distortion rate.k The power factor offset, ΔZ k (t) represents the impedance offset; the threshold vector Θ = [θ] is set. I ,θ THD ,θ cosφ ] T When ||F k An exception flag is triggered when ||2>||Θ||2;

[0054] The fault feature vector F k The calculation method for (t) is as follows:

[0055] Current mutation rate:

[0056] Change in harmonic distortion rate:

[0057] Power factor offset: Δcosφ k (t)=|cosφ k (t)-cosφ ref |

[0058] Impedance offset:

[0059] in:

[0060] I k (t): The effective value of the current at node k at time t;

[0061] T: Length of the sliding time window, ranging from 1 to 5 minutes;

[0062] I rated,k : Rated current of node k;

[0063] THD k (t): Total harmonic distortion of the current at node k at time t;

[0064] cosφ k (t): The power factor of node k at time t;

[0065] cosφ ref Rated power factor, typically taken as 0.95 to 1.0.

[0066] As a preferred embodiment of the charging pile power network fault monitoring and location method based on digital twins described in this invention, the specific implementation of the fusion function Φ(·) in step S4 is as follows:

[0067] Based on the OPCUA (IEC62541) communication protocol, a real-time data channel is established between the physical charging pile equipment and the digital twin node; real-time data is then transmitted. Mapped to topology model nodes;

[0068] Use the topology G defined in step S1 as the spatial skeleton;

[0069] The dynamic model state defined in step S2 Bind to the corresponding node v k Uploaded in real time;

[0070] The fault feature vector F defined in step S3 k (t) is a node v k Additional real-time attributes;

[0071] When any node's ||F k When (t)||2>||Θ||2, that is, when the fault detection model outputs an abnormality, the dynamic model of this node is automatically triggered to switch to the preset fault simulation mode, which includes short circuit, open circuit, and insulation degradation models.

[0072] As a preferred embodiment of the charging pile power network fault monitoring and location method based on digital twins described in this invention, wherein: the real-time monitoring in step S5 adopts a multi-time-scale adaptive sampling mechanism: the dynamic model is updated every 5 seconds under normal operating conditions; the default sampling period T normal =5 seconds;

[0073] When any node satisfies ||F k When (t)||2>0.5||Θ||2, switch to 1-second high-frequency sampling;

[0074] When any node satisfies ||F k When (t)||2>0.7||Θ||2, the global sampling period switches to T. alert = 2 seconds;

[0075] When any node satisfies ||F k When (t)||2>||Θ||2, the sampling period of the node and its direct upstream neighbor nodes switches to T. fault = 0.5 seconds.

[0076] As a preferred embodiment of the charging pile power network fault monitoring and location method based on digital twins described in this invention, the calculation process of the fault location data model in step S6 is as follows:

[0077] Identify the first node v that triggers the fault flag. src , satisfying ||F k (t)||2>||Θ||2 and the timestamp is the earliest;

[0078] From the first abnormal node v src Begin a breadth-first search (BFS) along the upstream direction (towards the root node) in the topological graph G;

[0079] For each edge e ij (connect node v) i and v j Calculate the fault propagation weights

[0080]

[0081] Z ij For line e ij The impedance magnitude; ΔI i ,ΔI j For node v i ,v j Current mutation rate (from F) i ,F j ); α is the empirical attenuation coefficient (with a value of 1 to 3);

[0082] Calculate v for each node k Failure probability score:

[0083]

[0084] Where: d(v) k ,v src ): Node v k to v src The number of hops for the shortest path in the topology graph;

[0085] λ: Path attenuation factor, with a value ranging from 0.5 to 2.0;

[0086] Path(v src ,v k ): From v src to v k The set of edges on the shortest path;

[0087] ∏ (i,j)∈Path w ij : The product of the weights of all edges on the path;

[0088] Update the failure probability distribution:

[0089] in For v k The set of neighboring nodes;

[0090] Will satisfy node v k It was determined to be the most likely source of the fault.

[0091] As a preferred embodiment of the charging pile power network fault monitoring and location method based on digital twins described in this invention, it further includes an online model calibration step:

[0092] per T calibrate =Execute once every 24 hours;

[0093] During steady-state operation (all ||F) k (t)||2<0.3||Θ||2), collect actual measurement data of each charging pile.

[0094] y k =[I k V k THD k ,cosφ k ] T

[0095] Obtain the simulation output of the corresponding node of the digital twin.

[0096] Gradient descent correction is applied to the dynamic model parameters R and L:

[0097]

[0098] Where η is the learning rate, ranging from 0.01 to 0.1; and N is the number of nodes participating in the calibration.

[0099] In step S4, the digital twin performs model calibration every 24 hours.

[0100] Compare the actual measured value y k Compared with simulation values

[0101] Adjusting dynamic model parameters:

[0102] Where η is the learning rate.

[0103] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the method for monitoring and locating faults in a charging pile power network based on digital twins as described in the first aspect of the present invention.

[0104] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for monitoring and locating faults in a power network for charging piles based on digital twins as described in the first aspect of the present invention.

[0105] The beneficial effects of this invention are as follows: It can significantly improve the management efficiency and risk control level of clustered charging piles. Its core lies in constructing a physical-virtual twin model of the charging piles and injecting operational data in real time, enabling the model to continuously and accurately simulate and predict the actual operating status of the charging piles. This dynamic simulation based on real data provides strong technical support for in-depth understanding of the current and future health status of the equipment, allowing managers to go beyond simple real-time monitoring and anticipate potential performance degradation or abnormal trends, thereby transforming passive response into proactive management.

[0106] This solution further integrates a guided fault diagnosis engine, using the state predictions output by the twin model as key inputs. The engine utilizes advanced analytical algorithms to deeply mine and intelligently analyze this prediction data, efficiently and accurately identifying potential fault modes and generating specific fault prediction results. This diagnostic approach is not performed in isolation but within the rich context provided by the twin model, ensuring the relevance and reliability of the diagnostic conclusions, effectively reducing the possibility of false alarms and missed alarms, and providing a solid data foundation for subsequent maintenance decisions.

[0107] Most importantly, this invention creatively employs a block-based grid of nodes as the boundary delineation basis for management and analysis. This structured organization, combined with data as the core driving element, tightly integrates digital twin technology, algorithm training and optimization, and intelligent diagnostic models into a cohesive whole. The engine's diagnostic results not only provide timely feedback on specific fault information but also intelligently generate and recommend the optimal maintenance plan based on a comprehensive analysis of fault type, location, impact range, and historical maintenance data. This significantly optimizes the allocation of operational resources, improves maintenance efficiency and focus, and avoids blind investment and waste of resources.

[0108] Ultimately, this invention enables intelligent and digital full lifecycle management of large-scale, clustered charging piles. It not only provides high-precision fault warnings but also forms a closed-loop response mechanism covering early warning, diagnosis, decision-making, and response. By moving the risk identification point forward and providing optimized handling solutions, this technical solution significantly enhances the risk warning and proactive defense capabilities of the entire charging infrastructure network, effectively reducing downtime losses and safety risks caused by equipment failures, ensuring the reliability and safety of charging services, and improving user experience and operational efficiency. Attached Figure Description

[0109] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0110] Figure 1 This is a flowchart of the charging pile power network fault monitoring and location method based on digital twin in Example 1.

[0111] Figure 2 This is a flowchart of the charging pile power network fault monitoring and location method based on digital twin in Example 2. Detailed Implementation

[0112] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0113] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0114] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0115] Example 1

[0116] Reference Figure 1 This is one embodiment of the present invention, which provides a method for monitoring and locating faults in the power network of charging piles based on digital twins, including the following steps:

[0117] Step S1: Construct a static model of the charging pile power network topology;

[0118] Step S2: Establish a dynamic model of the charging pile equipment Step S3: Construct a charging pile fault detection data model;

[0119] Step S4: Merge and construct a digital twin;

[0120] The static topology model, dynamic device model, and fault detection model are integrated through a unified data interface to establish a dynamic virtual mapping relationship:

[0121] Where G is a directed graph, DT(t) is the state of the digital twin at time t, Φ(·) is the fusion function driven by real-time data, and F k (t) represents the fault feature vector;

[0122] Step S5: Real-time monitoring and fault identification;

[0123] Collect real-time operating data from each node Input fault detection model calculation ||F k (t)||2, if Satisfy ||F k If (t)||2>||Θ||2, where Θ is the threshold vector, then a fault is determined to have occurred;

[0124] Step S6: Fault location;

[0125] Calling the fault location data model L(G,F) k Based on the topological adjacency matrix A and the fault signal propagation path, the fault probability distribution is calculated:

[0126]

[0127] Where d(v) k ,v src ) represents the path from node k to the first abnormal node v src The shortest path hop count, where λ is the decay factor; argmax k P f (v k This was identified as the source of the fault.

[0128] In this embodiment, step S1, which involves constructing a static topology model of the charging pile power network, specifically includes:

[0129] Based on graph theory, a directed graph G = (V, E) is constructed, where the vertex set V represents charging pile nodes and power distribution nodes, and the edge set E represents power connection lines.

[0130] Obtain the electrical wiring diagram of the power distribution network area where the charging pile is located through the power grid monitoring and data acquisition SCADA system or the power distribution management system DMS.

[0131] The physical devices in the wiring diagram are abstracted as nodes: charging piles are abstracted as load nodes, and transformers, switches, and busbars are abstracted as distribution nodes;

[0132] Based on the electrical connection relationship, the charging pile group is divided into a tree topology, and a tree topology structure with the power distribution node as the root node is constructed. The tree topology structure is a radial power distribution network topology, wherein: the root node is the outgoing switch or bus of the upper substation; the intermediate nodes are the power distribution branch box or ring network cabinet; and the leaf nodes are the charging pile equipment.

[0133] For each node v k Add attribute vector Q k = [ID, latitude and longitude, rated power];

[0134] Define the adjacency matrix A = [α] ij ] n×n Where: n is the total number of nodes, and α is the number of nodes when there is a direct power connection between node i and node j. ij =1, when there is no direct connection between node i and node j. ij =0, diagonal element α ii =0.

[0135] In this embodiment, step S2, which establishes a dynamic model of the charging pile equipment, specifically includes the following operations:

[0136] For each charging pile node v k Establish the state-space equations in the dq rotating coordinate system;

[0137] Phase synchronization of grid voltage is achieved through a phase-locked loop (PLL):

[0138]

[0139] Where: θ(t) is the real-time phase angle, in rad;

[0140] w nom The nominal angular frequency is 314 rad / s for a 50 Hz system;

[0141] (The control objective is to bring the q-axis voltage to zero);

[0142] K p ,K i This refers to the proportional-integral gain coefficient.

[0143] Add a device noise model:

[0144]

[0145] Where: σ k The white noise intensity factor is taken as 0.5-2% of the rated value; h m f is the amplitude of the m-th harmonic, set according to the limits of IEC 61000-3-12 standard; f1 is the fundamental frequency of 50Hz or 60Hz; f sThe sampling frequency is ≥10kHz.

[0146] In this embodiment, step S2 involves establishing a dynamic model of the charging pile equipment. The state-space equations are:

[0147]

[0148] Where x k (t) is the state vector, including output voltage, current harmonic distortion rate, u k (t) is the input control vector, ξ k (t) represents equipment noise;

[0149] For each charging pile node v k The state-space equations in the dq rotating coordinate system are established as follows:

[0150]

[0151] Among them: I dq (t)=[I d (t),I q (t)] T ,I dq (t) is the current vector of the charging pile in the dq rotating coordinate system, where the d-axis is the active component and the q-axis is the reactive component.

[0152] V dq (t) represents the voltage vector on the grid side in the dq coordinate system;

[0153] R is the equivalent AC side resistance of the charging pile;

[0154] L is the equivalent AC side inductance of the charging pile;

[0155] w is the angular frequency of the power grid synchronization, w = 2π × 50Hz or 60Hz.

[0156] In this embodiment, step S3, which involves constructing a charging pile fault detection data model, specifically includes the following operations:

[0157] Define a multidimensional fault feature vector F k (t);

[0158] Calculate the dynamic baseline value: Where X∈{I,THD}, I is the current, THD is the total harmonic distortion of the current; T w A 15-minute sliding time window;

[0159] Set the fault thresholds for different levels: θ = [θ1, θ2, θ3, θ4] T =[5.0,3.0,0.2,0.3] T ;

[0160] Establish fault determination rules: when ||F k (t)|| ∞ A warning is triggered when the value is >1.0; when Confirm the fault in time; where ||·|| ∞ Let ||·||1 be the L1 norm, and take the maximum value component. Let ||·||1 be the L1 norm, and take the sum of the absolute values ​​of the components.

[0161] In this embodiment, the specific steps of constructing the charging pile fault detection data model in step S3 are as follows:

[0162] Define the fault feature vector F k (t)=[ΔI k (t),ΔTHD k (t),Δcosφ k (t),ΔZ k (t)] T , where ΔI k ΔTHD is the rate of change of current. k Δcosφ is the change in harmonic distortion rate. k The power factor offset, ΔZ k (t) represents the impedance offset; the threshold vector Θ = [θ] is set. I ,θ THD ,θ cosφ ] T When ||F k An exception flag is triggered when ||2>||Θ||2;

[0163] Fault feature vector F k The calculation method for (t) is as follows:

[0164] Current mutation rate:

[0165] Change in harmonic distortion rate:

[0166] Power factor offset: Δcosφ k (t)=|cosφ k (t)-cosφ ref |

[0167] Impedance offset:

[0168] in:

[0169] I k (t): The effective value of the current at node k at time t;

[0170] T: Length of the sliding time window, ranging from 1 to 5 minutes;

[0171] I rated,k : Rated current of node k;

[0172] THD k (t): Total harmonic distortion of the current at node k at time t;

[0173] cosφ k (t): The power factor of node k at time t;

[0174] cosφ ref Rated power factor, typically taken as 0.95 to 1.0.

[0175] In this embodiment, the specific implementation of the fusion function Φ(·) in step S4 is as follows:

[0176] Based on the OPCUA (IEC62541) communication protocol, a real-time data channel is established between the physical charging pile equipment and the digital twin node; real-time data is then transmitted. Mapped to topology model nodes;

[0177] Use the topology G defined in step S1 as the spatial skeleton;

[0178] The dynamic model state defined in step S2 Bind to the corresponding node v k Uploaded in real time;

[0179] The fault feature vector F defined in step S3 k (t) is a node v k Additional real-time attributes;

[0180] When any node's ||F k When (t)||2>||Θ||2, i.e., when the fault detection model outputs an abnormality, the dynamic model of that node is automatically triggered to switch to a preset fault simulation mode, including short circuit, open circuit, and insulation degradation models. The fusion function Φ(·) can be a weighted average model, a Kalman filter model, or a trained neural network (such as LSTM) used to fuse heterogeneous data from different models. The fault detection data model can be based on an expert rule base, for example: if the voltage drop exceeds 10% and lasts for more than 0.1 seconds, it is determined to be a sag; it can also be one or more machine learning classifiers (such as SVM, decision tree), whose input is a fault feature vector and output is the fault type. In this embodiment, the real-time monitoring in step S5 adopts a multi-timescale adaptive sampling mechanism:

[0181] The dynamic model is updated every 5 seconds under normal operating conditions; the default sampling period T normal =5 seconds;

[0182] When any node satisfies ||F k When (t)||2>0.5||Θ||2, switch to 1-second high-frequency sampling;

[0183] When any node satisfies ||F k When (t)||2>0.7||Θ||2, the global sampling period switches to T. alert = 2 seconds;

[0184] When any node satisfies ||F k When (t)||2>||Θ||2, the sampling period of the node and its direct upstream neighbor nodes switches to T. fault = 0.5 seconds.

[0185] In this embodiment, the calculation process of the fault location data model in step S6 is as follows:

[0186] Identify the first node v that triggers the fault flag. src , satisfying ||F k (t)||2>||Θ||2 and the timestamp is the earliest;

[0187] From the first abnormal node v src Begin a breadth-first search (BFS) along the upstream direction (towards the root node) in the topological graph G;

[0188] For each edge e ij (connect node v) i and v j Calculate the fault propagation weights

[0189]

[0190] Z ij For line e ij The impedance magnitude; ΔI i ,ΔI j For node v i ,v j Current mutation rate (from F) i ,F j ); α is the empirical attenuation coefficient (with a value of 1 to 3);

[0191] Calculate v for each node k Failure probability score:

[0192]

[0193] Where: d(v) k ,v src ): Node v k to v src The number of hops for the shortest path in the topology graph;

[0194] λ: Path attenuation factor, with a value ranging from 0.5 to 2.0;

[0195] Path(v src ,v k ): From v src to v k The set of edges on the shortest path;

[0196] ∏ (i,j)∈Path w ij : The product of the weights of all edges on the path;

[0197] Update the failure probability distribution:

[0198] in For v k The set of neighboring nodes;

[0199] Will satisfy node v k It was determined to be the most likely source of the fault.

[0200] It also includes an online model calibration step:

[0201] per T calibrate =Execute once every 24 hours;

[0202] During steady-state operation (all ||F) k (t)|| 2 <0.3||Θ||2), collect actual measurement data for each charging pile.

[0203] y k =[I k V k THD k ,cosφ k ] T

[0204] Obtain the simulation output of the corresponding node of the digital twin.

[0205] Gradient descent correction is applied to the dynamic model parameters R and L:

[0206]

[0207]

[0208] Where η is the learning rate, ranging from 0.01 to 0.1; and N is the number of nodes participating in the calibration.

[0209] In this embodiment, the digital twin performs model calibration every 24 hours in step S4:

[0210] Compare the actual measured value y k Compared with simulation values

[0211] Adjusting dynamic model parameters:

[0212] Where η is the learning rate.

[0213] Example 2

[0214] Reference Figure 2 This is the second embodiment of the present invention, which provides the workflow of a charging pile power network fault monitoring and location method based on digital twins:

[0215] Phase 1: Data Fusion and Dynamic Mapping of Twin Models

[0216] The workflow begins with the real-time data acquisition layer of the charging pile's power network. Sensors distributed throughout the charging pile cluster continuously capture key operating parameters, including data on current, voltage, temperature, insulation resistance, and switch status. This raw data is preprocessed (e.g., filtering, outlier removal) by edge computing nodes and then transmitted to the cloud data center via IoT communication protocols (e.g., MQTT). The system injects the cleaned data into the digital twin model of the charging pile in real time. This model is pre-built based on the physical equipment structure, electrical topology, and material properties, and includes a three-dimensional virtual mapping of the charging module, cooling system, and relay components. The twin model uses real-time data to drive a dynamic simulation engine, updating the virtual charging pile's operating status every second and generating high-precision predictive indicators such as power fluctuation trends, component thermal stress distribution, and insulation performance degradation curves.

[0217] Phase Two: Collaborative Diagnosis and Fault Pre-location

[0218] The state prediction results output by the twin model are automatically pushed to the guided fault diagnosis engine. The engine first performs multi-dimensional correlation analysis on the prediction data: for example, correlating the current harmonic distortion rate with abnormal cooling fan speed, or matching the DC bus voltage drop with the capacitor aging model. The diagnostic engine's built-in rule base (based on historical fault cases) and machine learning models (such as LSTM time series prediction and random forest classifier) ​​operate in parallel: the rule base quickly screens for explicit fault patterns (such as over-temperature alarms), while the AI ​​model uncovers hidden risks (such as predicting the breakdown probability based on small changes in capacitor ESR). When a potential fault signal is detected, the engine initiates a collaborative localization process. Based on the physical coordinates of the charging pile in the block-based grid nodes (such as area A - node 3 - pile position 7), combined with the grid topology relationship (such as the load rate of the upstream transformer and the power interaction of adjacent piles), it initially delineates the fault impact range and generates fault type labels with probability weights (such as "DC / DC module fault in the pile (confidence 82%)" or "voltage sag on the grid side (confidence 67%)").

[0219] Phase 3: Spatial Positioning and Closed-Loop Processing

[0220] After fault confirmation, the system enters the precise location phase. The system retrieves full-domain data for the grid node containing the faulty charging pile, including the synchronous current phase, zero-sequence current difference, and grounding resistance monitoring values ​​for all charging piles at that node. By comparing the spatial distribution differences of fault feature vectors (such as specific harmonic spectra and leakage current amplitude) within the node, and combining this with fault propagation paths simulated by the digital twin (e.g., simulation of short-circuit current diffusion in a virtual power grid), the fault source is ultimately pinpointed to a specific equipment component (e.g., the IGBT drive circuit of charging pile at node 5 in area B). The location results automatically trigger closed-loop management: the maintenance work order system instantly receives the fault location coordinates, a predicted list of faulty components, and maintenance plans (e.g., "replace the cooling fan (emergency)" or "calibrate the voltage sampling circuit within 72 hours"), while simultaneously sending power adjustment commands to surrounding charging piles to isolate risks. Data throughout the entire process (from initial alarm to maintenance verification) is fed back to the twin model and diagnostic engine, enabling continuous optimization of algorithm parameters.

[0221] Phase 4: Dynamic Optimization and Knowledge Iteration

[0222] After each fault handling is completed, the system automatically initiates a knowledge base update process. The root causes of the fault, repair measures, and component photos recorded by maintenance personnel on-site are structured and entered into the database, and compared and verified with the simulation results of the twin model. The diagnostic engine adjusts algorithm weights based on false alarms / missed alarms (e.g., improving the detection sensitivity for temperature sensor failure scenarios), while the twin model corrects the component lifespan prediction model based on actual component wear data (e.g., the capacitance decay curve of disassembled capacitors). This incremental knowledge is horizontally synchronized to similar equipment nodes through a block-based grid node system, enabling intelligent accumulation of cross-regional fault experience and continuously improving the accuracy and efficiency of network-wide early warnings.

[0223] This workflow achieves intelligent management of the entire charging pile power network fault, from early warning to precise handling, through a four-stage progressive mechanism of "data-driven simulation, intelligent diagnosis, spatial positioning, and closed-loop optimization".

[0224] This embodiment also provides a computer device applicable to the charging pile power network fault monitoring and location method based on digital twins, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the charging pile power network fault monitoring and location method based on digital twins as proposed in the above embodiment.

[0225] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0226] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the method for monitoring and locating charging pile power network faults based on digital twins as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0227] In summary, this invention significantly improves the management efficiency and risk control level of clustered charging piles. Its core lies in constructing a physical-virtual twin model of the charging piles and injecting operational data in real time, enabling the model to continuously and accurately simulate and predict the actual operating status of the charging piles. This dynamic simulation based on real data provides strong technical support for in-depth understanding of the current and future health status of the equipment, allowing managers to move beyond simple real-time monitoring and anticipate potential performance degradation or abnormal trends, thereby transforming passive response into proactive management.

[0228] This solution further integrates a guided fault diagnosis engine, using the state predictions output by the twin model as key inputs. The engine utilizes advanced analytical algorithms to deeply mine and intelligently analyze this prediction data, efficiently and accurately identifying potential fault modes and generating specific fault prediction results. This diagnostic approach is not performed in isolation but within the rich context provided by the twin model, ensuring the relevance and reliability of the diagnostic conclusions, effectively reducing the possibility of false alarms and missed alarms, and providing a solid data foundation for subsequent maintenance decisions.

[0229] Most importantly, this invention creatively employs a block-based grid of nodes as the boundary delineation basis for management and analysis. This structured organization, combined with data as the core driving element, tightly integrates digital twin technology, algorithm training and optimization, and intelligent diagnostic models into a cohesive whole. The engine's diagnostic results not only provide timely feedback on specific fault information but also intelligently generate and recommend the optimal maintenance plan based on a comprehensive analysis of fault type, location, impact range, and historical maintenance data. This significantly optimizes the allocation of operational resources, improves maintenance efficiency and focus, and avoids blind investment and waste of resources.

[0230] Ultimately, this invention enables intelligent and digital full lifecycle management of large-scale, clustered charging piles. It not only provides high-precision fault warnings but also forms a closed-loop response mechanism covering early warning, diagnosis, decision-making, and response. By moving the risk identification point forward and providing optimized handling solutions, this technical solution significantly enhances the risk warning and proactive defense capabilities of the entire charging infrastructure network, effectively reducing downtime losses and safety risks caused by equipment failures, ensuring the reliability and safety of charging services, and improving user experience and operational efficiency.

[0231] Example 3, referring to Table 1, is the third embodiment of the present invention. To further verify the technical solution of the present invention, experimental simulation data of the charging pile power network fault monitoring and location method based on digital twin is provided. The following are the experimental simulation data and analysis of Example 3. The charging pile power network fault monitoring and location method based on digital twin is verified based on the key technical indicators listed in Table 1. The experiment simulated a cluster network containing 120 charging piles (divided into 6 block grid nodes). The running data was collected from the historical database of real charging stations and overlaid with fault scenarios.

[0232] Table 1. Results of core experimental simulation data

[0233]

[0234]

[0235] The “block grid node” mentioned in this invention is a logical management unit division based on the power distribution network topology and physical spatial location. For example, all charging piles under the same transformer can be divided into a block grid node.

[0236] Data analysis and indicator validation:

[0237] 1. Improved fault detection and early warning capabilities

[0238] The experiment introduced 35 types of faults (including power grid fluctuations, component aging, short circuits, etc.), and the digital twin model captured abnormal trends in advance through real-time simulation. For example:

[0239] Capacitor pre-failure case: The twin model, based on the bus voltage ripple growth rate (simulation prediction value > measured value 15%) and ESR temperature rise model, triggered an early warning 3.2 hours before the actual explosion of the charging pile capacitor at node 2. Traditional methods only trigger an alarm after the pile is shut down.

[0240] Voltage dip in the grid: When the load on adjacent transformers suddenly increases, the model combines the topology relationship to predict the voltage drop (error < 5%), and the diagnostic engine adjusts the charging power of the 6 affected nodes in a coordinated manner to avoid the downtime of the entire charging pile group.

[0241] 2. Positioning accuracy and collaborative verification

[0242] Spatial location verification: In the "multi-pile grounding fault" scenario, the system compared the zero-sequence current phase difference (simulated twin value vs. measured value) of the four charging piles in node 3 of area A, and identified the fault source as a loose PE line connection at pile 7 (confirmed by actual maintenance). Traditional methods, due to the lack of correlation with node data, misjudged it as grid-side interference.

[0243] Component-level location: For the "IGBT open circuit fault", the diagnostic engine integrates the simulation waveform (drive signal distortion rate 92%) and the current harmonic distribution within the grid node to accurately locate the power module of node 5 in area B. After repair and replacement, the fault was resolved.

[0244] 3. False alarm suppression and response efficiency

[0245] False alarm control: The traditional threshold method will falsely report "transient current of charging gun plugging and unplugging" as short circuit (12 times), while this method simulates the current curve of the plugging and unplugging process (duration < 0.1 seconds) through a twin model and combines it with a rule base to eliminate false alarms.

[0246] Real-time performance guarantee: The diagnostic engine adopts a hierarchical processing mechanism. The rule base prioritizes high-confidence alarms (response time < 2 seconds), while the AI ​​model asynchronously analyzes latent faults (maximum 15 seconds). The average response time of 8.7 seconds meets the requirements of real-time operation and maintenance. In this invention, latent faults refer to early fault states where equipment parameters have deteriorated and deviated from the normal model, but have not yet triggered traditional fixed threshold alarms, such as a slight increase in relay contact resistance.

[0247] 4. Latent Fault Identification and Closed-Loop Optimization

[0248] Case of hidden fault: The system identified 9 relay contacts with "micro-melting and sticking" (contact resistance rising to 1.8Ω). Traditional detection did not trigger an alarm as the resistance was below the threshold (<2Ω). The twin model predicted failure within 3 weeks based on historical contact resistance degradation curves, allowing for early repairs and avoiding fire risks.

[0249] Knowledge iteration effect: In the later stage of the experiment (the 10th batch of fault injection), the false alarm rate dropped to 1.9% because the engine continuously learns maintenance feedback data (such as marking "fan stops but heat dissipation is normal" as a non-fault mode).

[0250] Conclusion: The simulation data of Example 3 fully verify the superiority of the technical indicators listed in Table 1:

[0251] Predictability: Early warning time has been significantly improved, and the rate of identifying hidden faults has doubled;

[0252] Precision: Component-level positioning reduces maintenance costs by 40% (avoiding the need to replace the entire pile);

[0253] Reliability: False alarm rate reduced by 75%, response speed improved by 5 times;

[0254] Self-evolution: The closed-loop optimization mechanism enables the false alarm rate to continuously decrease over the operation and maintenance cycle.

[0255] This method significantly addresses three major pain points: delayed fault detection, ambiguous location, and high false alarm rate in cluster charging piles, providing core digital support for power network risk prevention and control.

[0256] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for monitoring and locating faults in the power network of charging piles based on digital twins, characterized in that, Includes the following steps: Step S1: Construct a static model of the charging pile power network topology; Step S2: Establish a dynamic model of the charging pile equipment. ; Step S3: Construct a charging pile fault detection data model; Step S4: Integrate and construct a digital twin; Integrate the topology static model, equipment dynamic model, and fault detection model through a unified data interface to establish a dynamic virtual mapping relationship: ， in It is a directed graph. Let t be the state of the digital twin. This is a fusion function driven by real-time data. For fault feature vectors, the fusion function The specific implementation method is as follows: Based on the OPCUAIEC62541 communication protocol, a real-time data channel is established between the physical charging pile equipment and the digital twin node; real-time data is then transmitted. Mapped to topology model nodes; Use the topology G defined in step S1 as the spatial skeleton; The dynamic model state defined in step S2 Bind to the corresponding node Uploaded in real time; The fault feature vector defined in step S3 As a node Additional real-time attributes; When any node When the fault detection model outputs an abnormality, the dynamic model of that node is automatically switched to a preset fault simulation mode, including short circuit, open circuit, and insulation degradation models. Step S5: Real-time monitoring and fault identification; collect real-time operating data from each node. Input fault detection model calculation ,like satisfy ,in If the threshold vector is used, a fault is determined to have occurred; Step S6: Fault location; Call the fault location data model. Based on the topological adjacency matrix A and the fault signal propagation path, the fault probability distribution is calculated: in From node k to the first abnormal node Shortest path hop count, As the attenuation factor; Once identified as a fault source, the calculation process for the fault location data model is as follows: Identify the first node that triggers the fault flag. ,satisfy And the earliest timestamp; From the first abnormal node Begin a breadth-first search (BFS) along the upstream direction on the topological graph G; For each edge (Connecting nodes) and Calculate the fault propagation weights in ; , For nodes , The rate of change of current; This is an empirical attenuation coefficient, with a value ranging from 1 to 3; Calculate each node Failure probability score: in: :node arrive The number of hops for the shortest path in the topology graph; Path attenuation factor, with a value ranging from 0.5 to 2.0; :from arrive The set of edges on the shortest path; : The product of the weights of all edges on the path; Update the failure probability distribution: in for The set of neighboring nodes; Will satisfy nodes It was determined to be the most likely source of the fault.

2. The method for monitoring and locating charging pile power network faults based on digital twins as described in claim 1, characterized in that... Step S1, which involves constructing a static topology model of the charging pile power network, specifically includes: Based on graph theory, a directed graph G=(V,E) is constructed, where the vertex set V represents charging pile nodes and power distribution nodes, and the edge set E represents power connection lines. Obtain the electrical wiring diagram of the power distribution network area where the charging pile is located through the power grid monitoring and data acquisition SCADA system or the power distribution management system DMS. The physical devices in the wiring diagram are abstracted as nodes: charging piles are abstracted as load nodes, and transformers, switches, and busbars are abstracted as distribution nodes; Based on the electrical connections, the charging pile group is divided into a tree topology, constructing a tree topology structure with distribution nodes as root nodes; the tree topology structure is a radial distribution network topology, wherein: the root node is the outgoing switch or busbar of the upstream substation; the intermediate nodes are distribution branch boxes or ring main units; and the leaf nodes are the charging pile devices; For each node Add attribute vector =[ID, latitude and longitude, rated power]; Define adjacency matrix Where: n is the total number of nodes, and when there is a direct power connection between node i and node j. When there is no direct connection between node i and node j diagonal elements .

3. The method for monitoring and locating charging pile power network faults based on digital twins as described in claim 2, characterized in that... Step S2, which establishes a dynamic model of the charging pile equipment, specifically includes the following operations: For each charging station node Establish the state-space equations in the dq rotating coordinate system; Phase synchronization of grid voltage is achieved through a phase-locked loop (PLL): in: This is the real-time phase angle, in rad. The nominal angular frequency is 314 rad / s for a 50 Hz system; (The control objective is to bring the q-axis voltage to zero); , This refers to the proportional-integral gain coefficient. Add a device noise model: in: The intensity factor is the white noise level, taken as 0.5-2% of the rated value; The mth harmonic amplitude is set according to the limit value of IEC61000-3-12 standard. The fundamental frequency is 50Hz or 60Hz. The sampling frequency is ≥10kHz.

4. The method for monitoring and locating charging pile power network faults based on digital twins as described in claim 3, characterized in that... In step S2, a dynamic model of the charging pile equipment is established. The state-space equations are: in This is a state vector, including output voltage, current harmonic distortion rate. For the input control vector, For equipment noise; The above for each charging pile node The state-space equations in the dq rotating coordinate system are established as follows: in: , Let be the current vector of the charging pile in the dq rotating coordinate system, where the d-axis represents the active component and the q-axis represents the reactive component. This represents the voltage vector on the grid side in the dq coordinate system. R is the equivalent AC side resistance of the charging pile; L is the equivalent AC side inductance of the charging pile; The angular frequency of the power grid synchronization. =2π×50Hz or 60Hz.

5. The method for monitoring and locating charging pile power network faults based on digital twins as described in claim 4, characterized in that, The step S3, which involves constructing a charging pile fault detection data model, specifically includes the following operations: Define a multidimensional fault feature vector ; Calculate the dynamic baseline value: , in , For current, The total harmonic distortion of the current; A 15-minute sliding time window; Set the fault threshold for different levels: ; Establish fault diagnosis rules: when When an alert is triggered; Confirm the fault in time; among which Given an infinite norm, take the maximum value component. It is the L1 norm, the sum of the absolute values ​​of the components.

6. The method for monitoring and locating charging pile power network faults based on digital twins as described in claim 5, characterized in that, The specific steps for constructing the charging pile fault detection data model in step S3 are as follows: Define fault feature vector ,in The current mutation rate, This represents the change in harmonic distortion rate. This is the power factor offset. For impedance offset; set threshold vector ,when When an exception is triggered; The fault feature vector The calculation method is as follows: Current mutation rate: Change in harmonic distortion rate: Power factor offset: Impedance offset: in: The effective value of the current at node k at time t; T: Length of the sliding time window, ranging from 1 to 5 minutes; : Rated current of node k; The total harmonic distortion rate of the current at node k at time t; The power factor of node k at time t; Rated power factor, typically taken as 0.95~1.

0.

7. The method for monitoring and locating charging pile power network faults based on digital twins as described in claim 1, characterized in that... In step S5, the real-time monitoring employs a multi-time-scale adaptive sampling mechanism: The dynamic model is updated every 5 seconds under normal operating conditions; the default sampling period is [not specified]. =5 seconds; At that time, switch to 1-second high-frequency sampling; When any node satisfies At that time, the global sampling period is switched to =2 seconds; When any node satisfies At that time, the sampling period of this node and its direct upstream neighbor nodes is switched to =0.5 seconds.

8. The method for monitoring and locating charging pile power network faults based on digital twins as described in claim 1, characterized in that, It also includes an online model calibration step: Every =Execute once every 24 hours; During steady-state operation, all Collect actual measurement data from each charging station. Obtain the simulation output of the corresponding node of the digital twin. ; Gradient descent correction is applied to the dynamic model parameters R and L: in The learning rate is 0.01 to 0.

1. The number of nodes participating in the calibration; In step S4, the digital twin performs model calibration every 24 hours. Compare actual measured values Compared with simulation values ; Adjusting dynamic model parameters: in This is the learning rate.

Images