Power system transient protection method based on fault information fusion

Abstract

According to the invention, the internal association relationship between the fault working condition attribute and the fault feature is mined, and the fault information fusion-based power system transient protection method is provided. The fault information fusion method adopts a Kalman filtering algorithm and an improved Kalman filtering algorithm, or adopts a particle filtering algorithm or an improved particle filtering algorithm. And the fault can be reflected more accurately by fusing the standard fault characteristics of the fault working condition attributes. According to the power system transient protection method based on the fault working condition attribute, the adverse effect of the fault initial phase angle and the transition resistance on protection is overcome, and the transient protection reliability is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system relay protection, and in particular relates to a power system transient quantity protection method based on fault condition attributes. Background Art

[0002] Accurate and rapid fault clearance is essential to improving the transient stability of power systems and the transmission capacity of transmission lines. As the voltage level of power grids continues to increase, the requirements for protection performance are also getting higher and higher. For ultra-high and extra-high voltage power grids, it is an urgent task to study line protection with ultra-high-speed action performance, safety and reliability. Line protection based on power frequency quantity is affected by power system oscillation, overload, and short-circuit point transition resistance, so traditional power frequency quantity protection is difficult to meet the needs of power grid development. Transient signals are generated at the moment of power system faults. The transient signals generated by the faults carry rich fault information, such as fault direction, fault location, fault type and fault time, especially in the high-frequency band. The fault characteristics are more significant. Therefore, transient protection can detect faults in a very short time to achieve ultra-high-speed protection by quickly capturing these high-frequency components of the fault signal. Transient quantity-based protection is not affected by power frequency oscillation, CT saturation, etc. and has ultra-high-speed action performance, becoming a research hotspot in the field of relay protection at home and abroad.

[0003] Various modern signal processing technologies such as Hilbert-Huang transform, S transform, wavelet transform, mathematical morphology and neural network are widely used in transient protection of transmission lines to improve the performance of identifying fault direction using fault transient information. Hilbert-Huang transform has advantages in processing nonlinear and non-stationary signals, and has the advantages of high time-frequency resolution, high sensitivity, strong adaptability and strong anti-interference ability. S transform combines the advantages of short-time Fourier transform (STFT) and wavelet transform, and has the advantages of high time-frequency resolution, support for multi-scale analysis and no need for preset mother wavelet. The advantages are mainly reflected in its flexible time-frequency analysis ability, adaptability to non-stationary signals and high-precision fault feature extraction. The advantages of wavelet transform are reflected in its powerful time-frequency analysis ability, adaptability to non-stationary signals, rapid fault detection and location, etc. This makes wavelet transform widely used in power systems, especially for the detection and processing of transient faults.

[0004] The performance of traditional transient protection is affected by the initial phase angle of the fault or the transition resistance. When the initial phase angle of the fault is small or the transition resistance is large, the transient signal of the fault is significantly weakened, the fault characteristics are not obvious, the protection performance is seriously affected, and may not meet the requirements. The reliability of traditional traveling wave protection is questioned because it relies on the wave head detection of the traveling wave.

[0005] Technical solution and shortcomings of prior art 1

[0006] In the case of weak faults (small initial fault phase angle and high transition resistance), the transient signal is very weak and the fault feature quantity extracted will be very small, which is very different from the fault feature value in other cases, resulting in low protection sensitivity and low reliability. Chinese invention patent CN201310065694.2, named Adaptive single-ended transient protection of transmission lines based on transition resistance and fault angle reduction, proposes a method and invents a reduction formula that can unify other transition resistances and fault initial phase angles to the same fault initial phase angle and transition resistance, solving the impact of transition resistance and fault initial phase angle on transient protection. However, this method relies on the bypass attenuation effect of the busbar system's distributed capacitance to the ground on the high-frequency components of the fault. In fact, because the line boundary effect of the busbar system is not obvious, its effect is not ideal and difficult to be practical. Summary of the invention

[0007] The purpose of the present invention is to solve the defects of the above-mentioned prior art, apply the blocking and attenuation effect of line surge arresters on fault high-frequency signals, and consider the influence of fault condition attributes on fault characteristics, so as to provide a power system transient protection method based on fault information fusion.

[0008] The present invention adopts the following technical solution:

[0009] S1. When the protection object is a circuit and the busbar connected to the circuit.

[0010] The power system transient protection method based on fault information fusion includes:

[0011] Step 1. Collect the line voltage u1 from the head end of the protected line and collect the bus voltage u2 from the protected bus.

[0012] Step 2. Extract the high-frequency component u of voltage u1 and u2 h1 and u h2 .

[0013] Step 3. Calculate u h1 and u h2 The transient quantity TQ1, TQ2. The transient quantity can be one of energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude, singular value and other transient high-frequency component processing quantity; or the instantaneous integral of the transient quantity, or the instantaneous cumulative sum of the transient quantity.

[0014] Step 4. Calculate the fault condition attributes: Calculate the initial phase angle θ of the fault f With transition resistance R f .

[0015] Step 5. Fault information fusion: transient quantities TQ1 and TQ2 are fused to the initial fault phase angle θ f and transition resistance R fThe standard transient quantities NTQ1 and NTQ2 are obtained. The transient quantity fusion fault condition attribute adopts the Kalman filtering algorithm proposed in the present invention.

[0016] Step 6. Calculate the difference NTQD between the standard transient quantities NTQ1 and NTQ2 12 NTQD 12 =NTQ1-NTQ2.

[0017] Step 7. Use the standard transient quantity difference in step 6 as the fault feature to determine the fault. Take line L3 and bus M as the protected objects, and NTQD 12 As fault characteristics, protection criteria are constructed:

[0018] When NTQD 12 >Aset, it is judged as an internal fault of line L3;

[0019] When NTQD 12 <-Bset, it is judged as a bus fault;

[0020] Otherwise, it is determined that there is no fault inside the protection area, i.e., line L3 and bus M;

[0021] Among them, Aset and Bset are positive real numbers of protection settings.

[0022] S2. When the protection object is a single circuit.

[0023] The power system transient protection method based on fault information fusion includes:

[0024] Step 1. Collect the line voltage u1 and the bus voltage u2 from the head end of the protected line.

[0025] Step 2. Extract the high-frequency component u of voltage u1 and u2 h1 and u h2 .

[0026] Step 3. Calculate u h1 and u h2 The transient quantity TQ1 and TQ2. The transient quantity is one of signal energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude, singular value, transient quantity gradient, transient quantity relative gradient; or the instantaneous integral of these transient quantities, or the instantaneous cumulative sum of these transient quantities.

[0027] Step 4. Calculate the fault condition attributes: Calculate the initial phase angle θ of the fault f With transition resistance R f .

[0028] Step 5. Fault information fusion: transient quantities TQ1 and TQ2 are fused to the initial fault phase angle θ f and transition resistance Rf The standard transient quantities NTQ1 and NTQ2 are obtained. The transient quantity fusion fault condition attribute adopts the Kalman filtering algorithm proposed in the present invention.

[0029] Step 6. Calculate the difference NTQD between the standard transient quantities NTQ1 and NTQ2 12 NTQD 12 =NTQ1-NTQ2.

[0030] Step 7. Use the standard transient quantity difference in step 6 as the transient characteristic of the fault direction to determine the fault direction. 12 >Aset, it is judged as the equivalent information of "positive direction fault" from bus M to line L3; and the equivalent information of "positive direction fault" is sent to the other end. Otherwise, it is judged that there is no fault in the positive direction from bus M to line L3;

[0031] Step 8. If the local end determines that it is a positive direction fault and receives equivalent information of "positive direction fault" from the other end, it is determined that it is an internal fault of line L3, otherwise it is determined that line L3 has no fault.

[0032] Where Aset is a positive real number for protection setting. Note that Aset here is not the same as Aset in method 1, and there is no connection between them. The same applies to the following.

[0033] S3. When the protection object is a circuit.

[0034] The power system transient protection method based on fault information fusion includes:

[0035] Step 1. Collect the line voltage u1 from the head end of the protected line.

[0036] Step 2. Extract the high-frequency component u of voltage u1 h1 .

[0037] Step 3. Calculate u h1 The transient quantity TQ1. The transient quantity is one of signal energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude integral, instantaneous amplitude, instantaneous amplitude accumulation, singular value, transient quantity gradient and transient quantity relative gradient, etc.

[0038] Step 4. Calculate the fault condition attributes: Calculate the initial phase angle θ of the fault f and transition resistance R f .

[0039] Step 5. Fault information fusion: transient quantity TQ1 fusion fault initial phase angle θ f and transition resistance R fThe standard transient quantity NTQ1 is obtained. The transient quantity is integrated with the fault condition attribute using the Kalman filtering algorithm proposed in the invention.

[0040] Step 6. Use the standard transient quantity NTQ1 as the fault feature to judge the fault. Use NTQ1 as the fault feature to construct the protection judgment criterion:

[0041] When NTQ1>Aset, it is judged that line L3 has an internal fault;

[0042] Otherwise, it is determined that line L3 has no fault;

[0043] Where Aset is a positive real number for protection setting.

[0044] S4. When one busbar is taken as the protection object.

[0045] The power system transient protection method based on fault information fusion includes:

[0046] Step 1. Collect the bus voltage u2 from the protected bus.

[0047] Step 2. Extract the high-frequency component u of voltage u2 h2 .

[0048] Step 3. Calculate u h2 The transient quantity TQ2. The transient quantity is one of signal energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude, singular value, transient quantity gradient and transient quantity relative gradient, etc.; or the instantaneous integral of these transient quantities, or the instantaneous cumulative sum of these transient quantities.

[0049] Step 4. Calculate the fault condition attributes: Calculate the initial phase angle θ of the fault f and transition resistance R f .

[0050] Step 5. Fault information fusion: transient quantity TQ2 fusion fault phase angle θ f and transition resistance R f The standard transient quantity NTQ2 is obtained. The Kalman filtering algorithm proposed in the present invention is used to fuse the transient quantity and the fault condition attribute information.

[0051] Step 6. Use the standard transient quantity NTQ2 as the fault feature to judge the fault. Use NTQ2 as the fault feature to construct the protection judgment criterion:

[0052] When NTQ2>Aset, it is judged as bus M fault;

[0053] Otherwise, it is judged that busbar M has no fault;

[0054] Where Aset is a positive real number for protection setting.

[0055] Beneficial effects of the present invention:

[0056] (1) The present invention makes a fault judgment based on the fact that when the fault transient quantity passes through the line surge arrester, the fault transient quantity is greatly attenuated, and the fault condition attributes are taken into consideration. The protection method based on this principle is not affected by the busbar-to-ground stray capacitance, the distributed capacitance and current of the transmission line, the saturation of the current transformer, the oscillation of the power system, the transition resistance and other conditions, and has ultra-high-speed action performance.

[0057] (2) The present invention integrates the fault condition attributes into the quasi-fault characteristic quantity, eliminates the influence of the fault condition attributes on transient protection, and greatly improves the protection sensitivity and protection reliability.

[0058] (3) The present invention utilizes a machine learning algorithm to judge faults based on quasi-fault characteristics and fault condition attributes, which can realize automatic fault classification, make fault judgments more intelligently, and improve protection reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a schematic diagram of the power system wiring.

[0060] Figure 2 Figure 2 is the line fault voltage and bus fault voltage diagram, (a) is the line fault voltage u1, (b) is the bus fault voltage u2.

[0061] Figure 3 is the high frequency component of the line fault voltage and the high frequency component of the bus fault voltage, (a) is the high frequency component of the line fault voltage u h1 , (b) is the high frequency component u of bus fault voltage h2 .

[0062] Figure 4 are the transient energy of the high-frequency component of the line fault voltage and the high-frequency component of the bus fault voltage, (a) is the transient energy TQ1 of the high-frequency component of the line fault voltage, and (b) is the transient energy TQ2 of the high-frequency component of the bus fault voltage.

[0063] Figure 5 are the standard transient energies of the high-frequency components of the line fault voltage and the bus voltage, (a) is the standard transient energy NTQ1 of the high-frequency components of the line fault voltage, and (b) is the standard transient energy NTQ2 of the high-frequency components of the bus fault voltage.

[0064] Figure 6 is the standard transient energy difference.

[0065] Figure 7 is the bus fault voltage.

[0066] Figure 8 It is the high frequency component of bus fault voltage.

[0067] Fig. 9 It is the transient energy of the high-frequency component of the bus fault voltage.

[0068] Fig.10 It is the standard transient energy of the high-frequency component of the bus fault voltage.

[0069] Fig.11 is the transient energy difference. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention is described clearly and completely below. 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.

[0071] Terminology explanation:

[0072] Transient quantity: Transient quantity refers to the electrical quantity calculated from the high-frequency components in the fault voltage, fault current or fault traveling wave.

[0073] Fault condition attributes: Fault condition attributes include fault initial phase angle and transition resistance.

[0074] Quasi-fault characteristics: Quasi-fault characteristics refer to the physical quantities that will be used for fault information fusion in the next step, and then directly or indirectly used for fault judgment. Indirect use for fault judgment includes the use of quasi-fault characteristics for fault judgment after fault information fusion, summation, difference, or ratio. Quasi-fault characteristics can be power frequency quantities or transient quantities calculated from power frequency components or high frequency components of current or voltage.

[0075] The scheme of the present invention:

[0076] (1) The present invention generates standardized transient quantities by fusing transient quantities with fault condition attributes to make fault judgments, thereby effectively avoiding the impact of fault condition attributes on the protection performance. This fusion strategy improves the system's adaptability to complex conditions and the accuracy of protection.

[0077] (2). The fusion algorithm uses the Kalman filter algorithm and improved Kalman filter fault information fusion, particle filter and improved particle filter algorithm fault information fusion.

[0078] (3) The fault characteristics are constructed based on the transient data of voltage and current on both sides of the existing surge arrester of the transmission line. There is no need to install additional detection units, so the existing equipment can be used to quickly extract and analyze fault information, improving the economy and practicality of the system.

[0079] S1. When the protection object is a circuit and the busbar connected to the circuit.

[0080] In conjunction with the instruction manual Figure 1 Here, taking line L3 and bus M as an example, the transient unit protection method of the line and bus is explained. The wave arrester installed on the transmission line has a strong blocking effect on the high-frequency signal in the stop band, and does not affect the transmission of the power frequency current. Therefore, the transient high-frequency signal of the fault will be greatly attenuated through the wave arrester, so that the high-frequency signal before and after passing through the wave arrester has a significant difference. Based on this feature, a protection method for transmission lines and buses can be constructed. At the same time, the initial phase angle of the fault and the transition resistance have a great influence on the transient signal formed by the fault. The present invention utilizes the blocking and attenuation effect of the line wave arrester on the fault high-frequency signal, and considers the influence of the two fault condition attributes of the initial phase angle of the fault and the transition resistance on the transient high-frequency signal of the fault, and proposes a transient protection method based on fault information fusion and power system traveling waves.

[0081] like Figure 1 As shown, the power system transient protection method based on fault information fusion includes:

[0082] Step 1. Collect the line voltage u1 from the head end of the protected line and collect the bus voltage u2 from the protected bus.

[0083] Step 2. Extract the high-frequency component u of voltage u1 and u2 h1 and u h2 .

[0084] Step 3. Calculate u h1 and u h2 The transient quantity TQ1, TQ2. The transient quantity can be one of energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude, singular value and other transient high-frequency component processing quantity; or the instantaneous integral of these transient quantities, or the instantaneous cumulative sum of these transient quantities.

[0085] Step 4. Calculate the fault condition attributes: Calculate the initial phase angle θ of the fault f With transition resistance R f .

[0086] Step 5. Fault information fusion: transient quantities TQ1 and TQ2 are fused to the initial fault phase angle θ f and transition resistance R f The standard transient quantities NTQ1 and NTQ2 are obtained. The transient quantity fusion fault condition attribute adopts the Kalman filtering algorithm proposed in the present invention.

[0087] Step 6. Calculate the difference NTQD between the standard transient quantities NTQ1 and NTQ2 12 NTQD 12 =NTQ1-NTQ2.

[0088] Step 7. Use the standard transient quantity difference in step 6 as the fault feature to determine the fault. Take line L3 and bus M as the protected objects, and NTQD 12 As fault characteristics, protection criteria are constructed:

[0089] When NTQD 12 >Aset, it is judged as an internal fault of line L3;

[0090] When NTQD 12 <-Bset, it is judged as a bus fault;

[0091] Otherwise, it is determined that there is no fault inside the protection area, i.e., line L3 and bus M;

[0092] Among them, Aset and Bset are positive real numbers of protection settings.

[0093] S2. When the protection object is a single circuit.

[0094] In conjunction with the instruction manual Figure 1 Here, taking the protection line L3 as an example, the line transient direction longitudinal protection method is explained.

[0095] like Figure 1 As shown, the power system transient protection method based on fault information fusion includes:

[0096] Step 1. Collect the line voltage u1 and the bus voltage u2 from the head end of the protected line.

[0097] Step 2. Extract the high-frequency component u of voltage u1 and u2 h1 and u h2 .

[0098] Step 3. Calculate u h1 and u h2 The transient quantity TQ1 and TQ2. The transient quantity is one of signal energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude, singular value, transient quantity gradient and transient quantity relative gradient, etc.; or the instantaneous integral of these transient quantities, or the instantaneous cumulative sum of these transient quantities.

[0099] Step 4. Calculate the fault condition attributes: Calculate the initial phase angle θ of the fault f With transition resistance R f .

[0100] Step 5. Fault information fusion: transient quantities TQ1 and TQ2 are fused to the initial fault phase angle θ f and transition resistance R fThe standard transient quantities NTQ1 and NTQ2 are obtained. The transient quantity fusion fault condition attribute adopts the Kalman filtering algorithm proposed in the present invention.

[0101] Step 6. Calculate the difference NTQD between the standard transient quantities NTQ1 and NTQ2 12 NTQD 12 =NTQ1-NTQ2.

[0102] Step 7. Use the standard transient quantity difference in step 6 as the transient characteristic of the fault direction to determine the fault direction. 12 >Aset, it is judged as the equivalent information of "positive direction fault" from bus M to line L3; and the equivalent information of "positive direction fault" is sent to the other end. Otherwise, it is judged that there is no fault in the positive direction from bus M to line L3;

[0103] Step 8. If the local end determines that it is a positive direction fault and receives equivalent information of "positive direction fault" from the other end, it is determined that it is an internal fault of line L3, otherwise it is determined that line L3 has no fault.

[0104] Where Aset is a positive real number for protection setting. Note that Aset here is not the same as Aset in method 1, and there is no connection between them. The same applies to the following.

[0105] S3. When the protection object is a circuit. Figure 1 Here, the protection line L3 is taken as an example to illustrate the line single-ended transient protection method.

[0106] The power system transient protection method based on fault information fusion includes:

[0107] Step 1. Collect the line voltage u1 from the head end of the protected line.

[0108] Step 2. Extract the high-frequency component u of voltage u1 h1 .

[0109] Step 3. Calculate u h1 The transient quantity TQ1. The transient quantity is one of signal energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude integral, instantaneous amplitude, instantaneous amplitude cumulative sum, singular value, transient quantity gradient and transient quantity relative gradient.

[0110] Step 4. Calculate the fault condition attributes: Calculate the initial phase angle θ of the fault f and transition resistance R f .

[0111] Step 5. Fault information fusion: transient quantity TQ1 fusion fault initial phase angle θ f and transition resistance R fThe standard transient quantity NTQ1 is obtained. The transient quantity is integrated with the fault condition attribute using the Kalman filtering algorithm proposed in the invention.

[0112] Step 6. Use the standard transient quantity NTQ1 as the fault feature to judge the fault. Use NTQ1 as the fault feature to construct the protection judgment criterion:

[0113] When NTQ1>Aset, it is judged that line L3 has an internal fault;

[0114] Otherwise, it is determined that line L3 has no fault;

[0115] Where Aset is a positive real number for protection setting.

[0116] S4. When one busbar is taken as the protection object.

[0117] In conjunction with the instruction manual Figure 1 Here, bus M is taken as an example to illustrate the transient voltage bus protection method.

[0118] The power system transient protection method based on fault information fusion includes:

[0119] Step 1. Collect the bus voltage u2 from the protected bus.

[0120] Step 2. Extract the high-frequency component u of voltage u2 h2 .

[0121] Step 3. Calculate u h2 The transient quantity TQ2. The transient quantity is one of signal energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude, singular value, transient quantity gradient, transient quantity relative gradient; or the instantaneous integral of these transient quantities, or the instantaneous cumulative sum of these transient quantities.

[0122] Step 4. Calculate the fault condition attributes: Calculate the initial phase angle θ of the fault f and transition resistance R f .

[0123] Step 5. Fault information fusion: transient quantity TQ2 fusion fault phase angle θ f and transition resistance R f The standard transient quantity NTQ2 is obtained. The Kalman filtering algorithm proposed in the present invention is used to fuse the transient quantity and the fault condition attribute information.

[0124] Step 6. Use the standard transient quantity NTQ2 as the fault feature to judge the fault. Use NTQ2 as the fault feature to construct the protection judgment criterion:

[0125] When NTQ2>Aset, it is judged as bus M fault;

[0126] Otherwise, it is judged that busbar M has no fault;

[0127] Where Aset is a positive real number for protection setting.

[0128] In addition, the present invention uses machine learning (such as support vector machine, neural network, etc.) to complete the fault information fusion starting from step 5 in (1) to (4) and the fault judgment in step 6, using transient quantity, fault initial phase angle and transition resistance as input vectors, and outputs the corresponding line L3 fault, busbar M fault, no fault inside the protection zone, and line positive direction fault in (1) to (4). Before using machine learning (such as support vector machine, neural network, etc.) for fault judgment, the training fault sample set is used to train the machine learning, and then the trained machine learning is used for actual fault judgment.

[0129] The fault information fusion method of the present invention adopts a Kalman filter algorithm to fuse transient quantities and fault condition attribute information. The Kalman filter algorithm can be an extended Kalman filter (EKF) algorithm, or an unscented Kalman filter (UKF) algorithm, or a particle filter (PF) algorithm, or an extended Kalman particle filter (EPF) algorithm, or an unscented Kalman particle filter (UPF) algorithm. Taking the extended Kalman filter (EKF) algorithm as an example, the state equation and the observation equation are established as formula 1 and formula (2) respectively, and the Kalman recursive formula is constructed. Then, the initial state and covariance matrix are initialized, and the recursive calculation completes the fusion of the fault characteristics and the fault condition attribute information.

[0130] X(k) = X(k-1) + W(k) (1)

[0131] Y(k) = F(θ f , R f )X(k) + V(k) (2)

[0132] Where X(k) is the standard transient quantity after the fault quantity is fused with the fault condition attribute information; W(k) is the standard transient quantity noise, Q is the standard transient quantity noise variance; Y(k) is the fault quantity before fusion with the fault condition attribute information, i.e., the quasi-fault feature; V(k) is the quasi-fault feature noise, R is the quasi-fault feature noise variance; F(θ f ,R f ) is the fault condition attribute factor, which can be taken as F(θ f ,R f )=γ·(a·exp(b·R f )+c·exp(d·R f ))·(a0+a1·cos(θ f ·w)+b1·sin(θ f·w)+a2·cos(2·θ f ·w)+b2·sin(2·θ f ·w)). θ f ,R f are the calculated initial phase angle and transition resistance of the fault; γ is the adjustment constant, which can be taken as 1; a, b, c, d, a0, a1, b1, a2, b2, w are real constants.

[0133] The specific steps of the fusion process are as follows:

[0134] Step 1. Initialize the initial state X(0), Y(0), P(0). The initialization can be X(0) = 0.01, Y(0) = 0.01, P(0) = 1; where X(0) is the initial estimate, Y(0) is the initial observation, and P(0) is the initial covariance matrix.

[0135] Step 2. State prediction.

[0136] X(k|k-1)=X(k-1)

[0137] Step 3. Observe and predict.

[0138] Y(k|k-1)=F(θ f ,R f )X(k|k-1)

[0139] Step 4. Solve the state transfer matrix Φ(k).

[0140]

[0141] Step 5. Solve the observation transfer matrix H(k).

[0142]

[0143] Step 6. Find the covariance matrix prediction P(k|k-1).

[0144] P(k|k-1)=Φ(k)P(k-1|k-1)Φ T (k)+Q

[0145] Step 7. Calculate the Kalman filter gain.

[0146] K(k)=P(k|k-1)H T (k)(H(k)P(k|k-1)H T (k)+R) -1

[0147] Step 8. Request status update.

[0148] X(k)=X(k|k-1)+K(k)(Y(k)-H(k)X(k|k-1))

[0149] Step 9. Covariance update.

[0150] P(k)=(I n -K(k)H(k))P(k|k-1)

[0151] Example 1

[0152] The protection object is a circuit and the busbar connected to the circuit. Figure 1 , assuming that a single-phase grounding fault occurs on line L3, which is 5 km away from busbar M, and the initial phase angle of the fault is θ f =45°, transition resistance R f =5Ω. Taking the line L3 and bus M as an example, the transient unit protection method of the line and bus is described. The steps include:

[0153] Step 1. Collect the line voltage u1 from the head end of the protected line L3, and collect the bus voltage u2 from the protected bus M. The line voltage u1 and bus voltage u2 collected in this embodiment are as follows: Figure 2 shown.

[0154] Step 2. Extract the high-frequency component u of voltage u1 and u2 h1 and u h2 Here we use wavelet transform, which can of course be replaced by Fourier transform, classical mode decomposition, variational mode decomposition, S transform, filter bank, mathematical morphology, etc. to extract the high-frequency components u of u1 and u2 respectively. h1 and u h2 The high-frequency component u extracted in this embodiment h1 and u h2 as follows Figure 3 shown.

[0155] Step 3. Calculate u h1 and u h2 The transient energy TQ1 and TQ2 of u h1 and u h2 The Hilbert transform is performed to obtain their instantaneous amplitudes, and then integrated to obtain their transient energies TQ1 and TQ2. The standard transient energies TQ1 and TQ2 calculated in this embodiment are as follows: Figure 4 shown.

[0156] Step 4. Calculate the initial phase angle θ of the fault f and transition resistance R f . Fault initial phase angle θ f and transition resistance R f There are several ways to calculate this, two of which are listed here.

[0157] The first type, the initial phase angle of the fault θ f Calculation formula:

[0158] θ f =θ0-aL

[0159] Where θ0 is the voltage phase when the fault is detected at the protection installation, a is the phase coefficient of the transmission line, and L is the fault distance from the fault point to the protection installation. f Calculation formula:

[0160]

[0161] Where P is the active power supplied to the short-circuit point at the line protection installation, I a It is the fault phase current of the short-circuit point supplied to the line protection installation; λ is the power distribution coefficient of the power supply on the transmission side.

[0162] The second type, the initial phase angle of the fault is θ f Calculation formula:

[0163]

[0164] Where θ0 is the voltage phase when the fault is detected at the protection installation, T0, T a are the time when the phase current a,0 component first reaches the protection installation, T u For a voltage cycle, take T u =0.02s, x0 is the relative speed coefficient, x0=0.7. Transition resistance R f Calculation formula:

[0165]

[0166] In the formula They are the positive and negative sequence current special phases, They are positive and negative sequence voltage special phases, special phases are A, B or C phases, are the zero-sequence current phasor values ​​at the protection installations at both ends of the line, Z m1 , Z m2 , Z m0 =0.5 times the positive, negative and zero sequence impedance of the line. The initial fault phase angle θ calculated in this embodiment f =45°, transition resistance R f =5Ω.

[0167] Step 5. Fault information fusion: transient quantities TQ1 and TQ2 are fused with fault condition attributes θ f and R fThe standard transient quantities NTQ1 and NTQ2 are obtained. The transient quantity fusion fault condition attributes can adopt the Kalman filtering algorithm proposed in the present invention. The standard transient energy difference values ​​NTQ1 and NTQ2 calculated in this embodiment are as follows Figure 5 shown.

[0168] Step 6. Calculate the difference NTQD between the standard transient quantities NTQ1 and NTQ2 12 NTQD 12 =NTQ1-NTQ2. The standard transient energy difference NTQD calculated in this embodiment 12 as follows Figure 6 shown.

[0169] Step 7. Use the standard transient quantity difference as the fault feature to determine the fault. Take line L3 and bus M as the protected objects and NTQD 12 As fault characteristics, protection criteria are constructed:

[0170] When NTQD 12 >5, it is judged as an internal fault of line L3;

[0171] When NTQD 12 When <-50, it is judged as a busbar fault;

[0172] Otherwise, it is determined that there is no fault inside the protection area, i.e., line L3 and bus M;

[0173] In this embodiment, when the fault occurs, NTQD 12 The maximum value is 291, which is greater than 5, so it is judged that line L3 is faulty, which is correct.

[0174] Example 2

[0175] The protection object is a circuit. Figure 1 , assuming that a single-phase grounding fault occurs on line L3, which is 5 km away from busbar M, and the initial phase angle of the fault is θ f =45°, transition resistance R f =5Ω. Taking line L3 as an example, the line transient direction longitudinal protection method is described. The steps include:

[0176] Step 1. Collect the line voltage u1 from the head end of the protected line L3 and collect the bus voltage u2 of the bus M. The line voltage u1 and bus voltage u2 collected in this embodiment are the same as u1 and u2 in step 1 of embodiment 1.

[0177] Step 2. Extract the high-frequency component u of voltage u1 and u2 h1 and u h2Here we use wavelet transform, which can also be replaced by Fourier transform, classical mode decomposition, variational mode decomposition, S transform, filter bank, mathematical morphology and other methods to extract the high-frequency components u of u1 and u2 respectively. h1 and u h2 In this embodiment, the high frequency component u h1 and u h2 The same as u in step 2 of Example 1 h1 and u h2 same.

[0178] Step 3. Calculate u h1 and u h2 The transient energy TQ1 and TQ2 of u h1 and u h2 The Hilbert transform is performed to obtain their instantaneous amplitudes, and then integrated to obtain their transient energies TQ1 and TQ2. The transient energies TQ1 and TQ2 of this embodiment are the same as TQ1 and TQ2 in step 3 of the first embodiment.

[0179] Step 4. Calculate the initial phase angle θ of the fault f and transition resistance R f . The θ calculated in this embodiment is f =45°, R f =5Ω.

[0180] Step 5. Fault information fusion: transient quantities TQ1 and TQ2 are fused with fault condition attributes θ f and R f The standard transient quantities NTQ1 and NTQ2 are obtained. The transient quantity fusion fault condition attribute can adopt the Kalman filtering algorithm proposed in the present invention. The high-frequency component standard transient energy NTQ1 and NTQ2 of this embodiment are the same as NTQ1 and NTQ2 in step 5 of embodiment 1.

[0181] Step 6. Calculate the difference NTQD between the standard transient quantities NTQ1 and NTQ2 12 NTQD 12 =NTQ1-NTQ2. The standard transient energy difference NTQD in this embodiment 12 Same as NTQD in step 5 of Example 1 12 same.

[0182] Step 7. Use the standard transient quantity difference as the transient characteristic of the fault direction to determine the fault direction. 12 As the transient characteristic of fault direction, the protection criterion is: 12>1, it is judged that there is a positive direction fault from bus M to line L3; and the equivalent information of "positive direction fault" is sent to the other end; otherwise, it is judged that there is no fault in the positive direction from bus M to line L3. 12 When the maximum value is 291 and is greater than 1, it is determined to be a positive direction fault from the line L3 direction, so the "positive direction fault" equivalent information is sent to the opposite end N end protection. Similarly, the opposite end N end protection uses the same method as the M end to determine the fault direction. In this embodiment, the N end protection determines it as a positive direction fault, so it also sends the "positive direction fault" information to the M end protection.

[0183] Step 8. If the local end determines that it is a positive direction fault and receives the equivalent information of "positive direction fault" from the opposite end, it is determined that it is an internal fault of line L3, otherwise it is determined that line L3 has no fault. In this embodiment, the M-end protection determines that it is a "positive direction fault" and receives the "positive direction fault" information sent by the opposite end N-end protection, so it is determined that it is an internal fault of L3 line, and the judgment is correct.

[0184] Example 3

[0185] The protection object is a circuit. Figure 1 , assuming that a single-phase grounding fault occurs on line L3, which is 5 km away from busbar M, and the initial phase angle of the fault is θ f =45°, transition resistance R f =5Ω. Here, the protection line L3 is taken as an example to illustrate the line single-ended transient protection method.

[0186] The method for transient protection of a power system based on fault information fusion comprises the following steps:

[0187] Step 1. Collect the line voltage u1 from the head end of the protected line L3. The line voltage u1 collected in this embodiment is the same as the u1 in step 1 of embodiment 1.

[0188] Step 2. Extract the high-frequency component u of voltage u1 h1 Here we use wavelet transform, which can also be replaced by Fourier transform, classical mode decomposition, variational mode decomposition, S transform, filter bank, mathematical morphology and other methods to extract the high-frequency component u of u1. h1 The high frequency component u of this embodiment h1 The same as u in step 2 of Example 1 h1 same.

[0189] Step 3. Calculate u h1 The transient quantity TQ1. h1 The instantaneous amplitude is obtained by Hilbert transformation, and then integrated to obtain their transient energy TQ1. The transient energy TQ1 of this embodiment is the same as TQ1 in step 3 of embodiment 1.

[0190] Step 4. Calculate the initial phase angle θ of the fault f and transition resistance R f . The θ calculated in this embodiment f =45°, R f =5Ω.

[0191] Step 5. Fault information fusion: transient quantity TQ1 is fused with fault condition attribute θ f and R f The standard transient quantity NTQ1 is obtained. The transient quantity fusion fault condition attribute can adopt the Kalman filtering algorithm proposed in the present invention. The high-frequency component standard transient energy NTQ1 of this embodiment is the same as the NTQ1 in step 5 of embodiment 1.

[0192] Step 6. Use the standard transient quantity as the fault feature to judge the fault. Use NTQ1 as the fault feature to construct the protection criterion: when NTQ1>4, it is judged as an internal fault of line L3; otherwise, it is judged as no fault of line L3; where Aset is a positive real number set by the protection. In this embodiment, NTQ1 is 291, which is greater than 4, so it is judged as a fault of line L3, which is correct.

[0193] Example 4

[0194] The protection object is a busbar. Figure 1 , a single-phase grounding fault occurs on bus M, and the initial phase angle of the fault is θ f =45°, transition resistance R f =5Ω. Here, the bus M is taken as an example to illustrate the transient voltage bus protection method.

[0195] The method for fault information fusion and power system transient protection includes the following steps:

[0196] Step 1. Collect the bus voltage u2 from the protected bus M. The bus voltage u2 collected in this embodiment is as follows: Figure 7 shown.

[0197] Step 2. Extract the high-frequency component u of voltage u2 h2 Here we use wavelet transform, which can also be replaced by Fourier transform, classical mode decomposition, variational mode decomposition, S transform, filter bank, mathematical morphology and other methods to extract the high-frequency component u of u2. h2 The high frequency component u of this embodiment h2 as follows Figure 8 shown.

[0198] Step 3. Calculate u h2 The transient quantity TQ2. h2 Perform Hilbert transform to find its instantaneous amplitude, and then integrate to obtain their transient energy TQ2. The transient energy TQ2 of this embodiment is as follows: Fig. 9 shown.

[0199] Step 4. Calculate the initial phase angle θ of the fault f and transition resistance R f θ of this embodiment f =45°, R f =5Ω.

[0200] Step 5. Fault information fusion: transient quantity TQ2 fusion fault condition attribute θ f and R f The standard transient quantity NTQ2 is obtained. The Kalman filtering algorithm proposed in the present invention can be used to fuse the transient quantity and the fault condition attribute information. The standard transient quantity NTQ2 of this embodiment is the same as the NTQ2 in step 5 of embodiment 1. Fig.10 shown.

[0201] Step 6. Use the standard transient quantity as the fault feature to judge the fault. Use NTQ2 as the fault feature to construct the protection criterion: when NTQ2>50, it is judged that the bus M is faulty; otherwise, it is judged that the bus M is not faulty. In this embodiment, NTQ2 is 461, which is greater than 50, so it is judged that the line L3 is faulty, and the judgment is correct.

[0202] Example 5

[0203] The protection object is a circuit and the busbar connected to the circuit. Figure 1 , assuming that a single-phase grounding fault occurs on line L3, which is 5 km away from busbar M, and the initial phase angle of the fault is θ f =45°, transition resistance R f =5Ω. Taking the line L3 and bus M as the protected objects, the transient unit protection method of the line and bus using SVM is described. The steps include:

[0204] Step 1. Collect the line voltage u1 from the head end of the protected line L3, and collect the bus voltage u2 from the protected bus M. The line voltage u1 and bus voltage u2 collected in this embodiment are the same as u1 and u2 in step 1 of embodiment 1.

[0205] Step 2. Extract the high-frequency component u of voltage u1 and u2 h1 and u h2 Here we use wavelet transform, which can also be replaced by Fourier transform, classical mode decomposition, variational mode decomposition, S transform, filter bank, mathematical morphology and other methods to extract the high-frequency components u of u1 and u2 respectively. h1 and u h2 In this embodiment, the high frequency component u h1 and u h2 The same as u in step 2 of Example 1 h1 and u h2same.

[0206] Step 3. Calculate u h1 and u h2 The transient energy TQ1 and TQ2 of u h1 and u h2 The Hilbert transform is performed to obtain their instantaneous amplitudes, and then integrated to obtain their transient energies TQ1 and TQ2. The transient energies TQ1 and TQ2 of this embodiment are the same as TQ1 and TQ2 in step 3 of the first embodiment.

[0207] Step 4. Calculate the difference TQD between transient quantities TQ1 and TQ2 12 .TQD 12 =TQ1-TQ2. The transient energy difference TQD in this embodiment 12 as follows Fig.11 shown.

[0208] Step 5. Calculate the initial phase angle θ of the fault f and transition resistance R f . The θ calculated in this embodiment is f =45°, R f =5Ω.

[0209] Step 6. Fault information fusion and fault judgment: Use support vector machine (SVM) to judge the fault area. The fault area is represented by fz. When line L3 fails, it is represented by fz=1; when bus M fails, it is represented by fz=2; when line L3 and bus M are not faulty, it is represented by fz=0. 12 ,θ f and R f As the support vector machine SVM input vector, the SVM output is the fault area judgment result fz. The SVM is first trained with the training sample set, and the trained SVM can be used for fault area judgment. In this embodiment, the SVM is first trained with the training sample set to judge the fault area, and then the trained SVM is used to judge the fault of this embodiment, and it is judged to be a line L3 fault, which is correct.

[0210] Example 6

[0211] The protection object is a circuit. Figure 1 , assuming that a single-phase grounding fault occurs on line L3, which is 5 km away from busbar M, and the initial phase angle of the fault is θ f =45°, transition resistance R f =5Ω. Taking line L3 as an example, the line transient direction longitudinal protection method using SVM is described. The steps include:

[0212] Step 1. Collect the line voltage u1 from the head end of the protected line L3 and collect the bus voltage u2 of the bus M. The line voltage u1 and bus voltage u2 collected in this embodiment are the same as u1 and u2 in step 1 of embodiment 1.

[0213] Step 2. Extract the high-frequency component u of voltage u1 and u2 h1 and u h2 Here we use wavelet transform, which can also be replaced by Fourier transform, classical mode decomposition, variational mode decomposition, S transform, filter bank, mathematical morphology and other methods to extract the high-frequency components u of u1 and u2 respectively. h1 and u h2 In this embodiment, the high frequency component u h1 and u h2 The same as u in step 2 of Example 1 h1 and u h2 same.

[0214] Step 3. Calculate u h1 and u h2 The transient energy TQ1 and TQ2 of u h1 and u h2 The Hilbert transform is performed to obtain their instantaneous amplitudes, and then integrated to obtain their transient energies TQ1 and TQ2. The transient energies TQ1 and TQ2 of this embodiment are the same as TQ1 and TQ2 in step 3 of the first embodiment.

[0215] Step 4. Calculate the initial phase angle θ of the fault f and transition resistance R f . The θ calculated in this embodiment f =45°, R f =5Ω.

[0216] Step 5. Calculate the difference TQD between transient quantities TQ1 and TQ2 12 .TQD 12 =TQ1-TQ2. The transient energy difference TQD in this embodiment 12 Same as TQD in step 5 of Example 5 12 same.

[0217] Step 6. Fault information fusion and fault judgment: Use support vector machine (SVM) to judge the fault direction. The fault direction is represented by fz. When the line L3 faults in the positive direction, it is represented by fz=1; when the line L3 faults in the reverse direction, it is represented by fz=-1. 12 ,θ f and R fAs the support vector machine SVM input vector, the SVM output is the fault direction judgment result fz. The SVM is first trained with the training sample set, and the trained SVM can be used for fault direction judgment. In this embodiment, the SVM is first trained with the training sample set to judge the fault direction, and then the trained SVM is used to judge the fault direction of this embodiment, and it is judged to be a positive direction fault of line L3, which is correct.

[0218] Step 7. When the SVM determines that there is a positive direction fault, that is, fz=1, it sends equivalent information of "positive direction fault" to the opposite end. In this embodiment, it is determined that there is a positive direction fault in line L3, and the "positive direction fault" information is sent to the opposite end N. Similarly, the N-end protection uses the same method to determine the fault direction. In this embodiment, the N-end is also determined to be a positive direction fault, and the "positive direction fault" information is sent to the M-end protection.

[0219] Step 8. If the local end determines that it is a positive direction fault and receives the equivalent information of "positive direction fault" from the opposite end, it determines that it is an internal fault of line L3, otherwise it determines that line L3 has no fault. In this embodiment, the local end determines that it is a positive direction fault and receives the "positive direction fault" sent by the opposite end N, so it determines that there is a fault on line L3, which is correct.

[0220] Example 7

[0221] The protection object is a circuit. Figure 1 , assuming that a single-phase grounding fault occurs on line L3, which is 5 km away from busbar M, and the initial phase angle of the fault is θ f =45°, transition resistance R f =5Ω. Here, the protection line L3 is taken as an example to illustrate the line single-ended transient protection method using SVM.

[0222] The method for transient protection of a power system based on fault information fusion comprises the following steps:

[0223] Step 1. Collect the line voltage u1 from the head end of the protected line L3. The line voltage u1 collected in this embodiment is the same as the u1 in step 1 of embodiment 1.

[0224] Step 2. Extract the high-frequency component u of voltage u1 h1 Here we use wavelet transform, which can also be replaced by Fourier transform, classical mode decomposition, variational mode decomposition, S transform, filter bank, mathematical morphology and other methods to extract the high-frequency component u of u1. h1 In this embodiment, the high frequency component u h1 The same as u in step 2 of Example 1 h1 same.

[0225] Step 3. Calculate u h1 The transient quantity TQ1. h1The instantaneous amplitude is obtained by Hilbert transformation, and then integrated to obtain their transient energy TQ1. The transient energy TQ1 of this embodiment is the same as TQ1 in step 3 of embodiment 1.

[0226] Step 4. Calculate the initial phase angle θ of the fault f and transition resistance R f . The θ calculated in this embodiment f =45°, R f =5Ω.

[0227] Step 5. Fault information fusion and fault judgment: Use support vector machine (SVM) to judge the fault area. The fault area is represented by fz. When line L3 is faulty, fz=1; when line L3 is not faulty, fz=0. f and R f As the support vector machine SVM input vector, the SVM output is the fault area judgment result fz. The SVM is first trained with the training sample set, and the trained SVM can be used for fault area judgment. In this embodiment, the SVM is first trained with the training sample set to judge the fault area, and then the trained SVM is used to judge the fault area of ​​this embodiment, and it is judged to be a fault on line L3, which is correct.

[0228] Example 8

[0229] The protection object is a busbar. Figure 1 , assuming that a single-phase grounding fault occurs on bus M, the initial phase angle of the fault is θ f =45°, transition resistance R f =5Ω. Here, bus M is taken as an example to illustrate the transient voltage bus protection method using SVM.

[0230] The method for fault information fusion and power system transient protection includes the following steps:

[0231] Step 1. Collect the bus voltage u2 from the protected bus M. The bus voltage u2 collected in this embodiment is the same as the u2 in step 1 of embodiment 4.

[0232] Step 2. Extract the high-frequency component u of voltage u2 h2 Here we use wavelet transform, which can also be replaced by Fourier transform, classical mode decomposition, variational mode decomposition, S transform, filter bank, mathematical morphology and other methods to extract the high-frequency component u of u2. h2 In this embodiment, the high frequency component u h2 The same as u in step 2 of Example 4 h2 same.

[0233] Step 3. Calculate u h2 The transient quantity TQ2. h2Perform Hilbert transform to find its instantaneous amplitude, and then integrate to obtain their transient energy TQ2. The transient energy TQ2 of this embodiment is the same as TQ2 in step 3 of embodiment 4.

[0234] Step 4. Calculate the initial phase angle θ of the fault f and transition resistance R f . The θ calculated in this embodiment f =45°, R f =5Ω.

[0235] Step 5. Fault information fusion and fault judgment: Use support vector machine (SVM) to judge the fault area. The fault area is represented by fz. When bus M is faulty, fz=1; when bus M is not faulty, fz=0. f and R f As the support vector machine SVM input vector, the SVM output is the fault area judgment result fz. The SVM is first trained with the training sample set, and the trained SVM can be used for fault area judgment. In this embodiment, the SVM is first trained with the training sample set to judge the fault area, and then the trained SVM is used to judge the fault area of ​​this embodiment, and it is judged to be a bus M fault, which is correct.

[0236] Example 9

[0237] A fault information fusion method is an embodiment of a method for fusing transient quantity and fault condition attribute information using a Kalman filtering algorithm.

[0238] In this embodiment, the extended Kalman filter (EKF) algorithm is used to integrate the transient quantity TQ1 in step 5 of embodiment 1 with the fault condition attribute θ f and R f Take the standard transient quantity NTQ1 as an example.

[0239] The state equation and observation equation are established as equation (1) and equation (2) respectively, and the Kalman recursive formula is constructed. Then, the initial state and covariance matrix are initialized, and the recursive calculation is performed to complete the fusion of fault characteristics and fault condition attribute information.

[0240] X(k) = X(k-1) + W(k) (1)

[0241] Y(k) = F(θ f , R f )X(k) + V(k) (2)

[0242] Where X(k) is the standard transient quantity integrated with fault condition attribute information; W(k) is the standard transient quantity noise, whose mean is 0, Q is the standard transient quantity noise variance, which can be taken as Q=0.01; Y(k) is the transient quantity before integrating the fault condition attribute information, i.e., the quasi-fault feature; V(k) is the quasi-fault feature noise, whose mean is 0, R is the quasi-fault feature noise variance, which can be taken as R=0.01; θ f ,R f is the calculated initial phase angle and transition resistance of the fault; F(θ f ,R f ) is the fault condition attribute factor function, which includes θ f ,R f Fault information can be implemented in many specific ways, one of which can be F(θ f ,R f )=γ·(a·exp(b·R f )+c·exp(d·R f ))·(a0+a1·cos(θ f ·w)+b1·sin(θ f ·w)+a2·cos(2·θ f ·w)+b2·sin(2·θ f ·w)). γ is the adjustment constant, which can be taken as 1; a, b, c, d, a0, a1, b1, a2, b2, w are constants, which can be taken as a=128.3, b=-0.01102, c=163.6, d=-0.001644, a0=38.72, a1=-43.86, b1=255.5, a2=5.494, b2=1.328, w=0.01935 respectively.

[0243] The specific steps are as follows:

[0244] Step 1. Initialize the initial state X(0), Y(0), and P(0). The initialization can be X(0) = 0.01, Y(0) = 0.01, and P(0) = 1;

[0245] Step 2. State prediction:

[0246] X(k|k-1)=X(k-1)

[0247] Step 3. Observation prediction:

[0248] Y(k|k-1)=F(θ f ,R f )X(k|k-1)

[0249] Step 4. Solve the state transfer matrix φ(k):

[0250]

[0251] Step 5. Solve the observation transfer matrix H(k):

[0252]

[0253] Step 6. Find the covariance matrix prediction P(k|k-1):

[0254] P(k|k-1)=φ(k)P(k-1|k-1)φ T (k)+Q

[0255] Step 7. Calculate the Kalman filter gain:

[0256] K(k)=P(k|k-1)H T (k)(H(k)P(k|k-1)H T (k)+R) -1

[0257] Step 8. Request status update:

[0258] X(k)=X(k|k-1)+K(k)(Y(k)-H(k)X(k|k-1))

[0259] Step 9. Covariance update:

[0260] P(k)=(I n -K(k)H(k))P(k|k-1).

[0261] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A power system transient protection method based on fault information fusion, characterized in that: When the protection object is a circuit and a busbar connected to the circuit, the steps of transient unit protection include: Step 1. Collect the head-end voltage u1 of the protected line and the protected bus voltage u2; Step 2. Extract the high-frequency components u of voltage u1 and u2 respectively h1 and u h2 ; Step 3. Calculate the high-frequency components u separately h1 and u h2 The transient quantity TQ1 and TQ2 are selected from the group consisting of signal energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude, singular value, transient quantity gradient and transient quantity relative gradient; or the instantaneous integral of the transient quantity, or the instantaneous cumulative sum of the transient quantity; Step 4. Calculate the fault initial phase angle θ of the fault condition attribute f and transition resistance R f ; Step 5. Fault information fusion: transient quantities TQ1 and TQ2 fusion fault condition attributes fault initial phase angle θ f and transition resistance R f Obtain standard transient quantities NTQ1 and NTQ2; Step 6. Calculate the difference NTQD between the standard transient quantities NTQ1 and NTQ2 12 ,NTQD 12 =NTQ1-NTQ2; Step 7. Use the standard transient quantity difference as the fault feature to determine the fault; when NTQD 12 >Aset, it is judged as an internal line fault; When NTQD 12 <-Bset, it is judged as a busbar fault; otherwise, it is judged as no fault inside the protection area; where Aset and Bset are positive real numbers set by the protection.

2. The method according to claim 1 is characterized in that, in steps 5 to 7, a support vector machine (SVM) or a neural network can be used to complete fault information fusion and fault judgment. When a support vector machine is used, the steps are as follows: Calculate the difference TQD between the transient quantities TQ1 and TQ2 12 , TQD 12 =TQ1-TQ2, TQD 12 、Fault initial phase angle θ f and transition resistance R f As the input vector of the support vector machine SVM, the SVM output is the fault area judgment result fz, where: First, the SVM is trained with the training sample set, and then the trained SVM is used for fault area judgment.

3. A method for transient protection of power system based on fault information fusion, characterized in that: When the protection object is a single-circuit line, the transient directional pilot protection steps include: Step 1. Collect the voltage u1 of the head end of the protected line and the voltage u2 of the protected busbar; Step 2. Extract the high-frequency components u of voltage u1 and u2 respectively h1 and u h2 ; Step 3. Calculate the high-frequency components u separately h1 and u h2 The transient quantity TQ1 and TQ2 are selected from the group consisting of signal energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude, singular value, transient quantity gradient and transient quantity relative gradient; or the instantaneous integral of the transient quantity, or the instantaneous cumulative sum of the transient quantity; Step 4. Calculate the fault initial phase angle θ of the fault condition attribute f and transition resistance R f ; Step 5. Fault information fusion: transient quantities TQ1 and TQ2 fusion fault condition attributes fault initial phase angle θ f and transition resistance R f Obtain standard transient quantities NTQ1 and NTQ2; Step 6. Calculate the difference NTQD between the standard transient quantities NTQ1 and NTQ2 12 ,NTQD 12 =NTQ1-NTQ2; Step 7. Use the standard transient quantity difference as the transient characteristic of the fault direction to determine the fault direction; when NTQD 12 >Aset, it is judged that the line is faulty in the positive direction, and sends equivalent information of "fault in the positive direction" to the other end; otherwise, it is judged that there is no fault in the positive direction of the line; Step 8. If the local protection judges it as "positive direction fault" and receives equivalent information of "positive direction fault" from the other end protection, it is judged to be an internal line fault, otherwise it is judged that there is no line fault.

4. The method according to claim 3, characterized in that Steps 5 to 7 can use support vector machine SVM or neural network to complete fault information fusion and fault judgment. When using support vector machine SVM, the steps are as follows: Calculate the difference TQD between transient quantities TQ1 and TQ2 12 , TQD 12 =TQ1-TQ2, TQD 12 、Fault initial phase angle θ f and transition resistance R f As the input vector of the support vector machine SVM, the SVM output is the fault direction judgment result fz, wherein the SVM is first trained with the training sample set, and then the trained SVM is used for fault direction judgment; When the SVM determines that there is a positive direction fault, it sends the equivalent information of "positive direction fault" to the other end, and determines that the line is a positive direction fault, and sends the "positive direction fault" information to the other end; When the local end determines that there is a positive direction fault and receives equivalent information of "positive direction fault" from the other end, it determines that there is an internal line fault, otherwise it determines that there is no line fault.

5. A method for transient protection of a power system based on fault information fusion, characterized in that: When the protection object is a single-ended transient protection of a line, the steps include: Step 1. Collect the voltage u1 at the head end of the protected line; Step 2. Extract the high-frequency component u of voltage u1 h1 ; Step 3.u h1 Perform Hilbert transform to find its instantaneous amplitude, then integrate it to get its transient quantity TQ1, where the transient quantity is one of signal energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude integral, instantaneous amplitude, instantaneous amplitude cumulative sum, singular value, transient quantity gradient and transient quantity relative gradient; Step 4. Calculate the fault initial phase angle θ of the fault condition attribute f and transition resistance R f ; Step 5. Fault information fusion: transient quantity TQ1 fusion fault initial phase angle θ of fault condition attributes f and transition resistance R f Get the standard transient quantity NTQ1; Step 6. Use the standard transient quantity as the fault feature to judge the fault; use NTQ1 as the fault feature to construct the protection criterion: when NTQ1>Aset, it is judged as an internal line fault; otherwise, it is judged as no line fault; where Aset is a positive real number set by the protection.

6. The method according to claim 5, characterized in that Step 5-6 can use support vector machine SVM or neural network to complete fault information fusion and fault judgment. When using support vector machine, it specifically includes: taking TQ1, fault initial phase angle θ f and transition resistance R f As the support vector machine SVM input vector, the SVM output is the fault area judgment result fz, wherein the SVM is first trained with the training sample set, and then the trained SVM is used for fault area judgment.

7. A method for transient protection of a power system based on fault information fusion, characterized in that: When the protection object is a bus, the transient voltage bus protection includes the following steps: Step 1. Collect the protected bus voltage u2; Step 2. Extract the high-frequency component u of voltage u2 h2 ; Step 3. Calculate u h2 The transient quantity TQ2 is one of signal energy, entropy of fault signal, signal complexity, signal singularity, Li index, instantaneous amplitude, singular value, transient quantity gradient and transient quantity relative gradient; or the instantaneous integral of the transient quantity, or the instantaneous cumulative sum of the transient quantity; Step 4. Calculate the fault initial phase angle θ of the fault condition attribute f and transition resistance R f ; Step 5. Fault information fusion: transient quantity TQ2 fusion fault initial phase angle θ of fault condition attributes f and transition resistance R f Get the standard transient quantity NTQ2; Step 6. Use the standard transient quantity as the fault feature to judge the fault, and use NTQ2 as the fault feature to construct the protection criterion: when NTQ2>Aset, it is judged as a bus fault; otherwise, it is judged as a bus no fault; where Aset is a positive real number set by the protection.

8. The method according to claim 7, characterized in that Step 5-6 can use support vector machine or neural network to complete fault information fusion and fault judgment. When using support vector machine, the steps are as follows: TQ2, θ f and R f As the support vector machine SVM input vector, the SVM output is the fault area judgment result fz, wherein the SVM is first trained with the training sample set, and then the trained SVM is used for fault area judgment.

9. The method for transient protection of a power system based on fault information fusion according to claim 1 or 3 or 5 or 7, characterized in that: Step 5: Fault information fusion is implemented by using Kalman filter algorithm or improved Kalman filter algorithm or particle filter algorithm or improved particle filter algorithm.

10. The method according to claim 9, characterized in that: When the Kalman filter algorithm is used, the steps include: Step 1. Initialize the initial state X(0), Y(0), and P(0), and set X(0) = 0.01, Y(0) = 0.01, and P(0) = 1; Step 2. State prediction: X(k|k-1)=X(k-1) Step 3. Observation prediction: Y(k|k-1)=F(θ f ,R f )X(k|k-1) Step 4. Solve the state transfer matrix Φ(k): Step 5. Solve the observation transfer matrix H(k): Step 6. Find the covariance matrix prediction P(k|k-1): P(k|k-1)=Φ(k)P(k-1|k-1)Φ T (k)+Q Step 7. Calculate the Kalman filter gain: K(k)=P(k|k-1)H T (k)(H(k)P(k|k-1)H T (k)+R) -1 Step 8. Request status update: X(k)=X(k|k-1)+K(k)(Y(k)-H(k)X(k|k-1)) Step 9. Covariance update: P(k)=(I n -K(k)H(k))P(k|k-1); Among them, X(k) is the standard transient quantity after the fault quantity is integrated with the fault condition attribute information; Q is the standard transient quantity noise variance; Y(k) is the fault quantity before integrating the fault condition attribute information, that is, the quasi-fault feature; R is the quasi-fault feature noise variance; F(θ f ,R f ) is the fault condition attribute factor, θ f ,R f are the fault initial phase angle and transition resistance respectively.

Images