A fault positioning method and system for a power collection line of a new energy station

Abstract

The application discloses a new energy station power collection line fault positioning method and system, and relates to the field of power troubleshooting, comprising: a single-chip microcomputer, which is the main control end of the system and is used for controlling the start and stop of the system and issuing execution commands; a sensing module, which is used for collecting electrical parameters, environmental variables and topological state signals of the power collection line in real time to create a fault analysis original data set; through multi-dimensional feature fusion and a deep learning model, the application accurately identifies fault rules and calculates positioning parameters, and then, in combination with a topological database verification and terrain compensation correction which are updated in real time, the fault positioning accuracy is greatly improved, and finally, standardized information containing accurate coordinates is outputted and linked to maintenance dispatching, so that the line fault can be quickly responded, the troubleshooting time is reduced, the line operation and maintenance demand in a complex environment is adapted, and the safe and stable operation of the new energy station power collection line is ensured.

Description

Technical Field

[0001] This invention relates to the field of power fault diagnosis technology, specifically to a fault location method and system for power collection lines in new energy power plants. Background Technology

[0002] The power collection line of a new energy power station is a transmission line that connects new energy power generation equipment and the substation. It is responsible for collecting distributed power and transmitting it to the substation. It adopts the form of cable or overhead line, and needs to be adapted to the outdoor environment to ensure efficient and safe power transmission. It is the core channel for power collection in the power station.

[0003] Chinese invention patent application No. 202510029319.5 discloses a method for locating partial discharge faults in a power collection line, comprising: acquiring a first high-frequency partial discharge pulse signal at the bus end and a second high-frequency partial discharge pulse signal at the end of each branch line when a fault occurs in the power collection line of a new energy power station; obtaining a suspected fault range corresponding to the power collection line of the new energy power station based on the first high-frequency partial discharge pulse signal, the second high-frequency partial discharge pulse signal, and a power collection line fault range location model; verifying the suspected fault range according to the power collection line structure type corresponding to each suspected fault range; and determining the power collection line of the new energy power station based on the verification result. The application aims to address the problem that "because new energy power stations are mostly distributed in mountainous areas with large coverage areas, their power collection lines are radially distributed with multiple branches and short T-connections, resulting in the need for a large number of fault monitoring devices and reasonable installation points for fault ranging based on traveling wave to achieve accurate location of fault points along the entire power collection line. At the same time, the distributed power characteristics of new energy power station power collection lines and the variability of fault points may lead to the inability to accurately determine the fault range."

[0004] However, in the operation scenario of power collection lines in new energy power plants, the types and volume of faults are enormous. Designing a fault detection scheme based on only one fault perspective cannot comprehensively manage the faults of power collection lines in new energy power plants.

[0005] To address this, we propose a fault location method and system for the power collection lines of new energy power plants. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a fault location method and system for the power collection lines of new energy power plants, which can effectively solve the problems of the existing technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions; This invention discloses a fault location system for the power collection lines of a new energy power station, comprising: The microcontroller is the main control unit of the system, used to control the system to start and stop and issue execution commands; The system comprises four modules: a perception module for real-time acquisition of electrical parameters, environmental variables, and topology status signals of the power collection lines to create a raw dataset for fault analysis; a processing module for dynamic adjustment of filtering and normalization strategies based on line operating conditions to suppress noise and standardize the data in the raw dataset; an extraction module for receiving the noise-suppressed and standardized data, identifying the characteristic quantity variation patterns corresponding to different fault types, and generating a fault feature vector set; an analysis module for receiving the fault feature vector set, fusing multi-dimensional information from the fault feature vector set using a deep learning model, and outputting a location parameter set including fault distance and phase attributes; a verification module for calling a pre-set, real-time dynamically updated power collection line topology database to perform path matching verification on the location parameter set and applying a historical error model to correct the location; and an output module for receiving the corrected location results, converting the results into standardized information containing line segment identification and precise coordinates, synchronously pushing them to a pre-set monitoring terminal within the station, and triggering a maintenance dispatching mechanism.

[0008] Furthermore, the sensing module employs a multi-modal time-series synchronous acquisition mechanism, with the electrical parameter acquisition frequency dynamically adjusted according to the line operating conditions: Under normal operating conditions, the sampling frequency is 5kHz. When a transient inrush current peak is detected, the sampling frequency is automatically increased to 20kHz. The topology status signal adopts a dual-channel mode of fiber optic transmission and wireless backup, so that the signal transmission delay does not exceed 50ms; in, This indicates the rated current of the line.

[0009] Furthermore, the filtering strategy of the processing module is an adaptive weighted filtering that fuses transient features, and the filtered output... Calculated using the following formula: ; In the formula: This is the filtered output result for the k-th discrete time step; These are the weighting coefficients; This represents the raw data collected at the k-th time step. This is the filtered output result for the (k-1)th discrete time step, i.e., the historical filtered value from the previous time step; ; In the formula: This refers to the change in current. These are the real-time RMS values ​​of the line current at time k and time k-1; The raw data collected includes electrical parameters, environmental variables, and topology status signals. When a voltage surge is detected... When this occurs, the emergency filtering mode is triggered. Revised to , and constraints ≤0.9; It represents the amount of voltage change. This indicates the rated voltage of the line.

[0010] Furthermore, the normalization process of the processing module employs a differentiated mapping mechanism for different types of parameters. The electrical parameter normalization formula is as follows: ; The formula for normalizing environmental variables is: In the formula: The output range is mapped to the [0, 1] interval for the normalized electrical parameter values. This is the i-th data point in the original electrical parameter sequence; These are the historical maximum and minimum values ​​of the parameter; The current ambient temperature; The output range is mapped to the [0, 1] interval for the normalized environmental variable values. This refers to the i-th data point in the original sequence of environment variables; This represents the historical maximum value of the environmental variable. When the parameter value is 0, the normalization result is directly set to 0.1. The original electrical parameter sequence includes at least voltage, current, and power, and the original environmental variable sequence includes at least temperature, humidity, wind speed, and precipitation.

[0011] Furthermore, the extraction module identifies fault patterns through multi-scale feature fusion: for electrical parameter time series... Calculate the eigenenergy of the short-time Fourier transform respectively. High-frequency coefficients of wavelet packet decomposition The generated fault feature vector V includes: Moment of Change Energy entropy: ; Peak factor of high-frequency coefficients: ; In the formula: This represents the number of sampling points; Among them, when >0.8 and When the value is greater than 5, it is marked as a significant fault characteristic point.

[0012] Furthermore, the deep learning model of the analysis module is a topology-aware attention Bi-LSTM model, and the formula for calculating the fault distance D is: ; In the formula: This represents the total number of segments in the collector circuit; This is the actual length of the nth segment of the line; Use the Sigmoid activation function; This represents the total number of dimensions of the fault feature vectors; The topological attention weights of the m-th dimension feature on the n-th line segment are given by the m-th dimension feature. For the m-th eigenvector The mapping function; This is the terrain correction factor for the nth segment of the route; The expression for the Sigmoid activation function is: , Satisfy constraints , These are the normalized coefficients obtained by training a topology-aware attention Bi-LSTM model to assess the correlation between the m-th dimension fault feature and the topology attributes of the n-th line segment. The settings are dynamically adjusted based on the terrain type of the route, including 0.02 for flat terrain and 0.05 for mountainous terrain, to compensate for the impact of terrain differences on fault location accuracy. The route topology attributes include length, connectivity, and terrain parameters; ; In the formula: , The sensitivity coefficient and time-series change rate weighting coefficient of the m-th feature are respectively, with values ​​ranging from [0.6, 1.2] to [0.3, 0.8]. For feature vectors The peak value in; For feature vectors The baseline value; Let be the temporal rate of change of the m-th dimension feature; This is the feature stability correction factor, with a value range of [0.2, 0.5]. The smaller the fluctuation of the feature under normal operating conditions, the smaller the value. The real-time ambient temperature of the nth line segment; The reference temperature is set to 25℃.

[0013] Furthermore, the topology database update of the verification module adopts a prediction-verification two-stage mechanism: Prediction phase: An LSTM model trained on historical topology change data outputs the probability of topology changes in the next 10 minutes. ; Verification phase: when When the impedance value is greater than 0.7 or a switch state change is detected in real time, a topology update is initiated, and the updated line impedance matrix is ​​generated. satisfy: ; In the formula: This is the line impedance matrix before the update; This is the change in impedance; This is the time since the last update; in, Indicated by It is a diagonal matrix with diagonal elements, used to correct the diagonal elements of the impedance matrix.

[0014] Furthermore, the precise coordinate calculation of the output module incorporates terrain elevation compensation, as shown in the formula: ; In the formula: The precise planar coordinates of the fault point, i.e., the final standardized coordinate result output; These are the plane coordinates of the station's reference point, and the coordinates of the system's preset reference origin. This is the length of the line path from the fault point to the station's reference point. The azimuth angle of the line segment where the fault point is located; This represents the altitude difference between the fault location and the station's reference point. This represents the average altitude of the line segment where the fault point is located. Among them, 5 and 3 are the preset basic terrain compensation coefficients. When the elevation difference between the fault point and the benchmark point exceeds 50 meters, the compensation coefficients are adjusted to 6 and 4 respectively.

[0015] Furthermore, the sensing module is interconnected with the processing module and the extraction module via a wireless network, the extraction module is interconnected with the analysis module via a wireless network, and the analysis module is interconnected with the verification module and the output module via a wireless network. The sensing module, processing module, extraction module, analysis module, verification module, and output module are all integrated onto the microcontroller via dielectric electrical connection and interact with the microcontroller in real time via dielectric electrical connection or wireless network.

[0016] On the other hand, a fault location method for the power collection lines of a new energy power station includes: A multimodal time-series synchronous acquisition mechanism is used to collect electrical parameters, environmental variables, and topology status signals of the power collection line in real time to create a raw dataset for fault analysis. The acquisition frequency of electrical parameters is dynamically adjusted according to the line operating conditions, and the topology status signals adopt a dual-channel mode of fiber optic transmission and wireless backup. The data in the raw dataset for fault analysis is subjected to noise suppression using an adaptive weighted filtering strategy that fuses transient features, and a differentiated mapping mechanism is used for format standardization for different types of parameters. After receiving the data after noise suppression and format standardization, the feature quantity change patterns corresponding to different fault types are identified through multi-scale feature fusion, generating a fault feature vector set containing energy entropy at the moment of abrupt change and peak factor of high-frequency coefficients. Based on the topology-aware attention Bi-LSTM deep learning model, the multi-dimensional information in the fault feature vector set is fused to output a location parameter set containing fault distance and phase attributes. The preset real-time dynamically updated power collection line topology database is called to perform path matching verification on the location parameter set, and the historical error model is applied to correct the location results. The corrected location results are received and converted into standardized information containing line segment identification and accurate coordinates with terrain elevation compensation, which is then synchronously pushed to the preset monitoring terminal in the station and triggers the maintenance dispatch linkage mechanism.

[0017] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects: This invention provides a fault location method and system for power collection lines in new energy power plants. During execution, this method and system dynamically adjust the data acquisition frequency, flexibly switching acquisition accuracy between normal operating conditions and transient impacts. Combined with dual-channel signal transmission, it ensures real-time data transmission. Employing adaptive filtering and differentiated normalization strategies that fuse transient features, it effectively suppresses noise and standardizes various parameters, laying a reliable data foundation for fault analysis. Through multi-dimensional feature fusion and deep learning models, it accurately identifies fault patterns and calculates location parameters. Combined with real-time updated topology database verification and terrain compensation correction, it significantly improves fault location accuracy. Finally, it outputs standardized information containing precise coordinates and coordinates with maintenance dispatch, enabling rapid response to line faults, reducing troubleshooting time, adapting to the operational needs of lines in complex environments, and ensuring the safe and stable operation of power collection lines in new energy power plants. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0019] Figure 1This is a schematic diagram of the structure of a fault location system for the power collection lines of a new energy power station; Figure 2 This is a flowchart illustrating a fault location method for power collection lines in a new energy power plant. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] The present invention will be further described below with reference to embodiments.

[0022] Example 1:

[0023] This embodiment provides a fault location system for the power collection lines of a new energy power station, such as... Figure 1 As shown, it includes: The microcontroller is the main control unit of the system, used to control the system to start and stop and issue execution commands; The sensing module is used to collect electrical parameters, environmental variables and topology status signals of the power collection line in real time to create a raw dataset for fault analysis. The sensing module employs a multimodal time-series synchronous acquisition mechanism. This mechanism simultaneously acquires different types of data, such as electrical parameters, environmental variables, and topology status signals, and uses a unified clock reference to achieve time synchronization and alignment of the data across different modes. The electrical parameter acquisition frequency is dynamically adjusted according to the line operating conditions. Under normal operating conditions, the sampling frequency is 5kHz. When a transient inrush current peak is detected, the sampling frequency is automatically increased to 20kHz. The topology status signal adopts a dual-channel mode of fiber optic transmission and wireless backup, so that the signal transmission delay does not exceed 50ms; in, Indicates the rated current of the line; The processing module is used to dynamically adjust the filtering and normalization strategies based on line conditions to perform noise suppression and format standardization on the raw data of the fault analysis dataset. The processing module's filtering strategy is an adaptive weighted filtering that fuses transient features, and the filtered output... Calculated using the following formula: ; In the formula: This is the filtered output result for the k-th discrete time step; These are the weighting coefficients; This represents the raw data collected at the k-th time step; This is the filtered output result for the (k-1)th discrete time step, i.e., the historical filtered value from the previous time step; ; In the formula: This refers to the change in current. These are the real-time RMS values ​​of the line current at time k and time k-1; The raw data collected includes electrical parameters, environmental variables, and topology status signals. When a voltage surge is detected... When this occurs, the emergency filtering mode is triggered. Revised to , and constraints ≤0.9; It represents the amount of voltage change. Indicates the rated voltage of the line; The filter output in the above formula is calculated by fusing the current collected data with the historical filter values. The weight coefficient is dynamically adjusted with the change of current. The greater the current fluctuation, the more the weight is biased towards the current data, so that the filter can adapt to changes in line conditions. When a voltage change is detected, the weight coefficient is corrected and constrained to within 0.9. By strengthening the weight of real-time data, the system can quickly respond to transient impacts and take into account both noise suppression and the timeliness of fault signal capture. The normalization process of the processing module adopts a differentiated mapping mechanism for different types of parameters. The normalization formula for electrical parameters is: ; The formula for normalizing environmental variables is: In the formula: The output range is mapped to the [0, 1] interval for the normalized electrical parameter values. This is the i-th data point in the original electrical parameter sequence; These are the historical maximum and minimum values ​​of the parameter; The current ambient temperature; The output range is mapped to the [0, 1] interval for the normalized environmental variable values. This refers to the i-th data point in the original sequence of environment variables; This represents the historical maximum value of the environmental variable. When the parameter value is 0, the normalization result is directly set to 0.1. The original electrical parameter sequence includes at least voltage, current, and power, and the original environmental variable sequence includes at least temperature, humidity, wind speed, and precipitation. Specifically, the normalization result when the parameter value is 0 is set to 0.1 to avoid numerical errors of log(0) in subsequent energy entropy calculations, prevent division by zero errors in peak factor calculations and neural network training, and maintain the numerical stability of gradient propagation in deep learning models. 0.1, as a non-zero value close to zero, will not change the overall distribution characteristics of the data and can ensure the validity of all numerical calculations. The above formula uses historical extreme value ranges to linearly map electrical parameters, normalizing the original data to the [0,1] interval to ensure that electrical quantities of different magnitudes can be directly used in the analysis; environmental variables are differentiated by combining the current temperature and are also mapped to the [0,1] interval. When the parameter value is 0, it is specially treated as 0.1 to avoid the interference of zero values ​​on subsequent model calculations, thus achieving standardization and unification of electrical and environmental data; The extraction module is used to receive data after noise suppression and format standardization, identify the feature quantity change patterns corresponding to different fault types, and generate a fault feature vector set. The extraction module identifies fault patterns through multi-scale feature fusion: based on the time series of electrical parameters. Calculate the eigenenergy of the short-time Fourier transform respectively. High-frequency coefficients of wavelet packet decomposition The generated fault feature vector V includes: Moment of Change Energy entropy: ; Peak factor of high-frequency coefficients: ; In the formula: This represents the number of sampling points; Among them, when >0.8 and When the value is greater than 5, it is marked as a significant fault characteristic point; In the above formula, energy entropy reflects the fault intensity by quantifying the disorder of signal energy distribution at the fault moment, while the peak factor highlights transient impact characteristics by the ratio of signal peak value to effective value. Combining these two factors with multi-scale feature extraction results, when energy entropy is greater than 0.8 and peak factor is greater than 5, significant fault features can be accurately marked, providing a quantitative basis for fault type identification. The analysis module receives a set of fault feature vectors, fuses multi-dimensional information from the fault feature vector set through a deep learning model, and outputs a set of positioning parameters including fault distance and phase attributes. The deep learning model used in the analysis module is a topology-aware attention Bi-LSTM model. This model integrates route topology information into the attention mechanism of a bidirectional long short-term memory network. By introducing a topology attribute weight matrix, the model can dynamically adjust feature attention based on route connectivity and terrain parameters. The formula for calculating the fault distance D is: ; In the formula: This represents the total number of segments in the collector circuit; This is the actual length of the nth segment of the line; Use the Sigmoid activation function; This represents the total number of dimensions of the fault feature vectors; The topological attention weights of the m-th dimension feature on the n-th line segment are given by the m-th dimension feature. For the m-th eigenvector The mapping function; This is the terrain correction factor for the nth segment of the route; The expression for the Sigmoid activation function is: , Satisfy constraints , These are the normalized coefficients obtained by training a topology-aware attention Bi-LSTM model to assess the correlation between the m-th dimension fault feature and the topology attributes of the n-th line segment. The settings are dynamically adjusted based on the terrain type of the route, including 0.02 for flat terrain and 0.05 for mountainous terrain, to compensate for the impact of terrain differences on fault location accuracy. In the above formula, the fault distance is obtained by summing the product of the length of each line segment and the topological correlation degree. The topological correlation degree is calculated by an attention mechanism combined with a feature mapping function, and then activated by Sigmoid to ensure the rationality of the weights. At the same time, a terrain correction factor is introduced, which is 0.02 for flat land and 0.05 for mountainous land, to compensate for the impact of terrain differences on signal propagation and improve the positioning accuracy under different terrain conditions. The route topology attributes include length, connectivity, and terrain parameters; ; In the formula: , The sensitivity coefficient and time-series change rate weighting coefficient of the m-th feature are respectively, with values ​​ranging from [0.6, 1.2] to [0.3, 0.8]. For feature vectors The peak value in; For feature vectors The baseline value; Let be the temporal rate of change of the m-th dimension feature; The feature stability correction factor has a value range of [0.2, 0.5]. The feature stability correction factor is used to quantify the degree of fluctuation of the feature under normal operating conditions. It is determined by calculating the ratio of the standard deviation of the feature value to the mean. The smaller the fluctuation, the more stable the feature is, and the smaller the value. The real-time ambient temperature of the nth line segment; The reference temperature is set to 25℃. The above function comprehensively considers the ratio of characteristic peak value to baseline value, time series change rate, stability correction factor and temperature effect. It adjusts the proportion of each factor through sensitivity coefficient and time series weight coefficient. The coefficient value range is strictly constrained to avoid extreme value interference. This function realizes the organic integration of multi-dimensional features, making the feature vector more consistent with the actual working conditions of the line. The verification module is used to call the preset real-time dynamically updated collection line topology database, perform path matching verification on the positioning parameter set, and apply the historical error model to complete the positioning correction. The topology database update of the verification module adopts a two-stage prediction-verification mechanism: Prediction phase: An LSTM model trained on historical topology change data outputs the probability of topology changes in the next 10 minutes. ; Verification phase: when When the impedance value is greater than 0.7 or a switch state change is detected in real time, a topology update is initiated, and the updated line impedance matrix is ​​generated. satisfy: ; In the formula: This is the line impedance matrix before the update; This is the change in impedance; This is the time since the last update; in, Indicated by It is a diagonal matrix with diagonal elements, used to correct the impedance matrix with diagonal elements; Triggered when the probability of a topology change exceeds 0.7 or when a switch state changes, an impedance change is introduced through a diagonal matrix and corrected by the update time. This dynamic adjustment method ensures that the topology database can reflect line impedance changes in real time, providing accurate data support for location verification. The output module is used to receive the corrected positioning results, convert the positioning results into standardized information containing line segment identification and precise coordinates, and push it synchronously to the preset monitoring terminal in the station and trigger the maintenance dispatch linkage mechanism. The output module's precise coordinate calculation incorporates terrain elevation compensation, using the following formula: ; In the formula: The precise planar coordinates of the fault point, i.e., the final standardized coordinate result output; These are the plane coordinates of the station's reference point, and the coordinates of the system's preset reference origin. This is the length of the line path from the fault point to the station's reference point. The azimuth angle of the line segment where the fault point is located; This represents the altitude difference between the fault location and the station's reference point. This represents the average altitude of the line segment where the fault point is located. Among them, 5 and 3 are the preset basic terrain compensation coefficients. When the elevation difference between the fault point and the benchmark point exceeds 50 meters, the compensation coefficients are adjusted to 6 and 4 respectively. The above formula combines the path length of the fault point and the azimuth of the direction to determine the basic location. Then, terrain compensation is introduced by the ratio of the altitude difference to the average altitude. When the altitude difference exceeds 50 meters, the compensation coefficient is dynamically adjusted. This method effectively offsets the influence of terrain elevation on the path distance and outputs standardized information containing accurate coordinates. The sensing module interacts with the processing module and the extraction module via a wireless network. The extraction module interacts with the analysis module via a wireless network. The analysis module interacts with the verification module and the output module via a wireless network. The sensing module, processing module, extraction module, analysis module, verification module, and output module are all integrated and connected to the microcontroller via dielectric electrical connection, interacting with the microcontroller in real time through dielectric electrical connection or wireless network. It should be noted that dielectric electrical connection refers to a comprehensive connection method that utilizes the electrical transmission characteristics of different transmission media (such as optical fiber, cable, and wireless) to achieve signal connection.

[0024] In this embodiment, the microcontroller-controlled sensing module collects electrical parameters, environmental variables, and topology status signals of the power line in real time to create a raw dataset for fault analysis. The processing module then dynamically adjusts filtering and normalization strategies based on line conditions to perform noise suppression and format standardization on the raw dataset. The extraction module receives the data after noise suppression and format standardization, identifies the characteristic quantity change patterns corresponding to different fault types, and generates a fault feature vector set. The analysis module further receives the fault feature vector set, fuses multi-dimensional information from the fault feature vector set using a deep learning model, and outputs a location parameter set containing fault distance and phase attributes. The verification module calls a preset, dynamically updated power line topology database to perform path matching verification on the location parameter set and applies a historical error model to complete the location correction. Finally, the output module receives the corrected location result, converts it into standardized information containing line segment identification and precise coordinates, and synchronously pushes it to a preset monitoring terminal in the station to trigger a maintenance dispatch linkage mechanism.

[0025] In the above embodiments, the system captures line anomalies in real time and accurately, quickly locates the fault distance and phase by dynamically adjusting the acquisition frequency and optimizing data processing, and outputs fault information with accurate coordinates by combining topology information verification and correction. This significantly shortens the fault investigation time. At the same time, it coordinates maintenance dispatching, adapts to different terrains and environmental interference, improves fault location accuracy, reduces power outage losses, and ensures the safe and stable operation of the power collection lines of new energy power plants.

[0026] Example 2:

[0027] At the implementation level, based on Example 1, this example refers to... Figure 2 A further detailed description of the fault location system for a new energy power station's collection line in Example 1 is provided below: A fault location method for power collection lines in a new energy power plant includes: The electrical parameters, environmental variables, and topology status signals of the power collection line are collected in real time through a multimodal time-series synchronous acquisition mechanism to create a raw dataset for fault analysis. The acquisition frequency of electrical parameters is dynamically adjusted according to the line operating conditions, and the topology status signals adopt a dual-channel mode of fiber optic transmission and wireless backup. The data in the original dataset for fault analysis are subjected to noise suppression using an adaptive weighted filtering strategy that integrates transient features, and format standardization is performed using a differentiated mapping mechanism for different types of parameters. The system receives data that has been processed by noise suppression and format standardization, identifies the feature quantity change patterns corresponding to different fault types through multi-scale feature fusion, and generates a fault feature vector set containing energy entropy at the moment of abrupt change and peak factor of high-frequency coefficient. Based on the topology-aware attention Bi-LSTM deep learning model, multi-dimensional information from the fault feature vector set is fused to output a localization parameter set including fault distance and phase attributes; The system calls a pre-defined, real-time, dynamically updated collector line topology database to perform path matching verification on the positioning parameter set and applies a historical error model to correct the positioning results. The system receives the corrected positioning results, converts them into standardized information containing line segment identifiers and accurate coordinates with terrain elevation compensation, and simultaneously pushes them to the preset monitoring terminals in the station, triggering the maintenance dispatch linkage mechanism.

[0028] In summary, the methods and systems described in the above embodiments dynamically adjust the data acquisition frequency of the lines during execution, flexibly switching the acquisition accuracy between normal operating conditions and transient impacts. Combined with dual-channel signal transmission, they ensure real-time data transmission. Adaptive filtering and differentiated normalization strategies that fuse transient features effectively suppress noise and standardize various parameters, laying a reliable data foundation for fault analysis. Through multi-dimensional feature fusion and deep learning models, they accurately identify fault patterns and calculate location parameters. Combined with real-time updated topology database verification and terrain compensation correction, they significantly improve fault location accuracy. Finally, they output standardized information containing precise coordinates and coordinate maintenance dispatch, enabling rapid response to line faults, reducing troubleshooting time, adapting to the operational needs of lines in complex environments, and ensuring the safe and stable operation of power collection lines in new energy power plants.

[0029] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault location system for the power collection lines of a new energy power station, characterized in that, include: The microcontroller is the main control unit of the system, used to control the system to start and stop and issue execution commands; The sensing module is used to collect electrical parameters, environmental variables and topology status signals of the power collection line in real time to create a raw dataset for fault analysis. The processing module is used to dynamically adjust the filtering and normalization strategies based on line conditions to perform noise suppression and format standardization on the raw data of the fault analysis dataset. The extraction module is used to receive data after noise suppression and format standardization, identify the feature quantity change patterns corresponding to different fault types, and generate a fault feature vector set. The analysis module receives a set of fault feature vectors, fuses multi-dimensional information from the fault feature vector set through a deep learning model, and outputs a set of positioning parameters including fault distance and phase attributes. The verification module is used to call the preset real-time dynamically updated collection line topology database, perform path matching verification on the positioning parameter set, and apply the historical error model to complete the positioning correction. The output module is used to receive the corrected positioning results, convert the positioning results into standardized information containing line segment identification and precise coordinates, and push it synchronously to the preset monitoring terminal in the station and trigger the maintenance dispatch linkage mechanism.

2. The fault location system for the power collection lines of a new energy power station according to claim 1, characterized in that, The sensing module adopts a multi-modal time-series synchronous acquisition mechanism, and the electrical parameter acquisition frequency is dynamically adjusted according to the line operating conditions. Under normal operating conditions, the sampling frequency is 5kHz. When a transient inrush current peak is detected, the sampling frequency is automatically increased to 20kHz. The topology status signal adopts a dual-channel mode of fiber optic transmission and wireless backup, so that the signal transmission delay does not exceed 50ms; in, This indicates the rated current of the line.

3. The fault location system for the power collection lines of a new energy power station according to claim 2, characterized in that, The filtering strategy of the processing module is an adaptive weighted filtering that fuses transient features, and the filtered output... Calculated using the following formula: ； In the formula: This is the filtered output result for the k-th discrete time step; These are the weighting coefficients; This represents the raw data collected at the k-th time step. This is the filtered output result for the (k-1)th discrete time step, i.e., the historical filtered value from the previous time step; ； In the formula: This refers to the change in current. These are the real-time RMS values ​​of the line current at time k and time k-1; The raw data collected includes electrical parameters, environmental variables, and topology status signals. When a voltage surge is detected... When this occurs, the emergency filtering mode is triggered. Revised to , and constraints ≤0.9; It represents the amount of voltage change. This indicates the rated voltage of the line.

4. The fault location system for the power collection lines of a new energy power station according to claim 1, characterized in that, The normalization process of the processing module adopts a differentiated mapping mechanism for different types of parameters. The normalization formula for electrical parameters is: ； The formula for normalizing environmental variables is: In the formula: The output range is mapped to the [0, 1] interval for the normalized electrical parameter values. This is the i-th data point in the original electrical parameter sequence; These are the historical maximum and minimum values ​​of the parameter; The current ambient temperature; The output range is mapped to the [0, 1] interval for the normalized environmental variable values. This refers to the i-th data point in the original sequence of environment variables. This represents the historical maximum value of the environmental variable. When the parameter value is 0, the normalization result is directly set to 0.

1. The original electrical parameter sequence includes at least voltage, current, and power, and the original environmental variable sequence includes at least temperature, humidity, wind speed, and precipitation.

5. A fault location system for a new energy power station's power collection line according to claim 1, characterized in that, The extraction module identifies fault patterns through multi-scale feature fusion: it analyzes the time-series sequences of electrical parameters. Calculate the eigenenergy of the short-time Fourier transform respectively. High-frequency coefficients of wavelet packet decomposition The generated fault feature vector V includes: Moment of Change Energy entropy: ; Peak factor of high-frequency coefficients: ; In the formula: This represents the number of sampling points; Among them, when >0.8 and When the value is greater than 5, it is marked as a significant fault characteristic point.

6. The fault location system for the power collection lines of a new energy power station according to claim 1, characterized in that, The deep learning model of the analysis module is a topology-aware attention Bi-LSTM model, and the formula for calculating the fault distance D is: ； In the formula: This represents the total number of segments in the collector circuit; This is the actual length of the nth segment of the line; Use the Sigmoid activation function; This represents the total number of dimensions of the fault feature vectors; The topological attention weights of the m-th dimension feature on the n-th line segment are given by the m-th dimension feature. For the m-th eigenvector The mapping function; This is the terrain correction factor for the nth segment of the route; The expression for the Sigmoid activation function is: , Satisfy constraints , These are the normalized coefficients obtained by training a topology-aware attention Bi-LSTM model to assess the correlation between the m-th dimension fault feature and the topology attributes of the n-th line segment. The settings are dynamically adjusted based on the terrain type of the route, including 0.02 for flat terrain and 0.05 for mountainous terrain, to compensate for the impact of terrain differences on fault location accuracy. The route topology attributes include length, connectivity, and terrain parameters; ； In the formula: , The sensitivity coefficient and time-series change rate weighting coefficient of the m-th feature are respectively, with values ​​ranging from [0.6, 1.2] to [0.3, 0.8]. For feature vectors The peak value in; For feature vectors The baseline value; Let be the temporal rate of change of the m-th dimension feature; This is the feature stability correction factor, with a value range of [0.2, 0.5]. The smaller the fluctuation of the feature under normal operating conditions, the smaller the value. The real-time ambient temperature of the nth line segment; The reference temperature is set to 25℃.

7. The fault location system for the power collection lines of a new energy power station according to claim 1, characterized in that, The topology database update of the verification module adopts a prediction-verification two-stage mechanism: Prediction phase: An LSTM model trained on historical topology change data outputs the probability of topology changes in the next 10 minutes. ; Verification phase: when When the impedance value is greater than 0.7 or a switch state change is detected in real time, a topology update is initiated, and the updated line impedance matrix is ​​generated. satisfy: ； In the formula: This is the line impedance matrix before the update; This is the change in impedance; This is the time since the last update; in, Indicated by It is a diagonal matrix with diagonal elements, used to correct the diagonal elements of the impedance matrix.

8. A fault location system for a new energy power station's power collection line according to claim 1, characterized in that, The precise coordinate calculation of the output module incorporates terrain elevation compensation, and the formula is as follows: ； In the formula: The precise planar coordinates of the fault point, i.e., the final standardized coordinate result output; These are the plane coordinates of the station's reference point, and the coordinates of the system's preset reference origin. This is the length of the line path from the fault point to the station's reference point. The azimuth angle of the line segment where the fault point is located; This represents the altitude difference between the fault location and the station's reference point. This represents the average altitude of the line segment where the fault point is located. Among them, 5 and 3 are the preset basic terrain compensation coefficients. When the elevation difference between the fault point and the benchmark point exceeds 50 meters, the compensation coefficients are adjusted to 6 and 4 respectively.

9. A fault location system for a new energy power station's power collection line according to claim 1, characterized in that, The sensing module is interactively connected to the processing module and the extraction module via a wireless network; the extraction module is interactively connected to the analysis module via a wireless network; and the analysis module is interactively connected to the verification module and the output module via a wireless network. The sensing module, processing module, extraction module, analysis module, verification module, and output module are all integrated onto the microcontroller via dielectric electrical connection and interact with the microcontroller in real time via dielectric electrical connection or wireless network.

10. A method for fault location of a power collection line in a new energy power station, wherein the method is an implementation method of the fault location system for a power collection line in a new energy power station as described in any one of claims 1-9, characterized in that, include: The electrical parameters, environmental variables, and topology status signals of the power collection line are collected in real time through a multimodal time-series synchronous acquisition mechanism to create a raw dataset for fault analysis. The acquisition frequency of electrical parameters is dynamically adjusted according to the line operating conditions, and the topology status signals adopt a dual-channel mode of fiber optic transmission and wireless backup. The data in the original dataset for fault analysis are subjected to noise suppression using an adaptive weighted filtering strategy that integrates transient features, and format standardization is performed using a differentiated mapping mechanism for different types of parameters. The system receives data that has been processed by noise suppression and format standardization, identifies the feature quantity change patterns corresponding to different fault types through multi-scale feature fusion, and generates a fault feature vector set containing energy entropy at the moment of abrupt change and peak factor of high-frequency coefficient. Based on the topology-aware attention Bi-LSTM deep learning model, multi-dimensional information from the fault feature vector set is fused to output a localization parameter set including fault distance and phase attributes; The system calls a pre-defined, real-time, dynamically updated collector line topology database to perform path matching verification on the positioning parameter set and applies a historical error model to correct the positioning results. The system receives the corrected positioning results, converts them into standardized information containing line segment identifiers and accurate coordinates with terrain elevation compensation, and simultaneously pushes them to the preset monitoring terminals in the station, triggering the maintenance dispatch linkage mechanism.

Images