Power terminal positioning method based on Beidou multi-frequency combination

Abstract

The invention relates to the technical field of satellite navigation and positioning, and discloses a Beidou multi-frequency combination-based power terminal positioning method, which comprises the following steps of: acquiring an original Beidou multi-frequency combination signal flow, inputting the original Beidou multi-frequency combination signal flow into a pre-trained positioning parameter matching network to obtain a matching result set, and determining a positioning parameter matching result corresponding to each signal. Wherein the positioning parameter matching network is trained through historical positioning data, the historical data comprises signal quality parameters extracted from historical Beidou multi-frequency signals, positioning abnormal labels and weight coefficients, and the signal quality parameters are determined by combining a single signal and adjacent signals thereof. The method further comprises the steps of extracting and fusing first and second groups of signal quality parameters, determining a positioning anomaly label and a weight coefficient, training a network by using historical joint positioning records, and adjusting network parameters according to training errors. According to the invention, through multi-frequency signal fusion and adaptive learning, the positioning precision and anti-interference capability of the power terminal in a complex environment are improved, and the method is suitable for the precise positioning requirement of a smart power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite navigation and positioning, and specifically to a power terminal positioning method based on Beidou multi-frequency combination. Background Technique

[0002] In the power system, the precise positioning of power terminal equipment is crucial for power grid operation monitoring, fault repair, asset management and other links. Traditional power terminal positioning methods mainly rely on satellite signals of a single frequency band or other positioning technologies. However, these methods face many challenges in complex power environments.

[0003] First, the environment where power facilities are located often has various interference sources, such as electromagnetic interference generated by high-voltage transmission lines, multipath effects in urban building clusters, etc. Signals of a single frequency band are easily interfered, resulting in a decrease in positioning accuracy or even positioning failure. In the prior art, positioning algorithms based on a single frequency band are difficult to effectively cope with signal distortion problems in complex environments and cannot accurately analyze the effective information in the signals, thus affecting the reliability of positioning results.

[0004] Second, with the development of smart grids, higher requirements are put forward for the real-time performance, high precision and stability of power terminal positioning. When traditional positioning methods process multi-source heterogeneous data, they lack comprehensive analysis of signal time-series correlation characteristics and frequency band distribution characteristics, and it is difficult to make full use of the complementary advantages of multi-frequency signals. For example, Beidou signals of different frequency bands have different characteristics during propagation. High-frequency signals may carry more detailed information but have weak penetration ability, while low-frequency signals have strong anti-interference ability but low resolution. How to effectively fuse the characteristics of multi-frequency signals has become a key issue in improving positioning performance.

[0005] In addition, existing positioning parameter matching algorithms usually rely on fixed model parameters and lack adaptive learning ability. In the face of a constantly changing signal environment, they cannot dynamically adjust the matching strategy according to real-time signal characteristics, resulting in insufficient generalization ability of the positioning model. Especially in the power field, the electromagnetic environments in different regions vary greatly, and traditional methods are difficult to meet the positioning requirements in diverse scenarios. For example, in different geographical environments such as mountainous areas and city centers, the propagation characteristics of Beidou signals are significantly different, and positioning models with fixed parameters cannot flexibly adapt to these changes, resulting in unstable positioning accuracy. Summary of the Invention

[0006] The purpose of the present invention is to provide a power terminal positioning method based on Beidou multi-frequency combination to solve the problems raised in the above background technique.

[0007] To achieve the above purpose, the present invention provides the following technical solution: A power terminal positioning method based on Beidou multi-frequency combination, the method is applied to a power terminal positioning system, and the method includes: Obtain the original Beidou multi-frequency combined signal stream to be positioned; Input the original Beidou multi-frequency combined signal stream into a pre-trained positioning parameter matching network to obtain a positioning parameter matching result set; Based on the positioning parameter matching result set, determine the positioning parameter matching result corresponding to each Beidou multi-frequency signal in the original Beidou multi-frequency combined signal stream; The original Beidou multi-frequency combined signal stream includes target frequency band signal data that needs to be parsed; The positioning parameter matching network is a parameter matching network obtained by training an initial parameter matching network to be trained with historical positioning data. The historical positioning data is generated from a batch of historical signal quality parameters extracted from a batch of historical Beidou multi-frequency signals, as well as positioning anomaly labels and weight coefficients corresponding to each historical signal quality parameter. The positioning anomaly label reflects the interference status identifier of a historical signal quality parameter, and the weight coefficient reflects the correction factor corresponding to the interference status identifier. The batch of historical Beidou multi-frequency signals is a batch of continuously collected Beidou multi-frequency signals, and each historical signal quality parameter includes a signal quality parameter determined by a historical Beidou multi-frequency signal and its adjacent historical Beidou multi-frequency signal; The positioning parameter matching result is used to represent the parsing parameter identifier of the target frequency band signal data.

[0008] Preferably, the method further includes: Perform first feature parameter extraction on the batch of historical Beidou multi-frequency signals to obtain a first set of signal quality parameters, where one signal quality parameter in the first set of signal quality parameters corresponds to one historical Beidou multi-frequency signal in the batch of historical Beidou multi-frequency signals; Perform second feature parameter extraction on the batch of historical Beidou multi-frequency signals to obtain a second set of signal quality parameters, where the first signal quality parameter in the second set of signal quality parameters corresponds to the first historical Beidou multi-frequency signal in the historical Beidou multi-frequency signals. The adjacent historical Beidou multi-frequency signal of the first historical Beidou multi-frequency signal is the second historical Beidou multi-frequency signal, and the first signal quality parameter represents the timing correlation feature between the first historical Beidou multi-frequency signal and the second historical Beidou multi-frequency signal; Fuse the first set of signal quality parameters and the second set of signal quality parameters respectively to obtain the batch of historical signal quality parameters.

[0009] Preferably, the performing second feature parameter extraction on the batch of historical Beidou multi-frequency signals to obtain a second set of signal quality parameters includes: The second feature parameters of the batch of historical Beidou multi-frequency signals are extracted through the following steps to obtain a second set of signal quality parameters. Among them, each time the historical Beidou multi-frequency signal for the second feature parameter extraction is used as the current historical Beidou multi-frequency signal, and the obtained signal quality parameter is used as the current signal quality parameter. The second set of signal quality parameters includes the current signal quality parameter; Perform a frequency band analysis operation on the current historical Beidou multi-frequency signal to determine a first set of frequency band feature quantities, where the first set of frequency band feature quantities is used to describe the frequency band distribution of the target frequency band signal data in the current historical Beidou multi-frequency signal; Perform a frequency band analysis operation on the adjacent historical Beidou multi-frequency signal of the current historical Beidou multi-frequency signal to determine a second set of frequency band feature quantities, where the second set of frequency band feature quantities is used to describe the frequency band distribution of the target frequency band signal data in the adjacent historical Beidou multi-frequency signal; Determine the current signal quality parameter based on the first set of frequency band feature quantities and the second set of frequency band feature quantities, where the current signal quality parameter is used to represent the deviation amount between the corresponding frequency band feature quantities in the first set of frequency band feature quantities and the second set of frequency band feature quantities.

[0010] Preferably, the method further includes: Determine the positioning anomaly label and weight coefficient corresponding to each historical signal quality parameter in the batch of historical signal quality parameters through the following steps. Among them, each time the historical signal quality parameter for determining the positioning anomaly label and weight coefficient is used as the current historical signal quality parameter, and the positioning anomaly label corresponding to the current historical signal quality parameter is used as the current positioning anomaly label and the current weight coefficient: Input the current historical signal quality parameter into the initial parameter matching network to obtain the current positioning anomaly label, where the initial parameter matching network is a parameter matching network obtained by pre-executing initialization processing; Determine the current weight coefficient based on the current positioning anomaly label and the reference verification identifier corresponding to the current historical signal quality parameter, where the reference verification identifier carries the calibration label in the historical Beidou multi-frequency signal corresponding to the current historical signal quality parameter, and the calibration label includes the target parsing parameter identifier of the target frequency band signal data. The weight coefficient is used to represent whether the current positioning anomaly label is consistent with the target parsing parameter identifier.

[0011] Preferably, the inputting the current historical signal quality parameter into the initial parameter matching network to obtain the current positioning anomaly label includes: Input each parsing parameter identifier in the pre-stored set of parsing parameter identifiers into the first matching algorithm branch in sequence together with the current historical signal quality parameter to obtain a first set of matching degree indicators, where the first set of matching degree indicators includes the matching degree indicators corresponding to each parsing parameter identifier; Determine the candidate parsing parameter identifier corresponding to the highest numerical value in the first set of matching degree indicators as the current parsing parameter identifier corresponding to the current positioning anomaly label.

[0012] Preferably, before inputting each parsing parameter identifier in the pre-stored set of parsing parameter identifiers into the first matching algorithm branch in sequence together with the current historical signal quality parameter to obtain a first set of matching degree indicators, the method further includes: Determine the signal strength distribution quantity through a preset mode; Determine whether the signal strength distribution quantity conforms to the set strength distribution rule; Based on the signal strength distribution quantity conforming to the set strength distribution rule, dynamically select the candidate parsing parameter identifier in the pre-stored set of parsing parameter identifiers as the current positioning anomaly label; Based on the signal strength distribution quantity not conforming to the set strength distribution rule, input the current historical signal quality parameter into the initial parameter matching network to obtain the current positioning anomaly label.

[0013] Preferably, determining the candidate parsing parameter identifier corresponding to the highest numerical value in the first set of matching degree indicators as the current parsing parameter identifier corresponding to the current positioning anomaly label includes: Based on the candidate parsing parameter identifier corresponding to the first matching degree indicator being the same as the target parsing parameter identifier, determine the current weight coefficient as the first weight coefficient, where the first weight coefficient indicates that the initial parameter matching network successfully generates a matching identifier; Based on the candidate parsing parameter identifier corresponding to the first matching degree indicator being different from the target parsing parameter identifier, determine the current weight coefficient as the second weight coefficient, where the second weight coefficient indicates that the initial parameter matching network fails to generate a matching identifier.

[0014] Preferably, the method further includes: Screen a number of historical joint positioning records from the historical positioning data, where the j-th historical joint positioning record in the number of historical joint positioning records includes the j-th historical signal quality parameter, the j-th positioning anomaly label corresponding to the j-th historical signal quality parameter, the j-th weight coefficient, and the (j + 1)-th historical signal quality parameter, and j is a positive integer; Training the initial parameter matching network to be trained based on the several groups of historical joint positioning records to obtain the positioning parameter matching network. Among them, when the number of iterations for training the initial parameter matching network reaches the set training threshold, the initial parameter matching network is determined as the positioning parameter matching network. When the number of iterations for training the initial parameter matching network does not reach the set training threshold, the network parameters of the initial parameter matching network are updated according to a preset training error function. The input for each round of training process is a group of historical joint positioning records from the several groups of historical joint positioning records.

[0015] Preferably, the method further includes: Training the initial parameter matching network to be trained based on multiple groups of historical joint positioning records through the following steps to obtain the positioning parameter matching network, where the historical joint positioning record input into the initial parameter matching network is the j-th group of joint positioning records: Determining whether the historical Beidou multi-frequency signal corresponding to the (j + 1)-th historical signal quality parameter is the last historical Beidou multi-frequency signal in the historical Beidou multi-frequency signal stream; Based on the j-th weight coefficient, determining the training parameters generated by the initial parameter matching network when the historical Beidou multi-frequency signal corresponding to the (j + 1)-th historical signal quality parameter is the last historical Beidou multi-frequency signal; Based on the j-th weight coefficient and the output of the associated parameter matching network, determining the training parameters when the historical Beidou multi-frequency signal corresponding to the (j + 1)-th historical signal quality parameter is not the last historical Beidou multi-frequency signal, where the associated parameter matching network is a parameter matching network obtained by performing initialization processing in advance, and there are differences in the network parameters between the associated parameter matching network and the initial parameter matching network; Determining the error amount of the training error function based on the training parameters and the output of the initial parameter matching network, and adjusting the network parameters of the initial parameter matching network based on the error amount of the training error function; When the number of iterations for performing the above steps reaches the set training threshold, determining the initial parameter matching network as the positioning parameter matching network; The determining the training parameters based on the j-th weight coefficient and the output of the associated parameter matching network includes: Inputting each parsing parameter identifier in the pre-stored parsing parameter identifier set into the second matching algorithm branch in sequence together with the (j + 1)-th historical signal quality parameter to obtain a second set of matching degree indicators, where the second set of matching degree indicators includes the matching degree indicators corresponding to each parsing parameter identifier; Determine the weighted result of the second matching metric with the highest value among the second set of matching metrics and the j-th weight coefficient as the training parameter.

[0016] Preferably, the method further includes: Performing joint training operations on each set of records in the multiple sets of historical joint positioning records to generate optimization parameters for the positioning parameter matching network, where the joint training operation for each set of records includes adjusting the update amplitude of the network parameters based on the correlation features between adjacent historical signal quality parameters; After the joint training operation for each set of records is completed, verify the parsing parameter identification matching accuracy of the positioning parameter matching network for the target frequency band signal data, and re-execute the training process when the accuracy does not reach the set standard.

[0017] Compared with the prior art, the beneficial effects of the present invention are: By obtaining the original Beidou multi-frequency combined signal stream and inputting it into the pre-trained positioning parameter matching network, the comprehensive processing of multi-frequency signals is realized. The positioning parameter matching network is trained using historical positioning data, where the historical data contains signal quality parameters determined by combining adjacent signals, and can effectively capture the temporal correlation features and frequency band distribution features of the signals. For example, when extracting signal quality parameters, both the frequency band characteristic quantity of a single signal and the frequency band deviation quantity between adjacent signals are considered. This dual feature extraction method enables the model to more comprehensively understand the signal characteristics, thereby improving the accuracy of parsing the target frequency band signal data.

[0018] Positioning anomaly labels and weight coefficients are introduced into the historical positioning data, and through the comparison between the reference verification identifier and the target parsing parameter identifier, the dynamic calibration of the model output is realized. When the candidate parsing parameter identifier corresponding to the matching metric is consistent with the target parsing parameter identifier, a higher weight coefficient is given to strengthen the correct matching; when they are inconsistent, the weight is reduced to guide the model to optimize the matching strategy. This mechanism enhances the model's ability to identify interference signals, effectively suppresses the influence of environmental factors such as electromagnetic interference and multipath effects on the positioning result, and improves the anti-interference ability of positioning.

[0019] During the network training process, iterative training is performed using multiple sets of historical joint positioning records, and the training parameters are dynamically adjusted in combination with the output of the correlation parameter matching network. By analyzing the correlation features between adjacent historical signal quality parameters, the update amplitude of the network parameters is optimized, enabling the positioning parameter matching network to adaptively learn the variation law of signal features. For example, based on the judgment of whether adjacent signals are the last signals, different training parameter calculation methods are adopted to ensure that the model can achieve accurate matching in different signal sequence scenarios, improving the generalization ability of the model and its adaptability to diverse power environments.

[0020] This method realizes the joint analysis of the spatial and temporal characteristics of signals by fusing the first group and the second group of signal quality parameters. The first group of parameters reflects the independent characteristics of individual signals, and the second group of parameters reflects the temporal correlation of adjacent signals. The fusion of the two enables the model to analyze signals from multiple dimensions, effectively solving the limitations of single - feature analysis in traditional methods. This multi - dimensional feature fusion strategy not only improves the richness of signal quality parameters but also provides a more solid foundation for the accurate matching of positioning parameters, thereby enhancing the overall performance of power terminal positioning.

[0021] By setting training thresholds and error functions, it is ensured that the positioning parameter matching network is continuously optimized during training until the set accuracy standard is reached. This closed - loop training mechanism guarantees the stability and reliability of the model, enabling this positioning method to continuously provide high - precision positioning services in actual power scenarios and meet the requirements of the smart grid for real - time and accurate positioning of power terminal devices. Brief Description of the Drawings

[0022] Figure 1 It is the working principle diagram of the power terminal positioning method based on Beidou multi - frequency combination described in the present invention; Figure 2 It is the flow chart of historical signal quality parameter extraction; Figure 3 It is the design diagram of the second feature parameter extraction. Detailed Implementation Modes

[0023] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] Please refer to Figures 1-3 , the power terminal positioning method based on Beidou multi - frequency combination involved in the present invention is specifically implemented as follows: Obtain the original Beidou multi - frequency combination signal stream to be positioned, and this signal stream contains the target frequency band signal data to be parsed.

[0025] Input the original Beidou multi-frequency combined signal stream into the pre-trained positioning parameter matching network to obtain a set of positioning parameter matching results. Among them, the positioning parameter matching network is trained as follows: Input historical positioning data into the initial parameter matching network to be trained for training. The historical positioning data is generated from a batch of historical signal quality parameters respectively extracted from a batch of historical Beidou multi-frequency signals, as well as the positioning anomaly labels and weight coefficients corresponding to each historical signal quality parameter. The historical Beidou multi-frequency signals are continuously collected signals, and each historical signal quality parameter is determined by a historical Beidou multi-frequency signal and its adjacent historical Beidou multi-frequency signal. The positioning anomaly label reflects the interference status identifier of the historical signal quality parameter, and the weight coefficient reflects the correction factor corresponding to the interference status identifier.

[0026] Based on the set of positioning parameter matching results, determine the positioning parameter matching result corresponding to each Beidou multi-frequency signal in the original Beidou multi-frequency combined signal stream. This result is used to represent the parsing parameter identifier of the target frequency band signal data.

[0027] The present invention will be further described below in conjunction with Embodiments 1 to 5: Embodiment 1: In this embodiment, the extraction process of historical signal quality parameters is refined. The specific steps are as follows: Perform first feature parameter extraction on a batch of continuously collected historical Beidou multi-frequency signals to obtain a first set of signal quality parameters, where each signal quality parameter corresponds one-to-one to a corresponding single historical Beidou multi-frequency signal. The first feature parameter extraction process focuses on the independent features of a single signal and obtains its unique feature quantities through time-domain and frequency-domain analysis of each signal. For example, for each historical Beidou multi-frequency signal, basic attribute parameters such as the amplitude, frequency, and phase of the signal can be extracted through signal processing means such as Fourier transform to form the first set of signal quality parameters. In this process, the feature extraction of each signal is carried out independently, aiming to capture the state information of a single signal at a specific moment and provide basic data support for subsequent analysis.

[0028] Perform second feature parameter extraction on the same batch of historical Beidou multi-frequency signals to obtain a second set of signal quality parameters. The first signal quality parameter in the second set of signal quality parameters corresponds to the first historical Beidou multi-frequency signal, and the adjacent signal of the first historical Beidou multi-frequency signal is the second historical Beidou multi-frequency signal. The first signal quality parameter represents the timing correlation feature between the first signal and the second signal. Specifically, the second feature parameter extraction process needs to consider the correlation between adjacent signals to dynamically capture the change law of signals in the time series.

[0029] During the second feature parameter extraction process, the historical Beidou multi-frequency signal for each feature extraction is used as the current historical Beidou multi-frequency signal, and the obtained signal quality parameter is used as the current signal quality parameter. The second set of signal quality parameters includes all current signal quality parameters. The specific operations are as follows: Perform a frequency band analysis operation on the current historical Beidou multi-frequency signal to determine the first set of frequency band feature quantities, which are used to describe the frequency band distribution of the target frequency band signal data in the current signal. The frequency band analysis can be implemented through a spectrum analysis algorithm. For example, calculate parameters such as the energy proportion of the target frequency band in the current signal, frequency offset, and bandwidth change to characterize the distribution characteristics of the target frequency band in the current signal. Perform the same frequency band analysis operation on the adjacent historical Beidou multi-frequency signal of the current historical Beidou multi-frequency signal to determine the second set of frequency band feature quantities, which are used to describe the frequency band distribution of the target frequency band signal data in the adjacent signal. The selection of adjacent signals follows the continuity of the time series, that is, the previous or next continuously collected signal of the current signal, which can be specifically determined according to the time sequence direction of signal processing. Based on the first set of frequency band feature quantities and the second set of frequency band feature quantities, determine the current signal quality parameter. The specific method is to calculate the deviation amount between the corresponding frequency band feature quantities of the two, and this deviation amount is used as the current signal quality parameter to characterize the difference degree of the frequency band distribution between adjacent signals. For example, if the energy proportion of the target frequency band in the first set of frequency band feature quantities is A and the corresponding value in the second set is B, the deviation amount can be expressed as |A - B|; if the frequency offsets are and , the deviation amount can be expressed as . In this way, the frequency band distribution difference between adjacent signals is quantified into specific numerical parameters, thereby reflecting the stability or change trend of the signal in the time series.

[0030] Finally, perform fusion processing on the first set of signal quality parameters and the second set of signal quality parameters respectively to obtain a batch of complete historical signal quality parameters. The fusion process aims to combine the independent features of a single signal with the associated features of adjacent signals to form a more comprehensive signal quality description. The fusion methods can adopt data splicing, feature weighting, etc. For example, combine each signal quality parameter in the first set with the corresponding adjacent deviation parameter in the second set to form a multi-dimensional feature vector. Among them, data splicing arranges the feature parameters of different groups in sequence to form a feature space with more dimensions; feature weighting assigns corresponding weights according to the importance of different features and then performs a linear combination to highlight the contribution of key features to the signal quality description. Through the fusion processing, the historical signal quality parameters can more comprehensively reflect the characteristics of the Beidou multi-frequency signal, including both the static features of a single signal and the dynamic association features of adjacent signals, providing rich feature inputs for the training of the subsequent positioning parameter matching network, enabling it to learn the mapping relationship between more complex signal patterns and positioning parameters.

[0031] During the entire implementation process, it is necessary to ensure the continuous acquisition of historical Beidou multi-frequency signals to guarantee that the temporal correlation characteristics between adjacent signals have practical physical significance. At the same time, the specific algorithms for feature extraction and fusion can be selected and adjusted according to actual signal processing requirements, but all should be centered around the core goal of obtaining parameters that can accurately reflect signal quality, ensuring that the subsequent positioning parameter matching network can be effectively trained based on reliable historical data, thereby enhancing the analytical ability for signal data in the target frequency band in the original Beidou multi-frequency combined signal stream and the accuracy of positioning parameter matching.

[0032] Embodiment 2: During the construction of historical positioning data, for each current historical signal quality parameter among a batch of historical signal quality parameters, the operation of determining the positioning anomaly label and weight coefficient needs to be performed. The specific implementation method is as follows: Input the current historical signal quality parameter into the initially parameterized matching network that has been pre-initialized, and obtain the current positioning anomaly label. The initially parameterized matching network is a neural network model constructed based on a deep learning architecture. Its network structure includes an input layer, a hidden layer, and an output layer. Through learning a large amount of historical data, it can capture the potential mapping relationship between signal quality parameters and positioning anomaly labels.

[0033] When obtaining the current positioning anomaly label, a pre-stored set of parsing parameter identifiers is used. This set contains various possible parsing parameter identifiers, which are used to represent different parsing states of target frequency band signal data. For example, the parsing parameter identifiers can include different types of positioning anomaly situations such as signal normal state, multipath effect interference state, ionospheric scintillation interference state, etc. Input each parsing parameter identifier in the set into the first matching algorithm branch of the initially parameterized matching network in sequence together with the current historical signal quality parameter. The first matching algorithm branch is a sub-module in the initially parameterized matching network, and its internal structure includes multiple layers of neural networks, which are used for feature extraction and matching degree calculation of the input signal quality parameters and parsing parameter identifiers. Through the calculation of this branch, the first set of matching degree indicators can be obtained. Each set of matching degree indicators corresponds to a parsing parameter identifier, reflecting the matching degree between this identifier and the current signal quality parameter. The calculation of the matching degree indicator is based on the output result of the neural network, usually represented in the form of a probability value or a similarity score. The higher the value, the better the matching degree.

[0034] Select the first matching degree indicator with the highest value from the first set of matching degree indicators, and determine the candidate parsing parameter identifier corresponding to it as the current parsing parameter identifier corresponding to the current positioning anomaly label. For example, if the set of parsing parameter identifiers includes three identifiers A, B, and C, and after the calculation of the first matching algorithm branch, the matching degree of A is 0.8, B is 0.6, and C is 0.7, then select the one with the highest matching degree, A, as the current parsing parameter identifier, that is, the current positioning anomaly label.

[0035] Before performing the above matching operation, it is necessary to first determine the signal strength distribution quantity through a preset mode. The signal strength distribution quantity is a parameter that describes the distribution of the signal strength in the target frequency band in the time or frequency dimension, and can be obtained by statistically analyzing the signal strength data of the target frequency band in the current historical signal quality parameters. For example, calculate statistical quantities such as the mean, variance, maximum value, and minimum value of the signal strength, or draw a distribution map of the signal strength changing with time or frequency. Then, determine whether the signal strength distribution quantity conforms to the set strength distribution rule. The strength distribution rule is a preset judgment criterion used to evaluate the rationality of the signal strength distribution. For example, judge whether the signal strength is within the preset threshold range, whether there are abnormal mutations or fluctuations, and whether it conforms to a specific distribution pattern (such as a normal distribution), etc.

[0036] If the signal strength distribution quantity conforms to the set strength distribution rule, it indicates that the signal state is stable, and the candidate parsing parameter identifier in the pre-stored parsing parameter identifier set can be directly and dynamically selected as the current positioning anomaly label without performing complex calculations through the initial parameter matching network. In this case, according to the characteristics of the signal strength distribution, the corresponding parsing parameter identifier can be quickly matched, improving the processing efficiency. If it does not conform to the rule, the current historical signal quality parameters must be input into the initial parameter matching network, and the current positioning anomaly label is obtained through algorithm calculation to ensure the accuracy of the label. This is because when the signal strength distribution is abnormal, directly selecting the parsing parameter identifier may lead to incorrect positioning results, and through the calculation of the initial parameter matching network, various factors can be comprehensively considered to improve the reliability of the label.

[0037] Determine the current weight coefficient based on the current positioning anomaly label and the reference verification identifier corresponding to the current historical signal quality parameters. The reference verification identifier is a reference identifier preset for each historical signal quality parameter, carrying the calibration label in the corresponding historical Beidou multi-frequency signal of the current historical signal quality parameters. The calibration label contains the target parsing parameter identifier of the target frequency band signal data. The target parsing parameter identifier is a known and accurately reflects the true state of the signal parsing parameter identifier, which can be obtained through manual annotation or other reliable methods.

[0038] The specific judgment logic is as follows: If the candidate parsing parameter identifier corresponding to the first matching degree index (i.e., the current parsing parameter identifier) is the same as the target parsing parameter identifier, it indicates that the initial parameter matching network has successfully generated a matching identifier, and the current weight coefficient is determined as the first weight coefficient. The first weight coefficient is a pre-set value used to represent the state of successful matching, usually set to a relatively large value to emphasize the importance of this set of data for network training. If the two are different, it indicates that the matching fails, and the current weight coefficient is determined as the second weight coefficient. The second weight coefficient is also a pre-set value used to represent the state of failed matching, usually set to a relatively small value to reduce the impact of this set of data on network training.

[0039] When determining the weight coefficient, the adjustment effect of the signal strength distribution amount on the weight also needs to be considered. If the signal strength distribution amount conforms to the set strength distribution rule, it indicates that the signal quality is good. At this time, the weight coefficient can be appropriately increased to enhance the influence of this set of data in network training; if it does not conform to the rule, it indicates that the signal quality is poor and there is significant interference. At this time, the weight coefficient can be appropriately decreased to reduce the interference of noise data on network training.

[0040] In this way, a weight coefficient corresponding to its matching state and signal quality is assigned to each historical signal quality parameter, so that in the subsequent training process of the positioning parameter matching network, more attention can be paid to the data with successful matching and good signal quality, improving the learning efficiency and generalization ability of the network. At the same time, the introduction of the weight coefficient can also effectively handle noise data and abnormal data, enhancing the robustness of the network, enabling it to accurately analyze Beidou multi-frequency signals in a complex electromagnetic environment and provide reliable positioning services for power terminals.

[0041] During the entire implementation process, it is necessary to ensure the accuracy and reliability of the benchmark verification identifier, and its quality directly affects the rationality of the weight coefficient and the effect of network training. At the same time, the initialization process of the initial parameter matching network is also crucial. Appropriate initialization methods (such as random initialization, pre-training initialization, etc.) need to be adopted, and reasonable network parameters (such as learning rate, number of iterations, etc.) need to be set to ensure that the network can effectively learn the mapping relationship between signal quality parameters and positioning anomaly labels. In addition, the construction of the parsing parameter identifier set should comprehensively cover possible positioning anomaly situations to ensure that the signal state can be accurately identified in various scenarios.

[0042] Embodiment 3: In the training session of the positioning parameter matching network, several groups of historical joint positioning records need to be constructed based on historical positioning data, and the performance of the initial parameter matching network is optimized through iterative training. The specific implementation method is as follows: Filter several groups of historical joint positioning records from the historical positioning data. Each group of historical joint positioning records (the j-th group, where j is a positive integer) includes the j-th historical signal quality parameter, the j-th positioning anomaly label corresponding to this parameter, the j-th weight coefficient, and the (j + 1)-th historical signal quality parameter. The construction of these records follows the continuity of the time series, that is, the historical BDS multi-frequency signal corresponding to the (j + 1)-th historical signal quality parameter is the next consecutive acquisition signal of the j-th signal, ensuring that the temporal correlation characteristics between adjacent signals can be effectively captured. By filtering consecutive signal pairs, joint records containing the characteristics of the front and rear signals, matching labels, and weights are formed, providing sample data with temporal dependence for network training.

[0043] The training process uses several groups of historical joint positioning records as input and iteratively trains the initial parameter matching network. The specific steps are as follows: After inputting the j-th group of historical joint positioning records, first determine whether the historical BDS multi-frequency signal corresponding to the (j + 1)-th historical signal quality parameter is the last signal in the historical BDS multi-frequency signal stream. The judgment of the last signal is based on the time order of signal acquisition, that is, there are no subsequent consecutive acquisition signals after this signal. If it is the last signal, determine the training parameters generated by the initial parameter matching network based on the j-th weight coefficient. At this time, the determination of the training parameters only depends on the weight coefficient of the current group, and this weight coefficient already reflects the matching status (such as successful or failed matching) between the j-th signal quality parameter and the positioning anomaly label. For example, if the j-th weight coefficient is the first weight coefficient (successful matching), the training parameters can be set to a value positively correlated with this weight coefficient; if it is the second weight coefficient (failed matching), the training parameters can be set to a value negatively correlated with this weight coefficient to guide the network to update the parameters in the direction of reducing the matching error.

[0044] If the historical Beidou multi-frequency signal corresponding to the (j + 1)-th historical signal quality parameter is not the last signal, the training parameter is jointly determined based on the j-th weight coefficient and the output of the associated parameter matching network. The associated parameter matching network is another parameter matching network with the same network structure as the initial parameter matching network but different network parameters, and its parameters are set through an independent initialization process, aiming to provide different perspectives on feature matching. The specific operations are as follows: Each parsing parameter identifier in the pre-stored parsing parameter identifier set is input into the second matching algorithm branch of the associated parameter matching network in sequence together with the (j + 1)-th historical signal quality parameter. The structure of the second matching algorithm branch is similar to that of the first matching algorithm branch of the initial parameter matching network, and a second set of matching degree indicators is calculated through a multi-layer neural network. Each set of indicators corresponds to a parsing parameter identifier, reflecting the matching degree between the identifier and the (j + 1)-th signal quality parameter. The second matching degree indicator with the highest value is selected from the second set of matching degree indicators, and it is weighted with the j-th weight coefficient. The result obtained is used as the training parameter. The way of weighted calculation can be a linear combination. For example, training parameter = α × second matching degree indicator + β × j-th weight coefficient, where α and β are preset weighting coefficients used to balance the influence of the matching state of adjacent signals and the current signal weight on the training parameter.

[0045] Based on the training parameter and the output of the initial parameter matching network, the error amount of the training error function is determined. The training error function usually adopts a loss function (such as mean square error loss function, cross-entropy loss function, etc.), which is used to measure the difference between the predicted positioning anomaly label and the actual label (i.e., the positioning anomaly label) by the network. The calculation method of the error amount depends on the selected loss function. For example, the mean square error loss function quantifies the error by calculating the average of the squares of the differences between the predicted value and the actual value. According to the error amount, the network parameters (such as weight matrix, bias vector, etc.) of the initial parameter matching network are adjusted using the backpropagation algorithm with the goal of minimizing the error to optimize the feature matching ability of the network. The backpropagation algorithm passes the error layer by layer through the chain rule, calculates the gradient of each parameter, and updates the parameters according to the gradient direction, enabling the network to more accurately predict the positioning anomaly label in subsequent training.

[0046] Repeat the above steps of inputting records, judging signal positions, determining training parameters, calculating errors, and updating network parameters until the number of training iterations reaches the set training threshold. The set training threshold is the pre-determined maximum number of iterations used to control the termination condition of the training process. When the number of iterations reaches the threshold, if the network parameters have converged (i.e., the error amount no longer significantly decreases in several consecutive iterations), the initial parameter matching network is determined as the positioning parameter matching network, and the training is completed; if the threshold is not reached, continue with iterative training until the conditions are met.

[0047] During the training process, attention should be paid to the temporal coherence between adjacent historical joint positioning records to ensure that the network can learn the dynamic change law of signals in the time series. For example, the (j + 1)-th signal quality parameter in the previous set of records serves as the j-th signal quality parameter in the next set of records, forming a continuous sequence of signal pairs, enabling the network to capture the evolution trend of signal features. In addition, the introduction of the correlation parameter matching network increases the diversity of the training process. By processing adjacent signals from different perspectives of network parameters, the network can be prevented from falling into local optimal solutions and the generalization ability of the model can be improved.

[0048] During the training process, the weighting coefficients α and β also need to be reasonably set to balance the influence of the adjacent signal matching degree and the current signal weight. If more attention is paid to the temporal correlation features of adjacent signals, the value of α can be appropriately increased; if more attention is paid to the reliability of the current signal matching state, the value of β can be increased. At the same time, hyperparameters such as the learning rate need to be dynamically adjusted according to the training progress of the network to optimize the training efficiency and convergence speed.

[0049] The entire training process focuses on using the temporal correlation features and matching state information in the historical joint positioning records. Through iterative optimization, the initial parameter matching network gradually learns the mapping relationship between Beidou multi-frequency signal features and positioning parameters, and finally forms a positioning parameter matching network that can accurately analyze the original Beidou multi-frequency combined signal stream. This process does not rely on assumed experimental effect data, but through strict signal processing logic and network training mechanisms, it ensures the effectiveness and reliability of the model.

[0050] Embodiment 4: In the training and optimization stage of the positioning parameter matching network, joint training operations need to be performed on multiple sets of historical joint positioning records, and the network performance is ensured through an accuracy verification mechanism. The specific implementation method is as follows: Perform joint training operations on each set of historical joint positioning records, aiming to improve the rationality of network parameter updates through the correlation analysis of adjacent signal features. Each set of records contains two consecutive historical signal quality parameters (the j-th and the j+1-th), the corresponding positioning anomaly labels, and weight coefficients. The core of the joint training operation lies in analyzing the correlation features between adjacent historical signal quality parameters, such as the frequency band distribution deviation amount and the time series feature difference value of adjacent signals. These correlation features are obtained by calculating the difference or ratio of the corresponding feature quantities in the two signal quality parameters, such as calculating the difference in the target frequency band energy ratio and the change rate of the frequency offset in adjacent signals. Based on the above correlation features, adjust the update amplitude of the initial parameter matching network when updating network parameters. Specifically, if the correlation features of adjacent signals vary greatly (such as the frequency band distribution deviation exceeds the preset threshold), it indicates that the signal state has changed significantly. At this time, increase the update amplitude of the network parameters to accelerate the network's response speed to new features; if the difference is small (such as the deviation is within the preset threshold), then reduce the update amplitude to avoid excessive adjustment of network parameters due to minor fluctuations, which may affect the model stability. The adjustment of the parameter update amplitude can be achieved by introducing a dynamic learning rate in the backpropagation algorithm. For example, multiply the difference value of the correlation features by a preset scaling factor to amplify or reduce the learning rate, thereby controlling the step size of each parameter update.

[0051] After the joint training operation of each set of records is completed, it is necessary to verify the performance of the positioning parameter matching network, with a focus on verifying the matching accuracy of the parsing parameter identification for the target frequency band signal data. The verification process uses a validation data set independent of the training data set, which contains historical Beidou multi-frequency signal data with known target parsing parameter identifications. The specific steps are as follows: Input the original Beidou multi-frequency combined signal stream in the validation data set into the positioning parameter matching network in the current training stage to obtain the set of parsing parameter identification matching results output by the network; Compare each matching result in the result set with the corresponding target parsing parameter identification, and calculate the matching accuracy rate (i.e., the proportion of the number of successfully matched samples to the total number of samples). The calculation method of the matching accuracy rate is: Count the number of samples in all validation samples where the parsing parameter identification output by the network is consistent with the target parsing parameter identification, divide it by the total number of validation samples, and obtain the accuracy rate value in percentage form.

[0052] If the matching accuracy reaches the set standard (such as a preset accuracy threshold, for example, 90%), it is considered that the current network parameters meet the actual application requirements, and the training process terminates; if the standard is not reached, the re-training mechanism is triggered. When re-training, the following measures can be taken: adjust the training parameters, such as increasing the number of iterations, modifying the learning rate, or optimizing the weighting coefficient; optimize the network structure, such as increasing the number of hidden layers and adjusting the number of neurons; re-screen or expand the historical joint positioning records and supplement the sample data containing complex signal features. Through the above measures, the training process is re-executed until the matching accuracy of the network meets the requirements.

[0053] In the joint training operation, the analysis of adjacent signal correlation features needs to be combined with the physical meaning of signal processing. For example, a large deviation in frequency band distribution may indicate that the signal is subject to sudden interference. At this time, the network needs to quickly adjust the parameters to adapt to the interference change; while a small deviation may belong to normal signal fluctuations and does not require significant parameter adjustment. This dynamic adjustment mechanism based on signal features enables the network to learn signal patterns more intelligently and improve its adaptability to complex electromagnetic environments.

[0054] During the verification process, the rationality of the set standard is crucial. This standard needs to be determined according to the actual application scenario of power terminal positioning. For example, in scenarios with high requirements for positioning accuracy, the set standard can be correspondingly increased; in scenarios with complex interference environments, the standard can be appropriately relaxed to balance accuracy and robustness. In addition, the verification dataset needs to be representative, covering signal samples with different interference levels and different frequency band distributions to comprehensively evaluate the generalization ability of the network.

[0055] The entire optimization process uses a loop mechanism of "joint training - accuracy verification - adjustment and optimization" to ensure that while the positioning parameter matching network learns the timing correlation features of adjacent signals, it continuously improves the matching accuracy of the parsed parameter identifiers. This process does not rely on hypothetical experimental effect descriptions, but through specific feature analysis, parameter adjustment, and accuracy calculation, it gradually optimizes the network performance and provides a reliable model support for the power terminal positioning system.

[0056] Example 5: In the training and optimization stage of the positioning parameter matching network, joint training operations need to be performed on multiple groups of historical joint positioning records, and a verification mechanism is used to ensure the matching accuracy of the parsed parameter identifiers of the network for the target frequency band signal data. The specific implementation method is as follows: When performing the joint training operation on each set of historical joint positioning records, it is necessary to analyze the correlation features between adjacent historical signal quality parameters. These correlation features include, but are not limited to, the frequency band distribution deviation between adjacent signals, the change of temporal characteristics, etc. The frequency band distribution deviation can be obtained by calculating indicators such as the energy distribution difference of adjacent signals in the target frequency band and the change of frequency offset. For example, for the i-th and the (i + 1)-th historical signal quality parameters, calculate their energy spectral density functions in the target frequency band respectively, and then determine the frequency band distribution deviation by comparing the differences between these two functions. The change of temporal characteristics can be obtained by analyzing the characteristic evolution of adjacent signals in the time dimension, such as the change trend of signal strength, the continuity of phase, etc.

[0057] Based on the above correlation features, dynamically adjust the update amplitude of network parameters. When the difference in correlation features between adjacent signals is large, it indicates that the signal state has changed significantly. At this time, increase the update amplitude of network parameters to accelerate the network's response speed to new features. Specifically, in the backpropagation algorithm, for the case where the difference in correlation features exceeds the preset threshold, multiply the learning rate by an adjustment factor greater than 1, thereby increasing the step size of parameter update. On the contrary, when the difference in correlation features is small, it indicates that the signal state is relatively stable. At this time, reduce the update amplitude to avoid over-adjustment of network parameters. For the case where the difference in correlation features is less than the preset threshold, multiply the learning rate by an adjustment factor less than 1, thereby reducing the step size of parameter update.

[0058] After the joint training operation of each set of records is completed, it is necessary to verify the matching accuracy of the parsing parameter identification of the positioning parameter matching network for the signal data in the target frequency band. The verification process uses an independent verification dataset, which contains multiple historical Beidou multi-frequency signals with known parsing parameter identifications. Input these signals into the trained positioning parameter matching network, obtain the parsing parameter identification output by the network, and compare it with the actual label.

[0059] To more accurately evaluate the network performance, the verification process adopts a multi-level matching accuracy calculation method. First, calculate the basic matching accuracy rate, that is, the proportion of the number of samples where the parsing parameter identification output by the network is exactly the same as the actual label in the total number of samples. Secondly, considering that some parsing parameter identifications may have multiple sub-categories, introduce the sub-category matching accuracy rate, that is, for the main category of each parsing parameter identification, calculate the proportion of correctly matched sub-categories under it. Finally, comprehensively consider the importance differences of parsing parameter identifications, assign different weights to different types of parsing parameter identifications, and calculate the weighted matching accuracy rate.

[0060] If the weighted matching accuracy rate reaches the set standard, it is considered that the training of the positioning parameter matching network is completed and can be used for actual power terminal positioning applications. If the standard is not reached, the training process is executed again. When retraining, the training parameters can be adjusted, such as optimizing the learning rate scheduling strategy, modifying the batch size, or adjusting the regularization parameter, etc. The network structure can also be optimized, such as introducing an attention mechanism to enhance the attention to key features, or adopting residual connections to improve the depth expression ability of the network.

[0061] In the joint training operation, the analysis of the correlation features between adjacent historical signal quality parameters needs to combine the physical characteristics of signal processing. For example, a large deviation in frequency band distribution may correspond to the signal being affected by multipath effects or ionospheric disturbances. At this time, the network needs to quickly adapt to this change; while the change in temporal features is small, it may indicate that the signal is in a stable propagation environment, and the relative stability of the network parameters should be maintained at this time. Through this adjustment strategy based on physical characteristics, the network can better adapt to the complex electromagnetic environment in actual applications.

[0062] The construction of the validation dataset needs to follow the principles of representativeness and diversity. It should include signal samples collected at different geographical locations, different time periods, and different weather conditions to cover various scenarios that may be encountered in actual applications. At the same time, to avoid the one-sidedness of the verification results, the validation dataset should be independent of the training dataset to ensure that the generalization ability of the model on unknown data is truly evaluated.

[0063] In the retraining process, the parameter adjustment and network structure optimization need to follow the principle of gradual iteration. Each adjustment should focus on solving specific performance bottlenecks. For example, for the problem of low matching accuracy of some types of parsing parameter identifiers, the weights of relevant samples can be increased or the extraction method of corresponding features can be optimized. At the same time, to avoid falling into local optimal solutions, methods such as random search or Bayesian optimization can be used to explore a broader parameter space.

[0064] The entire process of the embodiment continuously optimizes the performance of the positioning parameter matching network through the joint training and verification mechanism. This mechanism does not rely on assumed experimental effect data, but through rigorous algorithm design and training processes, it ensures the effectiveness and stability of the network in actual applications. By analyzing the correlation features of adjacent signals to dynamically adjust the learning strategy, and adopting a multi-level verification and evaluation method, the positioning parameter matching network can accurately parse the target frequency band signal data in the original Beidou multi-frequency combined signal stream and provide reliable positioning services for power terminals.

[0065] In practical applications, this mechanism can be customized according to the specific requirements of the power system. For example, for scenarios with high requirements for positioning accuracy, the verification standard can be appropriately increased; for areas with complex interference environments, challenging samples can be added to enhance the robustness of the network. Through this flexible implementation method, it is ensured that the positioning parameter matching network can adapt to different power terminal positioning application scenarios, providing strong support for the safe and stable operation of the power system.

[0066] To more clearly describe the process of dynamically adjusting the learning rate, the formula is introduced:

[0067] Where is the original learning rate, is the adjusted learning rate, is the adjustment coefficient (which can be set according to the actual situation), is the difference value of the correlation features of adjacent signals. This formula indicates that when the difference value of the correlation features is large, the learning rate will increase accordingly, thereby accelerating the update speed of the network parameters; conversely, when the difference value is small, the learning rate will decrease to avoid over-adjustment. This formula is only used to describe the logical relationship of the adjustment strategy and does not involve specific experimental data or effect quantification.

[0068] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0069] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A power terminal positioning method based on Beidou multi-frequency combination, characterized in that The method is applied to a power terminal positioning system, and the method includes: Obtain the original Beidou multi-frequency combined signal stream to be positioned; Input the original Beidou multi-frequency combined signal stream into a pre-trained positioning parameter matching network to obtain a positioning parameter matching result set; Based on the positioning parameter matching result set, determine the positioning parameter matching result corresponding to each Beidou multi-frequency signal in the original Beidou multi-frequency combined signal stream; The original Beidou multi-frequency combined signal stream includes target frequency band signal data to be parsed; The positioning parameter matching network is a parameter matching network obtained by training an initial parameter matching network to be trained with historical positioning data. The historical positioning data is generated from a batch of historical signal quality parameters respectively extracted from a batch of historical Beidou multi-frequency signals, positioning anomaly labels and weight coefficients corresponding to each historical signal quality parameter. The positioning anomaly label reflects the interference status identifier of a historical signal quality parameter, and the weight coefficient reflects the correction factor corresponding to the interference status identifier. The batch of historical Beidou multi-frequency signals is a batch of continuously collected Beidou multi-frequency signals. Each historical signal quality parameter includes a signal quality parameter determined by a historical Beidou multi-frequency signal and its adjacent historical Beidou multi-frequency signals; The positioning parameter matching result is used to represent the parsing parameter identifier of the target frequency band signal data.

2. The power terminal positioning method based on Beidou multi-frequency combination according to claim 1, characterized in that, The method further includes: Perform first feature parameter extraction on the batch of historical Beidou multi-frequency signals to obtain a first set of signal quality parameters, where one signal quality parameter in the first set of signal quality parameters corresponds to one historical Beidou multi-frequency signal in the batch of historical Beidou multi-frequency signals; Perform second feature parameter extraction on the batch of historical Beidou multi-frequency signals to obtain a second set of signal quality parameters, where the first signal quality parameter in the second set of signal quality parameters corresponds to the first historical Beidou multi-frequency signal in the historical Beidou multi-frequency signals. The adjacent historical Beidou multi-frequency signal of the first historical Beidou multi-frequency signal is the second historical Beidou multi-frequency signal, and the first signal quality parameter represents the timing correlation feature between the first historical Beidou multi-frequency signal and the second historical Beidou multi-frequency signal; Fuse the first set of signal quality parameters and the second set of signal quality parameters respectively to obtain the batch of historical signal quality parameters.

3. The power terminal positioning method based on Beidou multi-frequency combination according to claim 2, characterized in that, The performing second feature parameter extraction on the batch of historical Beidou multi-frequency signals to obtain a second set of signal quality parameters includes: Perform second feature parameter extraction on the batch of historical Beidou multi-frequency signals through the following steps to obtain a second set of signal quality parameters. Each time the historical Beidou multi-frequency signal for the second feature parameter extraction is used as the current historical Beidou multi-frequency signal, and the obtained signal quality parameter is used as the current signal quality parameter. The second set of signal quality parameters includes the current signal quality parameter; Perform a frequency band analysis operation on the current historical Beidou multi-frequency signal to determine a first set of frequency band characteristic quantities, where the first set of frequency band characteristic quantities is used to describe the frequency band distribution of the target frequency band signal data in the current historical Beidou multi-frequency signal; Perform a frequency band analysis operation on the adjacent historical Beidou multi-frequency signal of the current historical Beidou multi-frequency signal to determine a second set of frequency band characteristic quantities, where the second set of frequency band characteristic quantities is used to describe the frequency band distribution of the target frequency band signal data in the adjacent historical Beidou multi-frequency signal; Determine the current signal quality parameter based on the first set of frequency band characteristic quantities and the second set of frequency band characteristic quantities, where the current signal quality parameter is used to represent the deviation amount between the corresponding frequency band characteristic quantities in the first set of frequency band characteristic quantities and the second set of frequency band characteristic quantities.

4. The power terminal positioning method based on Beidou multi-frequency combination according to claim 1, characterized in that, The method further includes: Determine a positioning anomaly label and a weight coefficient corresponding to each historical signal quality parameter in the batch of historical signal quality parameters through the following steps, where each historical signal quality parameter for determining the positioning anomaly label and the weight coefficient each time is used as the current historical signal quality parameter, and the positioning anomaly label corresponding to the current historical signal quality parameter is used as the current positioning anomaly label and the current weight coefficient: Input the current historical signal quality parameter into the initial parameter matching network to obtain the current positioning anomaly label, where the initial parameter matching network is a parameter matching network obtained by performing initialization processing in advance; Determine the current weight coefficient based on the current positioning anomaly label and the reference verification identifier corresponding to the current historical signal quality parameter, where the reference verification identifier carries a calibration label in the historical Beidou multi-frequency signal corresponding to the current historical signal quality parameter, and the calibration label includes a target analysis parameter identifier of the target frequency band signal data, and the weight coefficient is used to represent whether the current positioning anomaly label is consistent with the target analysis parameter identifier.

5. The power terminal positioning method based on Beidou multi-frequency combination according to claim 4, wherein The inputting the current historical signal quality parameter into the initial parameter matching network to obtain the current positioning anomaly label includes: Input each analysis parameter identifier in the pre-stored set of analysis parameter identifiers into the first matching algorithm branch in sequence together with the current historical signal quality parameter to obtain a first set of matching degree indicators, where the first set of matching degree indicators includes the matching degree indicators corresponding to each analysis parameter identifier; Determine the candidate analysis parameter identifier corresponding to the first matching degree indicator with the highest value in the first set of matching degree indicators as the current analysis parameter identifier corresponding to the current positioning anomaly label.

6. The power terminal positioning method based on Beidou multi-frequency combination according to claim 5, wherein Before inputting each analysis parameter identifier in the pre-stored set of analysis parameter identifiers into the first matching algorithm branch in sequence together with the current historical signal quality parameter to obtain a first set of matching degree indicators, the method further includes: Determine the signal strength distribution quantity through a preset mode; Determine whether the signal strength distribution quantity conforms to the set strength distribution rule; Based on the signal strength distribution quantity conforming to the set intensity distribution rule, dynamically select a candidate parsing parameter identifier from the set of pre-stored parsing parameter identifiers as the current positioning anomaly label; Based on the signal strength distribution quantity not conforming to the set intensity distribution rule, input the current historical signal quality parameter into the initial parameter matching network to obtain the current positioning anomaly label.

7. The power terminal positioning method based on Beidou multi-frequency combination according to claim 5, characterized in that, The determining the candidate parsing parameter identifier corresponding to the highest numerical value in the first set of matching degree indicators as the current parsing parameter identifier corresponding to the current positioning anomaly label includes: Based on the candidate parsing parameter identifier corresponding to the first matching degree indicator being the same as the target parsing parameter identifier, determine the current weight coefficient as the first weight coefficient, where the first weight coefficient indicates that the initial parameter matching network successfully generates a matching identifier; Based on the candidate parsing parameter identifier corresponding to the first matching degree indicator being different from the target parsing parameter identifier, determine the current weight coefficient as the second weight coefficient, where the second weight coefficient indicates that the initial parameter matching network fails to generate a matching identifier.

8. The power terminal positioning method based on Beidou multi-frequency combination according to claim 1, wherein The method further includes: Screen several groups of historical combined positioning records from the historical positioning data, where the j-th group of historical combined positioning records in the several groups of historical combined positioning records includes the j-th historical signal quality parameter, the j-th positioning anomaly label corresponding to the j-th historical signal quality parameter, the j-th weight coefficient, and the (j + 1)-th historical signal quality parameter, and j is a positive integer; Train the initial parameter matching network to be trained based on the several groups of historical combined positioning records to obtain the positioning parameter matching network. Based on the number of training iterations for the initial parameter matching network reaching the set training threshold, determine the initial parameter matching network as the positioning parameter matching network. Based on the number of training iterations for the initial parameter matching network not reaching the set training threshold, update the network parameters of the initial parameter matching network according to a preset training error function. The input for each round of training process is a group of historical combined positioning records in the several groups of historical combined positioning records.

9. The power terminal positioning method based on Beidou multi-frequency combination according to claim 8, wherein, The method further includes: Train the initial parameter matching network to be trained based on multiple groups of historical combined positioning records through the following steps to obtain the positioning parameter matching network, where the historical combined positioning record input into the initial parameter matching network is the j-th group of combined positioning records: Determine whether the historical Beidou multi-frequency signal corresponding to the (j + 1)-th historical signal quality parameter is the last historical Beidou multi-frequency signal in the historical Beidou multi-frequency signal stream; Based on the (j + 1)-th historical signal quality parameter corresponding to the historical Beidou multi-frequency signal being the last historical Beidou multi-frequency signal, determine the training parameters generated by the initial parameter matching network based on the j-th weight coefficient; Based on the fact that the historical Beidou multi-frequency signal corresponding to the (j + 1)-th historical signal quality parameter is not the last historical Beidou multi-frequency signal, determine the training parameter based on the j-th weight coefficient and the output of the associated parameter matching network, where the associated parameter matching network is a parameter matching network obtained by performing initialization processing in advance, and there are differences in the network parameters between the associated parameter matching network and the initial parameter matching network; Determine the error amount of the training error function based on the training parameter and the output of the initial parameter matching network, and adjust the network parameters of the initial parameter matching network based on the error amount of the training error function; Based on the fact that the number of iterations of performing the above steps reaches the set training threshold, determine the initial parameter matching network as the positioning parameter matching network; The determining the training parameter based on the j-th weight coefficient and the output of the associated parameter matching network includes: Input each parsing parameter identifier in the pre-stored parsing parameter identifier set into the second matching algorithm branch in sequence together with the (j + 1)-th historical signal quality parameter to obtain a second set of matching degree indicators, where the second set of matching degree indicators includes the matching degree indicators corresponding to each parsing parameter identifier; Determine the weighted result of the second matching degree indicator with the highest value in the second set of matching degree indicators and the j-th weight coefficient as the training parameter.

10. The power terminal positioning method based on Beidou multi-frequency combination according to claim 9, characterized in that, The method further includes: Perform joint training operations on each group of records in the multiple groups of historical joint positioning records to generate optimization parameters of the positioning parameter matching network, where the joint training operation for each group of records includes adjusting the update amplitude of the network parameters based on the association features between adjacent historical signal quality parameters; After the joint training operation for each group of records is completed, verify the parsing parameter identifier matching accuracy of the positioning parameter matching network for the target frequency band signal data, and re-execute the training process when the accuracy does not reach the set standard.

Images