Research on Cross-Circuitry Fault Identification Method for AC/DC Transmission System Based on Blind Signal Separation Algorithm

Abstract

The AC/DC transmission system is an important component of the power system, and the cross-circuitry Fault diagnosis of the AC/DC transmission system plays an important role in ensuring the normal operation of power equipment and personal safety. The traditional AC/DC transmission detection methods have the characteristics of complex detection processes and low fault line identification rates. Aiming at such problems, this paper proposes a new method of cross-circuitry Fault diagnosis based on the AC/DC transmission system based on a blind signal separation algorithm. Firstly, the method takes the typical cross-circuitry Fault scenario as an example to construct the topology diagram of the AC/DC power transmission system. Then, the electrical signals of the AC system and the DC system of the AC/DC power transmission system are collected, and the collected signals are extracted by the blind signal separation algorithm. Then, aiming at the cross-circuitry Fault problem of the DC system, the electrical quantities of the positive and negative poles on the rectifier side and the inverter side are collected, and the characteristics of the electrical quantities are analyzed by wavelet to determine the fault. At the same time, aiming at the problem of the cross-circuitry Fault of the AC system, three fault types of cross-circuitry Fault, ground fault, and intact fault are set up, and the electrical quantities of A, B, and C are collected on the same side, and the characteristics of three-phase electrical quantities are analyzed by wavelet. Finally, the cross-circuitry Fault judgment interval of the AC/DC system is set as the basis of fault judgment. After experimental verification, the relative error of the model is 1.4683%. The crossline fault identification method of the AC/DC transmission system based on the blind source separation algorithm proposed in this paper can accurately identify the crossline fault location and identify the fault type. It also provides theoretical and experimental support for power system maintenance personnel to maintain equipment.

1. Introduction

With the advantages of improving energy efficiency, supporting renewable energy integration, and optimizing power quality, the AC/DC transmission system has become a hotspot of research and application [1]. However, due to the increasing complexity of AC/DC transmission systems and the interaction between AC/DC components, the research and development of fault detection technology have become more complex [2].

Due to the variety of erection methods and fault types of parallel lines on the same tower, relay protection is prone to mis-operation or rejection [3]. In recent years, non-fault line protection tripping caused by line faults on the same tower has occurred from time to time, causing transmission corridor fracture accidents [4]. Because the AC/DC transmission system is erected on the same tower, it is possible to have AC/DC crossline fault and ground fault. The crossline fault of the system will lead to direct electrical connection between the two unrelated AC/DC systems through the contact point. Due to the installation of AC/DC systems on the same tower, the electromagnetic relationship between wires is unclear. Once a fault occurs, significant transient characteristic quantities will be generated on the faulty line, making the analysis of fault characteristics complex. Therefore, fault identification characteristics also pose a new challenge to the protection of AC/DC systems [5]. In-depth study of the fault characteristics of AC/DC transmission system lines and targeted improvement and optimization of the existing AC/DC transmission system will provide better theoretical support for future operation of AC/DC transmission systems.

The difficulty of fault detection in AC/DC hybrid transmission networks lies in the complexity of the system, which includes active and passive components, making it difficult to accurately capture fault characteristics [6]. The AC/DC transmission system puts forward higher requirements for the detection method because of the difficulty in identifying the interference fault characteristics of its own system and external environment. There are several methods for detecting AC/DC systems:

(1) Fault detection based on power grid system model [7]. This method relies on the accurate mathematical model of the power grid and identifies faults by comparing the difference between real-time monitoring data and model prediction values. This method can theoretically achieve high-precision fault detection, but its performance heavily depends on the accuracy of the model. In practical applications, changes in system parameters and model simplification may lead to inaccurate detection results [8].

(2) Fault detection based on artificial intelligence technology [9]. In recent years, with the breakthroughs in artificial intelligence technology, fault detection methods based on machine learning and deep learning have received extensive attention. Using training models, the behavior characteristics of power grids under different fault modes are identified, and fast and accurate fault detection is realized. The advantage of this type of method is that it can handle nonlinear and high-dimensional data, and as the amount of data increases, the detection performance continues to improve. However, this method requires a large amount of labelled data for training, and the generalization ability and interpretability of the model need to be improved [10].

(3) Fault detection based on edge computing and Internet of Things technology [11]. With the development of Internet of Things technology, some computing tasks of fault detection are dispersed to each node of the power grid through edge computing technology, which can reduce the computing burden of the central processor and achieve faster response times. By combining high-density deployment of IoT devices, real-time and accurate monitoring of the power grid status can be achieved, improving the accuracy and real-time nature of fault detection. However, this requires efficient communication protocols and network security guarantees [12].

(4) Fault detection based on signal processing technology [13]. Changes in the collected voltage or current signals are analyzed using advanced signal processing techniques, such as Fourier transform, wavelet transform, etc., to find out the fault characteristics [14]. However, signal processing methods usually require a lot of computing resources, and the method is sensitive to signal noise [5].

The AC/DC hybrid transmission network is composed of the AC transmission network and the DC transmission network. The advantages in expansibility make its construction more scalable, and the transmission network can be flexibly constructed to make the transmission network safer, more reliable, stable, and efficient [15]. Aiming at the shortcomings of signal processing methods, this paper proposes a new method for cross-circuitry Fault diagnosis of AC/DC transmission systems based on the blind source separation algorithm. Firstly, the method takes the typical cross-circuitry Fault scenario as an example to construct the topology diagram of the AC/DC power transmission system. At the same time, the electrical signals of the AC system and the DC system of the AC/DC power transmission system are collected, and the collected signals are extracted by the blind signal separation algorithm. Then, aiming at the cross-circuitry Fault problem of the DC system, the electrical quantities of the positive and negative poles on the rectifier side and the inverter side are collected, and the characteristics of the electrical quantities are analyzed by wavelet to determine the fault. At the same time, aiming at the problem of cross-circuitry Fault of AC system, three fault types—cross-circuitry Fault, ground fault, and intact fault—are set up and the electrical quantities of A, B, and C are collected on the same side, and the characteristics of three-phase electrical quantities are analyzed by wavelet. Finally, the cross-circuitry Fault judgment interval of the AC/DC system is set as the basis of fault judgment. Through experimental verification, the cross-circuitry Fault identification method for AC/DC transmission systems based on the blind signal separation algorithm proposed in this paper can accurately identify the cross-circuitry Fault location, providing theoretical and experimental support for power system maintenance personnel to maintain equipment.

2. Structural Design of AC/DC Transmission Systems

The emergence of AC/DC hybrid transmission networks is aimed at overcoming the limitations of traditional AC transmission networks and improving the efficiency and reliability of the power system. The AC/DC hybrid transmission network is expected to provide better support for energy transformation and sustainable development by optimizing power transmission and transmission methods [16]. At present, research on AC/DC hybrid transmission networks mainly focuses on system planning, design, and optimization scheduling. There is relatively little research on the fault characteristics analysis and protection of AC/DC hybrid transmission networks [17]. It is of critical and practical significance to conduct fault characteristics analysis and subsequent protection strategy research on AC/DC hybrid transmission networks.

2.1. System Architecture of AC/DC Transmission Systems

According to the actual AC and DC tower structure and structural parameters, this paper uses PSCAD/EMTDC software to design the topology diagram of the typical AC and DC transmission systems as shown in Figure 1. This topology diagram consists of two parts: the AC system and the DC system. The AC system is divided into two parts: AC M side and AC N side. Each side is connected with A, B, and C phases [18]. The voltage and current of the system are ±400 KV and ±4 kA, respectively. The sampling frequency of the system is set to 20 kHz, and the cable length is 100 km. The DC system is divided into DC positive and DC negative poles [19]. The left and right sides of the DC system are the rectifier side and the inverter side. The system voltage and current are ±200 kV and 1 kA, respectively. The positive and negative poles are represented as P and N poles, respectively. During operation, the power supply on the AC side is connected in series impedance to the bridge arm composed of various phase switching devices. The controller accurately controls the switching state of each IGBT according to the required output voltage and current, thereby controlling the voltage of each phase. The switching devices of each phase chop the combined PWM signal to achieve inversion, converting the current into DC [20].

Figure 2 shows the current waveforms of the AC system and the DC system of the AC/DC power transmission system. It can be seen from the figure that when the line is fault-free, the electrical signal of the AC/DC power transmission system collected by the sensor is stable.

As shown in the above figure, the orange shaded area represents the current waveform diagram during faults in the AC/DC transmission system, while the rest of the area represents the waveform diagram during normal operation of the AC/DC transmission system. From the figure, it can be seen that the current of the DC system (represented by Figure 2a,b) fluctuates greatly when the system fails. In the AC system, because the fault occurs on the P-Electrode of the DC system and the A-phase of the AC system, the A-phase of the AC system is greatly affected by the fault and the B-phase and C-phase cause fluctuations in the electrical quantity due to coupling [21]. Therefore, the characteristics of the currents of the AC and DC transmission systems can be used to determine the fault type and fault location.

2.2. Analysis of Cross-Circuitry Fault System Architecture

Based on the actual engineering structure, this paper takes the common typical cross-circuitry Fault scenario of DC P pole–AC A phase as an example. The fault point on the DC line is F₁, the fault point on the AC line is F₂, and the transition resistance is Rf. The length of the DC system line cable is 180 km, and the length of the AC system line cable is 100 km. In Figure 3, $U_a$, $U_b$, and $U_c$ represent the measured AC voltage; $I_a$, $I_b$, and $I_c$ represent the measured AC current; and subscripts 1 and 2 represent the M and N sides of the AC system, respectively. The typical topology diagram of AC/DC crossline faults is shown in Figure 3.

2.3. Principles of BSS Algorithm

When using sensors to collect signals in AC/DC transmission systems, they will be disturbed and affected by a variety of signal sources, including the normal operation interference of electrical equipment and the electromagnetic interference of the surrounding environment [22]. Therefore, in the process of measuring the electrical quantity of AC/DC transmission network, it is necessary to effectively process and analyze the acquired multi-source signals, the most important of which is the processing and extraction of characteristic signals.

In order to solve the problems of noise and interference, this paper uses the BSS algorithm [23] to process and analyze the collected AC and DC signals. The BSS algorithm refers to the process of recovering the source signal from the observed signal only, without knowing the type, characteristics, and transmission channel of the source signal [24]. A flowchart of the BSS algorithm is shown in Figure 4.

When the BSS algorithm is used to process and analyze the measured AC and DC signals, the AC and DC signals are separated from the measured mixed signals and the interference from noise signals is eliminated to improve the accuracy of AC and DC signal recognition [25]. The mathematical model of the BSS algorithm is shown in Equation (1).

$$S=F·X$$

(1)

where $X={\left[x_1,x_2,x_3,\dots ,x_n\right]}^T$ is the signal collected by mixing n original AC/DC signals and interference signals measured at various points on the observation line. $S={\left[s_1,s_2,s_3,\dots ,s_m\right]}^T$ is m independent signal sources, and F is a separation matrix of $n\times m$. The function of the BSS algorithm is to obtain a separation matrix through a certain calculation method when both the signal source and the mixing matrix are unknown. Using this separation matrix F, the original AC/DC signals can be separated from the mixed signal.

In this paper, the fast independent component analysis (Fast-ICA) algorithm is added to the traditional BSS algorithm to separate the mixed signal in order to increase the separation accuracy of the traditional BSS algorithm. Fast-ICA is mainly used in the field of blind source separation and feature extraction. The Fast-ICA algorithm is used to find the projection of the data in a specific direction to show the greatest degree of non-Gaussian characteristics, thereby improving the accuracy of extracting the original signal [26].

3. Design of a Fault Identification Scheme Based on the AC/DC Transmission System

This paper proposes a new method for crossline fault diagnosis in AC/DC transmission systems based on the blind source separation algorithm, which addresses the complex fault characteristics and low identification rate of AC/DC crossline faults. The overall flowchart of the method is shown in Figure 5.

Firstly, the method takes the typical cross-circuitry Fault scenario as an example to construct the topology diagram of the AC/DC power transmission system. At the same time, the electrical signals of the AC system and the DC system of the AC/DC power transmission system are collected, and the collected signals are extracted by the blind signal separation algorithm. Then, aiming at the cross-circuitry Fault problem of the DC system, the electrical quantities of the positive and negative poles on the rectifier side and the inverter side are collected, and the characteristics of the electrical quantities are analyzed by wavelet to determine the fault. At the same time, aiming at the problem of cross-circuitry Fault of the AC system, three fault types—cross-circuitry Fault, ground fault, and intact fault—are set up and the electrical quantities of A, B, and C are collected on the same side, and the characteristics of three-phase electrical quantities are analyzed by wavelet. Finally, the cross-circuitry Fault judgment interval of the AC/DC system is set as the basis of fault judgment.

3.1. Design of DC System Fault Detection Scheme

Figure 3 shows the typical AC/DC cross-circuitry Fault topology diagram. The system includes a DC bipolar network unit and an AC network unit. In the model, a cross-circuitry Fault occurs between the AC side A phase and any pole (P or N electrode) on the DC side. The fault unit is located on the line on the positive or negative pole of the DC system cable. Because of its large scale, complex structure, and strong electromagnetic interference, the DC network is associated with difficult detection and large errors during cross-circuitry Fault diagnosis. Figure 6 shows the flowchart of cross-circuitry Fault detection on the DC side of the AC/DC fault.

As shown in Figure 6, the DC bipolar network P pole–AC network A phase crossline is selected as the cross-circuitry Fault scenario. The fault point on the DC network line is F₁, and the fault point on the AC network line is F₂. In this fault scenario, first collect the positive and negative currents of the rectifier and inverter sides of the DC network. Record the current values collected on the left and right sides of the P Electrode of the DC side as I_P1 and I_P2, and the current values collected on the left and right sides of the N Electrode of the DC side as I_N1 and I_N2; perform wavelet decomposition on I_P1, I_P2, I_N1, and I_N2 to extract the d4 wavelet spectra of the four time–domain signals I_P1, I_P2, I_N1, and I_N2; subsequently, subtract the rectifier side and inverter side of the DC network system, denoted as A = d4(I_P1) − d4 (I_P2) and B = d4 (I_N1) − d4 (I_N2); finally, the difference between A and B is compared with 0.1. If it is greater than 0.1, it indicates that there is a cross-circuitry Fault in this section of the line. This method can accurately identify a cross-circuitry Fault on the positive and negative lines.

3.2. Design of AC System Fault Detection Scheme

The detection flowchart shown in Figure 7 is designed for fault identification on the AC side of typical AC/DC transmission systems. Because the electrical signals of cross-circuitry Fault and ground fault in the AC system are similar, three fault types—cross-circuitry Fault, ground fault, and no fault—are set up to identify the DC system experiment [5].

As shown in Figure 3, a cross-circuitry Fault occurs between the AC side A phase and the DC side P Electrode. The fault unit is located on the positive line of the bipolar cable of the DC system. The experiment synchronously sets the AC system A phase to have a cross-circuitry Fault. Because of its large scale and complex structure, the AC/DC network is associated with difficult detection and large errors during cross-circuitry Fault diagnosis.

This study selected the scenario of crossline fault between the P-Electrode of the DC bipolar network and the A-phase of the AC network. The fault point on the DC bipolar network line is F₁, and the fault point on the AC network line is F₂. In this fault scenario, the three-phase AC current values of A, B, and C in the AC network are collected and the characteristics of the current are analyzed to diagnose AC-side faults and locate AC/DC cross-circuitry Faults. After that, a metal grounding fault scenario is set up in phase A of the AC/DC transmission system, and the fault point and cross-circuitry Fault breakpoint are consistent as F₁. After setting the model, this paper first randomly collects the three-phase AC current values of A, B, and C in the AC network, which are recorded as I_i; then, the AC current value I_i is decomposed by wavelet, and the time–domain curve of the low frequency wavelet frequency band after decomposition is taken and the curve is recorded as L(I_i), and the smoothness analysis of the curve L(I_i) is recorded as S(I_i); finally, the interval division of S(I_i) measured. If $S(I_i)\in [0,0.1]$, it is recorded as no fault, if $S(I_i)\in (0.1,0.2]$, it is recorded as a phase grounding fault, and if $S(I_i)>0.2$, it is recorded as a phase cross-circuitry Fault.

3.3. BSS Algorithm Denoising Analysis

Fast-ICA is a fast and excellent iterative method based on negative entropy [27]. During the separation process, the separation of the observed signal X is completed when the negative entropy of the separated estimated signal Y reaches its maximum [28], as shown in Equation (2).

$$N(Y)={(E\left\{g(Y)\right\}-E\left\{g(Y_{Gauss})\right\})}^2$$

(2)

where E is the mean value operation; g is a nonlinear function; and $Y_{Gauss}$ is a Gaussian random variable with the same variance as Y.

Fast-ICA is used to find a direction $Y=W^{T}X$ with maximum non-Gaussianity, and non-Gaussianity is usually judged by the approximation of negative entropy $N(W^{T}X)$. Assuming that $w$ is a column vector of the separation matrix W, X is a whitened observation signal with zero mean and unit covariance, and $Y=W^{T}X$ is brought into the above equation to obtain Equation (3).

$$N(W^{T}X)={(E\left\{g(W^{T}X)\right\}-E\left\{g(Y_{Gauss})\right\})}^2$$

(3)

In order to find the optimal matrix $W_0$ and maximize the negative entropy of $Y=W^{T}X$, an independent component $s_i(t)=Y$ can be obtained. Under the constraint of $E\left\{Xg{(W^{T}X)}^2\right\}=\left|\right|W|{|}^2=1$, the optimal value of $E\left\{g(W^{T}X)\right\}$ can satisfy Equation (4).

$$F(W)=E\left\{Xg(W^{T}X)\right\}+\beta W=0$$

(4)

where $\beta =E\left\{{W^T}_{0}Xg(W^{T}X)\right\}$ is a constant value. $W_0$ is the optimized $W$ value.

Using F (W) in F representation (4), the Jacobian matrix $J_F(W)$ of F can be obtained as shown in the following Equation (5).

$$J_F(W)=E\left\{XX{g}^{\prime }(W^{T}X)\right\}-\beta I$$

(5)

Since the data are whitened, $E\left\{X{X}^T\right\}=I$, $E\left\{XX{g}^{\prime }(W^{T}X)\right\}\approx E\left\{X{X}^T\right\}·E\left\{g^{\prime }(W^{T}X)\right\}=E\left\{g^{\prime }(W^{T}X)\right\}I$, the approximate Newton iteration formula can be obtained as shown in (6).

$$\left\{\begin{array}{c}W*=W-{\displaystyle \frac{[E\left\{Xg(W^{T}X)\right\}-\beta W]}{E\left\{g^{\prime }(W^{T}X)\right\}-\beta }}\\ W=W*/\left|\right|W*\left|\right|\end{array}\right.$$

(6)

In the above formula, $W*$ is a new value of $W$, $\beta =E\left\{W^{T}Xg(W^{T}X)\right\}$. The following Equation (7) can be obtained.

$$W*=E\left\{Xg(W^{T}X)\right\}-E\left\{g^{\prime }(W^{T}X)\right\}W$$

(7)

According to Equation (7), the optimal w can be obtained by iteration, so as to obtain the independent component $Y=W^{T}X$, and then the blind source separation of AC/DC transmission system signals can be realized.

4. Experimental Analysis of AC/DC Transmission Systems

4.1. Analysis of the BSS Algorithm

The fault signal data collected by the A-phase sensor are used as the experimental data. This experiment uses these data to verify the accuracy of the BSS algorithm. There are problems such as electromagnetic interference in the complex environment of the substation. The following Figure 8 are shown as the original fault signal, noise interference signal 1 and 2, respectively.

The fault signal is randomly mixed with the other two interference signals through the mixing matrix, and the waveform is shown in Figure 9.

Figure 10 shows the signal obtained by blind source separation and unmixing. The top waveform is the fault waveform after the BSS algorithm. It can be seen from the figure that the algorithm used in this paper effectively separates the fault signal from the noise interference signal.

In order to verify the feasibility and effectiveness of the Fast-ICA algorithm, this paper uses the area of the waveform to evaluate the original fault signal and the separated fault signal. The original fault waveform, the separated fault waveform, and the difference waveform between the two signals are calculated. The data obtained by calculating the absolute area of each curve are shown in

Table 1. Results of the calculation of absolute value area difference of three waveforms.

Signal Name	Absolute Area of the Signal
The original signal	8.49248
Signal after separation	8.42338
The difference between the two signals	0.12473

.

The absolute error calculated from the data in the table is 1.4683%, which is within the effective range. It was further verified that the Fast-ICA algorithm can effectively extract the fault signal in the complex environment of AC/DC transmission networks.

4.2. Analysis of the DC Side

As shown in Figure 3, this article sets up crossline faults to address the issue of faults in AC/DC transmission systems. In the article, the positive and negative currents of the rectifier and inverter sides of the DC network are first collected. The current values collected on the left and right sides of the P Electrode of the DC side are denoted as IP1 and IP2, and the current values collected on the left and right sides of the N Electrode of the DC side are denoted as I_N1 and I_N2. Figure 11 shows the positive and negative signal diagrams of the DC system.

As shown in the above figure, when a cross-circuitry Fault occurs between the DC positive Electrode (P-Electrode) on the DC side and the AC system A phase in the AC/DC transmission system, the positive and negative Electrodes are affected by the AC fault and the signal fluctuates. However, there is a significant change in the sending and receiving signals at the fault Electrode, while the no fault Electrode has a consistent signal frequency, even though it is affected by the AC signal.

This paper first performs wavelet decomposition on I_P1, I_P2, I_N1, and I_N2 to extract the d4 wavelet spectra of four time–domain signals; subsequently, the rectifier side and the inverter side of the DC network system are subtracted, denoted as A = d4 (I_P1) − d4 (I_P2), and B = d4 (I_N1) − d4 (I_N2). Figure 12 shows the processed signal waveform.

It can be clearly seen from the above figure that when a crossline fault occurs in the AC/DC transmission system, although the positive and negative poles of the DC system fluctuate due to the influence of the AC system, the difference can be compared with the threshold of 0.1 to determine whether a fault has occurred [29].

4.3. Analysis of the AC Side

Grounding faults can generate direct current in the fault signal. This article introduces grounding fault types in AC system fault diagnosis to distinguish between crossline faults and grounding faults. As shown in Figure 13, this article sets three types of faults on the AC side for AC/DC transmission systems: crossline fault, ground fault, and no fault.

This paper selects the scenario of crossline fault between the P-Electrode of the DC bipolar network and the A-phase of the AC network. The fault point on the DC bipolar network line is F₁, and the fault point on the AC network line is F₂. In this fault scenario, the AC current values of three phases, A, B, and C, in the AC network are collected, and the current characteristics are analyzed to diagnose AC side faults and locate AC and DC cross-circuitry Faults. After this, a metal grounding fault scenario is set up in phase A of the AC/DC transmission system, and the fault point is still F₁. After setting the model, this paper first randomly collects the three-phase AC current values of A, B, and C in the AC network, which are recorded as Ii.

Wavelet decomposition on three types of AC current values was performed, and the time–domain curve of the low-frequency wavelet band after decomposition, which is denoted as L(I_i), was generated. Figure 14 shows the waveform after wavelet transformation.

In order to more accurately determine the fault type on the AC side of the AC/DC transmission system, this paper analyzes three processed signals using stationarity. Standard deviation is a statistical measure that describes the degree of dispersion of data, with larger data indicating greater dispersion [30].

Figure 13 and

Table 2. Range of stationarity.

Fault Number	Fault Type	Fault Interval
1	Cross-circuitry Fault	S(Ii) ∈ [0, 0.1]
2	Grounding Fault	S(Ii) ∈ (0.1, 0.2]
3	No Fault	S(Ii) > 0.2

show that S(I₁) = 0.27 is a cross-circuitry Fault, S(I₂) = 0.14 is a ground fault, and S(I₃) = 0.02 is a no fault.

4.4. Summary of the Experiment

Table 3. Comparison of the performance of different signal processing methods.

Authors	Algorithm	Error
Amirteimoury F et al. [31]	DWT	12% (MAE)
Wei J et al. [32]	LSTM	1.77% (MAE)
The methods provided in this article	BSS	1.4683% (Relative Error)

compares the signal processing methods of different algorithms in this paper and other studies in the literature; it can be seen that the algorithm used in this paper has high recognition accuracy for this system.

This paper first uses the BSS algorithm to separate mixed signals and extract fault signals. This is the proposed crossline fault identification method for AC/DC transmission systems. This method can accurately identify crossline faults on the DC and AC sides of AC/DC transmission systems, meeting the diagnostic needs of crossline faults in various scenarios of AC/DC transmission systems. However, the model involved in this article did not consider the influence of the external environment on signals. The experimental design scenario is relatively ideal. The next step in this experiment will consider the impact of environmental interference on signals to improve the universality of this research.

5. Conclusions

This paper proposes a new method for crossline fault diagnosis in traditional AC/DC transmission systems based on the BSS algorithm to address the deficiencies in signal processing. The specific conclusions are as follows:

1. This article proposes a blind source separation algorithm for extracting source signals from complex signals in response to the complexity of extracting electrical signals from sensors. The relative error of the model is 1.4683%, indicating that this method can extract source waveforms effectively.

2. This article proposes fault identification methods for both AC and DC systems, and through experimental verification, can accurately identify the type of fault.

3. The BSS-algorithm-based crossline fault identification method proposed in this article can accurately identify the location of crossline faults in AC/DC transmission systems, providing theoretical and experimental support for power system maintenance personnel to maintain equipment.