Pole tower offset monitoring method and system based on polling beidou RTK positioning and lstm auxiliary solution

Abstract

The application discloses a kind of tower deviation monitoring method and system based on polling Beidou RTK positioning and LSTM auxiliary solution, which comprises the following steps: receiving Beidou satellite message, and converting into original observation data format after demodulation and despreading;The identity information of current Beidou terminal, environmental sensor information and the incoming original observation data are packed and data are sent by communication protocol;Polling request is initiated to each Beidou terminal in turn;Single point positioning solution is carried out to the received data, and the float solution of RTK positioning is obtained based on Kalman filtering;LSTM neural network model is constructed and trained;The to-be-measured feature data is input into the trained LSTM neural network model, the coordinate information of corresponding Beidou terminal is obtained, the tower deviation is calculated by anti-accident threshold judgment, and the monitoring of tower deviation is realized.The application reduces the overall power consumption of tower deviation monitoring, reduces the solution error, and improves the precision and efficiency of tower deviation monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tower detection, and in particular to a tower deviation monitoring method and system based on polling Beidou RTK positioning and LSTM auxiliary solution. BACKGROUND

[0002] Power transmission towers play an important role in power transmission tasks. Most towers are distributed in areas prone to geological disasters, such as mountain tops and landslides. Therefore, the overturning inspection of power transmission towers is an important matter, and the deviation of towers within a certain time is an important indicator. Common monitoring methods such as unmanned aerial vehicle monitoring and manual inspection have problems such as high consumption of manpower and material resources, high cost, and low stability. Some enterprises use RTK positioning for monitoring, but integrated RTK terminals are expensive and have high power consumption, while low-cost RTK terminals have low fixing rates and are unstable. SUMMARY

[0003] In order to overcome the defects and deficiencies of the prior art, the present application provides a tower deviation monitoring method and system based on polling Beidou RTK positioning and LSTM auxiliary solution. The neural network model makes the solution logic suitable for the tower use scenario, reduces the solution error, and realizes low-cost and efficient monitoring of the tower deviation.

[0004] In order to achieve the above purpose, the present application adopts the following technical solutions:

[0005] A tower deviation monitoring method based on polling Beidou RTK positioning and LSTM auxiliary solution, comprising the following steps:

[0006] Receive the Beidou satellite message, and convert it into the original observation data format after demodulation and expansion;

[0007] After receiving the polling confirmation request, the identity information of the current Beidou terminal, the environmental sensor information and the inflow of the original observation data are packed and sent according to the communication protocol;

[0008] The polling request is initiated to each Beidou terminal in turn, and after receiving the Beidou terminal polling response, the polling confirmation request is initiated to the corresponding terminal again, and the data receiving channel is opened;

[0009] Single point positioning solution is performed on the received data to obtain the single point solution, and the floating point solution of the RTK positioning is obtained based on the Kalman filter;

[0010] The output signal features of the environmental sensor and the floating point features obtained based on the extended Kalman filter are taken as the input of the LSTM neural network model, the coordinates of the Beidou terminal are taken as the expected output of the LSTM neural network model, and the LSTM neural network model is trained;

[0011] The feature data to be measured is input into the trained LSTM neural network model, the coordinate information of the corresponding Beidou terminal is obtained, the tower offset condition is calculated through the anti-occurrence threshold judgment, and the monitoring of the tower offset is realized.

[0012] As a preferred technical solution, a polling request is initiated to each Beidou terminal in turn, and the polling protocol followed is:

[0013] When each Beidou terminal is connected to the network, a polling table is formed according to the time sequence of the allocated addresses, the Beidou terminal with the last allocated address is at the end of the polling table, and the Beidou terminal with the first allocated address is at the beginning of the polling table.

[0014] If the polling end flag is received, the polling confirmation end information is sent and the mark of the address of the Beidou terminal in the polling table is deleted, if no mark is detected in the polling table, the polling is ended, and the next polling is re-entered.

[0015] The polling table is represented as:

[0016] pol={y1:1,y2:2,…,y n : n}

[0017] Wherein, n represents the number of the polled Beidou terminals, y n represents the address of the polled Beidou terminal, and l n represents the mark corresponding to the address of the Beidou terminal.

[0018] As a preferred technical solution, the received data is subjected to single point positioning solution, and the single point solution after the solution is obtained, which specifically includes:

[0019] The single point positioning is performed by referring to the pseudo-range error equation constructed by the reference satellite and each observation satellite, and the formula is as follows:

[0020]

[0021]

[0022] Wherein, is the pseudo-range residual error, δx k , δy k , and δz k are the offset amounts relative to the single point positioning result of the previous epoch, respectively, The first three coefficients obtained by Taylor series expansion of the satellite-geometric distance at the approximate coordinates of the station, c is the speed of light, and δt is the clock difference between the monitoring node and the satellite k The clock difference between the monitoring node and the satellite p The clock difference between the monitoring node and the satellite The single-point positioning estimate of the previous epoch, The ionospheric error, The tropospheric error, if the receiver module is a single-frequency module, The ionospheric error is established by the global broadcast ionospheric model (Broadcast model), if the receiver module is a dual-frequency module, The ionospheric error is obtained by the iono-free linear combination (Iono Free LC) algorithm.

[0023] As a preferred technical solution, the Kalman filter adopts an extended Kalman filter, and is specifically represented as:

[0024]

[0025]

[0026]

[0027] wherein, and P k are the state vector to be estimated and the variance-covariance matrix thereof at the current epoch t k (-) and (+) represent the before and after of the measurement update of the extended Kalman filter, h(x), H(x), and R k are the conversion vector of the measurement model, the corresponding partial derivative matrix, and the variance-covariance matrix of the corresponding error, respectively, and y k represents the observable measurement value, represents the residual error obtained by converting the observable measurement value through the measurement model before the measurement update of the extended Kalman filter, and I is the unit matrix.

[0028] As a preferred technical solution, the tower offset is calculated after the adventitious threshold is judged, so as to realize the monitoring of the tower offset, and specifically includes:

[0029] The threshold and the duration of the environmental feature are judged, if the duration and the change amount of the environmental feature at different epochs are within the set threshold range, then the neural network output result of the epoch is subtracted from the previous result to form the tower offset, and if not, the epoch result is recorded and not used.

[0030] The application also provides a tower offset monitoring system based on a Beidou RTK positioning and LSTM auxiliary solution, comprising a Beidou terminal, a communication module, a central control module, a CORS module, a solution module, a neural network model construction module and a neural network model training module.

[0031] The Beidou terminal is used for receiving a Beidou satellite message, converting into a raw observation data format through demodulation and despreading, and transmitting to the communication module.

[0032] The communication module is used for, after receiving a polling confirmation request, packing the identity information of the Beidou terminal, the environmental sensor information and the inflow raw observation data into a communication protocol and sending data.

[0033] The central control module is used for sequentially initiating a polling request to each Beidou terminal, initiating a polling confirmation request to the corresponding Beidou terminal again after receiving a polling response of the Beidou terminal, and opening a data receiving channel.

[0034] The solution module is used for performing single point positioning solution on the received data to obtain a single point solution, and interconnecting the single point solution with the CORS module in the form of network RTK to obtain a floating point solution of RTK positioning based on Kalman filtering.

[0035] The neural network model construction module is used for constructing an LSTM neural network model.

[0036] The neural network model training module is used for taking the output signal features of the environmental sensor and the floating point features obtained based on extended Kalman filtering as inputs of the LSTM neural network model, taking the coordinates of the Beidou terminal as expected outputs of the LSTM neural network model, and training the constructed LSTM neural network model.

[0037] The LSTM neural network model is trained by using the output signal features of the environmental sensor, the floating point features obtained based on extended Kalman filtering and the coordinate features of the Beidou terminal.

[0038] The solution module is used for inputting the to-be-measured feature data into the trained LSTM neural network model, obtaining the coordinate information of the corresponding Beidou terminal, and further calculating the tower offset through the anti-occurrence threshold value to realize the monitoring of the tower offset.

[0039] As a preferred technical solution, the central control module is used for sequentially initiating a polling request to each Beidou terminal, and the polling protocol followed is as follows:

[0040] When each Beidou terminal is connected to the network, a polling table is formed according to the time sequence of the allocated addresses, the terminal with the last allocated address is arranged at the beginning of the polling table, and the Beidou terminal with the first allocated address is arranged at the end of the polling table.

[0041] If the end of polling flag is received, the polling end confirmation information is sent and the mark of the Beidou terminal address in the polling table is deleted, if no mark is detected in the polling table, the polling ends and the next polling is entered again;

[0042] The polling table is expressed as:

[0043] pol={y1:l1,y2:l2,…,y n :l n}

[0044] Wherein, n represents the number of the polled Beidou terminals, y n represents the polled Beidou terminal address, and l n represents the mark corresponding to the Beidou terminal address.

[0045] As a preferred technical solution, the solving module is used for single point positioning solving of the received data to obtain a single point solution, and specifically comprises:

[0046] Single point positioning is performed by referring to the pseudo-range error equation constructed by the reference satellite and each observation satellite, and the formula is as follows:

[0047]

[0048]

[0049] Wherein, is a pseudo-range residual error, δx k , δy k , and δz k are offset amounts relative to a previous epoch single point positioning result, are the first three coefficients obtained by Taylor series expansion of the satellite-geometric distance at the approximate coordinates of the station, c is the speed of light, δt k is a clock difference of the monitoring node receiving module, δt p is a satellite clock difference, represents a single point positioning estimate of a previous epoch, is an ionospheric error, is a tropospheric error, if a receiver module is a single frequency module, then is established by a global broadcast ionospheric model (Broadcast model), if the receiver module is a dual frequency module, then is obtained by an ionosphere free linear combination (Iono Free LC) algorithm.

[0050] As a preferred technical solution, the Kalman filter adopts an extended Kalman filter, and is specifically expressed as:

[0051]

[0052]

[0053]

[0054] wherein, and P k is the state vector to be estimated at current epoch t k and its variance-covariance matrix, (-) and (+) represent the pre and post of the measurement update of extended Kalman filter, h(x), H(x) and R k are the transition vector, the corresponding partial derivative matrix and the variance-covariance matrix of the corresponding error of the measurement model respectively.

[0055] As a preferred technical solution, the tower offset situation is calculated through the anti-accidental threshold judgment, and the tower offset is monitored, and specifically comprises:

[0056] The threshold and the duration of the environmental characteristics are judged, if the duration and the change amount of the environmental characteristics under different epochs are within the set threshold range, the difference between the epoch neural network output result and the previous result is formed to form the tower offset, and if not within the range, the epoch result is recorded and not used.

[0057] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0058] (1) The present application sequentially initiates a polling request to each Beidou terminal, performs single point positioning calculation on the received data, obtains a floating point solution of RTK positioning based on Kalman filtering, trains an LSTM neural network based on the output signal characteristics of the environmental sensor, the floating point characteristics obtained based on extended Kalman filtering and the Beidou terminal coordinate characteristics, obtains the coordinate information of the corresponding Beidou terminal, calculates the tower offset situation through anti-accidental threshold judgment, and makes the calculation logic suitable for the tower use scenario through the neural network model, reduces the calculation error caused by the low-cost receiver, and realizes low-cost and efficient monitoring of the tower offset.

[0059] (2) The present application independently divides the data receiving and calculation of the receiver into a Beidou terminal and a calculation module, and uses a polling mode to reduce the pressure of the communication module, greatly reduces the cost and power consumption of the Beidou terminal, obtains a fixed solution based on LSTM auxiliary calculation, makes the calculation logic suitable for the tower use scenario through the neural network model, reduces the calculation error caused by the low-cost receiver, and finally realizes low-cost and efficient monitoring of the tower offset through threshold judgment and data processing. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 FIG. 1 is a structural schematic diagram of a tower offset monitoring system based on polling Beidou RTK positioning and LSTM auxiliary calculation of the present application;

[0061] Figure 2 This is a schematic diagram illustrating the implementation process of the tower offset monitoring system based on polling BeiDou RTK positioning and LSTM-assisted calculation according to the present invention.

[0062] Figure 3 This is a flowchart illustrating the polling protocol of the present invention. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0064] Example 1

[0065] like Figure 1 , Figure 2 As shown, this embodiment provides a tower offset monitoring system based on polling BeiDou RTK positioning and LSTM-assisted calculation, including: a BeiDou terminal (corresponding to the BeiDou module in the figure), a communication module, a central control module, a CORS module, a calculation module, a neural network model construction module, and a neural network model training module, specifically including:

[0066] Step 1: The Beidou terminal receives Beidou satellite messages, demodulates and amplifies them into the original observation data format, and transmits them to the communication module for further processing;

[0067] In this embodiment, in step 1, the BeiDou terminal only receives satellite signals and demodulates and amplifies them, converting them into raw observation data to be processed in RTCM3.3 format protocol. Edge processing is not performed on the BeiDou terminal to reduce the manufacturing cost and power consumption required for the BeiDou terminal.

[0068] In this embodiment, in order to reduce the size of the data packets of the communication module and alleviate the communication pressure, most of the information in the NMEA protocol format attached to the receiving module is not retained, and only the data with information numbers 1042 and 1121 to 1127 in the RTCM3.3 protocol format is transmitted.

[0069] Step 2: After receiving the polling confirmation request, the communication module packages the Beidou terminal's identity information, environmental sensor information, and incoming raw observation data into a communication protocol and sends the data in real time.

[0070] In step 2, the BeiDou terminal identity information is the address automatically assigned by the program when the BeiDou terminal is placed. The environmental sensors include temperature sensors, humidity sensors, and barometric pressure sensors. The BeiDou terminal identity information and environmental information are inserted together at the beginning of the RTCM frame structure. Finally, the frame header and frame tail of the communication protocol are added before and after the overall data to form a complete communication frame.

[0071] The communication means used in step 2 should maximize the packet capacity and rate of a single transmission packet to ensure that a Beidou raw observation data is transmitted in one packet.

[0072] Step 3: The central control module initiates a polling request to each Beidou terminal in turn according to the agreed protocol, and transmits the data sent by the Beidou terminal responding to the request to the solution module;

[0073] The central control module in step 3 needs to initiate a polling request to the corresponding Beidou terminal when managing polling, and after receiving the polling response of the Beidou terminal, initiates a polling confirmation request to the corresponding Beidou terminal again, and opens the data receiving channel.

[0074] As shown in Figure 3 , the polling protocol followed in step 3 is:

[0075] When the central control module is powered on for the first time and each Beidou terminal is connected to the network, a polling table is formed according to the time of address allocation. The Beidou terminal with the last address allocation will be at the beginning of the polling table, and the Beidou terminal with the first address allocation will be at the end of the polling table, so as to facilitate the rapid check and debugging of the Beidou terminal when it is powered on again. Each polling unit in the polling table is composed of the corresponding Beidou terminal address and a marker. In this polling thread, if the central control module receives the polling end flag sent by the communication module, it sends a polling confirmation end information to the communication module and deletes the marker of the Beidou terminal address in the polling table. If no marker is detected in the polling table, the current polling is ended and all polling units are marked for the next polling. The polling table is as follows:

[0076] pol = {y1: l1, y2: l2, …, y n : l n}

[0077] Where n represents the number of polled Beidou terminals, y n represents the address of the polled Beidou terminal, and l n represents the marker corresponding to the address of the Beidou terminal.

[0078] Step 4: The solution module performs single point positioning solution on the received data, and the single point solution obtained after the solution is interconnected with the CORS module in the form of network RTK to obtain the differential information issued by the base station, and the floating point solution of RTK positioning is obtained based on Kalman filtering;

[0079] In this embodiment, the network RTK in step 4 adopts a virtual reference station mode, and a virtual reference station is established near the single point positioning coordinates of the Beidou terminal to form a short baseline with a length of less than 1 km, so as to reduce the error caused by the low-cost receiver.

[0080] Step 4: Single point positioning is performed by using the pseudo-range error equation constructed by the reference satellite and each observation satellite, as shown in the following formula:

[0081]

[0082]

[0083] wherein is the pseudo-range residual error, δx k , δy k , and δz k are the offset amounts relative to the single point positioning result of the previous epoch, are the first three coefficients obtained by performing Taylor series expansion on the geometric distance between the satellite and the station at the approximate coordinates of the station, c is the speed of light, and δt k is the clock difference of the monitoring node receiving module, δt p is the satellite clock difference. represents the single point positioning estimate of the previous epoch, is the ionospheric error, is the tropospheric error. If the receiver module is a single-frequency module, then is established by the global broadcast ionospheric model (Broadcast model), and if the receiver module is a dual-frequency module, then is obtained by the ionosphere-free linear combination (Iono Free LC) algorithm.

[0084] In this embodiment, the Kalman filter used in Step 4 is the extended Kalman filter, and the formula is as follows:

[0085]

[0086]

[0087]

[0088] wherein, and P k are the state vector to be estimated and the variance-covariance matrix thereof at the current epoch t k . (-) and (+) represent the before and after of the measurement update of the extended Kalman filter, h(x), H(x), and R k are the conversion vector of the measurement model, the corresponding partial derivative matrix, and the variance-covariance matrix of the corresponding error, respectively, and the nonlinearity of the Kalman filter is reflected in that h(x) is nonlinear, y k represents the observable measurement value, represents the residual error obtained by converting the observable measurement value through the measurement model before the measurement update of the extended Kalman filter, and I is the unit matrix.

[0089] In the case that the observation satellite is in good condition and the spatial distribution of the satellites is reasonable, the receiver self-determination orthogonality detection method can be used to judge and exclude all observed satellites one by one, and then the pseudo-range residual error is calculated again by single point positioning, and the result with the minimum residual error value is selected as the input for the next step.

[0090] Step 5: The neural network model construction module constructs the LSTM neural network model. In a time period during which the environment of different Beidou terminals remains stable for a long time, the LSTM neural network model is trained by using the output signal features of the environment sensor, the floating point features obtained based on the extended Kalman filter, and the accurate coordinate features of the Beidou terminal.

[0091] In a time period during which the observation state of the Beidou satellite is good and the environment of different Beidou terminals remains stable for a long time, the output signal features of the environment sensor and the floating point features obtained based on the extended Kalman filter are used as the input of the LSTM neural network, and the accurate coordinate of the Beidou terminal is used as the expected output of the LSTM training, so as to train the constructed LSTM neural network model.

[0092] In a time period during which the observation state of the Beidou satellite is good and the environment of different Beidou terminals remains stable for a long time, the above step is repeatedly trained until the value of the loss function is less than a preset threshold, and a trained LSTM neural network model is obtained.

[0093] Step 6: The solving module inputs the feature data into the trained LSTM neural network model to obtain the coordinate information of the corresponding Beidou terminal, and then calculates the tower offset through the anti-occurrence threshold judgment, so as to realize the monitoring of the tower offset.

[0094] In the embodiment, the anti-occurrence threshold judgment in step 6 needs to combine the information of the environment features and the time node, and judge the threshold and the continuous time length of the environment features. If the continuous time length and the change amount of the environment features under different ephemeris are within the set threshold range, the difference between the neural network output result of the ephemeris and the previous result is formed to form the tower offset, and if it is not within the range, the ephemeris result is recorded and not used. The reason for using the threshold judgment is that the tower offset is mostly formed slowly and stably under the accumulation of days and months, and the threshold of the environment features can effectively prevent the deviation caused by the unstable calculation caused by the sudden change of the environment. The continuous time length can help to judge whether the abnormal solution is temporarily caused by poor communication channel quality or poor satellite signal. For the extreme tower overturning situation, whether the geometric distance of the output result of the neural network and the threshold is too large and whether the continuous time length is too large can be judged.

[0095] The above embodiments are the preferred embodiments of the present application, but the embodiments of the present application are not limited to the above embodiments, and any changes, modifications, substitutions, combinations, simplifications, etc. made without departing from the spirit and principles of the present application should be equivalent replacement manners and should be included in the protection scope of the present application.

Claims

1. A tower deviation monitoring method based on polling Beidou RTK positioning and LSTM auxiliary solution, characterized in that, Comprise the following steps: Receiving Beidou satellite message, after demodulation and expansion into the original observation data format; After receiving the polling confirmation request, the identity information of the current Beidou terminal, the environmental sensor information and the inflow raw observation data are packed according to the communication protocol and data transmission is carried out; The polling request is initiated to each Beidou terminal in turn, and after receiving the Beidou terminal polling response, the polling confirmation request is initiated to the corresponding terminal again, and the data receiving channel is opened; The received data is subjected to single point positioning solution, and the single point solution after solution is obtained, and the RTK positioning float solution is obtained based on Kalman filtering; An LSTM neural network model is constructed, the output signal features of the environmental sensor and the float features obtained based on the extended Kalman filtering are taken as the input of the LSTM neural network model, and the coordinate of the Beidou terminal is taken as the expected output of the LSTM neural network model, and the constructed LSTM neural network model is trained; The feature data to be tested is input into the trained LSTM neural network model, the coordinate information of the corresponding Beidou terminal is obtained, the tower offset is calculated after the anti-occurrence threshold value is judged, and the tower offset is monitored.

2. The tower offset monitoring method based on the Beidou RTK positioning and LSTM auxiliary solution of claim 1, characterized in that, The polling request is initiated to each Beidou terminal in turn, and the polling protocol followed is: When each Beidou terminal enters the network, a polling table is formed according to the time sequence of the address allocation, the Beidou terminal with the last address allocation will be at the beginning of the polling table, and the Beidou terminal with the first address allocation will be at the end of the polling table; If the polling end flag is received, the polling confirmation end information is sent and the mark of the address of the Beidou terminal in the polling table is deleted, if no mark is detected in the polling table, the polling ends, and the next polling is re-entered; The polling table is represented as: pol = {yl:ll, y2:12,..., yn:ln} where yl, y2,..., yn are the polynomials and ll, 12,..., ln are the corresponding coefficients. n : n : Wherein, n represents the number of polling Beidou terminals, y n represents the address of the polled Beidou terminal, and l n represents the mark corresponding to the address of the Beidou terminal.

3. The tower offset monitoring method based on the Beidou RTK positioning and LSTM auxiliary solution of claim 1, characterized in that, The received data is subjected to single point positioning solution, and the single point solution after solution is obtained, and the RTK positioning float solution is obtained based on Kalman filtering, specifically including: Single point positioning is carried out through the pseudo-range error equation constructed by the reference satellite and each observation satellite, and the formula is as follows: wherein, is the pseudo-range residual, δx k , δy k , δz k are the offset of the relative to the previous epoch single point positioning result, are the first three coefficients obtained by Taylor series expansion of the satellite-geometric distance at the approximate coordinates of the station, c is the speed of light, δt k is the clock error of the monitoring node receiving module, δt p is the satellite clock error, represents the single point positioning estimate of the previous epoch, is the ionospheric error, is the tropospheric error, if the receiver module is a single frequency module, then is established by the global broadcast ionospheric model, if the receiver module is a dual frequency module, then is obtained by the ionosphere-free linear combination Iono Free LC algorithm.

4. The tower offset monitoring method based on the Beidou RTK positioning and LSTM auxiliary solution of claim 1, characterized in that, The Kalman filtering adopts the extended Kalman filtering, and is specifically represented as: wherein, and P k is the state vector to be estimated and its variance-covariance matrix at the current epoch t k (-) and (+) denote the pre- and post- measurement update of the extended Kalman filter, h(x), H(x) and R k are the transformed vector, the corresponding partial derivative matrix and the variance-covariance matrix of the corresponding error of the measurement model, respectively, y k denotes the observed measurement value, denotes the residual error obtained by transforming the observed measurement value through the measurement model before the measurement update of the extended Kalman filter, and I is the identity matrix.

5. The tower offset monitoring method based on the Beidou RTK positioning and LSTM auxiliary solution of claim 1, characterized in that, The tower offset is calculated after the anti-occurrence threshold value is judged, and the tower offset is monitored, specifically including: The threshold value and the continuous time length of the environmental features are judged, if the continuous time length and the change amount of the environmental features at different epochs are within the set threshold value range, the neural network output result of the epoch is subtracted from the previous result to form the tower offset, and if it is not within the range, the epoch result is recorded and not used.

6. A tower deviation monitoring system based on polling Beidou RTK positioning and LSTM auxiliary solution, characterized in that, Comprise: The Beidou terminal, the communication module, the central control module, the CORS module, the solution module, the neural network model construction module, and the neural network model training module; The Beidou terminal is used for receiving Beidou satellite message, and converting into the original observation data format after demodulation and expansion, and transmitting to the communication module; The communication module is used for packing the identity information of the Beidou terminal, the environmental sensor information and the inflow raw observation data according to the communication protocol after receiving the polling confirmation request, and carrying out data transmission; The central control module is used for initiating the polling request to each Beidou terminal in turn, and initiating the polling confirmation request to the corresponding Beidou terminal again after receiving the Beidou terminal polling response, and opening the data receiving channel; The central control module is used for initiating the polling request to each Beidou terminal in turn, and initiating the polling confirmation request to the corresponding Beidou terminal again after receiving the Beidou terminal polling response, and opening the data receiving channel; The solving module is configured to perform single point positioning calculation on the received data to obtain a single point solution, and the single point solution is interconnected with the CORS module in the form of network RTK to obtain a float solution of RTK positioning based on Kalman filtering; The neural network model construction module is configured to construct an LSTM neural network model; The neural network model training module is configured to take the output signal features of the environmental sensor and the float features obtained based on extended Kalman filtering as input of the LSTM neural network model, take the coordinates of the Beidou terminal as expected output of the LSTM neural network model, and train the constructed LSTM neural network model; The output signal features of the environmental sensor and the float features obtained based on extended Kalman filtering are used to train the LSTM neural network model together with the Beidou terminal coordinate features. The solving module is configured to input the to-be-tested feature data into the trained LSTM neural network model to obtain coordinate information of the corresponding Beidou terminal, and further calculate the tower offset based on an anti-accident threshold to realize monitoring of the tower offset.

7. The tower offset monitoring system based on the Beidou RTK positioning and LSTM auxiliary solution of claim 6, characterized in that, The central control module is configured to sequentially initiate a polling request to each Beidou terminal, and the polling protocol followed is as follows: When each Beidou terminal is connected to the network, a polling table is formed according to the time sequence of address allocation, and the terminal with the last address allocation is at the end of the polling table. If a polling end flag is received, a polling confirmation end information is sent and the mark of the address of the Beidou terminal in the polling table is deleted, and if no mark is detected in the polling table, the polling ends and the next polling is re-entered. The polling table is represented as follows: pol = {yl:ll, y2:12,..., yn:ln} where yl, y2,..., yn are the polynomials and ll, 12,..., ln are the corresponding coefficients. n : n} Wherein, n represents the number of polling Beidou terminals, y n represents the address of the polled Beidou terminal, and l n represents the mark corresponding to the address of the Beidou terminal.

8. The tower offset monitoring system based on the Beidou RTK positioning and LSTM auxiliary solution of claim 6, characterized in that, The solving module is configured to perform single point positioning calculation on the received data to obtain a single point solution, and the single point solution is interconnected with the CORS module in the form of network RTK to obtain a float solution of RTK positioning based on Kalman filtering; The Kalman filtering adopts extended Kalman filtering, and is specifically represented as follows: wherein, is the pseudo-range residual, δx k , δy k , δz k are the offset of the relative to the previous epoch single point positioning result, are the first three coefficients obtained by Taylor series expansion of the satellite-geometric distance at the approximate coordinates of the station, c is the speed of light, δt k is the clock error of the monitoring node receiving module, δt p is the satellite clock error, represents the single point positioning estimate of the previous epoch, is the ionospheric error, is the tropospheric error, if the receiver module is a single frequency module, then is established by the global broadcast ionospheric model, if the receiver module is a dual frequency module, then is obtained by the ionosphere-free linear combination Iono Free LC algorithm.

9. The tower offset monitoring system based on the Beidou RTK positioning and LSTM auxiliary solution of claim 6, wherein, Further calculation of the tower offset based on an anti-accident threshold realizes monitoring of the tower offset, and specifically includes: where and P k is the state vector to be estimated at the current epoch t k and its variance-covariance matrix, (-) and (+) denote the pre- and post- measurement update of the extended Kalman filter, h(x), H(x) and R k are the measurement model's transition vector, the corresponding partial derivative matrix and the variance-covariance matrix of the corresponding error, respectively.

10. The tower offset monitoring system based on the Beidou RTK positioning and LSTM auxiliary solution of claim 6, wherein, The threshold and duration of the environmental features are judged, if the duration and the change of the environmental features at different epochs are within the set threshold range, the neural network output result of the epoch is subtracted from the previous result to form the tower offset, and if not, the result of the epoch is recorded and not used. ​

Images