Method for identifying faults in a cable-type distribution network based on a convolution signal energy ratio curve

Abstract

The application discloses a cable type power distribution network fault identification method based on a convolution signal energy ratio curve, which comprises the following steps: acquiring a ground line current and a zero sequence current of a first section of a power distribution network; calculating a convolution signal of the ground line current and the zero sequence current of each section; calculating an energy ratio characteristic curve of the convolution signal of the first end of each section by using a sliding window; and then respectively judging whether the energy ratio curve of the first section is smaller than a set negative threshold Th1, whether a starting data point of the energy ratio curve of each section reaches a set threshold Th2, whether a high-energy state duration is smaller than a set threshold Th3, and whether a time difference between an end time of the high-energy state and an end time of the energy ratio curve is smaller than a set threshold Th4, so that precise distinction and reliable identification of a broken line fault, a disturbance, an early fault and a serious fault are realized. The application can accurately and reliably distinguish the broken line fault, the disturbance, the early fault and the serious fault, and improves the operation reliability of the power distribution network.

Description

Technical Field

[0001] This invention belongs to the field of distribution network fault identification technology, specifically referring to a cable-type distribution network fault identification method based on the energy ratio curve of convolutional signals. Background Technology

[0002] As a crucial link connecting the power transmission system and users, the safe and stable operation of the power distribution network is of great significance. In recent years, with the continuous advancement of urbanization and industrialization in my country, the scale of the power system has been expanding and its structure becoming increasingly complex. In urban power supply networks, the proportion of power cables has increased significantly, gradually replacing the original overhead lines.

[0003] Currently, fault identification in distribution networks can be mainly categorized into three types: time-domain methods, frequency-domain methods, and artificial intelligence methods. Time-domain methods detect faults by analyzing the time-domain changes in electrical quantities such as voltage and current before and after a fault. However, they face difficulties in identifying faults under conditions of high-resistance grounding and early-stage faults due to the insignificant changes in electrical quantities. Frequency-domain methods utilize the harmonic characteristics of fault signals in the high-frequency band to compensate for the weak fault characteristics in time-domain methods. However, this method is easily affected by disturbances under weak arc faults, making it difficult to accurately distinguish fault arcs. In recent years, with the development of artificial intelligence technology, data-driven methods have gradually gained attention. However, these methods rely on a large amount of fault data for model training, and the limited availability of data in practice restricts their feasibility for engineering applications.

[0004] Therefore, there is an urgent need to develop an accurate and reliable fault identification technology to support subsequent processing steps such as fault reconstruction, fault selection, fault location and clearing, and to further improve the reliability of power distribution network operation. Summary of the Invention

[0005] The purpose of this invention is to provide a fault identification method for cable distribution networks based on the energy ratio curve of convolutional signals. This method can reliably and accurately identify open-circuit faults, disturbances, early faults, and serious fault events, thereby improving the operational reliability of the distribution network.

[0006] To achieve the above objectives, the present invention provides a cable-type distribution network fault identification method based on convolutional signal energy ratio curves, comprising the following steps:

[0007] Step 1: Obtain the time-series data of the grounding wire current and zero-sequence current at the beginning of section i of the entire cable distribution network for a total of seven cycles, namely I, in the two cycles before the fault and the five cycles after the fault. g(i) (t), I 0(i) (t), where i = 1, 2, 3, ..., N R N R This represents the total number of network segments in the distribution network. The original signal sampling interval is Δt, and the total number of original signal sampling points is N.

[0008] Step 2: For the grounding wire current signal I at the beginning of section i... g(i) (t) and zero-sequence current signal I 0(i) (t) Perform a complete convolution calculation to obtain the convolution signal C of segment i. (i) (t);

[0009] Step 3: Convolve signal C with segment i (i) (t) The signal with the initial window length is used as a reference to calculate the reference energy value, and a sliding window is used to calculate the convolution signal C of segment i. (i) (t) is the energy ratio curve at the beginning of the segment;

[0010] Step 4: If any section's first-end energy ratio curve has three consecutive points less than the set negative threshold Th1, then the event is determined to be a disconnection fault; otherwise, it is not a disconnection fault, and further judgment is performed.

[0011] Step 5: If the starting point of the energy ratio curve at the beginning of each section reaches the set threshold Th2, then the event is judged to be a disturbance; otherwise, it is not a disturbance, and subsequent judgment is performed.

[0012] Step 6: Calculate the duration of the high-energy state in the energy ratio curve at the beginning of each section;

[0013] Step 7: If the duration of the high-energy state in each section is less than the set threshold Th3, then the event is determined to be a short-cycle early fault; otherwise, it is not a short-cycle early fault, and further judgment is performed.

[0014] Step 8: Calculate the time difference between the end of the high-energy state and the end of the energy ratio curve at the beginning of the section. If the time difference between the end of the high-energy state and the end of the energy ratio curve at the beginning of the section is less than the set threshold Th4, then the event is judged as a serious fault; otherwise, it is a long-cycle early fault.

[0015] As a further aspect of the present invention: In step 3, the length of the sliding window is set to N. w The sliding step size is S w ;

[0016] Extracting segment i of the convolution signal C (i) (t) Initial length N w Using the signal as a reference, calculate the reference energy value:

[0017]

[0018] Among them, C (i) (k) is the convolution signal C of segment i. (i) (n) The amplitude at the k-th sampling point;

[0019] For a length of NC The convolutional signal sequence C (i) The formula for calculating the total number of sliding windows J is:

[0020]

[0021] The starting position of the j-th window is pos j for:

[0022] (j=1, 2, 3, ..., J)

[0023] The formula for calculating the energy ratio of the j-th window in segment i is as follows:

[0024]

[0025] Among them, C j(i) =[C (i) (pos j ), C (i) (pos j +1), C (i) (pos j +2), ..., C (i) (pos j +N w -1)];

[0026] Each energy ratio ER j(i) The corresponding time point is the moment when the center point of the window is located. The specific calculation formula is as follows:

[0027]

[0028] Finally, the signal C is convolved by traversing segment i. (i) (t) Construct the complete segment head-end energy ratio curve ER at all window positions. (i) (t)=[ER 1(i) ER 2(i) ER 3(i) , ..., ER J(i) ](t=t ER1(i) , t ER2(i) , t ER3(i) , ..., t ERJ(i) ).

[0029] As a further aspect of the present invention: the fault criterion for disconnection in step 4 is as follows:

[0030] The energy ratio curve ER at the beginning of the i-th segment (i) (i=1, 2, ..., N) R Set a negative threshold Th1 and determine the result according to the following procedure:

[0031] The disconnection fault flag F for the i-th segment break(i) The calculation is as follows:

[0032]

[0033] Among them, Th1 takes a value between -0.2 and -0.1;

[0034] Distribution network open circuit fault diagnosis:

[0035]

[0036] In this context, "1" represents a power distribution network outage fault, while "0" represents no outage fault.

[0037] As a further aspect of the present invention: In step 5, regarding the disturbance criterion:

[0038] The disturbance flag F of the i-th segment dis(i) The calculation is as follows:

[0039]

[0040] The formula for calculating Th2 is as follows:

[0041] TH2=β*max(ER (i) (t))

[0042] Where max(ER) (i) (t) represents the maximum value of the energy ratio curve at the beginning of the i-th segment, and β takes values ​​from 0.2 to 0.5;

[0043] Distribution network disturbance determination:

[0044]

[0045] In this context, "2" indicates that a disturbance has occurred in the power distribution network, while "0" indicates that no disturbance has occurred.

[0046] As a further aspect of the present invention: In step 6, the duration of the high-energy state of the energy ratio curve at the beginning of the i-th segment is calculated using a sliding window mechanism, and the specific steps are as follows:

[0047] The time window length is W h The sliding step size is S h ;

[0048] The starting index of the z-th time window is:

[0049] Start z(i) =t ER1(i) +(z-1)*S h

[0050] The energy ratio curve data at the beginning of the i-th segment contained in the z-th time window is:

[0051] Window z(i) ={ER (i) (t)|Start z(i) ≤ t ≤Start z(i) +W h}

[0052] Calculate the average value μ of the energy ratio curve data at the beginning of the i-th segment within the z-th time window. z(i) ;

[0053] High-energy state determination:

[0054] If the average value of the z-th time window is μ z(i) Greater than E th If , then the time window is determined to be a high-energy state time window, where E th The calculation formula is as follows:

[0055] E th =α*max(ER (i) (t))

[0056] Where α takes values ​​from 0.7 to 0.8;

[0057] Collect all high-energy state time windows. If the difference between the starting indices of two adjacent time windows is less than W... h Then, these two time windows are merged into a single continuous time window, with its starting time t. ER_st(i) The end time is t ER_end(i) ;

[0058] The formula for calculating the duration of the high-energy state of the energy ratio curve at the beginning of the i-th segment is:

[0059] T high(i) =t ER_end(i) -t ER_st(i) .

[0060] As a further aspect of the present invention: In step 7, the short-cycle early fault criterion is as follows:

[0061] The short-cycle early fault flag F of the i-th segment short(i) The calculation is as follows:

[0062]

[0063] Among them, Th3 is calculated based on the simulation system. The simulation system is used to simulate the short-cycle arc fault of the cable, calculates the duration of the high-energy state in each section, and calculates its average value.

[0064] Distribution network short-cycle early fault flag Fshort The calculation is as follows:

[0065]

[0066] In this context, "3" represents a short-cycle early fault that occurred in the distribution network, while "0" represents no short-cycle early fault that occurred.

[0067] As a further aspect of the present invention: in step 8, for the i-th segment, the time difference between the end time of the high-energy state and the end time of the energy ratio curve at the beginning of the segment is calculated, and the specific formula is as follows:

[0068] Δt (i) =t ERJ(i) - t ER_end(i)

[0069] The critical fault flag F of the i-th segment spg(i) The calculation is as follows:

[0070]

[0071] Among them, Th4 takes a value between 0.01 and 0.02 s;

[0072] Distribution network serious fault indicator F g The calculation is as follows:

[0073]

[0074] Among them, "4" represents a serious fault in the distribution network, and "5" represents an early-stage fault with a long cycle.

[0075] Compared with existing technologies, this invention amplifies the transient characteristics of faults through the convolution operation of grounding current and zero-sequence current, and constructs an energy ratio curve using a sliding window. It introduces multiple fault criteria and integrates them to construct an identification logic, achieving accurate differentiation and reliable identification of open-circuit faults, disturbances, early faults, and severe fault events. This method utilizes only conventional waveform data, requiring no additional hardware or complex transformations, effectively overcoming the engineering challenges of low sensitivity and easy misjudgment in traditional methods when dealing with high-resistance grounding faults and early faults, thus improving the operational reliability of the distribution network. Attached Figure Description

[0076] Figure 1 This is a flowchart illustrating the method of the present invention.

[0077] Figure 2 This is the topology diagram of the 10 kV distribution network fault simulation model used in this embodiment of the invention.

[0078] Figure 3These are amplitude diagrams of the convolution signal between zero-sequence current and grounding current in each section of this invention, from top to bottom: amplitude diagrams of cable L1 to cable L9.

[0079] Figure 4 These are energy ratio curves for each section in this embodiment of the invention, from top to bottom: energy ratio curves for cable L1 to cable L9. Detailed Implementation

[0080] The invention will now be further described with reference to the accompanying drawings.

[0081] like Figure 1 As shown, the fault identification method for cable distribution networks based on the energy ratio curve of convolutional signals includes the following steps: Step 1: Obtain the time-series data of the grounding wire current and zero-sequence current at the beginning of section i of the entire cable distribution network for a total of seven cycles, namely I, in the two cycles before the fault and the five cycles after the fault. g(i) (t), I 0(i) (t)(i=1, 2, 3, ..., N) R (N) R (Represents the total number of network segments in the distribution network), the original signal sampling interval is Δt, and the total number of original signal sampling points is N.

[0082] Step 2: For the grounding wire current signal I at the beginning of section i... g(i) (t) and zero-sequence current signal I 0(i) (t) Perform a complete convolution calculation to obtain the convolution signal C of segment i. (i) (t).

[0083] The specific formula for calculating convolution is as follows:

[0084] (n=0,1,2…2N-2)

[0085] The time axis corresponding to the convolution signal is determined by the following formula:

[0086]

[0087] Step 3: Convolve signal C with segment i (i) (t) The signal with the initial window length is used as a reference to calculate the reference energy value, and a sliding window is used to calculate the convolution signal C of segment i. (i) The energy ratio curve at the beginning of the section (t).

[0088] Let the length of the sliding window be N. w The sliding step size is S w .

[0089] Extracting segment i of the convolution signal C (i) (t) Initial length N wUsing the signal as a reference, calculate the reference energy value:

[0090]

[0091] Among them, C (i) (k) is the convolution signal C of segment i. (i) (n) is the amplitude at the kth sampling point.

[0092] For a length of N C The convolutional signal sequence C (i) The formula for calculating the total number of sliding windows J is:

[0093]

[0094] The starting position of the j-th window is pos j for:

[0095] (j=1, 2, 3, ..., J)

[0096] The formula for calculating the energy ratio of the j-th window in segment i is as follows:

[0097]

[0098] Among them, C j(i) =[C (i) (pos j ), C (i) (pos j +1), C (i) (pos j +2), ..., C (i) (pos j +N w -1)].

[0099] Each energy ratio ER j(i) The corresponding time point is the moment when the center point of the window is located. The specific calculation formula is as follows:

[0100]

[0101] Finally, the signal C is convolved by traversing segment i. (i) (t) Construct the complete segment head-end energy ratio curve ER at all window positions. (i) (t)=[ER 1(i) ER 2(i) ER 3(i) , ..., ER J(i) ](t=t ER1(i) , t ER2(i) , t ER3(i) , ..., t ERJ(i) ).

[0102] Step 4: If any section's first-end energy ratio curve has three consecutive points less than the set negative threshold Th1, then the event is determined to be a disconnection fault; otherwise, it is not a disconnection fault, and further judgment is performed.

[0103] Criteria for determining wire breakage faults:

[0104] The energy ratio curve ER at the beginning of the i-th segment (i) (i=1, 2, ..., N) R Set a negative threshold Th1 and determine the result according to the following procedure:

[0105] The disconnection fault flag F for the i-th segment break(i) The calculation is as follows:

[0106]

[0107] The value of Th1 is typically between -0.2 and -0.1.

[0108] Distribution network open circuit fault diagnosis:

[0109]

[0110] In this context, "1" represents a power distribution network outage fault, while "0" represents no outage fault.

[0111] Step 5: If the starting point of the energy ratio curve at the beginning of each section reaches the set threshold Th2, then the event is judged to be a disturbance; otherwise, it is not a disturbance, and subsequent judgment is performed.

[0112] Regarding the perturbation criterion:

[0113] The disturbance flag F of the i-th segment dis(i) The calculation is as follows:

[0114]

[0115] The formula for calculating Th2 is as follows:

[0116] TH2=β*max(ER (i) (t))

[0117] Where max(ER) (i) (t) represents the maximum value of the energy ratio curve at the beginning of the i-th segment, and β is generally taken as 0.2 to 0.5.

[0118] Distribution network disturbance determination:

[0119]

[0120] In this context, "2" indicates that a disturbance has occurred in the power distribution network, while "0" indicates that no disturbance has occurred.

[0121] Step 6: Calculate the duration of the high-energy state of the energy ratio curve at the beginning of each section.

[0122] The high-energy state duration of the energy ratio curve at the beginning of the i-th segment is calculated using a sliding window mechanism. The specific steps are as follows:

[0123] The time window length is W h The sliding step size is S h .

[0124] The starting index of the z-th time window is:

[0125] Start z(i) =t ER1(i) +(z-1)*S h

[0126] The energy ratio curve data at the beginning of the i-th segment contained in the z-th time window is:

[0127] Window z(i) ={ER (i) (t)|Start z(i) ≤ t ≤Start z(i) +W h}

[0128] Calculate the average value μ of the energy ratio curve data at the beginning of the i-th segment within the z-th time window. z(i) .

[0129] High-energy state determination:

[0130] If the average value of the z-th time window is μ z(i) Greater than E th If so, then the time window is determined to be a high-energy state time window. Where E th The calculation formula is as follows:

[0131] E th =α*max(ER (i) (t))

[0132] In this case, α is typically taken as 0.7 to 0.8.

[0133] Collect all high-energy state time windows. If the difference between the starting indices of two adjacent time windows is less than W... h Then, these two time windows are merged into a single continuous time window, with its starting time t. ER_st(i) The end time is t ER_end(i) .

[0134] The formula for calculating the duration of the high-energy state of the energy ratio curve at the beginning of the i-th segment is:

[0135] T high(i) =t ER_end(i) -t ER_st(i)

[0136] Step 7: If the duration of the high-energy state in each section is less than the set threshold Th3, then the event is determined to be a short-cycle early fault; otherwise, it is not a short-cycle early fault, and further judgment is performed.

[0137] For short-cycle early failure criteria:

[0138] The short-cycle early fault flag F of the i-th segment short(i) The calculation is as follows:

[0139]

[0140] Among them, Th3 can be calculated based on a simulation system. A short-cycle arc fault simulation of the cable is performed using the simulation system to calculate the duration of the high-energy state in each section, and then the average value is obtained.

[0141] Distribution network short-cycle early fault flag F short The calculation is as follows:

[0142]

[0143] In this context, "3" represents a short-cycle early fault that occurred in the distribution network, while "0" represents no short-cycle early fault that occurred.

[0144] Step 8: Calculate the time difference between the end of the high-energy state and the end of the energy ratio curve at the beginning of the section. If the time difference between the end of the high-energy state and the end of the energy ratio curve at the beginning of the section is less than the set threshold Th4, then the event is judged as a serious fault; otherwise, it is a long-cycle early fault.

[0145] For the i-th segment, calculate the time difference between the end of the high-energy state and the end of the energy ratio curve at the beginning of the segment, using the following formula:

[0146] Δt (i) =t ERJ(i) - t ER_end(i)

[0147] The critical fault flag F of the i-th segment spg(i) The calculation is as follows:

[0148]

[0149] Among them, Th4 typically takes a value between 0.01 and 0.02 s.

[0150] Distribution network serious fault indicator F g The calculation is as follows:

[0151]

[0152] Among them, "4" represents a serious fault in the distribution network, and "5" represents an early-stage fault with a long cycle.

[0153] To verify the effectiveness of the proposed fault identification method, a system was built in PSCAD / EMTDC as follows: Figure 2 The diagram shows a low-resistance grounded distribution network. The system comprises nine cable feeders (L1 to L9), using the YJV22-3x70 cable as the reference cable type, all employing a frequency-dependent model. Cable lengths and loads are labeled in the diagram. Switch S1 is initially open, and switch S2 is initially closed. The sheaths of each cable are grounded at both ends through a 2Ω resistor, and the neutral point grounding resistance is 10Ω. Specific cable parameters are shown in Table 1. The sampling rate is 4kHz. Early faults are simulated using a Cassie model and a fixed resistor in series.

[0154] Table 1 Parameters of 10 kV Three-Core Cable

[0155]

[0156] Set cable L2 to experience an early fault, with the fault location 3km from the beginning, the duration 0.005s, and the fixed resistance 50Ω;

[0157] After the convolution calculation in step 2, the convolution signals of the zero-sequence current and the grounding wire current in each section are obtained as follows: Figure 3 As shown.

[0158] The energy ratio curves for each section obtained from step 3 are as follows: Figure 4 As shown.

[0159] Set TH1 to -0.2; β to 0.5; set TH3 to 0.05; set TH4 to 0.01.

[0160] According to step 4, no section of the energy ratio curve in this event has three consecutive points below the set negative threshold. Therefore, this event is not a disconnection fault.

[0161] According to step 5, the starting points of the energy ratio curves for each section are 0.774254, 0.716105, 0.99938, 0.962042, 0.961128, 0.955087, 0.885391, 0.888284, and 0.881848 (from cable L1 to cable L9, respectively). Comparing these points with TH2 for each section (7.8, 7.7, 9.05, 8.7, 8.7, 8.9, 8.7, 8.7, 8.8, from cable L1 to cable L9, respectively), it can be seen that the starting point of the energy ratio curve at the beginning of each section has not reached the set threshold Th2. Therefore, this event is not a disturbance.

[0162] Following step 6, the durations of the high-energy states of the energy ratio curves at the beginning of each section are calculated to be 0.0350s, 0.0450s, 0.0300s, 0.0300s, 0.0350s, 0.0300s, 0.0350s, 0.0300s, and 0.0350s, respectively.

[0163] According to step 7, the duration of the high-energy state of the energy ratio curves at the beginning of each section obtained in step 6 is less than the set threshold TH3. Therefore, this event is a short-cycle early fault.

[0164] The final conclusion was that this incident was a short-cycle, early-stage failure.

[0165] This invention addresses the problem of insufficient accuracy in traditional fault identification methods due to the weak early fault characteristics and complex and diverse fault types in distribution networks. It proposes a fault identification method for cable-type distribution networks based on the energy ratio curve of convolutional signals. The method amplifies the transient characteristics of faults through the convolution operation of grounding current and zero-sequence current, and constructs an energy ratio curve using a sliding window. It innovatively introduces multiple fault criteria and integrates them to construct the identification logic, achieving accurate differentiation and identification of events such as disturbances, severe faults, early faults, and line breakages. This invention utilizes only conventional waveform data, requiring no additional hardware or complex transformations. It effectively overcomes the engineering challenges of low sensitivity and easy misjudgment in traditional methods when dealing with high-resistance grounding faults and early faults. The principle is clear, the steps are well-defined, and the computational load is small, demonstrating outstanding engineering practicality and field application value.

Claims

1. A method for fault identification in cable-type distribution networks based on convolutional signal energy ratio curves, characterized in that, Includes the following steps: Step 1: Obtain the time-series data of the grounding wire current and zero-sequence current at the beginning of section i of the entire cable distribution network for a total of seven cycles, namely I, in the two cycles before the fault and the five cycles after the fault. g(i) (t), I 0(i) (t), where i = 1, 2, 3, ..., N R N R This represents the total number of network segments in the distribution network. The original signal sampling interval is Δt, and the total number of original signal sampling points is N. Step 2: For the grounding wire current signal I at the beginning of section i... g(i) (t) and zero-sequence current signal I 0(i) (t) Perform a complete convolution calculation to obtain the convolution signal C of segment i. (i) (t); Step 3: Convolve signal C with segment i (i) (t) The signal with the initial window length is used as a reference to calculate the reference energy value, and a sliding window is used to calculate the convolution signal C of segment i. (i) (t) is the energy ratio curve at the beginning of the segment; Step 4: If any section's first-end energy ratio curve has three consecutive points less than the set negative threshold Th1, then the event is determined to be a disconnection fault; otherwise, it is not a disconnection fault, and further judgment is performed. Step 5: If the starting point of the energy ratio curve at the beginning of each section reaches the set threshold Th2, then the event is determined to be a disturbance; otherwise, it is not a disturbance, and subsequent judgments are made. Step 6: Calculate the duration of the high-energy state in the energy ratio curve at the beginning of each section; Step 7: If the duration of the high-energy state in each section is less than the set threshold Th3, then the event is determined to be a short-cycle early fault; otherwise, it is not a short-cycle early fault, and further judgment is performed. Step 8: Calculate the time difference between the end of the high-energy state and the end of the energy ratio curve at the beginning of the section. If the time difference between the end of the high-energy state and the end of the energy ratio curve at the beginning of the section is less than the set threshold Th4, then the event is judged as a serious fault; otherwise, it is a long-cycle early fault.

2. The cable-type distribution network fault identification method based on convolutional signal energy ratio curve according to claim 1, characterized in that, In step 3, let the length of the sliding window be N. w The sliding step size is S w ; Extracting segment i of the convolution signal C (i) (t) Initial length N w Using the signal as a reference, calculate the reference energy value: Among them, C (i) (k) is the convolution signal C of segment i. (i) (n) The amplitude at the k-th sampling point; For a length of N C The convolutional signal sequence C (i) The formula for calculating the total number of sliding windows J is: The starting position of the j-th window is pos j for: (j=1, 2, 3, ..., J) The formula for calculating the energy ratio of the j-th window in segment i is as follows: Among them, C j(i) =[C (i) (pos j ), C (i) (pos j +1), C (i) (pos j +2), ..., C (i) (pos j +N w -1)]; Each energy ratio ER j(i) The corresponding time point is the moment when the center point of the window is located. The specific calculation formula is as follows: Finally, the signal C is convolved by traversing segment i. (i) (t) Construct the complete segment head-end energy ratio curve ER at all window positions. (i) (t)=[ER 1(i) ER 2(i) ER 3(i) , ..., ER J(i) ](t=t ER1(i) , t ER2(i) , t ER3(i) , ..., t ERJ(i) ).

3. The cable-type distribution network fault identification method based on convolutional signal energy ratio curves according to claim 1, characterized in that, The fault criterion for disconnection in step 4 is as follows: The energy ratio curve ER at the beginning of the i-th segment (i) (i=1, 2, ..., N) R Set a negative threshold Th1 and determine the result according to the following procedure: The disconnection fault flag F for the i-th segment break(i) The calculation is as follows: Among them, Th1 takes a value between -0.2 and -0.1; Distribution network open circuit fault diagnosis: In this context, "1" represents a power distribution network outage, and "0" represents no outage.

4. The cable-type distribution network fault identification method based on convolutional signal energy ratio curve according to claim 1, characterized in that, In step 5, the disturbance criterion is as follows: The disturbance flag F of the i-th segment dis(i) The calculation is as follows: The formula for calculating Th2 is as follows: TH2=β*max(ER (i) (t)) Where max(ER) (i) (t) represents the maximum value of the energy ratio curve at the beginning of the i-th segment, and β takes values ​​from 0.2 to 0.5; Distribution network disturbance determination: In this context, "2" indicates that a disturbance has occurred in the power distribution network, while "0" indicates that no disturbance has occurred.

5. The cable-type distribution network fault identification method based on convolutional signal energy ratio curve according to claim 1, characterized in that, In step 6, the duration of the high-energy state of the energy ratio curve at the beginning of the i-th segment is calculated using a sliding window mechanism. The specific steps are as follows: The time window length is W h The sliding step size is S h ; The starting index of the z-th time window is: Start z(i) =t ER1(i) +(z-1)*S h The energy ratio curve data at the beginning of the i-th segment contained in the z-th time window is: Window z(i) ={ER (i) (t)|Start z(i) ≤ t ≤Start z(i) +W h } Calculate the average value μ of the energy ratio curve data at the beginning of the i-th segment within the z-th time window. z(i) ; High-energy state determination: If the average value of the z-th time window is μ z(i) Greater than E th If , then the time window is determined to be a high-energy state time window, where E th The calculation formula is as follows: E th =α*max(IS (i) (t)) Where α takes values ​​from 0.7 to 0.8; Collect all high-energy state time windows. If the difference between the starting indices of two adjacent time windows is less than W... h Then, these two time windows are merged into a single continuous time window, with its starting time t. ER_st(i) The end time is t ER_end(i) ; The formula for calculating the duration of the high-energy state of the energy ratio curve at the beginning of the i-th segment is: T high(i) =t ER_end(i) -t ER_st(i) 。 6. The cable-type distribution network fault identification method based on convolutional signal energy ratio curve according to claim 1, characterized in that, In step 7, the short-cycle early fault criterion is as follows: The short-cycle early fault flag F of the i-th segment short(i) The calculation is as follows: Among them, Th3 is calculated based on the simulation system. The simulation system is used to simulate the short-cycle arc fault of the cable, calculates the duration of the high-energy state in each section, and calculates its average value. Distribution network short-cycle early fault flag F short The calculation is as follows: In this context, "3" represents a short-cycle early fault that occurred in the distribution network, while "0" represents no short-cycle early fault that occurred.

7. The cable-type distribution network fault identification method based on convolutional signal energy ratio curve according to claim 1, characterized in that, In step 8, for the i-th segment, the time difference between the end of the high-energy state and the end of the energy ratio curve at the beginning of the segment is calculated, and the specific formula is as follows: Δt (i) =t ERJ(i) - t ER_end(i) The critical fault flag F of the i-th segment spg(i) The calculation is as follows: Among them, Th4 takes a value between 0.01 and 0.02 s; Distribution network serious fault indicator F g The calculation is as follows: In this context, "4" represents a serious fault in the distribution network, and "5" represents an early-stage fault with a long cycle.

Images