Distributed intelligent terminal fault detection system and method based on Beidou positioning

Abstract

The invention discloses a distributed intelligent terminal fault detection system and method based on Beidou positioning, and the system comprises a collection module which is used for collecting the Beidou information of a distributed intelligent terminal; the fault detection module is used for collecting operation information of the distributed intelligent terminal, training a distributed intelligent terminal fault detection model based on a Transform architecture, and performing fault detection by using the distributed intelligent terminal fault detection model; the terminal positioning module is used for calculating coordinate information of the distributed intelligent terminal with the fault according to the fault detection result and positioning the distributed intelligent terminal; and the navigation correction module is used for navigating to the distributed intelligent terminal with the fault according to the positioning result and based on the distributed intelligent terminal Beidou navigation technology corrected by the multi-modal model. The fault detection level and the operation and maintenance efficiency of the distributed intelligent terminal are effectively improved, and the safe operation level of a power distribution network is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment, and in particular to a distributed intelligent terminal fault detection system and method based on Beidou positioning. Background Art

[0002] Effectively detecting the faults of distributed intelligent terminals and carrying out operation and maintenance in a timely manner can effectively improve the safe operation level of the distribution network. However, the existing methods still have the following shortcomings: 1) Single data dependence. The fault detection method of traditional distributed intelligent terminals usually relies on a single data source, which may lead to limitations in fault detection, especially in the face of complex and changing environments. It is impossible to fully capture the characteristics of the fault and is prone to missed detection or misdetection. 2) Poor real-time performance. Traditional fault detection methods are often difficult to achieve real-time monitoring. Especially in the case of large data volume and complex analysis algorithms, the real-time performance of fault detection may be affected, resulting in the inability to discover and respond to faults in a timely manner. 3) Limited communication. Distributed intelligent terminals usually transmit fault information to the operation and maintenance center through wireless communication networks (such as Wi-Fi, cellular networks, etc.). However, in the case of poor network coverage or communication interruption, fault information may not be transmitted in time, resulting in the inability of the operation and maintenance center to respond in time. 4) Multipath effect and signal interference. In urban environments, due to building reflection, occlusion, etc., distributed intelligent terminals may be affected by multipath effects, resulting in increased positioning errors. In addition, the terminal may also be interfered by signals from other wireless devices, further reducing the accuracy of fault detection and positioning. Summary of the invention

[0003] In order to solve the above problems, the present invention proposes a distributed intelligent terminal fault detection system and method based on Beidou positioning, which can improve the fault detection level of distributed intelligent terminals and accurately locate the faulty distributed intelligent terminals.

[0004] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0005] A distributed intelligent terminal fault detection system based on Beidou positioning of the present invention comprises:

[0006] The acquisition module is used to construct a pseudo-range model between Beidou satellites and distributed intelligent terminals, calculate the pseudo-range between a single Beidou satellite and a distributed intelligent terminal, use the pseudo-range model between multiple Beidou satellites and distributed intelligent terminals to lock the location information of the distributed intelligent terminal, build a Beidou satellite geographic coordinate conversion model, convert the location information of the distributed intelligent terminal into geographic coordinates, and realize the collection of Beidou information of the distributed intelligent terminal;

[0007] The fault detection module is used to collect the operation information of the distributed intelligent terminal, train the distributed intelligent terminal fault detection model based on the Transformer architecture, and perform fault detection using the distributed intelligent terminal fault detection model;

[0008] The terminal positioning module is used to calculate the coordinate information of the faulty distributed intelligent terminal and locate it according to the fault detection result;

[0009] The navigation correction module is used to navigate to the faulty distributed intelligent terminal based on the positioning results and the distributed intelligent terminal Beidou navigation technology corrected by the multimodal model.

[0010] A further improvement of the present invention is that the specific operations performed by the acquisition module include:

[0011] Construct a pseudo-range model between Beidou satellites and distributed intelligent terminals, the expression is:

[0012]

[0013] Among them, ρ i is the measured pseudo-range between the i-th Beidou satellite and the distributed intelligent terminal, (X i ,Y i ,Z i ) is the known position of the i-th BeiDou satellite in the Earth-centered Earth-fixed coordinate system, (X0, Y0, Z0) is the position of the distributed intelligent terminal, c is the speed of light, δt is the clock deviation of the distributed intelligent terminal, δt i is the satellite clock bias;

[0014] Using the pseudo-range model between multiple BeiDou satellites and distributed intelligent terminals, the position (X0, Y0, Z0) of the distributed intelligent terminal and the clock deviation δt of the distributed intelligent terminal are solved;

[0015] A Beidou satellite geographic coordinate conversion model is constructed to convert the location information of distributed intelligent terminals into geographic coordinates. The expression of the Beidou satellite geographic coordinate conversion model is:

[0016]

[0017] Where, φ is the latitude of the distributed intelligent terminal, λ is the longitude of the distributed intelligent terminal, h is the altitude of the distributed intelligent terminal, R E is the average radius of the Earth.

[0018] A further improvement of the present invention is that the specific operations performed by the fault detection module include:

[0019] Collect the operation information of distributed intelligent terminals, including:

[0020] The collected operating information of the distributed intelligent terminal includes temperature, voltage, current, and power, and the expression is:

[0021] X(t)=[T(t),V(t),I(t),P(t)] (3);

[0022] Wherein, X(t) is the operating information of the distributed intelligent terminal at time t, T(t) is the temperature collected at time t, V(t) is the voltage collected at time t, I(t) is the current collected at time t, and P(t) is the power collected at time t;

[0023] The distributed intelligent terminal fault detection model is trained using input samples consisting of the operation information of the distributed intelligent terminal. The distributed intelligent terminal fault detection model includes a Transformer architecture position encoding model, a self-attention model, a multi-head attention feature extraction model, and a distributed intelligent terminal fault prediction model, specifically including:

[0024] The expression of the input sample consisting of the operation information of the distributed intelligent terminal is:

[0025]

[0026] Among them, X is the input sample of the Transformer architecture, and X(N) is the operation information of the distributed intelligent terminal at time N;

[0027] Construct the position encoding model of the Transformer architecture. The expression of the position encoding model of the Transformer architecture is:

[0028]

[0029]

[0030] X input =X+PE (7);

[0031] Among them, PE is the position encoding information, t is the time step, g is the dimension index, d is the total dimension of the feature, X input The input matrix for adding position encoding;

[0032] Construct a self-attention model and calculate the relationship between different time steps. The expression is:

[0033]

[0034]

[0035] Where: Q, K, V are query, key and value vector models respectively, W Q , W K, W V are the weights of the query, key, and value vector models, respectively. Attention(Q,K,V) is the attention weight. k is the dimension of the key vector, K T is the relationship coefficient between the core key and the related target in the self-attention mechanism;

[0036] Construct a multi-head attention feature extraction model, the expression is:

[0037]

[0038] Among them: MultiHead(Q,K,V) is a multi-head attention feature extraction model, Concat() is a string concatenation function, head j is the j-th sample attention weight, Q j , K j 、V j are the query, key and value vector models of the jth sample, respectively;

[0039] Construct a distributed intelligent terminal fault prediction model based on the Transformer architecture, the expression is:

[0040] y=σ(W out h+b out ) (11);

[0041] Among them, y is the fault prediction result of the distributed intelligent terminal, h is the feature vector processed by the Transformer architecture, and W out and b out is the weight and bias of the output layer, σ is the activation function;

[0042] According to the output distributed intelligent terminal fault prediction result y, a threshold τ is set to determine whether there is a fault and issue an early warning. If the distributed intelligent terminal fault prediction result y is greater than the threshold τ, 1 is output to determine that the distributed intelligent terminal has a fault. The expression is:

[0043]

[0044] The further improvement of the present invention is that the distributed intelligent terminal Beidou navigation technology based on multimodal model correction specifically includes:

[0045] Based on the basic coordinate system and the state vector, the relative position and relative speed of the distributed intelligent terminal at the distance fault are obtained, and the expression is:

[0046]

[0047] Where: Q z(t) is the state vector of the distributed intelligent terminal with relative fault at time t,

[0048] r z (t) = [x z (t),y z (t),Z z (t)] T is the relative position from the distributed intelligent terminal at time t,

[0049] V z (t) = [v x (t),v y (t),v z (t)] T is the speed relative to the distributed intelligent terminal at time t;

[0050] M z (t)=[M G (t),M H (t),M D (t)] is the multimodal data measured at time t, where M G (t) is the satellite signal, indicating whether the distributed terminal has a fault or not. A value of 1 indicates a fault, and a value of 0 indicates no fault. M H (t) is the spatial electromagnetic environment data, which affects the signal transmission quality; M D (t) is terrain information, including contour lines and terrain slope;

[0051] Construct the kinetic model, the expression is:

[0052] Q z (t+1)=FQ z (t)+Bu(t)+w(t) (14);

[0053] Where: Q z (t+1) is the state vector of the distributed intelligent terminal relative to the fault at time t+1, F is the state transfer matrix, B is the control matrix, u(t) is the control input, and w(t) is the process noise;

[0054] Construct a multimodal data fusion model, the expression is:

[0055]

[0056] Among them: G(t+1) is the state updated at time t+1, r GNSS (t+1) is the GNSS position, z vision (t+1) is the visual feature at time t+1, and H is the observation matrix;

[0057] The relative position and speed update model of the fused multimodal data is constructed, and the relative position and relative speed of the distributed intelligent terminal are estimated. The expression of the relative position and speed update model of the fused multimodal data is:

[0058]

[0059] in: is the state vector predicted at time t+1,

[0060] are the relative position and relative speed from the distributed intelligent terminal at time t+1, K(t+1) is the Kalman gain at time t+1, H T is the transpose of the observation matrix, P(t+1|t) is the covariance matrix of the state estimate at time t for time t+1, P(t|t-1) is the covariance matrix of the state estimate at time t for time t-1, R is the observation noise covariance, I is the identity matrix, the values ​​on the diagonal are 1, and the values ​​elsewhere are 0.

[0061] A distributed intelligent terminal fault detection method based on Beidou positioning of the present invention includes the following operations:

[0062] Collect Beidou information of distributed intelligent terminals, including building a pseudo-range model between Beidou satellites and distributed intelligent terminals, calculating the pseudo-range between a single Beidou satellite and the distributed intelligent terminal, using the pseudo-range model between multiple Beidou satellites and distributed intelligent terminals, locking the location information of the distributed intelligent terminal, building a Beidou satellite geographic coordinate conversion model, and converting the location information of the distributed intelligent terminal into geographic coordinates;

[0063] Collect the operation information of distributed intelligent terminals, train the distributed intelligent terminal fault detection model based on the Transformer architecture, and use the distributed intelligent terminal fault detection model to perform fault detection;

[0064] If the fault detection result is a fault, a fault signal is sent to the Beidou navigation satellite. The Beidou navigation satellite receives the signal and calculates the coordinate information of the distributed intelligent terminal where the fault occurs. The operation and maintenance center analyzes the coordinate information of the distributed intelligent terminal where the fault occurs and converts it into longitude, latitude, and altitude coordinates.

[0065] The Beidou navigation technology for distributed intelligent terminals based on multimodal model correction can be used to navigate to the faulty distributed intelligent terminal.

[0066] A further improvement of the present invention is that the distributed intelligent terminal Beidou information is collected, specifically including:

[0067] Construct a pseudo-range model between Beidou satellites and distributed intelligent terminals, the expression is:

[0068]

[0069] Among them, ρ i is the measured pseudo-range between the i-th Beidou satellite and the distributed intelligent terminal, (X i ,Y i ,Z i ) is the known position of the i-th BeiDou satellite in the Earth-centered Earth-fixed coordinate system, (X0, Y0, Z0) is the position of the distributed intelligent terminal, c is the speed of light, δt is the clock deviation of the distributed intelligent terminal, δt i is the satellite clock bias;

[0070] Using the pseudo-range model between multiple BeiDou satellites and distributed intelligent terminals, the position (X0, Y0, Z0) of the distributed intelligent terminal and the clock deviation δt of the distributed intelligent terminal are solved;

[0071] A Beidou satellite geographic coordinate conversion model is constructed to convert the location information of distributed intelligent terminals into geographic coordinates. The expression of the Beidou satellite geographic coordinate conversion model is:

[0072]

[0073] Where, φ is the latitude of the distributed intelligent terminal, λ is the longitude of the distributed intelligent terminal, h is the altitude of the distributed intelligent terminal, R E is the average radius of the Earth.

[0074] A further improvement of the present invention is to collect the operation information of the distributed intelligent terminal, train the distributed intelligent terminal fault detection model based on the Transformer architecture, and perform fault detection using the distributed intelligent terminal fault detection model, which specifically includes:

[0075] Collect the operating information of the distributed intelligent terminal, where the collected operating information of the distributed intelligent terminal includes temperature, voltage, current, and power, and the expression is:

[0076] X(t)=[T(t),V(t),I(t),P(t)] (3);

[0077] Wherein, X(t) is the operating information of the distributed intelligent terminal at time t, T(t) is the temperature collected at time t, V(t) is the voltage collected at time t, I(t) is the current collected at time t, and P(t) is the power collected at time t;

[0078] The distributed intelligent terminal fault detection model is trained using input samples consisting of the operation information of the distributed intelligent terminal. The distributed intelligent terminal fault detection model includes a Transformer architecture position encoding model, a self-attention model, a multi-head attention feature extraction model, and a distributed intelligent terminal fault prediction model, specifically including:

[0079] The expression of the input sample consisting of the operation information of the distributed intelligent terminal is:

[0080]

[0081] Among them, X is the input sample of the Transformer architecture, and X(N) is the operation information of the distributed intelligent terminal at time N;

[0082] Construct the position encoding model of the Transformer architecture. The expression of the position encoding model of the Transformer architecture is:

[0083]

[0084]

[0085] X input =X+PE (7);

[0086] Among them, PE is the position encoding information, t is the time step, g is the dimension index, d is the total dimension of the feature, X input The input matrix for adding position encoding;

[0087] Construct a self-attention model and calculate the relationship between different time steps. The expression is:

[0088]

[0089]

[0090] Where: Q, K, V are query, key and value vector models respectively, W Q , W K , W V are the weights of the query, key, and value vector models, respectively. Attention(Q,K,V) is the attention weight. k is the dimension of the key vector, K T is the relationship coefficient between the core key and the related target in the self-attention mechanism;

[0091] Construct a multi-head attention feature extraction model, the expression is:

[0092]

[0093] Among them: MultiHead(Q,K,V) is a multi-head attention feature extraction model, Concat() is a string concatenation function, head j is the j-th sample attention weight, Q j , K j 、V j are the query, key and value vector models of the jth sample, respectively;

[0094] Construct a distributed intelligent terminal fault prediction model based on the Transformer architecture, the expression is:

[0095] y=σ(W out h+b out ) (11);

[0096] Among them, y is the fault prediction result of the distributed intelligent terminal, h is the feature vector processed by the Transformer architecture, and W out and b out is the weight and bias of the output layer, σ is the activation function;

[0097] According to the output distributed intelligent terminal fault prediction result y, a threshold τ is set to determine whether there is a fault and issue an early warning. If the distributed intelligent terminal fault prediction result y is greater than the threshold τ, 1 is output to determine that the distributed intelligent terminal has a fault. The expression is:

[0098]

[0099] The further improvement of the present invention is that the distributed intelligent terminal Beidou navigation technology based on multimodal model correction specifically includes:

[0100] Based on the basic coordinate system and the state vector, the relative position and relative speed of the distributed intelligent terminal at the distance fault are obtained, and the expression is:

[0101]

[0102] Where: Q z (t) is the state vector of the distributed intelligent terminal with relative fault at time t,

[0103] r z (t) = [x z (t),y z (t),Z z (t)] T is the relative position from the distributed intelligent terminal at time t,

[0104] V z (t) = [v x (t),vy (t),v z (t)] T is the speed relative to the distributed intelligent terminal at time t;

[0105] M z (t)=[M G (t),M H (t),M D (t)] is the multimodal data measured at time t, where M G (t) is the satellite signal, indicating whether the distributed terminal has a fault or not. A value of 1 indicates a fault, and a value of 0 indicates no fault. M H (t) is the spatial electromagnetic environment data, which affects the signal transmission quality; M D (t) is terrain information, including contour lines, terrain slope, etc.

[0106] Construct the kinetic model, the expression is:

[0107] Q z (t+1)=FQ z (t)+Bu(t)+w(t) (14);

[0108] Where: Q z (t+1) is the state vector of the distributed intelligent terminal relative to the fault at time t+1, F is the state transfer matrix, B is the control matrix, u(t) is the control input, and w(t) is the process noise;

[0109] Construct a multimodal data fusion model, the expression is:

[0110]

[0111] Among them: G(t+1) is the state updated at time t+1, r GNSS (t+1) is the GNSS position, z vision (t+1) is the visual feature at time t+1, and H is the observation matrix;

[0112] The relative position and speed update model of the fused multimodal data is constructed, and the relative position and relative speed of the distributed intelligent terminal are estimated. The expression of the relative position and speed update model of the fused multimodal data is:

[0113]

[0114] in: is the state vector predicted at time t+1,

[0115] are the relative position and relative speed from the distributed intelligent terminal at time t+1, K(t+1) is the Kalman gain at time t+1, H T is the transpose of the observation matrix, P(t+1|t) is the covariance matrix of the state estimate at time t for time t+1, P(t|t-1) is the covariance matrix of the state estimate at time t for time t-1, R is the observation noise covariance, I is the identity matrix, the values ​​on the diagonal are 1, and the values ​​elsewhere are 0.

[0116] The beneficial effect of the present invention is that the present invention improves the overall vision and accuracy of distributed intelligent terminal fault detection through a distributed intelligent terminal fault detection model based on the Transformer architecture.

[0117] The present invention proposes a distributed terminal Beidou navigation technology based on multimodal model correction, which realizes the dynamic correction of the relative position and speed of the distributed intelligent terminal, and improves the positioning accuracy and navigation performance of the distributed intelligent terminal. BRIEF DESCRIPTION OF THE DRAWINGS

[0118] Figure 1 is a schematic diagram of the system principle in an embodiment of the present invention;

[0119] Figure 2 is a flow chart of a method in an embodiment of the present invention. DETAILED DESCRIPTION

[0120] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0121] like Figure 2 As shown, a distributed intelligent terminal fault detection method based on Beidou positioning in this embodiment includes the following steps:

[0122] Step 1, collect Beidou information of distributed intelligent terminals. The reference coordinate system used by the Beidou satellite navigation system is usually an earth-centered earth-fixed coordinate system. This embodiment parses the geographical coordinate information such as longitude, latitude and altitude of distributed intelligent terminals from a number of Beidou satellites by constructing a model. Specifically, it includes: constructing a pseudo-range model between Beidou satellites and distributed intelligent terminals, calculating the pseudo-range between a single Beidou satellite and the distributed intelligent terminal; using the pseudo-range model between multiple Beidou satellites and distributed intelligent terminals to lock the location information of the distributed intelligent terminal; constructing a Beidou satellite geographic coordinate conversion model to convert the location information of the distributed intelligent terminal into geographic coordinates, including longitude, latitude and altitude, to realize the collection of Beidou information of distributed intelligent terminals.

[0123] Step 2: Collect the operating information of the distributed intelligent terminal, including temperature, voltage, current, power and other information, and train the distributed intelligent terminal fault detection model based on the Transformer architecture. The distributed intelligent terminal fault detection model includes the Transformer architecture position encoding model, self-attention model, multi-head attention feature extraction model, and Transformer-based distributed intelligent terminal fault prediction model. According to the predicted fault result, set the threshold to determine whether the distributed intelligent terminal has a fault.

[0124] Step 3: If the fault detection result is that a fault has occurred, a fault signal is sent to the Beidou navigation satellite. The Beidou navigation satellite receives the signal, performs pseudo-range measurement according to expressions (1)-(3) and calculates the coordinate information of the distributed intelligent terminal where the fault has occurred. The operation and maintenance center parses the coordinate information of the distributed intelligent terminal where the fault has occurred and converts it into longitude, latitude, and altitude coordinates to achieve accurate positioning of the faulty distributed intelligent terminal.

[0125] Step 4: Navigate to the faulty distributed intelligent terminal based on Beidou navigation technology of distributed intelligent terminal corrected by multimodal model. This embodiment uses multimodal model to conduct training analysis for multiple data sources (such as satellite signal data, distributed intelligent terminal location data, space electromagnetic environment data, terrain information, etc.), and predicts the relative position and speed with the distributed intelligent terminal at the next moment based on the relative position and speed with the distributed intelligent terminal at the previous moment, so as to realize dynamic correction of the relative position and speed with the distributed intelligent terminal, and improve the positioning accuracy and navigation performance of the distributed intelligent terminal.

[0126] In step 1, the reference coordinate system used by the Beidou satellite navigation system is usually the Earth-centered Earth-fixed coordinate system. A pseudo-range model is constructed between the Beidou satellite and the distributed intelligent terminal. The pseudo-range model represents the clock error between the Beidou satellite and the distributed intelligent terminal, and the expression is:

[0127]

[0128] Among them, ρ i is the measured pseudo-range between the i-th Beidou satellite and the distributed intelligent terminal, (X i ,Y i ,Z i ) is the known position of the i-th BeiDou satellite in the Earth-centered Earth-fixed coordinate system, (X0, Y0, Z0) is the position of the distributed intelligent terminal, c is the speed of light, δt is the clock deviation of the distributed intelligent terminal, δt i is the satellite clock bias.

[0129] In order to solve the position (X0, Y0, Z0) of the distributed intelligent terminal, it is necessary to use the pseudo-range model between multiple Beidou satellites and the distributed intelligent terminal for solution. This embodiment uses four satellites (i=1, 2, 3, 4) to establish the nonlinear equation group shown in expression (2) to solve the position of the distributed intelligent terminal.

[0130]

[0131] Wherein: ρ1, ρ2, ρ3, ρ4 are the measured pseudoranges between the 1st, 2nd, 3rd, and 4th Beidou satellites and the distributed intelligent terminal respectively, (X1, Y1, Z1), (X2, Y2, Z2), (X3, Y3, Z3), and (X4, Y4, Z4) are the known positions of the 1st, 2nd, 3rd, and 4th Beidou satellites in the Earth-centered Earth-fixed coordinate system respectively, and δt1, δt2, δt3, and δt4 are the clock biases of the 1st, 2nd, 3rd, and 4th Beidou satellites respectively.

[0132] By solving the nonlinear equation group of expression (2), the three-dimensional coordinates (X0, Y0, Z0) and clock deviation δt of the distributed intelligent terminal are obtained. This embodiment constructs a Beidou satellite geographic coordinate conversion model, as shown in expression (3). According to expression (3), the location information of the distributed intelligent terminal is converted into geographic coordinates to obtain the longitude, latitude and altitude of the distributed terminal.

[0133]

[0134] Where, φ is the latitude of the distributed intelligent terminal, λ is the longitude of the distributed intelligent terminal, h is the altitude of the distributed intelligent terminal, R E is the average radius of the Earth.

[0135] In step 2, the collected operating information of the distributed intelligent terminal includes temperature, voltage, current, and power, and the expression is:

[0136] X(t)=[T(t),V(t),I(t),P(t)] (4);

[0137] Among them, X(t) is the operating information of the distributed intelligent terminal at time t, T(t) is the temperature collected at time t, V(t) is the voltage collected at time t, I(t) is the current collected at time t, and P(t) is the power collected at time t.

[0138] The expression of the input sample consisting of the operation information of the distributed intelligent terminal is:

[0139]

[0140] Among them, X is the input sample of the Transformer architecture, X(N) is the operation information of the distributed intelligent terminal at time N, and N is the length of the time series.

[0141] Construct the position encoding model of the Transformer architecture. Since the Transformer architecture does not have intrinsic time series information, it is necessary to embed the position information of the time series into the input data through position encoding. The expression of the position encoding model of the Transformer architecture is:

[0142]

[0143]

[0144] X input =X+PE (8);

[0145] Among them, PE is the position encoding information, t is the time step, g is the dimension index, d is the total dimension of the feature, X input The input matrix for adding position encoding.

[0146] Construct a self-attention model and calculate the relationship between different time steps. The expression is:

[0147]

[0148]

[0149] Where: Q, K, V are query, key and value vector models respectively, W Q , W K , W V are the weights of the query, key, and value vector models, respectively. Attention(Q,K,V) is the attention weight. k is the dimension of the key vector, K T is the relationship coefficient between the core key and the related target in the self-attention mechanism.

[0150] Construct a multi-head attention feature extraction model. The multi-head attention feature extraction model of the Transformer architecture can capture the patterns of different features in the data. This embodiment extracts the features of the distributed intelligent terminal through the multi-head attention feature extraction model. The expression is:

[0151]

[0152] Among them: MultiHead(Q,K,V) is a multi-head attention feature extraction model, Concat() is a string concatenation function, head j is the j-th sample attention weight, Q j , K j 、V j are the j-th sample query, key, and value vector models, respectively.

[0153] Construct a distributed intelligent terminal fault prediction model based on the Transformer architecture, the expression is:

[0154] y=σ(W out h+b out ) (12);

[0155] Among them, y is the fault prediction result of the distributed intelligent terminal, h is the feature vector processed by the Transformer architecture, and W out and b out is the weight and bias of the output layer, σ is the activation function;

[0156] According to the output distributed intelligent terminal fault prediction result y, a threshold τ is set to determine whether there is a fault and issue an early warning. If the distributed intelligent terminal fault prediction result y is greater than the threshold τ, 1 is output to determine that the distributed intelligent terminal has a fault. The expression is:

[0157]

[0158] In step 4, the proposed distributed terminal Beidou navigation technology based on multimodal model correction is shown in Expressions (14) to (17).

[0159] Expression (14) is the basic coordinate system and state vector, including the relative position of the distributed intelligent terminal away from the fault and the relative distributed intelligent terminal speed.

[0160]

[0161] Where: Q z (t) is the state vector of the distributed intelligent terminal with relative fault at time t,

[0162] r z (t) = [xz (t),y z (t),Z z (t)] T is the relative position from the distributed intelligent terminal at time t,

[0163] V z (t) = [v x (t),v y (t),v z (t)] T is the speed relative to the distributed intelligent terminal at time t;

[0164] M z (t)=[M G (t),M H (t),M D (t)] is the multimodal data measured at time t, where M G (t) is the satellite signal, indicating whether the distributed terminal has a fault or not. A value of 1 indicates a fault, and a value of 0 indicates no fault. M H (t) is the spatial electromagnetic environment data, which affects the signal transmission quality; M D (t) is terrain information, including contour lines, terrain slope, etc.

[0165] The dynamic model is constructed and the dynamic equation of motion is described using the inertial measurement unit (IMU) data. The expression is:

[0166] Q z (t+1)=FQ z (t)+Bu(t)+w(t) (15);

[0167] Where: Q z (t+1) is the state vector of the distributed intelligent terminal relative to the fault at time t+1, F is the state transfer matrix, B is the control matrix, u(t) is the control input (such as IMU data such as acceleration), and w(t) is the process noise, which represents the uncertainty of the model.

[0168] Construct a multimodal data fusion model to update the state estimate by fusing data from multimodal sensors (such as vision, GNSS, etc.). The expression is:

[0169]

[0170] Among them: G(t+1) is the state updated at time t+1, r GNSS (t+1) is the GNSS position, z vision (t+1) is the visual feature at time t+1, and H is the observation matrix;

[0171] Construct a relative position and velocity update model for fusion multimodal data. The expression of the relative position and velocity update model for fusion multimodal data is:

[0172]

[0173] in: is the state vector predicted at time t+1, are the relative position and relative speed from the distributed intelligent terminal at time t+1, K(t+1) is the Kalman gain at time t+1, H T is the transpose of the observation matrix, P(t+1|t) is the covariance matrix of the state estimate at time t for time t+1, P(t|t-1) is the covariance matrix of the state estimate at time t for time t-1, R is the observation noise covariance, I is the identity matrix, the values ​​on the diagonal are 1, and the values ​​elsewhere are 0.

[0174] The relative position and relative speed of the distributed terminal are estimated through the above steps. Realize autonomous navigation of faulty distributed intelligent terminals.

[0175] The above method is implemented by a distributed intelligent terminal fault detection system based on Beidou positioning in this embodiment. Figure 1 As shown, the system comprises:

[0176] The acquisition module is used to construct a pseudo-range model between Beidou satellites and distributed intelligent terminals, calculate the pseudo-range between a single Beidou satellite and a distributed intelligent terminal, use the pseudo-range model between multiple Beidou satellites and distributed intelligent terminals to lock the location information of the distributed intelligent terminal, build a Beidou satellite geographic coordinate conversion model, convert the location information of the distributed intelligent terminal into geographic coordinates, and realize the collection of Beidou information of the distributed intelligent terminal;

[0177] The fault detection module is used to collect the operation information of the distributed intelligent terminal, train the distributed intelligent terminal fault detection model based on the Transformer architecture, and perform fault detection using the distributed intelligent terminal fault detection model;

[0178] The terminal positioning module is used to calculate the coordinate information of the faulty distributed intelligent terminal and locate it according to the fault detection result;

[0179] The navigation correction module is used to navigate to the faulty distributed intelligent terminal based on the positioning results and the distributed intelligent terminal Beidou navigation technology corrected by the multimodal model.

[0180] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have the same meaning as in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless defined similarly here.

[0181] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A distributed intelligent terminal fault detection system based on Beidou positioning, characterized in that: The system comprises: The acquisition module is used to construct a pseudo-range model between Beidou satellites and distributed intelligent terminals, calculate the pseudo-range between a single Beidou satellite and a distributed intelligent terminal, use the pseudo-range model between multiple Beidou satellites and distributed intelligent terminals to lock the location information of the distributed intelligent terminal, build a Beidou satellite geographic coordinate conversion model, convert the location information of the distributed intelligent terminal into geographic coordinates, and realize the collection of Beidou information of the distributed intelligent terminal; The fault detection module is used to collect the operation information of the distributed intelligent terminal, train the distributed intelligent terminal fault detection model based on the Transformer architecture, and perform fault detection using the distributed intelligent terminal fault detection model; The terminal positioning module is used to calculate the coordinate information of the faulty distributed intelligent terminal and locate it according to the fault detection result; The navigation correction module is used to navigate to the faulty distributed intelligent terminal based on the positioning results and the distributed intelligent terminal Beidou navigation technology corrected by the multimodal model.

2. According to claim 1, a distributed intelligent terminal fault detection system based on Beidou positioning is characterized in that: The specific operations performed by the acquisition module include: Construct a pseudo-range model between Beidou satellites and distributed intelligent terminals, the expression is: Among them, ρ i is the measured pseudo-range between the i-th Beidou satellite and the distributed intelligent terminal, (X i ,Y i ,Z i ) is the known position of the i-th BeiDou satellite in the Earth-centered Earth-fixed coordinate system, (X0, Y0, Z0) is the position of the distributed intelligent terminal, c is the speed of light, δt is the clock deviation of the distributed intelligent terminal, δt i is the satellite clock bias; Using the pseudo-range model between multiple BeiDou satellites and distributed intelligent terminals, the position (X0, Y0, Z0) of the distributed intelligent terminal and the clock deviation δt of the distributed intelligent terminal are solved; A Beidou satellite geographic coordinate conversion model is constructed to convert the location information of distributed intelligent terminals into geographic coordinates. The expression of the Beidou satellite geographic coordinate conversion model is: Where, φ is the latitude of the distributed intelligent terminal, λ is the longitude of the distributed intelligent terminal, h is the altitude of the distributed intelligent terminal, R E is the average radius of the Earth.

3. The distributed intelligent terminal fault detection system based on Beidou positioning according to claim 1 is characterized in that: The specific operations performed by the fault detection module include: Collect the operation information of distributed intelligent terminals, including: The collected operating information of the distributed intelligent terminal includes temperature, voltage, current, and power, and the expression is: X(t)=[T(t),V(t),I(t),P(t)] (3); Wherein, X(t) is the operating information of the distributed intelligent terminal at time t, T(t) is the temperature collected at time t, V(t) is the voltage collected at time t, I(t) is the current collected at time t, and P(t) is the power collected at time t; The distributed intelligent terminal fault detection model is trained using input samples consisting of the operation information of the distributed intelligent terminals. The distributed intelligent terminal fault detection model includes a Transformer architecture position encoding model, a self-attention model, a multi-head attention feature extraction model, and a distributed intelligent terminal fault prediction model, specifically including: The expression of the input sample consisting of the operation information of the distributed intelligent terminal is: Among them, X is the input sample of the Transformer architecture, and X(N) is the operation information of the distributed intelligent terminal at time N; Construct the position encoding model of the Transformer architecture. The expression of the position encoding model of the Transformer architecture is: X input =X+PE (7); Among them, PE is the position encoding information, t is the time step, g is the dimension index, d is the total dimension of the feature, X input The input matrix for adding position encoding; Construct a self-attention model and calculate the relationship between different time steps. The expression is: Where: Q, K, V are query, key and value vector models respectively, W Q , W K , W V are the weights of the query, key, and value vector models, respectively. Attention(Q,K,V) is the attention weight. k is the dimension of the key vector, K T is the relationship coefficient between the core key and the related target in the self-attention mechanism, and softmax() is the normalized exponential function; Construct a multi-head attention feature extraction model, the expression is: Among them: MultiHead(Q,K,V) is a multi-head attention feature extraction model, Concat() is a string concatenation function, head j is the j-th sample attention weight, Q j , K j 、V j are the query, key and value vector models of the jth sample, respectively; Construct a distributed intelligent terminal fault prediction model based on the Transformer architecture, the expression is: y=σ(W out h+b out ) (11); Among them, y is the fault prediction result of the distributed intelligent terminal, h is the feature vector processed by the Transformer architecture, and W out and b out is the weight and bias of the output layer, σ is the activation function; According to the output distributed intelligent terminal fault prediction result y, a threshold τ is set to determine whether there is a fault and issue an early warning. If the distributed intelligent terminal fault prediction result y is greater than the threshold τ, 1 is output to determine that the distributed intelligent terminal has a fault. The expression is:

4. The distributed intelligent terminal fault detection system based on Beidou positioning according to claim 1 is characterized in that: The distributed intelligent terminal Beidou navigation technology based on multimodal model correction specifically includes: Based on the basic coordinate system and the state vector, the relative position and relative speed of the distributed intelligent terminal at the distance fault are obtained, and the expression is: Where: Q z (t) is the state vector of the distributed intelligent terminal with relative fault at time t, r z (t) = [x z (t),y z (t),Z z (t)] T is the relative position from the distributed intelligent terminal at time t, V z (t) = [v x (t),v y (t),v z (t)] T is the speed relative to the distributed intelligent terminal at time t; M z (t)=[M G (t),M H (t),M D (t)] is the multimodal data measured at time t, where M G (t) is the satellite signal, indicating whether the distributed terminal has a fault or not. A value of 1 indicates a fault, and a value of 0 indicates no fault. M H (t) is the spatial electromagnetic environment data, which affects the signal transmission quality; M D (t) is terrain information, including contour lines and terrain slope; Construct the kinetic model, the expression is: Q z (t+1)=FQ z (t)+Bu(t)+w(t) (14); Where: Q z (t+1) is the state vector of the distributed intelligent terminal relative to the fault at time t+1, F is the state transfer matrix, B is the control matrix, u(t) is the control input, and w(t) is the process noise; Construct a multimodal data fusion model, the expression is: Among them: G(t+1) is the state updated at time t+1, r GNSS (t+1) is the GNSS position, z vision (t+1) is the visual feature at time t+1, and H is the observation matrix; The relative position and speed update model of the fused multimodal data is constructed, and the relative position and relative speed of the distributed intelligent terminal are estimated. The expression of the relative position and speed update model of the fused multimodal data is: Where: Q z (t+1) is the state vector predicted at time t+1, are the relative position and relative speed from the distributed intelligent terminal at time t+1, K(t+1) is the Kalman gain at time t+1, H T is the transpose of the observation matrix, P(t+1|t) is the covariance matrix of the state estimate at time t for time t+1, P(t|t-1) is the covariance matrix of the state estimate at time t for time t-1, R is the observation noise covariance, I is the identity matrix, the values ​​on the diagonal are 1, and the values ​​elsewhere are 0.

5. A distributed intelligent terminal fault detection method based on Beidou positioning, characterized in that: The following operations are included: Collect Beidou information of distributed intelligent terminals, including building a pseudo-range model between Beidou satellites and distributed intelligent terminals, calculating the pseudo-range between a single Beidou satellite and the distributed intelligent terminal, using the pseudo-range model between multiple Beidou satellites and distributed intelligent terminals, locking the location information of the distributed intelligent terminal, building a Beidou satellite geographic coordinate conversion model, and converting the location information of the distributed intelligent terminal into geographic coordinates; Collect the operation information of distributed intelligent terminals, train the distributed intelligent terminal fault detection model based on the Transformer architecture, and use the distributed intelligent terminal fault detection model to perform fault detection; If the fault detection result is a fault, a fault signal is sent to the Beidou navigation satellite. The Beidou navigation satellite receives the signal and calculates the coordinate information of the distributed intelligent terminal where the fault occurs. The operation and maintenance center analyzes the coordinate information of the distributed intelligent terminal where the fault occurs and converts it into longitude, latitude, and altitude coordinates. The Beidou navigation technology for distributed intelligent terminals based on multimodal model correction can be used to navigate to the faulty distributed intelligent terminal.

6. The method for distributed intelligent terminal fault detection based on Beidou positioning according to claim 5 is characterized in that: The collecting of Beidou information of distributed intelligent terminals specifically includes: Construct a pseudo-range model between Beidou satellites and distributed intelligent terminals, the expression is: Among them, ρ i is the measured pseudo-range between the i-th Beidou satellite and the distributed intelligent terminal, (X i ,Y i ,Z i ) is the known position of the i-th BeiDou satellite in the Earth-centered Earth-fixed coordinate system, (X0, Y0, Z0) is the position of the distributed intelligent terminal, c is the speed of light, δt is the clock deviation of the distributed intelligent terminal, δt i is the satellite clock bias; Using the pseudo-range model between multiple BeiDou satellites and distributed intelligent terminals, the position (X0, Y0, Z0) of the distributed intelligent terminal and the clock deviation δt of the distributed intelligent terminal are solved; A Beidou satellite geographic coordinate conversion model is constructed to convert the location information of distributed intelligent terminals into geographic coordinates. The expression of the Beidou satellite geographic coordinate conversion model is: Where, φ is the latitude of the distributed intelligent terminal, λ is the longitude of the distributed intelligent terminal, h is the altitude of the distributed intelligent terminal, R E is the average radius of the Earth.

7. The method for detecting faults of distributed intelligent terminals based on Beidou positioning according to claim 5 is characterized in that: The collecting of the operation information of the distributed intelligent terminal, training the distributed intelligent terminal fault detection model based on the Transformer architecture, and performing fault detection using the distributed intelligent terminal fault detection model specifically include: Collect the operating information of the distributed intelligent terminal, where the collected operating information of the distributed intelligent terminal includes temperature, voltage, current, and power, and the expression is: X(t)=[T(t),V(t),I(t),P(t)] (3); Wherein, X(t) is the operating information of the distributed intelligent terminal at time t, T(t) is the temperature collected at time t, V(t) is the voltage collected at time t, I(t) is the current collected at time t, and P(t) is the power collected at time t; The distributed intelligent terminal fault detection model is trained using input samples consisting of the operation information of the distributed intelligent terminals. The distributed intelligent terminal fault detection model includes a Transformer architecture position encoding model, a self-attention model, a multi-head attention feature extraction model, and a distributed intelligent terminal fault prediction model, specifically including: The expression of the input sample consisting of the operation information of the distributed intelligent terminal is: Among them, X is the input sample of the Transformer architecture, and X(N) is the operation information of the distributed intelligent terminal at time N; Construct the position encoding model of the Transformer architecture. The expression of the position encoding model of the Transformer architecture is: X input =X+PE (7); Among them, PE is the position encoding information, t is the time step, g is the dimension index, d is the total dimension of the feature, X input The input matrix for adding position encoding; Construct a self-attention model and calculate the relationship between different time steps. The expression is: Where: Q, K, V are query, key and value vector models respectively, W Q , W K , W V are the weights of the query, key, and value vector models, respectively. Attention(Q,K,V) is the attention weight. k is the dimension of the key vector, K T is the relationship coefficient between the core key and the related target in the self-attention mechanism; Construct a multi-head attention feature extraction model, the expression is: Among them: MultiHead(Q,K,V) is a multi-head attention feature extraction model, Concat() is a string concatenation function, head j is the j-th sample attention weight, Q j , K j 、V j They are the j-th sample query, key and value vector models respectively; a distributed intelligent terminal fault prediction model based on the Transformer architecture is constructed, and the expression is: y=σ(W out h+b out ) (11); Among them, y is the fault prediction result of the distributed intelligent terminal, h is the feature vector processed by the Transformer architecture, and W out and b out is the weight and bias of the output layer, σ is the activation function; According to the output distributed intelligent terminal fault prediction result y, a threshold τ is set to determine whether there is a fault and issue an early warning. If the distributed intelligent terminal fault prediction result y is greater than the threshold τ, 1 is output to determine that the distributed intelligent terminal has a fault. The expression is:

8. The method for distributed intelligent terminal fault detection based on Beidou positioning according to claim 5 is characterized in that: The distributed intelligent terminal Beidou navigation technology based on multimodal model correction specifically includes: Based on the basic coordinate system and the state vector, the relative position and relative speed of the distributed intelligent terminal at the distance fault are obtained, and the expression is: Where: Q z (t) is the state vector of the distributed intelligent terminal with relative fault at time t, r z (t) = [x z (t),y z (t),Z z (t)] T is the relative position from the distributed intelligent terminal at time t, V z (t) = [v x (t),v y (t),v z (t)] T is the speed relative to the distributed intelligent terminal at time t; M z (t)=[M G (t),M H (t),M D (t)] is the multimodal data measured at time t, where M G (t) is the satellite signal, indicating whether the distributed terminal has a fault or not. A value of 1 indicates a fault, and a value of 0 indicates no fault. M H (t) is the spatial electromagnetic environment data, which affects the signal transmission quality; M D (t) is terrain information, including contour lines and terrain slope; Construct the kinetic model, the expression is: Q z (t+1)=FQ z (t)+Bu(t)+w(t) (14); Where: Q z (t+1) is the state vector of the distributed intelligent terminal relative to the fault at time t+1, F is the state transfer matrix, B is the control matrix, u(t) is the control input, and w(t) is the process noise; Construct a multimodal data fusion model, the expression is: Among them: G(t+1) is the state updated at time t+1, r GNSS (t+1) is the GNSS position, z vision (t+1) is the visual feature at time t+1, and H is the observation matrix; The relative position and speed update model of the fused multimodal data is constructed, and the relative position and relative speed of the distributed intelligent terminal are estimated. The expression of the relative position and speed update model of the fused multimodal data is: Where: Q z (t+1) is the state vector predicted at time t+1, are the relative position and relative speed from the distributed intelligent terminal at time t+1, K(t+1) is the Kalman gain at time t+1, H T is the transpose of the observation matrix, P(t+1|t) is the covariance matrix of the state estimate at time t for time t+1, P(t|t-1) is the covariance matrix of the state estimate at time t for time t-1, R is the observation noise covariance, I is the identity matrix, the values ​​on the diagonal are 1, and the values ​​elsewhere are 0.

Images