Direct current transmission line fault location method and device based on neural network and medium

Abstract

The application discloses a kind of neural network-based direct current transmission line fault location method, device and medium.Therein, method includes: detect whether the fault of direct current line occurs, and in the case where the fault of direct current line is detected, store the single-ended voltage sampling value in the predetermined time period before and after direct current line fault;Distinguish whether the negative pole fault of direct current line occurs, in the case where the fault of direct current line is negative pole fault, single-ended voltage sampling value is multiplied-1;Single-ended voltage sampling value is smoothed, and voltage smooth sampling value is determined;According to voltage smooth sampling value, extract the first negative peak time, the first negative peak and delay time during direct current line fault;The first negative peak time, the first negative peak and delay time are input into the regression neural network model constructed in advance, and the predicted fault position of direct current transmission line is output.

Description

Technical Field

[0001] The present invention relates to the technical field of power system relay protection, and more particularly to a method, device and medium for locating a DC transmission line fault based on a neural network. Background Art

[0002] Currently, DC transmission systems are typically equipped with fast-acting protection devices and hybrid DC switches to quickly isolate DC faults within a few milliseconds. Therefore, only a very short period of fault data can be obtained before the DC switch is disconnected, which increases the difficulty of accurately locating faults in the DC transmission network.

[0003] Several typical fault location methods have been proposed, but they present the following problems. Fault location methods based on the arrival time of a fault traveling wave at both ends of the line require a sampling frequency of several MHz, and the accuracy of the location is highly dependent on the time synchronization of the signals at both ends, which places very high demands on communication and timing at both ends. Location methods based on the arrival time difference of traveling waves of different modes at the endpoints decrease in accuracy as the transition resistance increases. Fault location methods based on electromagnetic time reversal voltage energy and those based on convolutional neural networks and Hilbert transforms require data with time windows that are too long and are not suitable for fast-acting DC power grids. Fault location methods based on the centralized RL model of DC lines, while relatively simple in algorithm, are difficult to apply to long lines. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method, device and medium for locating faults in a DC transmission line based on a neural network.

[0005] According to one aspect of the present invention, a method for locating a DC transmission line fault based on a neural network is provided, comprising:

[0006] Detecting whether a DC line fault exists, and if a DC line fault is detected, storing single-ended voltage sampling values ​​within a predetermined time period before and after the DC line fault;

[0007] Determine whether the DC line fault is a negative pole fault. If the DC line fault is a negative pole fault, multiply the single-ended voltage sampling value by -1.

[0008] Smoothing the single-ended voltage sampling value to determine a voltage smoothing sampling value;

[0009] Calculate the first negative peak time, the first negative peak value, and the delay time during the DC line fault period based on the voltage smoothing sampling value;

[0010] The first negative peak time, the first negative peak, and the delay time are input into a pre-built regression neural network model to output the predicted fault location of the DC transmission line.

[0011] Optionally, the predetermined time period is from 0.1 ms before the DC line fault to 2.5 ms after the fault.

[0012] Optionally, the calculation formula for the voltage smoothing sampling value is:

[0013]

[0014] Where U fm is the single-ended voltage sampling value, U sfm [t] is the voltage sampling smoothing value at any time t, and the sliding average window of the smoothing process is 30us.

[0015] Optionally, calculating the first negative peak time, the first negative peak value, and the delay time during the DC line fault period based on the voltage smoothing sampling value includes:

[0016] Calculate the average pre-fault voltage of the DC line fault based on the voltage smoothing sampling value;

[0017] Extract the first negative peak time and the first negative peak corresponding to the moment in the voltage smoothing sampling value;

[0018] The delay time of the voltage smoothing sampling value of the DC line fault is estimated according to the time of the average value of the voltage before the fault and the average value of the first negative peak.

[0019] Optionally, the process of building a regression neural network model is as follows:

[0020] Simulate different fault conditions of the DC line, including single-pole and double-pole faults at different fault locations and fault resistances;

[0021] Extract voltage characteristics under different fault conditions, where the voltage characteristics include: the first negative peak of the fault, the time of the first negative peak, and the delay time;

[0022] A regression neural network model is trained based on voltage characteristic quantities under different fault conditions and corresponding fault locations, wherein all weights and biases of the regression neural network model are solved according to the minimization objective function, and the hyperparameters of the regression neural network are determined according to a preset optimization algorithm, wherein the hyperparameters include: an activation function f, a regularization coefficient λ, and one or more of the number of fully connected layers and the number of output nodes in each layer.

[0023] Optionally, the data processing process of the regression neural network model is:

[0024]

[0025] Where, ω ij and ω kj’They are the first layer of the fully connected layer h 1j and the kth layer h kj’ The weight, ω j’ is the weight of the output layer x; θ 1j and θ kj’ are the biases of the first and kth layers of the fully connected layer, respectively, and θ1 is the bias of the output layer x; the fully connected layer has K layers, the first layer has N output nodes, and the kth layer has M output nodes; f represents the activation function; the prediction variables of the three inputs are defined as O = (o1, o2, o3) T ; The output vector of the kth fully connected layer is H k =(h k1 ,h k2 ,h k3 ,…,h kj ) T ;The output variable of the output layer is the fault location x.

[0026] According to another aspect of the present invention, a device for locating a DC transmission line fault based on a neural network is provided, comprising:

[0027] a detection and storage module, configured to detect whether a DC line fault exists and, if a DC line fault is detected, store single-ended voltage sampling values ​​within a predetermined time period before and after the DC line fault;

[0028] A discrimination and adjustment module is used to discriminate whether the DC line has a negative pole fault. If the DC line fault is a negative pole fault, the single-ended voltage sampling value is multiplied by -1.

[0029] A smoothing module is used to smooth the single-ended voltage sampling value and determine the voltage smoothing sampling value;

[0030] a calculation module, configured to calculate the first negative peak time, the first negative peak value, and the delay time during a DC line fault period based on the voltage smoothing sampling value;

[0031] The prediction module is used to input the first negative peak time, the first negative peak and the delay time into a pre-built regression neural network model and output the predicted fault location of the DC transmission line.

[0032] Optionally, the predetermined time period in the detection storage module is 0.1 ms before the DC line fault to 2.5 ms after the fault.

[0033] Optionally, the calculation formula of the voltage smoothing sampling value in the smoothing module is:

[0034]

[0035] Where U fm is the single-ended voltage sampling value, U sfm[t] is the voltage sampling smoothing value at any time t, and the sliding average window of the smoothing process is 30us.

[0036] Optionally, the computing module includes:

[0037] The average value calculation submodule is used to calculate the pre-fault voltage average value of the DC line fault based on the voltage smoothing sampling value;

[0038] The extraction submodule is used to extract the first negative peak time and the first negative peak corresponding to the moment in the voltage smoothing sampling value;

[0039] The estimation submodule is used to estimate the delay time of the voltage smoothing sampling value of the DC line fault according to the time of the average value of the voltage before the fault and the average value of the first negative peak.

[0040] Optionally, the process of building a regression neural network model in the prediction module is as follows:

[0041] A simulation submodule is used to simulate different fault conditions of the DC line, including single-pole and double-pole faults at different fault locations and fault resistances.

[0042] The extraction submodule is used to extract voltage characteristics under different fault conditions, wherein the voltage characteristics include: the first negative peak value of the fault, the time of the first negative peak value, and the delay time;

[0043] A training submodule is used to train a regression neural network model based on the voltage characteristic quantities under different fault conditions and the corresponding fault locations, wherein all weights and biases of the regression neural network model are solved according to the minimization objective function, and the hyperparameters of the regression neural network are determined according to a preset optimization algorithm, wherein the hyperparameters include: an activation function f, a regularization coefficient λ, and one or more of the number of fully connected layers and the number of output nodes in each layer.

[0044] Optionally, the data processing process of the regression neural network model is:

[0045]

[0046] Where, ω ij and ω kj’ They are the first layer of the fully connected layer h 1j and the kth layer h kj’ The weight, ω j’ is the weight of the output layer x; θ 1j and θ kj’are the biases of the first and kth layers of the fully connected layer, respectively, and θ1 is the bias of the output layer x; the fully connected layer has K layers, the first layer has N output nodes, and the kth layer has M output nodes; f represents the activation function; the prediction variables of the three inputs are defined as O = (o1, o2, o3) T ; The output vector of the kth fully connected layer is H k =(h k1 ,h k2 ,h k3 ,…,h kj ) T ;The output variable of the output layer is the fault location x.

[0047] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the storage medium stores a computer program, and the computer program is used to execute the method according to any one of the above aspects of the present invention.

[0048] According to another aspect of the present invention, an electronic device is provided, comprising: a processor; a memory for storing instructions executable by the processor; and the processor for reading the executable instructions from the memory and executing the instructions to implement the method described in any one of the above aspects of the present invention.

[0049] Therefore, the present application analyzes the fault response characteristics of the single-ended transient voltage after a DC transmission line fault, and concludes that the first negative peak time, the first negative peak and the delay time of the fault transient voltage have a clear mathematical relationship with the fault location and the transition resistance. Based on this conclusion, the present invention uses the first negative peak time, the first negative peak and the delay time of the single-ended transient voltage of the DC line as characteristic quantities, and inputs them into a regression neural network to estimate the fault location. It has the following beneficial effects: (1) For single-pole and bipolar faults with different fault locations and different transition resistances, the present invention has high fault location accuracy and strong transition resistance resistance resistance. (2) The required data time window is short, which can be applied to flexible DC power grids with fast protection. (3) The sampling frequency is low, only in the kHz range. Only single-ended voltage data is required, and no two-end communication is required. (4) It has good generalization ability and still has high positioning accuracy under line parameter deviations within a reasonable range. (5) It has strong anti-noise ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0051] Figure 1 1 is a flow chart of a DC transmission line fault location method based on a neural network provided by an exemplary embodiment of the present invention;

[0052] Figure 21 is a topological diagram of a four-terminal bipolar flexible DC grid provided by an exemplary embodiment of the present invention;

[0053] Figure 3 is a typical topological structure diagram of a recurrent neural network provided by an exemplary embodiment of the present invention;

[0054] Figure 4 This is a schematic diagram of a specific process of a DC transmission line fault location method based on a neural network provided by an exemplary embodiment of the present invention;

[0055] Figure 5 is a parameter configuration diagram of a DC cable provided by an exemplary embodiment of the present invention;

[0056] Figure 6 is another topological structure diagram of a four-terminal bipolar flexible DC grid provided by an exemplary embodiment of the present invention;

[0057] Figure 7 1 is a schematic structural diagram of a DC transmission line fault location device based on a neural network provided by an exemplary embodiment of the present invention;

[0058] Figure 8 This is a structure of an electronic device provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0059] Below, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0060] It should be noted that the relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0061] Those skilled in the art will understand that the terms "first" and "second" in the embodiments of the present invention are only used to distinguish different steps, devices or modules, and neither represent any specific technical meaning nor indicate the necessary logical order between them.

[0062] It should also be understood that, in the embodiments of the present invention, “a plurality of” may refer to two or more than two, and “at least one” may refer to one, two or more than two.

[0063] It should also be understood that any component, data or structure mentioned in the embodiments of the present invention can generally be understood as one or more, unless explicitly limited or otherwise indicated in the context.

[0064] In addition, the term "and / or" in this invention merely describes an association relationship between related objects, indicating that three possible relationships exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Furthermore, the character " / " in this invention generally indicates that the related objects are in an "or" relationship.

[0065] It should also be understood that the description of the various embodiments of the present invention focuses on the differences between the various embodiments, and the same or similar aspects thereof can be referenced with each other. For the sake of brevity, they will not be described one by one.

[0066] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0067] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0068] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0069] It should be noted that like reference numerals and letters refer to like items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0070] Embodiments of the present invention can be applied to electronic devices such as terminal devices, computer systems, and servers, and can operate in conjunction with numerous other general-purpose or specialized computing system environments or configurations. Examples of well-known terminal devices, computing systems, environments, and / or configurations suitable for use with terminal devices, computer systems, servers, and other electronic devices include, but are not limited to, personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network personal computers, minicomputer systems, mainframe computer systems, and distributed cloud computing technology environments including any of the above.

[0071] Electronic devices such as terminal devices, computer systems, and servers can be described in the general context of computer system-executable instructions (such as program modules) executed by a computer system. Generally, program modules can include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in a distributed cloud computing environment, where tasks are performed by remote processing devices linked via a communication network. In a distributed cloud computing environment, program modules can be located on local or remote computing system storage media, including storage devices.

[0072] Exemplary Methods

[0073] Figure 1 This is a flow chart of a DC transmission line fault location method based on a neural network provided by an exemplary embodiment of the present invention. This embodiment can be applied to electronic devices, such as Figure 1 As shown, the DC transmission line fault location method 100 based on a neural network includes the following steps:

[0074] Step 101: Detect whether a DC line fault exists, and if a DC line fault is detected, store single-ended voltage sampling values ​​within a predetermined time period before and after the DC line fault.

[0075] Step 102 , determining whether the DC line fault is a negative pole fault, and if the DC line fault is a negative pole fault, multiplying the single-ended voltage sampling value by -1;

[0076] Step 103, smoothing the single-ended voltage sampling value to determine a voltage smoothing sampling value;

[0077] Step 104 , calculating the first negative peak time, the first negative peak value, and the delay time during the DC line fault period based on the voltage smoothing sampling value;

[0078] Step 105 : Input the first negative peak time, the first negative peak value, and the delay time into a pre-built regression neural network model, and output a predicted fault location of the DC transmission line.

[0079] Specifically, this application addresses the problems existing in existing DC transmission line fault location and proposes an ultra-fast fault location algorithm for DC transmission lines with a short required data time window, high positioning accuracy, strong resistance to transition resistance and noise, and adaptability to different line parameters and different system structures.

[0080] In the RLC centralized equivalent circuit of the DC line and the modular multilevel converter station, the present invention analyzes the fault response characteristics of the single-end transient voltage of the DC line after the fault and obtains the first negative peak time t of the fault transient voltage based on the control response theory of the second-order system. p , the first negative peak δ and the delay time t d With the fault location x and the transition resistance R f There is a clear mathematical relationship between them. And through simulation of a four-terminal flexible DC grid built with a frequency-dependent distributed parameter line model and a modular multi-level converter station detailed model, it can be seen that t p , δ and t d All can be extracted within 2.5ms after the fault. p , δ and t d The relationship between the transient voltage characteristic and the fault location is a complex irrational equation. Therefore, the fault location can be estimated by inputting the above three voltage characteristic quantities into a recurrent neural network. The present invention proposes a DC transmission line protection method based on transient voltage characteristic quantities and a recurrent neural network. The overall technical concept is as follows:

[0081] 1) Measure the single-ended voltage of the DC line and perform DC line fault detection.

[0082] 2) If a DC line fault is detected, the single-ended voltage sampling value U from 0.1ms before the fault to 2.5ms after the fault is stored fm [t1,t2].

[0083] 3) Check whether it is a negative pole fault. If a negative pole fault occurs in the DC line, the voltage sampling value U fm Multiply by -1.

[0084] 4) The voltage sampling value U fm [t1, t2] is smoothed with a 30us sliding average window, as shown in formula (1):

[0085]

[0086] U sfm [t] is the voltage sampling smoothing value at any time t. For the sampling values ​​at the beginning and end of the data window, the length of the smoothing window is shortened to the existing sampling points.

[0087] 5) With U sfm The average value of the voltage before the fault is calculated as U pre , as shown in formula (2).

[0088]

[0089] 6) Extract U sfm The first negative peak time t of [t1,t2]p and the negative peak δ at that moment. This can be extracted using the findpeak function in MATLAB.

[0090] 7) With U pre The time at which the average value of and δ is estimated is U sfm The delay time t of [t1,t2] d .

[0091] 8) Before using this fault location method, Figure 2 The simulated DC line in the four-terminal bipolar ±500kV flexible DC grid is shown in the figure. AB Unipolar and bipolar faults at different fault locations (10% interval from the line start to the line end) and different fault resistances (0.01Ω to 1000.01Ω with 10Ω interval) are sampled at a frequency of 100kHz.

[0092] 9) Extract t under 1111 different single-pole faults in step 8) p ,δ,t d As the predictor variable of the training data set, the corresponding fault location is used as the response variable of the training data set to train the regression neural network for single-pole fault location. Similarly, the t p ,δ,t d As the predictor variable of the training dataset, the corresponding fault location is used as the response variable of the training dataset to train the regression neural network for bipolar fault location.

[0093] The regression neural network used in the present invention includes an input layer, several fully connected layers and an output layer. The three input prediction variables are defined as O = (o1, o2, o3) T The output vector of the kth fully connected layer is H k =(h k1 ,h k2 ,h k3 ,…,h kj ) T The output variable of the output layer is the fault location x. The data processing process of the regression neural network is shown in formula (3).

[0094]

[0095] ω ij and ω jk are the weights of the fully connected layer and the output layer respectively. θ j and θ k are the biases of the fully connected layer and the output layer respectively. f represents the activation function. A typical neural network structure is as follows Figure 3As shown. All weights, biases and activation functions of the regression neural network are determined by the prediction variables and response variables of the training input. Specifically, the objective function is minimized by the limited memory Broyden-Fletcher-Goldfarb-Shanno quasi-Newton (LBFGS) algorithm to solve all weights and biases of the regression neural network. The objective function consists of a mean square error loss function between the calculated output value and the true value and a ridge penalty term, which can improve the generalization ability of the regression neural network. In addition, the activation function f of the regression neural network, the regularization coefficient λ, the number of fully connected layers, the number of output nodes in each layer, and other hyperparameters involved can be determined by a Bayesian optimization algorithm, where the preset optimization algorithm is not limited to the Bayesian optimization algorithm, and can also be other optimization algorithms.

[0096] The recurrent neural network trained according to the above algorithm has two fully connected layers, which output 293 and 13 variables respectively. The output layer outputs 1 variable. The activation function is a linear rectifier function, as shown in formula (4):

[0097]

[0098] The regularization coefficient λ is 5.07×10 -6 With the trained regression neural network, the estimated value of the objective function is 0.09. The training time is 21.84 seconds on a computer with 12GB of RAM and an i7-10510U CPU.

[0099] 10) Input the voltage characteristic quantity t of the unknown fault location into the trained recurrent neural network p ,δ,t d , the recurrent neural network outputs the fault location.

[0100] Typical simulation test:

[0101] Built on PSCAD / EMTDC Figure 2 The detailed model of the four-terminal bipolar ±500kV flexible DC grid is shown in Figure 2. The specific algorithm flow chart is as follows: Figure 4 The solid blocks at both ends of the line represent the fault location devices FL ij and DC circuit breakers. The DC line adopts the cable frequency-dependent distributed parameter model, such as Figure 5 As shown. Simulated DC line l AB Unipolar and bipolar faults under different fault locations (5% to 95% from the line head end, with an interval of 10%) and different fault resistances (5Ω to 1005Ω, with an interval of 10Ω) are sampled at a frequency of 100kHz. p ,δ,t dAs a test data set, the trained regression neural network is input for fault location. The average error percentage μ of the fault location of the present invention is e and standard deviation σ e As shown in Table 1.

[0102] Table 1 μ under different fault locations e and σ e Value

[0103]

[0104]

[0105] Table 1 shows that for different fault locations and transition resistances, the average error percentages for unipolar and bipolar faults are 0.31% and 0.36%, respectively, with standard deviations of 0.03% and 0.02%, respectively. Therefore, the present invention achieves high positioning accuracy for various fault locations across the entire line and transition resistances up to 1005Ω, demonstrating strong tolerance to transition resistances. For a trained recurrent neural network, the total computation time for the present invention is approximately 1ms on the same computer as described above.

[0106] Generalization ability test:

[0107] Changes in the DC grid voltage level and transmission power do not affect the positioning accuracy of the present invention, as the relationship between the fault location and the three voltage characteristics depends solely on the parameters of the DC transmission line and converter station, as well as the DC grid topology. The present invention is applicable to different voltage levels and transmission powers.

[0108] To verify the generalization ability of the present invention to line parameter deviation, the distance between the two poles and the installation depth of the DC cable are increased by 30%, and the cable length is increased by 10%. The above 2020 fault conditions are simulated in these two system parameters and t p ,δ,t d As a training data set, it is input into the above-trained regression neural network, and the test results are shown in Table 2.

[0109] Table 2 μ under different line parameter deviations e and σ e Value

[0110]

[0111] Table 2 shows that when the DC cable pole spacing and depth increase by 30% or the cable length increases by 10%, the μ of single-pole and double-pole faults increases by 10%. e and σ e Therefore, the present invention can tolerate a relatively large degree of line parameter deviation, and the fault location algorithm has a sufficiently strong generalization capability.

[0112] If the line parameter deviation is greater than this, the fault locator's recurrent neural network needs to be retrained using a new training dataset for that system. This is equivalent to applying the fault locator to a new system, which requires the algorithm's adaptability rather than generalization.

[0113] Adaptive ability test:

[0114] The adaptive capability of the present invention is tested from two aspects: different line lengths and different topological structures.

[0115] In order to test the adaptability of the present invention to different line lengths, Figure 2 The length of all DC lines in the simulation is increased from 200km to 600km. The above 2222 different fault conditions are simulated again, i.e., the DC line l AB Unipolar and bipolar faults at different fault locations (10% interval from the line head to the line end) and different fault resistances (0.01Ω to 1000.01Ω interval 10Ω) are used to generate new training data sets for input into the recurrent neural network and retrained with the same algorithm. Then the above 2020 fault conditions are re-simulated, i.e., the DC line l AB Unipolar and bipolar faults at different fault locations (5% to 95% from the line head end, with an interval of 10%) and different fault resistances (5Ω to 1005Ω with an interval of 10Ω) are used to generate test data sets. The test results are: the average error percentages for unipolar and bipolar faults are 0.91% and 0.94% respectively, and the standard deviations are 0.76% and 0.80% respectively. Since the intervals between fault locations in the same number of training data sets will increase with the increase of line length, the positioning error will increase slightly with the length of the line. However, with 2222 training data sets, the positioning error of the present invention for a 600km transmission line is still less than 1%. If the line length is greater than 600km, the positioning accuracy can be improved by increasing the training data set and reducing the intervals between fault locations in the training data set.

[0116] Secondly, add a 250km transmission line between converter station A and converter station D of the test system, such as Figure 6 As shown in Figure 2 , the 2222 training fault conditions were re-simulated in the test system to train the recurrent neural network, and then the 2020 test fault conditions were simulated for testing. The test results showed that the average errors for single-pole and double-pole faults were 0.34% and 0.32%, respectively, with standard deviations of 0.05% and 0.03%, respectively. Therefore, the present invention is well adapted to systems with different topologies.

[0117] Noise immunity test:

[0118] In order to verify the anti-noise capability of the present invention, white noise with a signal-to-noise ratio of 40 dB was added to the transient voltage of the above 2020 test fault sets, and t p ,δ,t d , is input into the regression neural network trained with the noise-free training data set. The test results are shown in Table 3. It can be seen that under 40dB noise, the positioning error μ e and σ e It is still less than 1%. Therefore, the fault location algorithm has good anti-noise ability.

[0119] Table 3 μ under 40dB noise e and σ e Value

[0120]

[0121] In summary, the present invention has the following characteristics: (1) For single-pole and double-pole faults with different fault locations and different transition resistances, the average error of the fault location algorithm is 0.34% and the standard deviation is 0.03%, which has high accuracy for different fault locations and transition resistances up to 1005Ω. (2) It only requires transient voltage data 2.5ms after the fault, and can be used to protect flexible DC power grids with fast action. (3) The sampling frequency is only 100kHz, and only a single-end voltage signal is required, without the need for two-end communication. (4) It has good generalization ability and still has high positioning accuracy under line parameter deviations within a reasonable range. (5) It has strong noise resistance and can tolerate 40dB white noise.

[0122] Therefore, the present application has the following beneficial effects: (1) For single-pole and double-pole faults with different fault locations and different transition resistances, the present invention has high fault location accuracy and strong transition resistance resistance resistance. (2) The required data time window is short, which can be applied to protect flexible DC power grids with fast action. (3) The sampling frequency is low, only in the kHz range. Only single-end voltage data is required, and no two-end communication is required. (4) It has good generalization ability and still has high positioning accuracy under line parameter deviations within a reasonable range. (5) It has strong anti-noise ability.

[0123] Exemplary devices

[0124] Figure 7 FIG is a schematic diagram of a DC transmission line fault location device based on a neural network provided by an exemplary embodiment of the present invention. Figure 7 As shown, the apparatus 700 includes:

[0125] A detection and storage module 710 is used to detect whether a DC line fault exists, and if a DC line fault is detected, store single-ended voltage sampling values ​​within a predetermined time period before and after the DC line fault;

[0126] The determination and adjustment module 720 is used to determine whether the DC line fault is a negative pole fault. If the DC line fault is a negative pole fault, the single-ended voltage sampling value is multiplied by -1.

[0127] The smoothing module 730 is used to smooth the single-ended voltage sampling value to determine a voltage smoothing sampling value;

[0128] A calculation module 740 is configured to calculate the first negative peak time, the first negative peak value, and the delay time during the DC line fault period based on the voltage smoothing sampling value;

[0129] The prediction module 750 is used to input the first negative peak time, the first negative peak and the delay time into a pre-built regression neural network model, and output the predicted fault location of the DC transmission line.

[0130] Optionally, the predetermined time period in the determination and adjustment module 720 is 0.1 ms before the DC line fault to 2.5 ms after the fault.

[0131] Optionally, the calculation formula of the voltage smoothing sampling value in the smoothing module 730 is:

[0132]

[0133] Where U fm is the single-ended voltage sampling value, U sfm [t] is the voltage sampling smoothing value at any time t, and the sliding average window of the smoothing process is 30us.

[0134] Optionally, the calculation module 740 includes:

[0135] The average value calculation submodule is used to calculate the pre-fault voltage average value of the DC line fault based on the voltage smoothing sampling value;

[0136] The extraction submodule is used to extract the first negative peak time and the first negative peak corresponding to the moment in the voltage smoothing sampling value;

[0137] The estimation submodule is used to estimate the delay time of the voltage smoothing sampling value of the DC line fault according to the time of the average value of the voltage before the fault and the average value of the first negative peak.

[0138] Optionally, the process of constructing the regression neural network model in the prediction module 750 is as follows:

[0139] A simulation submodule is used to simulate different fault conditions of the DC line, including single-pole and double-pole faults at different fault locations and fault resistances.

[0140] The extraction submodule is used to extract voltage characteristics under different fault conditions, wherein the voltage characteristics include: the first negative peak value of the fault, the time of the first negative peak value, and the delay time;

[0141] A training submodule is used to train a regression neural network model based on the voltage characteristic quantities under different fault conditions and the corresponding fault locations, wherein all weights and biases of the regression neural network model are solved according to the minimization objective function, and the hyperparameters of the regression neural network are determined according to a preset optimization algorithm, wherein the hyperparameters include: an activation function f, a regularization coefficient λ, and one or more of the number of fully connected layers and the number of output nodes in each layer.

[0142] Optionally, the data processing process of the regression neural network model is:

[0143]

[0144] Where, ω ij and ω kj’ They are the first layer of the fully connected layer h 1j and the kth layer h kj’ The weight, ω j’ is the weight of the output layer x; θ 1j and θ kj’ are the biases of the first and kth layers of the fully connected layer, respectively, and θ1 is the bias of the output layer x; the fully connected layer has K layers, the first layer has N output nodes, and the kth layer has M output nodes; f represents the activation function; the prediction variables of the three inputs are defined as O = (o1, o2, o3) T ; The output vector of the kth fully connected layer is H k =(h k1 ,h k2 ,h k3 ,…,h kj ) T ;The output variable of the output layer is the fault location x.

[0145] Exemplary electronic devices

[0146] Figure 8 This is the structure of an electronic device provided by an exemplary embodiment of the present invention. Figure 8 As shown, the electronic device 80 includes one or more processors 81 and a memory 82 .

[0147] The processor 81 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0148] The memory 82 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may, for example, include read-only memory (ROM), a hard disk, a flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 81 may execute the program instructions to implement the methods of the software programs of the various embodiments of the present invention described above and / or other desired functions. In one example, the electronic device may further include: an input device 83 and an output device 84, which are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0149] In addition, the input device 83 may also include, for example, a keyboard, a mouse, etc.

[0150] The output device 84 can output various information to the outside. The output device 84 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto.

[0151] Of course, to simplify, Figure 8 Only some of the components related to the present invention in the electronic device are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, the electronic device may further include any other appropriate components according to specific application conditions.

[0152] Exemplary computer program products and computer-readable storage media

[0153] In addition to the above-mentioned methods and devices, an embodiment of the present invention may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to perform the steps of the method according to various embodiments of the present invention described in the above "Exemplary Method" section of this specification.

[0154] The computer program product may be written in any combination of one or more programming languages ​​to implement the operations of embodiments of the present invention, including object-oriented programming languages ​​such as Java, C++, and conventional procedural programming languages ​​such as C or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's computing device, as a stand-alone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0155] In addition, an embodiment of the present invention may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, causes the processor to execute the steps of the method according to various embodiments of the present invention described in the above "Exemplary Method" section of this specification.

[0156] The computer-readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can, for example, include but is not limited to a system, system or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0157] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in the present invention are merely illustrative and non-limiting, and should not be construed as necessarily possessed by each embodiment of the present invention. Furthermore, the specific details disclosed above are provided for illustrative purposes and to facilitate understanding, and are not intended to be limiting. These details do not necessarily limit the present invention to being implemented using these specific details.

[0158] Each embodiment in this specification is described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. References to the same or similar parts between the various embodiments are sufficient. For system embodiments, since they largely correspond to method embodiments, their description is relatively simple. For relevant parts, references to the description of the method embodiments are sufficient.

[0159] The block diagrams of the devices, systems, equipment, and systems involved in the present invention are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, systems, equipment, and systems can be connected, arranged, or configured in any manner. Words such as "including," "comprising," "having," and the like are open-ended words, meaning "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or" and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0160] The method and system of the present invention may be implemented in many ways. For example, the method and system of the present invention may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above sequence of steps for the method is for illustration only, and the steps of the method of the present invention are not limited to the sequence specifically described above, unless otherwise specified. In addition, in some embodiments, the present invention may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present invention. Thus, the present invention also covers recording media that store programs for executing the method according to the present invention.

[0161] It should also be noted that, in the system, device and method of the present invention, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. The above description of the disclosed aspects is provided to enable any technician in this field to make or use the present invention. Various modifications to these aspects will be very obvious to those skilled in the art, and the general principles defined here can be applied to other aspects without departing from the scope of the present invention. Therefore, the present invention is not intended to be limited to the aspects shown here, but according to the widest scope consistent with the principles disclosed here and novel features.

[0162] The above description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the present invention to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A DC transmission line fault location method based on neural network, characterized in that: include: Detecting whether a DC line fault exists, and if a DC line fault is detected, storing single-ended voltage sampling values ​​within a predetermined time period before and after the DC line fault; determining whether the DC line fault is a negative pole fault, and if the DC line fault is a negative pole fault, multiplying the single-ended voltage sample value by -1; Smoothing the single-ended voltage sampling value to determine a voltage smoothing sampling value; Calculating the first negative peak time, the first negative peak value, and the delay time during the DC line fault period according to the voltage smoothing sampling value; The first negative peak time, the first negative peak value, and the delay time are input into a pre-built regression neural network model, and a predicted fault location of the direct current transmission line is output.

2. The method according to claim 1, characterized in that The predetermined time period is from 0.1 ms before the DC line fault to 2.5 ms after the fault.

3. The method according to claim 1, characterized in that The calculation formula of the voltage smoothing sampling value is: Where U fm is the single-ended voltage sampling value, U sfm [t] is the voltage sampling smoothing value at any time t, and the sliding average window of the smoothing process is 30us.

4. The method according to claim 1, wherein Calculating the first negative peak time, the first negative peak value, and the delay time during the DC line fault period according to the voltage smoothing sampling value, including: Calculating a pre-fault voltage average value of the DC line fault according to the voltage smoothing sampling value; Extracting the first negative peak time and the first negative peak corresponding to the time in the voltage smoothing sampling value; The delay time of the voltage smoothing sampling value of the DC line fault is estimated according to the time at which the average value of the voltage before the fault and the average value of the first negative peak value are located.

5. The method according to claim 1, wherein The construction process of the regression neural network model is as follows: Simulating different fault conditions of a DC line, wherein the different fault conditions include: single-pole and double-pole faults at different fault locations and different fault resistances; Extracting voltage characteristics under different fault conditions, wherein the voltage characteristics include: the first negative peak value of the fault, the time of the first negative peak value, and the delay time; The regression neural network model is trained according to the voltage characteristic quantities and the corresponding fault locations under the different fault conditions, wherein all weights and biases of the regression neural network model are solved according to the minimization objective function, and the hyperparameters of the regression neural network are determined according to a preset optimization algorithm, wherein the hyperparameters include: an activation function f, a regularization coefficient λ, and one or more of the number of fully connected layers and the number of output nodes in each layer.

6. The method according to claim 5, characterized in that The data processing process of the regression neural network model is as follows: Where, ω ij and ω kj’ They are the first layer of the fully connected layer h 1j and the kth layer h kj’ The weight, ω j’ is the weight of the output layer x; θ 1j and θ kj’ are the biases of the first and kth layers of the fully connected layer, respectively, and θ1 is the bias of the output layer x; the fully connected layer has K layers, the first layer has N output nodes, and the kth layer has M output nodes; f represents the activation function; the prediction variables of the three inputs are defined as O = (o1, o2, o3) T ; The output vector of the kth fully connected layer is H k =(h k1 ,h k2 ,h k3 ,…,h kj ) T ;The output variable of the output layer is the fault location x.

7. A DC transmission line fault location device based on neural network, characterized in that: include: a detection and storage module, configured to detect whether a DC line fault exists, and, if a DC line fault is detected, store single-ended voltage sampling values ​​within a predetermined time period before and after the DC line fault; a determination and adjustment module, configured to determine whether the DC line fault is a negative pole fault, and multiply the single-ended voltage sampling value by -1 if the DC line fault is a negative pole fault; a smoothing module, configured to smooth the single-ended voltage sampling value to determine a voltage smoothing sampling value; a calculation module, configured to calculate the first negative peak time, the first negative peak value, and the delay time during the DC line fault period according to the voltage smoothing sampling value; The prediction module is used to input the first negative peak time, the first negative peak value and the delay time into a pre-built regression neural network model, and output the predicted fault location of the direct current transmission line.

8. The device according to claim 7, characterized in that The predetermined time period in the detection storage module is 0.1 ms before the DC line fault to 2.5 ms after the fault.

9. The device according to claim 7, characterized in that The calculation formula of the voltage smoothing sampling value in the smoothing module is: Where U fm is the single-ended voltage sampling value, U sfm [t] is the voltage sampling smoothing value at any time t, and the sliding average window of the smoothing process is 30us.

10. The device according to claim 7, characterized in that Computing module, including: an average value calculation submodule, configured to calculate the pre-fault voltage average value of the DC line fault according to the voltage smoothing sampling value; An extraction submodule, configured to extract the time of the first negative peak value and the first negative peak value corresponding to the time in the voltage smoothing sampling value; The estimation submodule is configured to estimate the delay time of the voltage smoothing sampling value of the DC line fault according to the time at which the average value of the voltage before the fault and the average value of the first negative peak value are located.

11. The device according to claim 7, characterized in that The construction process of the regression neural network model in the prediction module is as follows: A simulation submodule, configured to simulate different fault conditions of a DC line, wherein the different fault conditions include: single-pole and double-pole faults at different fault locations and different fault resistances; An extraction submodule is used to extract voltage characteristics under different fault conditions, wherein the voltage characteristics include: the first negative peak value of the fault, the time of the first negative peak value, and the delay time; A training submodule is used to train the regression neural network model according to the voltage characteristic quantities under the different fault conditions and the corresponding fault locations, wherein all weights and biases of the regression neural network model are solved according to the minimization objective function, and the hyperparameters of the regression neural network are determined according to a preset optimization algorithm, wherein the hyperparameters include: an activation function f, a regularization coefficient λ, and one or more of the number of fully connected layers and the number of output nodes in each layer.

12. The device according to claim 11, characterized in that The data processing process of the regression neural network model is as follows: Where, ω ij and ω kj ' are the first layer of the fully connected layer h 1j and the kth layer h kj’ The weight, ω j ' is the weight of the output layer x; θ 1j and θ kj’ are the biases of the first and kth layers of the fully connected layer, respectively, and θ1 is the bias of the output layer x; the fully connected layer has K layers, the first layer has N output nodes, and the kth layer has M output nodes; f represents the activation function; the prediction variables of the three inputs are defined as O = (o1, o2, o3) T ; The output vector of the kth fully connected layer is H k =(h k1 ,h k2 ,h k3 ,…,h kj ) T ;The output variable of the output layer is the fault location x.

13. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and the computer program is used to execute the method according to any one of claims 1 to 6.

14. An electronic device, characterized in that: The electronic device comprises: a processor, a memory for storing instructions executable by the processor; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method according to any one of claims 1 to 6.

Images