Active power distribution network differential protection method and system

Abstract

The invention discloses a differential protection method and system for an active power distribution network. The method comprises the following steps: collecting a local-side current sequence signal Ip and an opposite-side current sequence signal Iq of a protection section; detecting an abnormal point of the current sequence signal Iq by adopting a K nearest neighbor algorithm, and correcting the detected abnormal point to obtain a corrected current sequence signal; performing trajectory translation on the current sequence signal Ip to obtain a new current sequence signal, and calculating the similarity degree DTW (Ip,-Iq) between the new current sequence signal and the corrected current sequence signal; when it is determined that the protection section is an internal fault based on comparison of the similarity degree DTW (Ip,-Iq) and the similarity degree D0 of the current signals on the two sides of the protection section during normal operation, differential protection action is carried out; according to the invention, the access of the unmeasurable load branches in the protection section is considered, and the accuracy of differential protection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system control and protection, and in particular to an active distribution network differential protection method and system. Background Art

[0002] With the widespread access of distributed renewable energy to the distribution network, the access of high-penetration distributed power sources has also brought many adverse effects on the operation, control and protection of traditional distribution networks: after a high proportion of distributed power sources are connected to the grid, the power supply mode of the distribution system changes from the traditional single power supply to a multi-terminal, multi-source network with system power supply and distributed power supply. The active nature of the distribution network has put forward higher requirements for the protection of the distribution network, which is specifically manifested as follows:

[0003] (1) Inverter-interfaced Distributed Generator (IIDG) is connected to the distribution network through an inverter. The fault response characteristics are affected by its own control strategy. The nonlinearity of control switching causes the internal potential to be unstable after the fault, making it impossible for the traditional fault analysis method based on constant voltage source or constant current source to accurately characterize the fault characteristics of renewable energy power sources. The access of IIDG will affect the amplitude and phase characteristics of the fault current. Traditional differential protection is difficult to ensure rapid and reliable fault identification, threatening the safe operation of the power system. (2) The access of unmeasurable load branches in the protection section causes unbalanced current to exist before and after the fault, affecting the reliability and sensitivity of traditional differential protection.

[0004] In summary, traditional current differential protection is no longer suitable for distribution networks with load branches and high proportion of distributed power access. For this reason, the document "Research on Line Protection of Distribution Network Based on 5G Communication Technology, Long Wei, Master's Thesis" proposes a distribution network protection method based on the dynamic time warping distance algorithm (Dynamic Time Warping, DTW). The document points out that the DTW distance algorithm is an algorithm used to compare the similarity of two sequence fragments, which is consistent with the principle of differential protection. When an in-zone fault occurs, the current on both sides of the fault point are independent, which will cause a sudden increase in the DTW distance. By setting a threshold, it can be determined whether the line has an in-zone fault; while in the case of an out-of-zone fault or normal operation, the current waveforms on both sides are not much different. In theory, the DTW distance is small and cannot reach the threshold. This difference can be used to form a differential protection based on the DTW distance. The document "Differential protection of distribution network considering communication conditions, Xiong Yuwei, master's thesis" proposes a distribution network differential protection algorithm based on dynamic time warping. The document points out that the differential protection algorithm based on dynamic time warping can compare the similarity of two time series of different lengths. In a distribution network containing DG, if the fault occurs outside the protection zone, the currents at both ends of the line are the same in magnitude but opposite in direction. If the current on one side is reversed, the currents at both ends of the line are almost exactly the same due to the short distribution network line. At this time, even if some data from the protection device on the opposite side fails to reach this side due to delay jitter, the waveform similarity of the two current sampling sequences in the time window is still high, so the DTW similarity value is small; if the fault occurs in the protection zone, the currents at both ends of the line have no correlation, so the waveforms are not similar, so the DTW similarity value is large. Based on this fault characteristic, differential protection of the distribution network containing DG under 5G communication conditions is realized.

[0005] In order to solve the communication delay and delay jitter problems existing in 5G communication, the above-mentioned related literature introduces the traditional DTW algorithm. However, the traditional DTW algorithm has defects such as high overhead, distorted current sequence waveform, and reliance on experience to select parameters. In addition, the traditional DTW algorithm is greatly affected by data abnormal values ​​when used for distribution network differential protection. For this reason, a distribution network differential protection algorithm based on 5G communication and improved DTW is proposed in the patent application document with publication number CN114865602A. On the one hand, the document combines the DTW algorithm with the lower bound function, the sorting and discarding strategy and the particle swarm optimization strategy to overcome the defects of the traditional DTW algorithm such as high overhead, distorted current sequence waveform, and reliance on experience to select parameters; on the other hand, the elements on the optimal regular path are arranged in ascending order and only the first b items are retained to minimize the impact of abnormal data values ​​on the calculation. However, since only the first b items are retained, that is, the items exceeding the limit are directly discarded, some data will always be discarded, resulting in incomplete data affecting the accuracy of differential protection action. Summary of the invention

[0006] The technical problem to be solved by the present invention is how to consider the access of unmeasurable load branches within the protection section and improve the accuracy of differential protection of the active distribution network.

[0007] The present invention solves the above technical problems by the following technical means:

[0008] On the one hand, a method for active distribution network differential protection is proposed, the method comprising:

[0009] Collect the current sequence signal I on the side of the protection zone p and the opposite side current sequence signal I q ;

[0010] The K nearest neighbor algorithm is used to detect the opposite side current sequence signal I q The abnormal point is detected and corrected to obtain the corrected current sequence signal I q ′;

[0011] The current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), and calculate the new current sequence signal F(I p ) and the corrected current sequence signal I q The similarity between DTW(I p ,-I q );

[0012] Based on the similarity DTW (I p ,-I q ) is compared with the similarity D0 of the current signals on both sides of the protection zone during normal operation to determine that the protection zone is an internal fault, and differential protection is performed.

[0013] Furthermore, the K nearest neighbor algorithm is used to detect the opposite side current sequence signal I q The abnormal point is detected and corrected to obtain the corrected current sequence signal I q ' q ,include:

[0014] Calculate the current sequence signal I on the side of the protection zone p and the opposite side current sequence signal I q The Euclidean distance between the current values ​​at the corresponding moments in the equation is calculated, and a distance matrix E is constructed based on the calculated Euclidean distance;

[0015] Searching for an optimal curved path in the distance matrix E, where the sum of the Euclidean distances between the current values ​​at all corresponding moments on the optimal curved path is the smallest;

[0016] The Euclidean distances between the current values ​​at corresponding moments on the optimal curved path are combined into a sequence D = {d1, d2, ..., d l}, calculate each distance element d in the sequence D i The distance ΔD between the k distance elements with the smallest distance i ;

[0017] At a distance ΔD i When it is greater than the upper threshold or less than the lower threshold, the current sequence signal I q The distance element d i The corresponding current value is the abnormal point, and the abnormal point is corrected to obtain the corrected current sequence signal I′ q .

[0018] Furthermore, the calculation formula of the Euclidean distance is:

[0019]

[0020] Where, Ip={p1,p2,…,p i ,…,p n}, Iq={q1,q2,…,q j , …, q m}, d(p i ,q j ) is the current value p i With the current value q j The Euclidean distance between .

[0021] Furthermore, when searching for the optimal curved path in the distance matrix E, the constraints of the DTW algorithm on the search path include boundary constraints and monotonicity constraints, where the starting point in the boundary constraint is d(p1,q1) and the end point is d(p n ,q m );

[0022] The monotonicity constraint is that the curved path is continuous and monotonically non-decreasing: p k ≤p k+1 ,q k ≤q k+1 .

[0023] Furthermore, the distance ΔD i When it is greater than the upper threshold or less than the lower threshold, the current sequence signal I q The distance element d i The corresponding current value is the abnormal point, and the abnormal point is corrected to obtain the corrected current sequence signal I′ q ,include:

[0024] The distance ΔD iCompared with the upper and lower thresholds, at a distance ΔD i When it is greater than the upper threshold or less than the lower threshold, the current sequence signal I q The distance element d i The corresponding current value is the abnormal point, where the upper threshold Fmax = X3 + 1.5 (X3 - X1), the lower threshold Fmin = X1 - 1.5 (X3 - X1), X1 is the value corresponding to the 25% quantile of the data in the sequence ΔD, X3 is the value corresponding to the 75% quantile of the data in the sequence ΔD, and the sequence ΔD is the distance ΔD i Arrange from small to large;

[0025] When the abnormal point is a non-boundary point, the average value of the two sampling values ​​before and after the abnormal point is used to correct the abnormal point. When the abnormal point is a boundary point, the adjacent current value of the abnormal point is directly used to replace it to obtain the corrected current sequence signal I q ′.

[0026] Furthermore, the current sequence signal I p After trajectory translation, a new current sequence signal F(I p ),include:

[0027] The current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), where I p ={(t1,p1),(t2,p2),…,(t i ,p i ),…,(t n ,p n )},F(I p )={(t1+δ,p1), (t2+δ,p2),…, (t i +δ,p i ),…,(t n +δ,p n )}, δ represents the translation distance.

[0028] Furthermore, the calculation of the new current sequence signal F(I p ) and the corrected current sequence signal I q The similarity between DTW(I p ,-I q ), the formula is:

[0029]

[0030] Where δ represents the translation distance, and δ1 and δ2 are set values ​​used to limit the range of waveform translation.

[0031] Further, the similarity degree DTW (I p ,-I q ) is compared with the similarity D0 of the current signals on both sides of the protection zone during normal operation to determine that the protection zone is an internal fault, and the differential protection action is performed, including:

[0032] Calculate the similarity DTW(I p ,-I q ) and the similarity D0. If the similarity DTW(I p ,-I q ) and the similarity D0 are both greater than the action threshold D set , the protection section is determined to be an internal fault, and differential protection is performed;

[0033] Among them, the action threshold value D set The protection is adjusted to prevent false operation during normal system operation and out-of-zone faults.

[0034] Furthermore, the current sequence signal I p and the opposite side current sequence signal I q The sampling time is t.

[0035] In addition, the present invention also proposes an active distribution network differential protection system, the system comprising:

[0036] Signal acquisition module, used to collect the current sequence signal I on the protection zone side p and the opposite side current sequence signal I q ;

[0037] The abnormal point correction module is used to detect the opposite side current sequence signal I using the K nearest neighbor algorithm q The abnormal point is detected and the detected abnormal point is corrected to obtain the corrected current sequence signal I q ′;

[0038] DTW distance optimization module is used to convert the local current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), and calculate the new current sequence signal F(I p ) and the corrected current sequence signal I q The similarity between DTW(I p ,-I q );

[0039] A differential protection module is used for determining the similarity DTW(I p ,-I q) is compared with the similarity D0 of the current signals on both sides of the protection zone during normal operation to determine that the protection zone is an internal fault, and differential protection is performed.

[0040] The advantages of the present invention are:

[0041] (1) Based on the full current signal sampling values ​​of the line on both sides of the protection section, the present invention calculates the similarity of the current waveforms on both sides through an improved dynamic time warping (DTW) algorithm, and uses the difference in the similarity of the waveforms on both sides of the line when the fault is inside and outside the zone to realize the identification of faults inside and outside the zone, which can adapt to the impact caused by the access of a high proportion of distributed power sources and load branches; and in view of the problem that the existence of data abnormal values ​​in the measurement data may cause the DTW distance to increase, resulting in protection malfunction during normal operation or fault outside the zone, the present invention adopts a K nearest neighbor algorithm to detect abnormal points, and realizes abnormal point processing through average value correction, thereby improving the anti-interference ability of the protection; in addition, in view of the fact that the protection section contains unmeasurable load branches, there is a certain difference between the currents on both sides during normal operation, and the output value of the DTW algorithm is relatively large, resulting in the weakening of the similarity between the sequences, and the problem that the protection refuses to operate when the transition resistance of the fault in the zone is large, the present invention realizes optimization by introducing trajectory translation, effectively reducing the DTW distance of the current signals on both sides during normal operation of the system, and is suitable for distribution networks with unmeasurable load branches, thereby improving the reliability and sensitivity of differential protection.

[0042] (2) The present invention calculates each point d i The mean of the distances to the nearest k points represents the distance to this point. The collection of these distances sets the upper and lower thresholds according to the box plot. Values ​​exceeding the threshold are judged as outliers and replaced, avoiding unnecessary data discard.

[0043] (3) The present invention adopts a box diagram setting, that is, the upper and lower thresholds are obtained according to the calculation results, which is more suitable for the scenario where the distribution network differential protection load and the power supply output power are variable.

[0044] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a flow chart of an active distribution network differential protection method proposed in one embodiment of the present invention;

[0046] Figure 2 It is a topological structure diagram of an active distribution network including load branches in one embodiment of the present invention;

[0047] Figure 3 is a schematic diagram of distance matrix path optimization in one embodiment of the present invention;

[0048] Figure 4 is a schematic diagram of processing abnormal points by average value correction in one embodiment of the present invention;

[0049] Figure 5 is a schematic diagram of trajectory translation optimization in one embodiment of the present invention;

[0050] Figure 6 is a comparison diagram of the protection performance of the fault occurring in the upstream f1 outside the zone in one embodiment of the present invention, wherein (a) the transition resistance is 100Ω, and (b) the transition resistance is 200Ω;

[0051] Figure 7 is a comparison diagram of the protection performance of the fault occurring in the zone f2 in one embodiment of the present invention, wherein (a) the transition resistance is 100Ω, and (b) the transition resistance is 200Ω;

[0052] Figure 8 is a comparison diagram of the protection performance of the downstream f3 outside the zone in one embodiment of the present invention, wherein (a) the transition resistance is 100Ω, and (b) the transition resistance is 200Ω;

[0053] Fig. 9 It is a schematic diagram of the overall process of an active distribution network differential protection method proposed in one embodiment of the present invention;

[0054] Fig.10 It is a structural schematic diagram of an active distribution network differential protection system proposed in one embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described in combination with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] like Figure 1 As shown, the first embodiment of the present invention proposes an active distribution network differential protection method, the method comprising the following steps:

[0057] S10, collect the current sequence signal I on the side of the protection zone p and the opposite side current sequence signal I q ;

[0058] It should be noted that, in this embodiment, the current sequence signal I on the side of the protection zone is collected at a frequency f. p and the opposite side current sequence signal I q , and the sampling time is the same, that is, the current sequence signal I with equal time step length is obtained pand the current sequence signal I q .

[0059] S20, using K nearest neighbor algorithm to detect the opposite side current sequence signal I q The abnormal point is detected and the detected abnormal point is corrected to obtain the corrected current sequence signal I q ′;

[0060] It should be noted that this embodiment uses the K nearest neighbor algorithm to detect abnormal points, and implements abnormal point processing through average value correction, thereby improving the anti-interference ability of the protection.

[0061] S30, the current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), and calculate the new current sequence signal F(I p ) and the corrected current sequence signal I q The similarity between DTW(I p ,-I q );

[0062] It should be noted that, for the protection section containing unmeasurable load branches, there is a certain difference in the current on both sides during normal operation, and the output value of the DTW algorithm is relatively large, which leads to the weakening of the similarity between sequences and the problem of protection refusal to operate when the fault transition resistance in the zone is large. The present invention achieves optimization by introducing trajectory translation, which effectively reduces the DTW distance of the current signals on both sides during normal operation of the system, and is suitable for distribution networks connected to unmeasurable load branches, thereby improving the reliability and sensitivity of differential protection.

[0063] S40, based on the similarity DTW (I p ,-I q ) is compared with the similarity D0 of the current signals on both sides of the protection zone during normal operation to determine that the protection zone is an internal fault, and differential protection is performed.

[0064] This embodiment uses the difference between the transient current waveforms of IIDG and traditional power sources to achieve differential protection. When an in-zone fault occurs, the fault current flowing through the protection installation on one side is mainly provided by the synchronous generator, and the other side is mainly provided by the IIDG, with a significant difference. When an out-of-zone fault occurs, a through current flows through the line, and the difference in the transient current waveforms on both sides is very small; when there is a fault inside or outside the zone, the difference in the waveforms on both sides of the protection is large. The differential protection criterion constructed in this way can be effectively operated and is suitable for distribution networks with a high proportion of distributed power sources.

[0065] As a further preferred technical solution, step S20: using the K nearest neighbor algorithm to detect the opposite side current sequence signal I q The abnormal point is detected and the detected abnormal point is corrected to obtain the corrected current sequence signal I′q , specifically including the following steps:

[0066] S21, calculate the current sequence signal I on the side of the protection zone p and the opposite side current sequence signal I q The Euclidean distance between the current values ​​at the corresponding moments in the equation is calculated, and a distance matrix E is constructed based on the calculated Euclidean distance;

[0067] Specifically, the sampling time is t, and the current sequence signal I is obtained. p ={p1, p2, ..., p i ,…,p n} and I q ={q1, q2, ..., q j ,…,q m}, the Euclidean distance between the current values ​​at corresponding moments in the two current time series is calculated as:

[0068]

[0069] An n×m distance matrix E is established based on the calculated Euclidean distance.

[0070] S22, searching for an optimal curved path in the distance matrix E, wherein the sum of the Euclidean distances between the current values ​​at all corresponding moments on the optimal curved path is the smallest;

[0071] Specifically, when searching for the optimal curved path in the distance matrix E, the constraints of the DTW algorithm on the search path include boundary constraints and monotonicity constraints, where the starting point in the boundary constraint is d(p1,q1) and the end point is d(p n ,q m ); The monotonicity constraint is that the curved path is continuous and monotonically non-decreasing: p k ≤p k+1 ,q k ≤q k+1 .

[0072] It should be noted that this embodiment finds a path in the distance matrix E so that the sum of the distances of the current values ​​at corresponding moments on the path is minimized. The cumulative minimum distance is the corresponding DTW distance. This path is called the optimal curved path, that is:

[0073]

[0074] Through the optimal curved path, from d(p1,q1) to d(p n ,q m ) has the smallest cumulative Euclidean distance, denoted as D(I p ,-I q ).

[0075] S23, forming a sequence D = {d1, d2, ..., d l}, calculate each distance element d in the sequence D i The distance ΔD between the k distance elements with the smallest distance i ;

[0076] It should be noted that when the DTW algorithm is applied to distribution network protection, the K nearest neighbor algorithm is used to detect abnormal points, and the upper and lower thresholds are set using the box plot, and the abnormal point processing is realized through average value correction to address the problem that the data outliers in the measured data may increase the DTW distance and cause the protection to malfunction during normal operation or out-of-zone faults. i The sequence D = {d1, d2, ..., d l}, for each point d in the sequence D i , select k points with the smallest distance and calculate d i The distance between these k points is recorded as ΔD i .

[0077] S24, at a distance ΔD i When it is greater than the upper threshold or less than the lower threshold, the current sequence signal I q The distance element d i The corresponding current value is the abnormal point, and the abnormal point is corrected to obtain a corrected current sequence signal I2′q.

[0078] Furthermore, in this embodiment, the calculated ΔD i Arranged from small to large, the sequence ΔD={ΔD1, ΔD2,…, ΔD l}, the value corresponding to the 25% quantile of the data in the sequence is X1, and the value corresponding to the 75% quantile of the data in the sequence is X3; the calculation formula for the upper threshold is Fmax=X3+1.5(X3-X1); the calculation formula for the lower threshold is Fmin=X1-1.5(X3-X1). When ΔD i When it is greater than the upper threshold or less than the lower threshold, the current sequence signal I q The distance element d i The corresponding current value is the abnormal point.

[0079] It should be noted that this embodiment uses a box plot as a statistical diagram to display a set of data dispersion information. In the box plot, X1, X2, and X3 are used to represent the upper quartile, the median, and the lower quartile.

[0080] Furthermore, for the data outliers at the non-boundary point, the average of the two previous and next sampling values ​​is used to correct the outlier. Assuming that point m is the detected outlier, the replacement is performed as follows:

[0081]

[0082] For the outliers q1 and q at the boundary m , then directly use the adjacent current values ​​of the abnormal point to replace, that is, use q2 and q m-1 Make a replacement.

[0083] It should be noted that I q is the sampling value of the current signal on the opposite side, which is transmitted to this side for data processing. The sequence D = {d1, d2, ..., d i ,…,d l Each value in} is obtained through the corresponding I p and I q Calculated, so the detection d i The current sequence signal I can be determined q The abnormal points in .

[0084] It should be noted that, in this embodiment, each point d is first calculated i The mean of the distances to the nearest k points represents the distance to this point. The upper and lower thresholds are set for the collection of these distances according to the box plot. Values ​​exceeding the threshold are determined as outliers, and the values ​​of the outliers are corrected to avoid unnecessary data discard.

[0085] As a further preferred technical solution, in step S30, the current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), specifically:

[0086] The current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), where I p ={(t1,p1),(t2,p2),…,(t i ,p i ),…,(t n ,p n )},F(I p )={(t1+δ,p1), (t2+δ,p2),…, (t i +δ,p i ),…,(t n +δ,p n )}, δ represents the translation distance.

[0087] Considering that the protection zone contains unmeasurable load branches, there is a certain difference in the current on both sides during normal operation. The output value of the DTW algorithm is relatively large, which leads to the weakening of the similarity between sequences. When the fault transition resistance in the zone is large, the protection refuses to operate. The optimization is achieved by introducing trajectory translation. p ,I q For a time series with equal step length, F(I p ) can be regarded as a series of time clusters. In a series of time series, we look for a group with the smallest cumulative Euclidean distance. Then the trajectory I p ,I q The DTW distance optimization calculation between is shown as follows:

[0088]

[0089] In the formula, δ represents the translation distance, δ1 and δ2 are the setting values ​​used to limit the range of waveform translation. Generally, -T / 2≤δ1,δ2≤T / 2 is set, and T is the waveform sampling time window length.

[0090] It should be noted that δ1 and δ2 are set empirical values, that is, the allowable translation range. For example, if the sampling window time is 10ms, then δ1 = -5ms and δ2 = 5ms can be set to allow the waveform to be translated within this range. The size of the value range will affect the calculation efficiency. A larger range can better align the waveform, but the calculation time will be longer; a smaller range will shorten the calculation time, but may not be able to align the waveform well.

[0091] As a further preferred technical solution, Fig. 9 As shown, in step S40, based on the similarity DTW (I p ,-I q ) is compared with the similarity D0 of the current signals on both sides of the protection zone during normal operation to determine that the protection zone is an internal fault, and the differential protection action is performed, including:

[0092] Calculate the similarity DTW(I p ,-I q ) and the similarity D0, if within two consecutive time periods t, it satisfies:

[0093]

[0094] If the fault is within the protection zone, the protection zone is determined to be an internal fault and the differential protection action is performed; otherwise, the protection zone is determined to be an external fault and no protection action is performed;

[0095] Among them, the action threshold value D set The protection is adjusted to prevent false operation during normal system operation and out-of-zone faults.

[0096] Below Figure 2The active distribution network with load branch is shown as an example for explanation. The voltage level of the distribution network is 10kV, and line BC is used as the protection zone. The lengths of lines AB, BC, CD, and AE are 4, 4, 2, and 2km respectively, and the unit length parameter is z = 0.17 + j0.38Ω / km. The inverter-type distributed power source DG is connected at bus D, with a capacity of 5MVA. The line load L1 is a branch line in the protection zone, with a rated power of 1.5MVA, and the access point is the midpoint of the BC section. The line load L2 is 2MVA, and L3 is 5MVA. The power factor of the three lines is 0.8.

[0097] Assume that the unmeasurable load branch is connected to the midpoint of line BC, the system sampling frequency is 10kHz, and the sampling window is half a cycle, i.e. 10ms. Set three fault points, f1 is located at the midpoint of the upstream AB section of the protection, f2 is located 3km away from point B in the protection section, and f3 is located at the midpoint of the downstream CD section of the protection. In the example, the corresponding DTW (I) is calculated according to the above steps when the system is operating normally, when the upstream f1 outside the zone is faulty, when the f2 inside the zone is faulty, and when the downstream f3 outside the zone is faulty. p ,-I q ):

[0098] (1) Collect the current signal I on both sides of the protection zone at frequency f and duration t p and I q , and obtain the current signal Ip={p1,p2,…,p i ,…,p n} and Iq={q1,q2,…,q j ,…,q m}.

[0099] (2) Calculate the current sequence signal I on the protection zone side p and the opposite side current sequence signal I q The Euclidean distance between the current values ​​at the corresponding moments is calculated, and the distance matrix E is constructed according to the calculated Euclidean distance.

[0100] (3) Find the optimal curved path in the distance matrix E. The sum of the Euclidean distances between the current values ​​at all corresponding moments on the optimal curved path is the smallest. The schematic diagram of path optimization is as follows: Figure 3 shown.

[0101] (4) The Euclidean distances between the current values ​​at corresponding moments on the optimal curved path are combined into a sequence D = {d1, d2, ..., d l}, calculate each distance element d in the sequence D i The distance ΔD between the k distance elements with the smallest distance i ; Calculated D i Arranged from small to large, the sequence ΔD={ΔD1, ΔD2,…, ΔDl}, the value corresponding to the 25% quantile of the data in the sequence is X1=0.0138, and the value corresponding to the 75% quantile of the data is X3=0.229; the calculation formula for the upper threshold is Fmax=X3+1.5(X3-X1)=0.5518; the calculation formula for the lower threshold is Fmin=X1-1.5(X3-X1)=-0.309.

[0102] (5) At a distance ΔD i When it is greater than the upper threshold or less than the lower threshold, the current sequence signal I q The distance element d i The corresponding current value is the abnormal point, and the abnormal point is corrected to obtain the corrected current sequence signal I q ′q.

[0103] The average value replacement diagram is as follows Figure 4 As shown. Before correction, d(p7,q7)=2.612, d(p 20 ,q 20 )=2.622,D(P,Q)=7.6;After correction d(p7,q7)=0.189,d(p 20 ,q 20 )=1.029, D(P,Q)=3.3.

[0104] (6) The current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), and calculate the new current sequence signal F(I p ) and the corrected current sequence signal I q The similarity between DTW(I p ,-I q ), find the group with the smallest cumulative Euclidean distance in a series of time series, the trajectory translation optimization diagram is as follows Figure 5 shown.

[0105] First, calculate the DTW (I p ,-I q ), considering the impact of unmeasurable load branches in the area on the DTW distance, the results are shown in Table 1:

[0106] Table 1 Impact of unmeasured load branch power

[0107]

[0108]

[0109] It can be seen from Table 1 that there are unmeasurable load branches in the differential protection zone. As the branch power increases, the DTW distance also increases. p ,-I q ) / D0 is used as the protection criterion. When the system is operating normally, the ratio is 1 and is not affected by changes in load branch power.

[0110] Then calculate the DTW (I p ,-I q ), considering the impact of noise and fault type on DTW distance during data transmission, the results are shown in Table 2 and Table 3:

[0111] Table 2 Simulation results of AB phase short circuit fault at f1

[0112]

[0113] Table 3 Simulation results of three-phase fault at f1

[0114]

[0115] From Table 2 and Table 3, we can see that DTW(I p ,-I q ) / D0 has a maximum value of only 1.13, which is close to the value when the system is operating normally. When a fault occurs at f1, regardless of the capacity of the unmeasurable load branch, the current amplitudes on the B and C sides of the line are not much different and the phases are close, with a high degree of similarity.

[0116] Then calculate the DTW (I p ,-I q ), considering the influence of noise and fault type on DTW distance during data transmission, the results are shown in Table 4 and Table 5:

[0117] Table 4 Simulation results of AB phase short circuit fault at f3

[0118]

[0119] Table 5 Simulation results of three-phase fault at f3

[0120]

[0121] From Table 4 and Table 5, we can see that some DTW (I p ,-I q) / D0 increases significantly, with a maximum value of 2.89. Due to the high penetration rate of distributed photovoltaic power in the distribution network, the rated power of photovoltaic power is 5MW, and the load L1+L2 is only 3.5MW, and the power is reversed. When a fault occurs at f3 downstream of the protection, when the transition resistance is relatively large, the current of the load branch L1 may be provided by the grid-side synchronous generator and IIDG, and the current waveforms on both sides may differ to a certain extent.

[0122] When the system is operating normally, DTW / D0 is 1. When there is a fault outside the zone, the maximum value of DTW / D0 is 2.89, and the reliability coefficient is 1.2. Therefore, the action threshold D set =3.468.

[0123] Finally, the DTW (I p ,-I q ), the results are shown in Tables 6 and 7:

[0124] Table 6 Simulation results of AB phase short circuit fault at f2

[0125]

[0126] Table 7 Simulation results of three-phase fault at f2

[0127]

[0128] It can be seen from Table 6 and Table 7 that when a fault occurs at f2 in the area, DTW(I p ,-I q ) / D0 minimum value is 11.21, which is greater than the setting value, and the protection can operate correctly. When the transition resistance is small, the DTW distance increases significantly because the fault current output by the IIDG inverter has a limiting link and the fault current amplitude provided by the synchronous generator is very different. When the transition resistance is large, the fault characteristics are not obvious. When there is an intra-area fault, the current waveforms on both sides of the line are provided by the IIDG and the synchronous generator respectively. There is a significant difference in the fault current between the two. The improved DTW algorithm can effectively identify intra-area faults.

[0129] The improved DTW algorithm proposed in this embodiment is compared with the traditional DTW algorithm. Figure 6 Comparison of protection performance when the fault occurs upstream f1 outside the zone, where (a) the transition resistance is 100Ω and (b) the transition resistance is 200Ω; Figure 7 Comparison of protection performance when fault f2 occurs in zone, where (a) transition resistance is 100Ω, (b) transition resistance is 200Ω; Figure 8The comparison diagram of protection performance when the downstream f3 outside the zone occurs fault, where (a) the transition resistance is 100Ω and (b) the transition resistance is 200Ω. The results show that the traditional DTW algorithm performs poorly when the transition resistance of the zone fault is relatively large in the distribution network with unmeasurable load branches, while the proposed improved DTW algorithm can correctly distinguish between faults inside and outside the zone, has a strong ability to tolerate transition resistance, and can also ensure that there is no false operation when there is a fault outside the zone.

[0130] This embodiment uses the difference between the transient current waveforms of IIDG and traditional power supplies when a line fault occurs in the distribution network to analyze it. On this basis, the dynamic time warping algorithm is used to calculate the similarity of the waveforms of the full current signal sampling values ​​on both sides of the line. The difference in the similarity of the waveforms on both sides of the line when there are faults inside and outside the zone is used to identify faults inside and outside the zone. The measurement effect of the DTW algorithm is optimized by the idea of ​​trajectory translation, the K nearest neighbor algorithm is used to detect abnormal points, and the average value correction is used to realize the processing of abnormal points during signal transmission. The protection criterion is constructed based on the cumulative Euclidean distance during normal operation. Simulation analysis shows that the proposed protection scheme can correctly distinguish between internal and external faults, has strong adaptability, and has good tolerance to transition resistance. The present invention is suitable for distribution networks with a high proportion of distributed power sources connected, and can maintain good action characteristics in response to the situation where unmeasurable load branches are connected in the protection section.

[0131] In addition, if Fig.10 As shown, another embodiment of the present invention provides an active distribution network differential protection system, the system comprising:

[0132] The signal acquisition module 10 is used to collect the current sequence signal I on the side of the protection zone. p and the opposite side current sequence signal I q ;

[0133] The abnormal point correction module 20 is used to detect the opposite side current sequence signal I using the K nearest neighbor algorithm q The abnormal point is detected and the detected abnormal point is corrected to obtain the corrected current sequence signal I q ′;

[0134] The DTW distance optimization module 30 is used to convert the local current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), and calculate the new current sequence signal F(I p ) and the corrected current sequence signal I q The similarity between DTW(I p ,-I q );

[0135] The differential protection module 40 is used to determine the similarity based on the DTW (I p ,-Iq ) is compared with the similarity D0 of the current signals on both sides of the protection zone during normal operation to determine that the protection zone is an internal fault, and differential protection is performed.

[0136] As a further preferred technical solution, the outlier correction module 20 includes:

[0137] The distance matrix construction unit is used to calculate the current sequence signal I on the side of the protection zone p and the opposite side current sequence signal I q The Euclidean distance between the current values ​​at the corresponding moments in the equation is calculated, and a distance matrix E is constructed based on the calculated Euclidean distance;

[0138] An optimal curved path search unit, used to search for an optimal curved path in the distance matrix E, where the sum of the Euclidean distances between the current values ​​at all corresponding moments on the optimal curved path is the smallest;

[0139] A distance sequence construction unit is used to form a sequence D = {d1, d2, ..., d l}, calculate each distance element d in the sequence D i The distance ΔD between the k distance elements with the smallest distance i ;

[0140] Anomaly correction unit for the distance ΔD i When it is greater than the upper threshold or less than the lower threshold, the current sequence signal I q The distance element d i The corresponding current value is the abnormal point, and the abnormal point is corrected to obtain the corrected current sequence signal I q ′q.

[0141] As a further preferred technical solution, the abnormality correction unit is specifically used to:

[0142] When the abnormal point is a non-boundary point, the average value of the two sampling values ​​before and after the abnormal point is used to correct the abnormal point. When the abnormal point is a boundary point, the adjacent current value of the abnormal point is directly used to replace it to obtain the corrected current sequence signal I q ′.

[0143] As a further preferred technical solution, the DTW distance optimization module 30 includes:

[0144] The translation unit is used to convert the current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), where I p ={(t1,p1),(t2,p2),…,(ti ,p i ),…,(t n ,p n )},F(I p )={(t1+δ,p1), (t2+δ,p2),…, (t i +δ,p i ),…,(t n +δ,p n )}, δ represents the translation distance;

[0145] Distance optimization unit, used to calculate the new current sequence signal F(I p ) and the corrected current sequence signal I q The similarity between DTW(I p ,-I q ), the formula is:

[0146]

[0147] Where δ represents the translation distance, and δ1 and δ2 are set values ​​used to limit the range of waveform translation.

[0148] As a further preferred technical solution, the differential protection module 40 is specifically used for:

[0149] Calculate the similarity DTW(I p ,-I q ) and the similarity D0. If the similarity DTW(I p ,-I q ) and the similarity D0 are both greater than the action threshold D set , the protection section is determined to be an internal fault, and differential protection is performed;

[0150] Among them, the action threshold value D set The protection is adjusted to prevent false operation during normal system operation and out-of-zone faults.

[0151] It should be noted that other embodiments or implementation methods of the active distribution network differential protection system of the present invention can refer to the above-mentioned method embodiments, which will not be repeated here.

[0152] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0153] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0154] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for differential protection of an active distribution network, characterized in that: The method comprises: Collect the current sequence signal I on the side of the protection zone p and the opposite side current sequence signal I q ; The K nearest neighbor algorithm is used to detect the opposite side current sequence signal I q The abnormal point is detected and the detected abnormal point is corrected to obtain the corrected current sequence signal I q ′; The current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), and calculate the new current sequence signal F(I p ) and the corrected current sequence signal I q The similarity between DTW(I p ,-I q ); Based on the similarity DTW (I p ,-I q ) is compared with the similarity D0 of the current signals on both sides of the protection zone during normal operation to determine that the protection zone is an internal fault, and differential protection is performed.

2. The active distribution network differential protection method according to claim 1, characterized in that: The K nearest neighbor algorithm is used to detect the opposite side current sequence signal I q The abnormal point is detected and the detected abnormal point is corrected to obtain the corrected current sequence signal I q ',include: Calculate the current sequence signal I on the side of the protection zone p and the opposite side current sequence signal I q The Euclidean distance between the current values ​​at the corresponding moments in the equation is calculated, and a distance matrix E is constructed based on the calculated Euclidean distance; Searching for an optimal curved path in the distance matrix E, where the sum of the Euclidean distances between all current values ​​at corresponding moments on the optimal curved path is the smallest; The Euclidean distances between the current values ​​at corresponding moments on the optimal curved path are combined into a sequence D = {d1, d2, ..., d l }, calculate each distance element d in the sequence D i The distance ΔD between the k distance elements with the smallest distance i ; At a distance ΔD i When it is greater than the upper threshold or less than the lower threshold, the current sequence signal I q The distance element d i The corresponding current value is the abnormal point, and the abnormal point is corrected to obtain the corrected current sequence signal I q ′.

3. The active distribution network differential protection method according to claim 2, characterized in that: The calculation formula of the Euclidean distance is: Where, Ip={p1,p2,…,p i ,…,p n }, Iq={q1,q2,…,q j , …, q m }, d(p i ,q j ) is the current value p i With the current value q j The Euclidean distance between .

4. The active distribution network differential protection method according to claim 2, characterized in that: When searching for the optimal curved path in the distance matrix E, the constraints of the DTW algorithm on the search path include boundary constraints and monotonicity constraints, where the starting point in the boundary constraint is d(p1,q1) and the end point is d(p n ,q m ); The monotonicity constraint is that the curved path is continuous and monotonically non-decreasing: p k ≤p k+1 ,q k ≤q k+1 .

5. The active distribution network differential protection method according to claim 2, characterized in that: The distance ΔD i When it is greater than the upper threshold or less than the lower threshold, the current sequence signal I q The distance element d i The corresponding current value is the abnormal point, and the abnormal point is corrected to obtain the corrected current sequence signal I′ q ,include: The distance ΔD i Compared with the upper and lower thresholds, at a distance ΔD i When it is greater than the upper threshold or less than the lower threshold, the current sequence signal I q The distance element d i The corresponding current value is the abnormal point, where the upper threshold Fmax = X3 + 1.5 (X3 - X1), the lower threshold Fmin = X1 - 1.5 (X3 - X1), X1 is the value corresponding to the 25% quantile of the data in the sequence ΔD, X3 is the value corresponding to the 75% quantile of the data in the sequence ΔD, and the sequence ΔD is the distance ΔD i Arrange from small to large; When the abnormal point is a non-boundary point, the average value of the two sampling values ​​before and after the abnormal point is used to correct the abnormal point. When the abnormal point is a boundary point, the adjacent current value of the abnormal point is directly used to replace it to obtain the corrected current sequence signal I q ′.

6. The active distribution network differential protection method according to claim 1, characterized in that: The current sequence signal I p After trajectory translation, a new current sequence signal F(I p ),include: The current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), where I p ={(t1,p1),(t2,p2),…,(t i ,p i ),…,(t n ,p n )},F(I p )={(t1+δ,p1), (t2+δ,p2),…, (t i +δ,p i ),…,(t n +δ,p n )}, δ represents the translation distance.

7. The active distribution network differential protection method according to claim 1, characterized in that: The calculation of the new current sequence signal F(I p ) and the corrected current sequence signal I q The similarity between DTW(I p ,-I q ), the formula is: Where δ represents the translation distance, and δ1 and δ2 are set values ​​used to limit the range of waveform translation.

8. The active distribution network differential protection method according to any one of claims 1 to 7, characterized in that: The similarity DTW (I p ,-I q ) is compared with the similarity D0 of the current signals on both sides of the protection zone during normal operation to determine that the protection zone is an internal fault, and the differential protection action is performed, including: Calculate the similarity DTW(I p ,-I q ) and the similarity D0, if within two consecutive time periods t, the similarity DTW(I p ,-I q ) and the similarity D0 are both greater than the action threshold D set , the protection section is determined to be an internal fault, and differential protection is performed; Among them, the action threshold value D set The protection is adjusted to prevent false operation during normal system operation and out-of-zone faults.

9. The active distribution network differential protection method according to claim 1, characterized in that: The local current sequence signal I p and the opposite side current sequence signal I q The sampling time is t.

10. An active distribution network differential protection system, characterized in that: The system comprises: Signal acquisition module, used to collect the current sequence signal I on the protection zone side p and the opposite side current sequence signal I q ; The abnormal point correction module is used to detect the opposite side current sequence signal I using the K nearest neighbor algorithm q The abnormal point is detected and the detected abnormal point is corrected to obtain the corrected current sequence signal I q ′; DTW distance optimization module is used to convert the local current sequence signal I p After trajectory translation, a new current sequence signal F(I p ), and calculate the new current sequence signal F(I p ) and the corrected current sequence signal I q The similarity between DTW(I p ,-I q ); A differential protection module is used for determining the similarity DTW(I p ,-I q ) is compared with the similarity D0 of the current signals on both sides of the protection zone during normal operation to determine that the protection zone is an internal fault, and differential protection is performed.

Images