Intelligent power consumption management method and device for vehicle-mounted connected terminals based on multi-mode fusion

Abstract

This application provides a method and apparatus for intelligent power consumption management of vehicle-mounted connected terminals based on multi-mode fusion. Through state fusion and scene representation, it achieves precise power consumption monitoring. A management mechanism is constructed, combining predictive analysis and multi-objective optimization to establish a reliable power consumption control strategy. Decision optimization is introduced, ensuring continuous improvement in management through scheme selection and instruction execution. This method effectively addresses the shortcomings of traditional technologies in state monitoring, predictive optimization, and decision execution, providing technical support for power consumption management of vehicle-mounted connected terminals.

Description

Technical Field

[0001] This application relates to the field of data processing, specifically to a method and device for intelligent power consumption management of vehicle-mounted connected terminals based on multi-mode fusion. Background Technology

[0002] Existing power consumption management methods for vehicle-mounted connected terminals have significant shortcomings. Traditional systems perform poorly in power consumption acquisition and state fusion, failing to effectively monitor multi-mode power consumption accurately and thus affecting management performance.

[0003] Furthermore, existing technologies suffer from bottlenecks in predictive analysis and optimization solutions. Most systems lack robust joint conditional coding mechanisms and multi-objective optimization strategies, resulting in suboptimal power consumption control accuracy.

[0004] The existing system has technical shortcomings in decision-making and execution. It lacks in-depth analysis of operating modes, making it difficult to achieve efficient power consumption regulation through configuration commands, thus affecting management effectiveness. Solving these problems is crucial for improving the power consumption management capabilities of in-vehicle terminals. Summary of the Invention

[0005] To address the problems in the existing technology, this application provides a method and device for intelligent power management of vehicle-mounted connected terminals based on multi-mode fusion, which can effectively solve the shortcomings of traditional technologies in terms of status monitoring, predictive optimization and decision execution, and provide technical support for power management of vehicle-mounted connected terminals.

[0006] To solve at least one of the above problems, this application provides the following technical solution:

[0007] Firstly, this application provides a method for intelligent power consumption management of a vehicle-mounted connected terminal based on multi-mode fusion, comprising:

[0008] The instantaneous power consumption value and working mode encoding of each communication module of the vehicle-mounted connected terminal are collected according to a preset sampling period to obtain multi-mode power consumption sampling data. The multi-mode power consumption sampling data is timestamped and assembled into a multi-mode power consumption state vector. The multi-mode power consumption state vector is stored in a buffer according to the time sequence to obtain a multi-mode power consumption state vector sequence. The multi-mode power consumption state vector sequence is fused and encoded with vehicle motion state information and network coverage quality information to obtain a scene representation vector.

[0009] The multi-mode power consumption state vector sequence is input into a shared encoder to obtain a historical time-series encoded vector. The historical time-series encoded vector and the scene representation vector are concatenated and transformed to obtain a joint condition vector. The joint condition vector is input into a dual-branch decoder to generate a service demand prediction sequence and a power consumption trend prediction sequence. The service demand prediction sequence and the power consumption trend prediction sequence are input into a multi-objective optimization solver to perform Pareto optimization to obtain the optimal solution set.

[0010] The optimal solution set is selected based on the current vehicle operating mode and power status to obtain a target decision scheme. The target decision scheme is then decoded to generate a configuration instruction sequence for each communication module. The configuration instruction sequence is then sent to each communication module to perform power consumption control.

[0011] Furthermore, it also includes: collecting the instantaneous current and instantaneous voltage values ​​of the power supply input terminals of the cellular mobile communication module, dedicated short-range communication module, satellite positioning module, wireless local area network module, and Bluetooth module of the vehicle-mounted connected terminal through independent power consumption sampling circuits at a preset sampling period, and reading the current working mode code of each module to obtain the power consumption sampling record of each module; and adding a sampling timestamp generated based on a unified system clock to the power consumption sampling record of each module to obtain multi-mode power consumption sampling data.

[0012] The power consumption sampling records of each module in the multi-mode power consumption sampling data are aligned according to the sampling timestamp to obtain a synchronous power consumption sampling record set. The instantaneous current value and instantaneous voltage value of each module in the synchronous power consumption sampling record set are multiplied to obtain the instantaneous power consumption value of each module. The instantaneous power consumption value of each module and the corresponding working mode code are assembled according to the preset vector dimension order to obtain a multi-mode power consumption state vector.

[0013] Furthermore, it also includes: writing the multi-mode power consumption state vector into the power consumption state circular buffer in the order of sampling timestamps to obtain a multi-mode power consumption state vector sequence; reading the vehicle speed, acceleration, and vehicle working mode marker from the vehicle bus interface to obtain vehicle motion state information; and reading the signal strength indication value, channel busyness measurement value, and connection status marker from the status interface of each communication module to obtain network coverage quality information.

[0014] Normalization is performed on each vector in the multi-mode power consumption state vector sequence to obtain a normalized power consumption sequence. Normalization is performed on the numerical fields in the vehicle motion state information and embedding encoding is performed on the categorical fields to obtain motion state feature vectors. Normalization and embedding encoding are performed on each field in the network coverage quality information to obtain network coverage feature vectors. The normalized power consumption sequence, the motion state feature vectors, and the network coverage feature vectors are concatenated and input into a multilayer perceptron fusion network to perform cross-modal feature interaction to obtain a scene representation vector.

[0015] Furthermore, it also includes: extracting local temporal pattern features from the temporal convolutional layer of the shared encoder to obtain a convolutional feature sequence, and inputting the convolutional feature sequence into the gated recurrent unit layer of the shared encoder to capture long-range temporal dependencies to obtain a historical temporal encoding vector;

[0016] The historical time-series encoding vector and the scene representation vector are concatenated according to a preset dimensional order to obtain a concatenated vector. The concatenated vector is then input into a linear transformation layer to perform dimensional mapping to obtain a joint conditional vector.

[0017] Furthermore, it also includes: the decoder network of the joint condition vector input service demand prediction branch generates the expected data transmission and reception volume and expected service request frequency of each communication module in an autoregressive manner to obtain a service demand prediction sequence; and the decoder network of the joint condition vector input power consumption trend prediction branch generates the expected power consumption mean and expected working mode distribution probability of each communication module in an autoregressive manner to obtain a power consumption trend prediction sequence.

[0018] The service demand prediction sequence and the power consumption trend prediction sequence are configured as input parameters of the fitness evaluation function into a multi-objective optimization solver. The multi-objective optimization solver uses the power consumption target value, the service quality constraint satisfaction value, and the mode switching overhead value as evaluation indicators to perform non-dominated sorting and congestion calculation on the working mode selection variable and the transmit power level selection variable of each communication module to obtain the optimal solution set.

[0019] Furthermore, it also includes: reading the current vehicle operating mode flag from the vehicle bus interface to obtain a vehicle operating mode determination value; reading the current battery power and charging status from the vehicle power system interface to obtain a power status determination value; and performing a matching operation on the vehicle operating mode determination value and the power status determination value according to a preset scenario classification rule to obtain a current scenario category identifier.

[0020] For each candidate solution in the optimal solution set, the corresponding target priority weight configuration is retrieved according to the current scenario category identifier. For each candidate solution, a weighted score is calculated based on the power consumption target value, service quality constraint satisfaction value, and mode switching overhead value according to the target priority weight configuration to obtain the comprehensive score of each candidate solution. The comprehensive scores of each candidate solution are sorted and the candidate solution with the highest score is selected to obtain the target decision scheme.

[0021] Furthermore, it also includes: parsing the working mode selection variables and transmission power level selection variables of each communication module in the target decision scheme according to the control time slot order to obtain the configuration parameters of each time slot module, and converting the configuration parameters of each time slot module into working mode switching instructions and transmission power adjustment instructions according to the preset instruction encoding rules to obtain a configuration instruction sequence;

[0022] The execution timing of each instruction in the configuration instruction sequence is determined according to the specified effective time slot and the current service status. The working mode switching instruction is sent through the driver interface of each communication module to execute the power status control of the module and the configuration of discontinuous reception parameters. The transmit power adjustment instruction is sent through the power control interface of each communication module to execute the power amplifier gain adjustment to complete the power consumption regulation.

[0023] Secondly, this application provides a power consumption intelligent management device for vehicle-mounted connected terminals based on multi-mode fusion, comprising:

[0024] The data acquisition module is used to collect instantaneous power consumption values ​​and working mode encodings of each communication module of the vehicle-mounted connected terminal according to a preset sampling period to obtain multi-mode power consumption sampling data, perform timestamp alignment on the multi-mode power consumption sampling data and assemble it into a multi-mode power consumption state vector, store the multi-mode power consumption state vector in a buffer according to the time sequence to obtain a multi-mode power consumption state vector sequence, and perform fusion encoding on the multi-mode power consumption state vector sequence with vehicle motion state information and network coverage quality information to obtain a scene representation vector;

[0025] The power consumption prediction module is used to input the multi-mode power consumption state vector sequence into a shared encoder to obtain a historical time-series encoding vector, perform a concatenation transformation on the historical time-series encoding vector and the scene representation vector to obtain a joint condition vector, input the joint condition vector into a dual-branch decoder to generate a service demand prediction sequence and a power consumption trend prediction sequence, and input the service demand prediction sequence and the power consumption trend prediction sequence into a multi-objective optimization solver to perform Pareto optimization to obtain the optimal solution set.

[0026] The dynamic adjustment module is used to make a decision selection based on the current vehicle operating mode and power status to obtain a target decision scheme from the optimal solution set, decode the target decision scheme to generate a configuration instruction sequence for each communication module, and send the configuration instruction sequence to each communication module to perform power consumption control.

[0027] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion.

[0028] Fourthly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion.

[0029] Fifthly, this application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the aforementioned intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion.

[0030] As described above, this application provides a method and device for intelligent power consumption management of vehicle-mounted connected terminals based on multi-mode fusion. Through state fusion and scene representation, it achieves precise power consumption monitoring. A management mechanism is constructed, combining predictive analysis and multi-objective optimization to establish a reliable power consumption control strategy. Decision optimization is introduced, ensuring continuous improvement in management through scheme selection and instruction execution. This method effectively addresses the shortcomings of traditional technologies in state monitoring, predictive optimization, and decision execution, providing technical support for power consumption management of vehicle-mounted connected terminals. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart illustrating the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion in the embodiments of this application. Figure 2 This is a structural diagram of the intelligent power consumption management device for vehicle-mounted connected terminals based on multi-mode fusion in the embodiments of this application. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] The acquisition, storage, use, and processing of data in this application comply with relevant laws and regulations.

[0035] To address the problems existing in current technologies, this application provides a method and device for intelligent power consumption management of vehicle-mounted connected terminals based on multi-mode fusion. Through state fusion and scene representation, it achieves precise power consumption monitoring. A management mechanism is constructed, combining predictive analysis and multi-objective optimization to establish a reliable power consumption control strategy. Decision optimization is introduced, ensuring continuous improvement in management through scheme selection and instruction execution. This method effectively solves the shortcomings of traditional technologies in state monitoring, predictive optimization, and decision execution, providing technical support for power consumption management of vehicle-mounted connected terminals.

[0036] To effectively address the shortcomings of traditional technologies in areas such as state monitoring, predictive optimization, and decision execution, and to provide technical support for power consumption management of vehicular connected terminals, this application provides an embodiment of a multi-mode fusion-based intelligent power consumption management method for vehicular connected terminals. See [link to embodiment]. Figure 1 The intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion specifically includes the following:

[0037] Step S101: Collect instantaneous power consumption values ​​and working mode encodings for each communication module of the vehicle-mounted connected terminal according to a preset sampling period to obtain multi-mode power consumption sampling data. Perform timestamp alignment on the multi-mode power consumption sampling data and assemble it into a multi-mode power consumption state vector. Store the multi-mode power consumption state vector in a buffer according to the time sequence to obtain a multi-mode power consumption state vector sequence. Perform fusion encoding on the multi-mode power consumption state vector sequence with vehicle motion state information and network coverage quality information to obtain a scene representation vector.

[0038] In this embodiment, independent power consumption sampling circuits are deployed for the cellular mobile communication module, dedicated short-range communication module, satellite positioning module, wireless LAN module, and Bluetooth module of the vehicle-mounted connected terminal. Each sampling circuit synchronously collects current and voltage data at the power supply input terminal of the corresponding module within a preset sampling period. The sampling circuit uses a high-precision current sensor to obtain instantaneous current values ​​and a voltage divider sampling method to obtain instantaneous voltage values. Simultaneously, it reads the current operating mode code from the status register of each module. The operating mode code distinguishes five states: full-power active mode, reduced-power active mode, low-power monitoring mode, deep sleep mode, and complete shutdown mode. Each module's power consumption sampling record is appended with a sampling timestamp based on a unified system clock during generation. The sampling records of all modules are aggregated to form multi-mode power consumption sampling data.

[0039] After the multi-mode power consumption sampling data is generated, this embodiment performs alignment processing on the power consumption sampling records of each module according to the sampling timestamp. The alignment processing merges the sampling records of each module within the same or adjacent timestamp range into a synchronous power consumption sampling record set, with the alignment accuracy controlled within one-tenth of the sampling period. The instantaneous current value and instantaneous voltage value of each module in the synchronous power consumption sampling record set are multiplied to obtain the instantaneous power consumption value of each module. In this embodiment, the instantaneous power consumption value of each module and the corresponding working mode code are assembled according to a preset vector dimension order. The vector dimension order is as follows: cellular module power consumption value, cellular module working mode code, dedicated short-range communication module power consumption value, dedicated short-range communication module working mode code, and so on up to the Bluetooth module. After assembly, a multi-mode power consumption state vector is obtained.

[0040] The multi-mode power consumption state vectors are written into the power consumption state circular buffer in the order of sampling timestamps. The buffer maintains vector records within the most recent complete monitoring period, forming a multi-mode power consumption state vector sequence. In this embodiment, vehicle speed, acceleration, and vehicle operating mode flags are simultaneously read from the vehicle bus interface. The vehicle operating mode flags distinguish between driving mode, parking mode, charging mode, and sleep / wake-up mode. This data constitutes the vehicle's motion state information. In this embodiment, signal strength indication values, channel busyness measurements, and connection status flags are read from the status interfaces of each communication module. This data constitutes network coverage quality information.

[0041] After the multi-mode power consumption state vector sequence, the vehicle motion state information, and the network coverage quality information are collected, this embodiment performs fusion encoding processing. Normalization is performed on each vector in the multi-mode power consumption state vector sequence, with normalization parameters configured based on historical data statistics, resulting in a normalized power consumption sequence. Normalization is performed on numerical fields such as vehicle speed and acceleration in the vehicle motion state information, while embedding encoding is performed on categorical fields such as vehicle operating mode labels, with the embedding vector dimension configured based on the number of categories, resulting in a motion state feature vector. Normalization and embedding encoding are performed on each field in the network coverage quality information for both numerical and categorical types, resulting in a network coverage feature vector.

[0042] Accordingly, this embodiment performs a concatenation operation on the normalized power consumption sequence, the motion state feature vector, and the network coverage feature vector. The concatenated feature vector is then input into a multilayer perceptron fusion network. The multilayer perceptron fusion network consists of alternating stacks of fully connected layers and layer normalization layers, performing cross-modal feature interaction and outputting a scene representation vector of a unified dimension. This scene representation vector will be concatenated with the historical temporal encoding vector in subsequent step S201, serving as the conditional input for the dual-branch decoder to generate the service demand prediction sequence and the power consumption trend prediction sequence.

[0043] Step S102: Input the multi-mode power consumption state vector sequence into the shared encoder to obtain the historical time-series encoding vector, perform a concatenation transformation on the historical time-series encoding vector and the scene representation vector to obtain the joint condition vector, input the joint condition vector into the dual-branch decoder to generate the service demand prediction sequence and the power consumption trend prediction sequence, and input the service demand prediction sequence and the power consumption trend prediction sequence into the multi-objective optimization solver to perform Pareto optimization to obtain the optimal solution set;

[0044] In this embodiment, the multi-mode power state vector sequence generated in step S101 is subjected to temporal feature encoding, which is completed through a shared encoder. The shared encoder adopts a hybrid architecture of temporal convolutional layers and gated recurrent unit layers. The multi-mode power state vector sequence is first input into the temporal convolutional layer. The temporal convolutional layer performs local feature extraction on the power state vectors at adjacent time steps through a sliding window, capturing short-range pattern features of power changes and outputting a convolutional feature sequence. The convolutional feature sequence is then input into the gated recurrent unit layer. The gated recurrent unit layer selectively memorizes and forgets the features at each time step in the sequence through update gates and reset gates, capturing long-range temporal dependencies of power evolution and outputting a historical temporal encoded vector.

[0045] After the historical time-series encoded vector is generated, this embodiment performs a concatenation transformation with the scene representation vector generated in step S101. The concatenation operation places the historical time-series encoded vector at the beginning and the scene representation vector at the end according to a preset dimensional order, forming a concatenated vector. The concatenated vector is input to a linear transformation layer for dimensional mapping. The linear transformation layer compresses and reorganizes the features of the concatenated high-dimensional vector using a learnable weight matrix, outputting a joint conditional vector. This joint conditional vector integrates historical power consumption evolution patterns and current scene state information, providing a unified conditional input for subsequent prediction tasks.

[0046] The joint condition vector is input to the service demand prediction branch and the power consumption trend prediction branch of the dual-branch decoder, respectively. The decoder network of the service demand prediction branch uses the joint condition vector as its initial input and employs an autoregressive method to generate the expected data transmission and reception volume and expected service request frequency of each communication module within the future prediction time window. The output of each prediction time serves as the input condition for the prediction at the next time moment. After all predictions are completed, a service demand prediction sequence is formed. Similarly, the decoder network of the power consumption trend prediction branch uses the joint condition vector as its initial input and employs an autoregressive method to generate the expected average power consumption and expected operating mode probability distribution of each communication module at each time moment, forming a power consumption trend prediction sequence.

[0047] After the service demand prediction sequence and the power consumption trend prediction sequence are generated, this embodiment configures them as input parameters to the multi-objective optimization solver as fitness evaluation functions. The multi-objective optimization solver defines three types of evaluation metrics: the first is the power consumption target value, calculated based on the operating mode selection variables and transmit power level selection variables of each communication module in each control time slot. The second is the service quality constraint satisfaction value, calculated based on the expected service load of each module in the service demand prediction sequence and the carrying capacity of the current configuration scheme. The third is the mode switching overhead value, calculated based on the number of operating mode changes between adjacent time slots and the power consumption cost of switching.

[0048] Accordingly, the multi-objective optimization solver performs non-dominated sorting and congestion calculation on the operating mode selection variables and transmit power level selection variables of each communication module. Non-dominated sorting divides candidate solutions into multiple frontier levels according to Pareto dominance, with solutions in the first frontier level not dominated by other solutions in any of the three evaluation metrics. Congestion calculation measures the sparsity of the distribution of each solution within the same frontier level in the target space, retaining representative solutions with uniform distribution. After multiple rounds of evolutionary iterations, the multi-objective optimization solver outputs an optimal solution set, which contains multiple Pareto optimal candidate schemes. In subsequent step S103, the optimal solution set will be used to make a decision selection based on the current vehicle operating mode and power status to determine the final target decision scheme.

[0049] Step S103: Based on the current vehicle operating mode and power status, a decision selection is performed on the optimal solution set to obtain a target decision scheme. The target decision scheme is decoded to generate a configuration instruction sequence for each communication module. The configuration instruction sequence is sent to each communication module to perform power consumption control.

[0050] This embodiment performs a decision selection on the optimal solution set generated in step S102, based on the current vehicle operating mode and power status. This embodiment reads the current vehicle operating mode flag from the vehicle bus interface. This vehicle operating mode flag distinguishes between driving mode, parking mode, charging mode, and sleep / wake-up mode. The reading result is used as the vehicle operating mode determination value. This embodiment reads the current battery level and charging status from the vehicle power system interface. The reading result is used as the power status determination value.

[0051] After obtaining the vehicle operating mode determination value and the power status determination value, this embodiment performs matching processing according to preset scenario classification rules. The scenario classification rules map the combination of the vehicle operating mode determination value and the power status determination value to a current scenario category identifier. For example, the combination of driving mode and high battery status maps to a normal driving scenario, the combination of parking mode and charging status maps to a parking and charging scenario, and the combination of driving mode and low battery status maps to a low battery driving scenario. The current scenario category identifier is used to retrieve the corresponding target priority weight configuration.

[0052] After each candidate solution in the optimal solution set retrieves its target priority weight configuration based on the current scenario category identifier, this embodiment performs a weighted score calculation on each candidate solution. The weighted score calculation multiplies the power consumption target value, the service quality constraint satisfaction value, and the mode switching overhead value by their respective priority weight coefficients, and then sums them to obtain the comprehensive score for each candidate solution. In a normal driving scenario, the weight coefficient for the service quality constraint satisfaction value is higher than the weight coefficient for the power consumption target value, ensuring communication service responsiveness. In a parking and charging scenario, the weight coefficient for the power consumption target value is increased, focusing on reducing terminal power consumption. In a low-battery driving scenario, the weight coefficient for the power consumption target value is further increased, while maintaining basic service quality constraint satisfaction. This embodiment sorts the comprehensive scores of each candidate solution in descending order and selects the candidate solution with the highest score as the target decision scheme.

[0053] After the target decision scheme is determined, this embodiment performs decoding processing to generate a configuration instruction sequence. The decoding process parses the operating mode selection variables and transmit power level selection variables of each communication module in the target decision scheme according to the control time slot order, obtaining the configuration parameters of each time slot module. The configuration parameters of each time slot module are converted into specific instructions according to preset instruction encoding rules; the operating mode selection variables are converted into operating mode switching instructions, and the transmit power level selection variables are converted into transmit power adjustment instructions. All instructions are arranged in time slot order to form a configuration instruction sequence.

[0054] The execution timing of each instruction in the configuration instruction sequence is determined based on the specified effective time slot and the current service status. If there is an ongoing high-priority service transmission, the working mode switching instruction is delayed until the service transmission is completed, with the delay duration not exceeding the preset maximum delay tolerance value. For cellular mobile communication modules, the working mode switching instruction is issued through the module driver interface. When switching to low-power listening mode, discontinuous reception parameters are configured; when switching to deep sleep mode, the module is triggered to enter power-saving mode and a periodic wake-up timer is configured. For dedicated short-range communication modules and satellite positioning modules, the working mode switching instruction is implemented by controlling the RF front-end switch and the module power state. The transmit power adjustment instruction is issued through the power control interface of each communication module to perform power amplifier gain adjustment. After all instructions are executed, each communication module operates according to the configuration specified by the target decision scheme to achieve power consumption control. The execution results of the configuration instruction sequence will be continuously collected by the power consumption monitoring module to form a new multi-mode power consumption state vector sequence, which will be read and used by step S101 of the next control cycle.

[0055] As described above, the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion provided in this application can achieve precise power consumption monitoring through state fusion and scene representation. A management mechanism is constructed, combining predictive analysis and multi-objective optimization to establish a reliable power consumption control strategy. Decision optimization is introduced, ensuring continuous improvement in management through scheme selection and instruction execution. This method effectively addresses the shortcomings of traditional technologies in state monitoring, predictive optimization, and decision execution, providing technical support for power consumption management of vehicle-mounted connected terminals.

[0056] In one embodiment of the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion in this application, the method may further include the following:

[0057] Step S201: For the cellular mobile communication module, dedicated short-range communication module, satellite positioning module, wireless local area network module, and Bluetooth module of the vehicle-mounted network terminal, the instantaneous current value and instantaneous voltage value of the power supply input terminal of each module are collected by an independent power consumption sampling circuit according to a preset sampling period, and the current working mode code of each module is read to obtain the power consumption sampling record of each module. The sampling timestamp generated based on the unified system clock is added to the power consumption sampling record of each module to obtain multi-mode power consumption sampling data.

[0058] Step S202: Align the power consumption sampling records of each module in the multi-mode power consumption sampling data according to the sampling timestamp to obtain a synchronous power consumption sampling record set. Perform a product operation on the instantaneous current value and instantaneous voltage value of each module in the synchronous power consumption sampling record set to obtain the instantaneous power consumption value of each module. Assemble the instantaneous power consumption value of each module and the corresponding working mode code according to the preset vector dimension order to obtain a multi-mode power consumption state vector.

[0059] In this embodiment, independent power consumption sampling circuits are deployed for each communication module in the power management unit of the vehicle-mounted connected terminal. The sampling circuit for the cellular mobile communication module is located at the module power supply input terminal, while the sampling circuits for the dedicated short-range communication module are located at the RF front-end power supply terminal and the baseband processor power supply terminal, respectively. The sampling circuits for the satellite positioning module, wireless LAN module, and Bluetooth module are configured at their respective power supply terminals according to the module integration method.

[0060] Each independent power consumption sampling circuit performs synchronous acquisition of current and voltage according to a preset sampling period. The sampling circuit uses a high-precision current sensor to measure the instantaneous current value at the power supply input terminal of each module, and uses a voltage divider resistor network to measure the instantaneous voltage value. The sampling period of the cellular mobile communication module is configured to be aligned with the cellular communication frame period in order to capture the power consumption peak characteristics of the transmission time slot.

[0061] While acquiring instantaneous current and voltage values, this embodiment reads the current operating mode code from the status register of each communication module. The operating mode code distinguishes five states: full-power active mode, reduced-power active mode, low-power monitoring mode, deep sleep mode, and complete shutdown mode. The instantaneous current and voltage values ​​of each module, combined with the operating mode code, form the power consumption sampling record of each module.

[0062] Each module's power consumption sampling record is appended with a sampling timestamp based on a unified system clock during generation. The unified system clock is maintained by the main controller of the vehicle-mounted connected terminal, providing a synchronization time reference for each sampling circuit. The power consumption sampling records of all modules are aggregated to form multi-mode power consumption sampling data, which will undergo timestamp alignment processing in subsequent step S202.

[0063] In this embodiment, the power consumption sampling records of each module in the multi-mode power consumption sampling data are aligned according to the sampling timestamp. The alignment process groups the sampling records of each module with the same timestamp or whose timestamp difference is within the alignment tolerance range into one group, and the alignment tolerance is configured to be one-tenth of the sampling period. After alignment, a synchronous power consumption sampling record set is obtained, and each group of records in the synchronous power consumption sampling record set corresponds to the power consumption status of each module at the same sampling time.

[0064] After the synchronous power consumption sampling record set is generated, this embodiment performs a product operation on the instantaneous current value and instantaneous voltage value of each module. The product operation multiplies the current value and voltage value of the same module at the same sampling time to obtain the instantaneous power consumption value of the module at that time. The unit of the instantaneous power consumption value of each module is uniformly set to milliwatts, which facilitates subsequent vector assembly and cross-module power consumption comparison.

[0065] Accordingly, in this embodiment, the instantaneous power consumption values ​​and corresponding operating mode codes of each module are assembled according to a preset vector dimension order. The vector dimension order is as follows: power consumption value of cellular mobile communication module, operating mode code of cellular mobile communication module, power consumption value of dedicated short-range communication module, operating mode code of dedicated short-range communication module, power consumption value of satellite positioning module, operating mode code of satellite positioning module, power consumption value of wireless local area network module, operating mode code of wireless local area network module, power consumption value of Bluetooth module, and operating mode code of Bluetooth module. After assembly, a multi-mode power consumption state vector is obtained, which will be written into the power consumption state circular buffer in a timing sequence in the subsequent step S301.

[0066] In one embodiment of the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion in this application, the method may further include the following:

[0067] Step S301: Write the multi-mode power consumption state vector into the power consumption state circular buffer according to the sampling timestamp order to obtain the multi-mode power consumption state vector sequence; read the vehicle speed, acceleration, and vehicle working mode mark from the vehicle bus interface to obtain vehicle motion state information; and read the signal strength indication value, channel busyness measurement value, and connection status mark from the status interface of each communication module to obtain network coverage quality information.

[0068] Step S302: Normalize each vector in the multi-mode power consumption state vector sequence to obtain a normalized power consumption sequence; normalize the numerical fields in the vehicle motion state information and perform embedding encoding on the categorical fields to obtain motion state feature vectors; normalize and embedding encode each field in the network coverage quality information to obtain network coverage feature vectors; concatenate the normalized power consumption sequence, the motion state feature vectors, and the network coverage feature vectors and input them into the multilayer perceptron fusion network to perform cross-modal feature interaction to obtain a scene representation vector.

[0069] In this embodiment, the multi-mode power state vectors generated in step S202 are written into the power state circular buffer in the order of sampling timestamps. The circular buffer adopts a first-in, first-out (FIFO) data management method. When the buffer capacity reaches its limit, the newly written vectors overwrite the earliest written vectors. The buffer maintains all vector records within the most recent complete monitoring period, forming a multi-mode power state vector sequence.

[0070] While the multi-mode power consumption state vector sequence is continuously updated, this embodiment reads vehicle motion state information from the vehicle bus interface. The read information includes the vehicle's current speed, acceleration, and a vehicle operating mode marker, which distinguishes between four states: driving mode, parking mode, charging mode, and sleep / wake-up mode. This data, after being parsed by the bus protocol, constitutes the vehicle motion state information.

[0071] In this embodiment, network coverage quality information is read from the status interfaces of each communication module. The cellular mobile communication module provides signal strength and signal quality indicators, the dedicated short-range communication module provides channel busyness measurements and estimates of the number of nearby nodes, and the wireless LAN module provides access point signal strength and connection status markers. The above data are aggregated to form the network coverage quality information.

[0072] After the multi-mode power consumption state vector sequence, the vehicle motion state information, and the network coverage quality information are collected, this embodiment performs normalization and encoding processing. The power consumption value field of each vector in the multi-mode power consumption state vector sequence is normalized. The normalization parameters are configured based on the mean and standard deviation of historical data. After processing, a normalized power consumption sequence is obtained.

[0073] The vehicle motion state information is processed by segmentation. Vehicle speed and acceleration are normalized as numerical fields, and vehicle operating mode labels are embedded and encoded as categorical fields. Embedding encoding maps discrete mode labels to dense vectors of fixed dimensions, with the embedding vector dimension configured according to the number of categories. The normalized numerical features and the embedded and encoded categorical features are concatenated to form the motion state feature vector.

[0074] The network coverage quality information is processed by dividing it into fields. The signal strength indicator and channel busyness measurement values ​​are normalized as numerical fields, while the connection status marker is embedded and encoded as a categorical field. After processing, the features of each field are concatenated to form a network coverage feature vector.

[0075] Accordingly, this embodiment performs a concatenation operation on the normalized power consumption sequence, the motion state feature vector, and the network coverage feature vector. The concatenation order is as follows: the flattened feature vector of the normalized power consumption sequence, the motion state feature vector, and the network coverage feature vector. The concatenated feature vector is then input into the multilayer perceptron fusion network.

[0076] The multilayer perceptron fusion network consists of alternating stacks of fully connected layers and layer normalization layers. The fully connected layers perform linear transformations and nonlinear activations on the input features, while the layer normalization layers adjust the distribution of the outputs of each layer to stabilize the training process. Multilayer processing enables cross-modal feature interaction, outputting a scene representation vector with a unified dimension. This scene representation vector will be concatenated with the historical temporal encoding vector in subsequent step S401.

[0077] In one embodiment of the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion in this application, the method may further include the following:

[0078] Step S401: Input the multi-mode power consumption state vector sequence into the temporal convolutional layer of the shared encoder to extract local temporal pattern features to obtain a convolutional feature sequence, and input the convolutional feature sequence into the gated recurrent unit layer of the shared encoder to capture long-range temporal dependencies to obtain a historical temporal coding vector;

[0079] Step S402: Perform vector concatenation on the historical time-series encoding vector and the scene representation vector according to a preset dimension order to obtain a concatenated vector, and perform dimension mapping on the concatenated vector by inputting it into a linear transformation layer to obtain a joint condition vector.

[0080] In this embodiment, the multi-mode power state vector sequence generated in step S301 is input into the shared encoder for temporal feature encoding. The first processing stage of the shared encoder is a temporal convolutional layer, which performs a sliding window convolution operation on the input sequence.

[0081] The convolutional kernels of temporal convolutional layers slide along the time dimension, performing weighted aggregation of multi-mode power state vectors at several adjacent time points. The weight parameters of the convolutional kernels are learned through training, enabling them to adaptively extract local pattern features of power consumption changes. For example, the power consumption jump pattern generated when a communication module switches from a low-power listening mode to a full-power active mode can be captured by a specific convolutional kernel.

[0082] The temporal convolutional layer outputs the corresponding local feature vector at each time point in the sequence, and the outputs at all positions are arranged temporally to form a convolutional feature sequence. The convolutional feature sequence preserves the temporal order structure of the input sequence, while converting the original power consumption state information into a local pattern feature representation.

[0083] After the convolutional feature sequence is generated, this embodiment inputs it into the gated recurrent unit layer of the shared encoder. The gated recurrent unit layer processes each feature vector in the convolutional feature sequence in chronological order, passing information between time steps through hidden states.

[0084] The gated loop unit layer internally employs two gating mechanisms: an update gate and a reset gate. The update gate controls how much information from the previous time step is retained in the current hidden state, while the reset gate controls the degree to which information from the previous time step is forgotten when calculating candidate hidden states. These two gating mechanisms work together to enable the gated loop unit layer to selectively remember long-range temporal dependencies.

[0085] After the gated recurrent unit layer processes the last time step of the convolutional feature sequence, it outputs the final hidden state as a historical temporal encoding vector. This historical temporal encoding vector compresses the power consumption evolution pattern throughout the entire monitoring period and will be concatenated with the scene representation vector in subsequent step S402.

[0086] In this embodiment, the historical temporal encoding vector and the scene representation vector generated in step S302 are concatenated according to a preset dimensional order. The concatenation order places the historical temporal encoding vector at the beginning and the scene representation vector at the end, resulting in a concatenated vector.

[0087] The dimension of the concatenated vector is the sum of the dimension of the historical temporal encoding vector and the dimension of the scene representation vector. In this embodiment, the concatenated vector is input into a linear transformation layer to perform dimension mapping. The linear transformation layer compresses and reassembles the high-dimensional concatenated vector using a learnable weight matrix.

[0088] The linear transform layer outputs a joint condition vector, the dimension of which is configured according to the input requirements of the subsequent dual-branch decoder. This joint condition vector integrates historical power consumption evolution patterns and current scene state information, and will serve as the conditional input for the dual-branch decoder to generate the prediction sequence in subsequent step S501.

[0089] In one embodiment of the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion in this application, the method may further include the following:

[0090] Step S501: The decoder network of the joint conditional vector input service demand prediction branch generates the expected data transmission and reception volume and expected service request frequency of each communication module in an autoregressive manner to obtain the service demand prediction sequence. The decoder network of the joint conditional vector input power consumption trend prediction branch generates the expected power consumption mean and expected working mode distribution probability of each communication module in an autoregressive manner to obtain the power consumption trend prediction sequence.

[0091] Step S502: Configure the service demand prediction sequence and the power consumption trend prediction sequence as input parameters of the fitness evaluation function to the multi-objective optimization solver. Use the power consumption target value, service quality constraint satisfaction value, and mode switching overhead value as evaluation indicators to perform non-dominated sorting and congestion calculation on the working mode selection variable and transmit power level selection variable of each communication module to obtain the optimal solution set.

[0092] In this embodiment, the joint conditional vector generated in step S402 is input into the service demand prediction branch of the dual-branch decoder. The decoder network of the service demand prediction branch uses the joint conditional vector as the initial hidden state and generates the prediction output time-by-time in an autoregressive manner.

[0093] The autoregressive generation process begins at the first moment of the prediction time window. At the first moment, the decoder network takes the joint condition vector as input and outputs the expected data transmission and reception volume and the expected service request frequency for each communication module. The expected data transmission and reception volume represents the number of uplink and downlink data bytes that each module is expected to carry at that moment, and the expected service request frequency represents the number of service requests that each module is expected to receive at that moment.

[0094] The predicted output at the first time step serves as the input condition for the prediction at the second time step, and the decoder network generates the predicted value for the second time step accordingly. This process is repeated time-by-time until the last time step of the prediction time window, and the predicted outputs at all times are arranged in chronological order to form a business demand prediction sequence.

[0095] In this embodiment, the joint conditional vector is simultaneously input into the power consumption trend prediction branch of the dual-branch decoder. The decoder network of the power consumption trend prediction branch adopts the same autoregressive generation method as the service demand prediction branch, but the output content is different.

[0096] The power consumption trend prediction branch outputs the expected average power consumption and the expected operating mode distribution probability for each communication module at each prediction time. The expected average power consumption represents the expected average power consumption level of each module at that time, and the expected operating mode distribution probability represents the probability distribution of each module being in each operating mode. The prediction outputs at all times are arranged in chronological order to form a power consumption trend prediction sequence.

[0097] After the business demand prediction sequence and the power consumption trend prediction sequence are generated, this embodiment configures them as input parameters to the multi-objective optimization solver as fitness evaluation functions. The fitness evaluation function calculates three types of evaluation indicators based on candidate decision schemes and prediction sequences.

[0098] The first type of evaluation metric is the target power consumption value. The target power consumption value is calculated based on the operating mode selection variables and transmit power level selection variables of each communication module in the candidate scheme for each control time slot, combined with the power consumption model parameter table to determine the expected total power consumption. The power consumption model parameter table records the typical power consumption values ​​of each module under various combinations of operating modes and transmit power levels.

[0099] The second type of evaluation metric is the service quality constraint satisfaction value. This value is calculated by comparing the expected service load of each module in the predicted service demand sequence with the carrying capacity of the candidate scheme configuration, thus determining the degree to which the service response latency constraint and data throughput constraint are satisfied.

[0100] The third type of evaluation metric is the mode switching overhead value. The mode switching overhead value is calculated based on the changes in the operating modes of each module between adjacent control slots in the candidate scheme, combined with the switching power consumption cost and switching delay cost recorded in the mode switching matrix to calculate the cumulative overhead.

[0101] The multi-objective optimization solver performs non-dominated sorting on the operating mode selection variables and transmit power level selection variables of each communication module. The non-dominated sorting divides all candidate solutions into multiple frontier levels according to Pareto dominance. If a candidate solution is not dominated by other candidate solutions in any of the three evaluation indicators, then the candidate solution is assigned to the first frontier level.

[0102] After the non-dominated sorting is completed, the multi-objective optimization solver performs crowding degree calculation on each candidate solution within the same frontier level. The crowding degree calculation measures the distance of each candidate solution to its neighboring solutions in the objective space. Candidate solutions with higher crowding degree values ​​are more sparsely distributed and have better representativeness.

[0103] After multiple rounds of evolutionary iterations, the multi-objective optimization solver outputs an optimal solution set. The optimal solution set contains multiple Pareto optimal candidate solutions located at the first frontier level and with uniform congestion distribution, which will be selected in subsequent step S601 based on the current vehicle operating mode and power status.

[0104] In one embodiment of the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion in this application, the method may further include the following:

[0105] Step S601: Read the current vehicle working mode flag from the vehicle bus interface to obtain the vehicle working mode determination value; read the current battery power and charging status from the vehicle power system interface to obtain the power status determination value; match the vehicle working mode determination value and the power status determination value according to the preset scenario classification rules to obtain the current scenario category identifier.

[0106] Step S602: For each candidate solution in the optimal solution set, retrieve the corresponding target priority weight configuration according to the current scenario category identifier. Perform weighted scoring calculation on each candidate solution according to the target priority weight configuration for power consumption target value, service quality constraint satisfaction value, and mode switching overhead value to obtain a comprehensive score for each candidate solution. Sort the comprehensive scores of each candidate solution and select the candidate solution corresponding to the highest score to obtain the target decision scheme.

[0107] This embodiment reads the current vehicle operating mode flag from the vehicle bus interface. The vehicle operating mode flag is maintained by the vehicle controller and broadcast via the bus. The flag content distinguishes four states: driving mode, parking mode, charging mode, and sleep / wake-up mode. After reading, the vehicle operating mode determination value is obtained.

[0108] This embodiment reads the current battery level and charging status from the vehicle's power system interface. Battery level is expressed as a remaining percentage, and charging status is categorized into three types: not charging, slow charging in progress, and fast charging in progress. A power status determination value is obtained after the reading is complete.

[0109] After obtaining the vehicle operating mode determination value and the power status determination value, this embodiment performs matching processing according to preset scenario classification rules. The scenario classification rules define the combination conditions of vehicle operating mode and power status and their corresponding scenario categories.

[0110] The scenario classification rules match the driving mode with a battery level above a preset threshold as a normal driving scenario. The scenario classification rules match the driving mode with a battery level below a preset threshold as a low-battery driving scenario. The scenario classification rules match the parking mode with charging in progress as a parking charging scenario. The scenario classification rules match the parking mode with not charging as a parking standby scenario. After matching, the current scenario category identifier is obtained.

[0111] After the current scenario category identifier is determined, this embodiment retrieves the corresponding target priority weight configuration for each candidate solution in the optimal solution set generated in step S502. The target priority weight configuration is pre-set for each scenario category and includes three parameters: power consumption target value weight coefficient, service quality constraint satisfaction value weight coefficient, and mode switching overhead value weight coefficient.

[0112] In the weighted configuration for regular driving scenarios, the service quality constraint satisfaction value has a high weight coefficient to ensure the responsiveness of communication services. In the weighted configuration for low-battery driving scenarios, the power consumption target value has an increased weight coefficient to reduce power consumption while maintaining basic service quality. In the weighted configuration for parking and charging scenarios, the power consumption target value has an even higher weight coefficient, focusing on reducing the overall power consumption of the terminal.

[0113] In this embodiment, a weighted score calculation is performed on each candidate solution according to the target priority weight configuration. The weighted score calculation involves multiplying the reciprocal of the power consumption target value of each candidate solution by the corresponding weight coefficient, directly multiplying the service quality constraint satisfaction value by the corresponding weight coefficient, and multiplying the reciprocal of the mode switching overhead value by the corresponding weight coefficient. The sum of these three weighted results yields the comprehensive score of the candidate solution.

[0114] In this embodiment, the comprehensive scores of the candidate solutions are sorted in descending order. After sorting, the candidate solution with the highest comprehensive score is selected as the target decision scheme. The target decision scheme will be decoded in subsequent step S701 to generate the configuration instruction sequence for each communication module.

[0115] In one embodiment of the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion in this application, the method may further include the following:

[0116] Step S701: The working mode selection variable and the transmission power level selection variable of each communication module in the target decision scheme are parsed according to the control time slot order to obtain the configuration parameters of each time slot module. The configuration parameters of each time slot module are converted into working mode switching instructions and transmission power adjustment instructions according to the preset instruction encoding rules to obtain the configuration instruction sequence.

[0117] Step S702: Determine the execution timing for each instruction in the configuration instruction sequence according to the specified effective time slot and the current service status; send the working mode switching instruction through the driver interface of each communication module to execute the power status control of the module and the configuration of discontinuous reception parameters; send the transmit power adjustment instruction through the power control interface of each communication module to execute the power amplifier gain adjustment to complete the power consumption regulation.

[0118] This embodiment performs analytical processing on the target decision scheme generated in step S602. The target decision scheme includes operating mode selection variables and transmission power level selection variables for each communication module in each time slot within the future control time window.

[0119] The parsing process extracts the configuration parameters of each communication module in the order of the control time slots. Each time slot corresponds to a set of module configuration parameters, including the operating mode selection values ​​and transmit power level selection values ​​for the cellular mobile communication module, dedicated short-range communication module, satellite positioning module, wireless LAN module, and Bluetooth module. After the configuration parameters of all time slots are extracted, the module configuration parameters for each time slot are formed.

[0120] After the configuration parameters for each time slot module are generated, this embodiment performs instruction conversion according to preset instruction encoding rules. The instruction encoding rules define the mapping relationship between the operating mode selection value and the operating mode switching instruction, as well as the mapping relationship between the transmit power level selection value and the transmit power adjustment instruction.

[0121] When a working mode selection value is converted into a working mode switching command, the command content includes the target module identifier, the target working mode code, and the effective timeslot marker. When a transmit power level selection value is converted into a transmit power adjustment command, the command content includes the target module identifier, the target power level code, and the effective timeslot marker. After all configuration parameters are converted, a configuration command sequence is formed.

[0122] This embodiment determines the execution timing for each instruction in the configuration instruction sequence. The execution timing is determined based on a combination of the effective time slot specified in the instruction and the current service status. If the current time has reached the effective time slot specified by the instruction and the corresponding module has no high-priority service transmission, then the instruction is executed immediately.

[0123] If the corresponding module is currently transmitting a high-priority service, the operating mode switching command will be executed with a delay. The delay duration will not exceed the preset maximum delay tolerance value, and the command will be executed immediately after the service transmission is completed. The transmit power adjustment command can be executed during service transmission without waiting for the service to complete.

[0124] In this embodiment, the operating mode switching command is issued and executed through the driver interfaces of each communication module. After receiving the operating mode switching command, the cellular mobile communication module performs module power state control according to the target operating mode code. When switching to low-power listening mode, discontinuous reception parameters are configured; when switching to deep sleep mode, the module is triggered to enter power-saving mode and a periodic wake-up timer is configured.

[0125] After receiving the operating mode switching command, the dedicated short-range communication module switches modes by controlling the RF front-end switch and the baseband processor power state. Similarly, after receiving the operating mode switching command, the satellite positioning module switches modes by controlling the module power state and configuring the positioning update cycle.

[0126] In this embodiment, the transmit power adjustment command is issued and executed through the power control interface of each communication module. After receiving the target power level code, the power control interface adjusts the gain configuration of the corresponding module's power amplifier. After the power adjustment takes effect, each module operates according to the new transmit power level, completing power consumption regulation. The execution results of the configuration command sequence will be continuously collected by the power consumption monitoring module to form new multi-mode power consumption sampling data for use in the next regulation cycle.

[0127] To effectively address the shortcomings of traditional technologies in state monitoring, predictive optimization, and decision execution, and to provide technical support for power consumption management of vehicle-mounted connected terminals, this application provides an embodiment of a multi-mode fusion-based intelligent power consumption management device for vehicle-mounted connected terminals, used to implement all or part of the aforementioned multi-mode fusion-based intelligent power consumption management method. See [link to embodiment]. Figure 2 The intelligent power consumption management device for vehicle-mounted connected terminals based on multi-mode fusion specifically includes the following components:

[0128] The data acquisition module 10 is used to collect instantaneous power consumption values ​​and working mode encodings of each communication module of the vehicle-mounted network terminal according to a preset sampling period to obtain multi-mode power consumption sampling data, perform timestamp alignment on the multi-mode power consumption sampling data and assemble it into a multi-mode power consumption state vector, store the multi-mode power consumption state vector in a buffer according to the time sequence to obtain a multi-mode power consumption state vector sequence, and perform fusion encoding on the multi-mode power consumption state vector sequence with vehicle motion state information and network coverage quality information to obtain a scene representation vector;

[0129] The power consumption prediction module 20 is used to input the multi-mode power consumption state vector sequence into a shared encoder to obtain a historical time-series encoding vector, perform a concatenation transformation on the historical time-series encoding vector and the scene representation vector to obtain a joint condition vector, input the joint condition vector into a dual-branch decoder to generate a service demand prediction sequence and a power consumption trend prediction sequence, and input the service demand prediction sequence and the power consumption trend prediction sequence into a multi-objective optimization solver to perform Pareto optimization to obtain the optimal solution set.

[0130] The dynamic adjustment module 30 is used to make a decision selection based on the current vehicle working mode and power status to obtain a target decision scheme from the optimal solution set, decode the target decision scheme to generate a configuration instruction sequence for each communication module, and send the configuration instruction sequence to each communication module to perform power consumption control.

[0131] As described above, the intelligent power consumption management device for vehicle-mounted connected terminals based on multi-mode fusion provided in this application embodiment can achieve precise power consumption monitoring through state fusion and scene representation. A management mechanism is constructed, combining predictive analysis and multi-objective optimization to establish a reliable power consumption control strategy. Decision optimization is introduced, ensuring continuous improvement in management through scheme selection and instruction execution. This method effectively solves the shortcomings of traditional technologies in state monitoring, predictive optimization, and decision execution, providing technical support for power consumption management of vehicle-mounted connected terminals.

[0132] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion.

[0133] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion.

[0134] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion.

[0135] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0136] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0137] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0138] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0139] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent power consumption management of vehicle-mounted connected terminals based on multi-mode fusion, characterized in that, The method includes: The instantaneous power consumption value and working mode encoding of each communication module of the vehicle-mounted connected terminal are collected according to a preset sampling period to obtain multi-mode power consumption sampling data. The multi-mode power consumption sampling data is timestamped and assembled into a multi-mode power consumption state vector. The multi-mode power consumption state vector is stored in a buffer according to the time sequence to obtain a multi-mode power consumption state vector sequence. The multi-mode power consumption state vector sequence is fused and encoded with vehicle motion state information and network coverage quality information to obtain a scene representation vector. The multi-mode power consumption state vector sequence is input into a shared encoder to obtain a historical time-series encoded vector. The historical time-series encoded vector and the scene representation vector are concatenated and transformed to obtain a joint condition vector. The joint condition vector is input into a dual-branch decoder to generate a service demand prediction sequence and a power consumption trend prediction sequence. The service demand prediction sequence and the power consumption trend prediction sequence are input into a multi-objective optimization solver to perform Pareto optimization to obtain the optimal solution set. The optimal solution set is selected based on the current vehicle operating mode and power status to obtain a target decision scheme. The target decision scheme is then decoded to generate a configuration instruction sequence for each communication module. The configuration instruction sequence is then sent to each communication module to perform power consumption control.

2. The intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion according to claim 1, characterized in that, The process of collecting instantaneous power consumption values ​​and encoding the operating mode of each communication module of the vehicle-mounted connected terminal according to a preset sampling period to obtain multi-mode power consumption sampling data, performing timestamp alignment on the multi-mode power consumption sampling data and assembling it into a multi-mode power consumption state vector, includes: The cellular mobile communication module, dedicated short-range communication module, satellite positioning module, wireless local area network module, and Bluetooth module of the vehicle-mounted network terminal are each sampled by an independent power consumption sampling circuit according to a preset sampling period. The instantaneous current value and instantaneous voltage value of the power supply input terminal of each module are collected, and the current working mode code of each module is read to obtain the power consumption sampling record of each module. The sampling timestamp generated based on the unified system clock is added to the power consumption sampling record of each module to obtain multi-mode power consumption sampling data. The power consumption sampling records of each module in the multi-mode power consumption sampling data are aligned according to the sampling timestamp to obtain a synchronous power consumption sampling record set. The instantaneous current value and instantaneous voltage value of each module in the synchronous power consumption sampling record set are multiplied to obtain the instantaneous power consumption value of each module. The instantaneous power consumption value of each module and the corresponding working mode code are assembled according to the preset vector dimension order to obtain a multi-mode power consumption state vector.

3. The intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion according to claim 1, characterized in that, The process involves storing the multi-mode power state vectors into a buffer in a time sequence to obtain a multi-mode power state vector sequence, and then performing fusion encoding on the multi-mode power state vector sequence with vehicle motion state information and network coverage quality information to obtain a scene representation vector, including: The multi-mode power consumption state vector is written into the power consumption state ring buffer in the order of sampling timestamps to obtain the multi-mode power consumption state vector sequence. The vehicle speed, acceleration, and vehicle working mode mark are read from the vehicle bus interface to obtain the vehicle motion state information. The signal strength indication value, channel busyness measurement value, and connection status mark are read from the status interface of each communication module to obtain the network coverage quality information. Normalization is performed on each vector in the multi-mode power consumption state vector sequence to obtain a normalized power consumption sequence. Normalization is performed on the numerical fields in the vehicle motion state information and embedding encoding is performed on the categorical fields to obtain motion state feature vectors. Normalization and embedding encoding are performed on each field in the network coverage quality information to obtain network coverage feature vectors. The normalized power consumption sequence, the motion state feature vectors, and the network coverage feature vectors are concatenated and input into a multilayer perceptron fusion network to perform cross-modal feature interaction to obtain a scene representation vector.

4. The intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion according to claim 1, characterized in that, The process of inputting the multi-mode power state vector sequence into the shared encoder to obtain a historical time-series encoded vector, and performing a concatenation transformation on the historical time-series encoded vector and the scene representation vector to obtain a joint condition vector includes: The multi-mode power consumption state vector sequence is input into the temporal convolutional layer of the shared encoder to extract local temporal pattern features to obtain a convolutional feature sequence. The convolutional feature sequence is then input into the gated recurrent unit layer of the shared encoder to capture long-range temporal dependencies and obtain a historical temporal encoding vector. The historical time-series encoding vector and the scene representation vector are concatenated according to a preset dimensional order to obtain a concatenated vector. The concatenated vector is then input into a linear transformation layer to perform dimensional mapping to obtain a joint conditional vector.

5. The intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion according to claim 1, characterized in that, The process of inputting the joint conditional vector into a dual-branch decoder to generate a service demand prediction sequence and a power consumption trend prediction sequence, and then inputting the service demand prediction sequence and the power consumption trend prediction sequence into a multi-objective optimization solver to perform Pareto optimization to obtain the optimal solution set includes: The decoder network of the joint conditional vector input service demand prediction branch generates the expected data transmission and reception volume and expected service request frequency of each communication module in an autoregressive manner to obtain the service demand prediction sequence. The decoder network of the joint conditional vector input power consumption trend prediction branch generates the expected power consumption mean and expected working mode distribution probability of each communication module in an autoregressive manner to obtain the power consumption trend prediction sequence. The service demand prediction sequence and the power consumption trend prediction sequence are configured as input parameters of the fitness evaluation function into a multi-objective optimization solver. The multi-objective optimization solver uses the power consumption target value, the service quality constraint satisfaction value, and the mode switching overhead value as evaluation indicators to perform non-dominated sorting and congestion calculation on the working mode selection variable and the transmit power level selection variable of each communication module to obtain the optimal solution set.

6. The intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion according to claim 1, characterized in that, The step of performing decision selection on the optimal solution set based on the current vehicle operating mode and power status to obtain the target decision scheme includes: The vehicle bus interface is used to read the current vehicle working mode flag to obtain the vehicle working mode determination value, and the current battery power and charging status are read from the vehicle power system interface to obtain the power status determination value. The vehicle working mode determination value and the power status determination value are matched according to the preset scenario classification rules to obtain the current scenario category identifier. For each candidate solution in the optimal solution set, the corresponding target priority weight configuration is retrieved according to the current scenario category identifier. For each candidate solution, a weighted score is calculated based on the power consumption target value, service quality constraint satisfaction value, and mode switching overhead value according to the target priority weight configuration to obtain the comprehensive score of each candidate solution. The comprehensive scores of each candidate solution are sorted and the candidate solution with the highest score is selected to obtain the target decision scheme.

7. The intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion according to claim 1, characterized in that, The process of decoding the target decision scheme to generate a configuration instruction sequence for each communication module, and then sending the configuration instruction sequence to each communication module to perform power consumption control, includes: The working mode selection variable and the transmit power level selection variable of each communication module in the target decision scheme are parsed according to the control time slot order to obtain the configuration parameters of each time slot module. The configuration parameters of each time slot module are converted into working mode switching instructions and transmit power adjustment instructions according to the preset instruction encoding rules to obtain the configuration instruction sequence. The execution timing of each instruction in the configuration instruction sequence is determined according to the specified effective time slot and the current service status. The working mode switching instruction is sent through the driver interface of each communication module to execute the power status control of the module and the configuration of discontinuous reception parameters. The transmit power adjustment instruction is sent through the power control interface of each communication module to execute the power amplifier gain adjustment to complete the power consumption regulation.

8. A power consumption intelligent management device for vehicle-mounted connected terminals based on multi-mode fusion, characterized in that, The device includes: The data acquisition module is used to collect instantaneous power consumption values ​​and working mode encodings of each communication module of the vehicle-mounted connected terminal according to a preset sampling period to obtain multi-mode power consumption sampling data, perform timestamp alignment on the multi-mode power consumption sampling data and assemble it into a multi-mode power consumption state vector, store the multi-mode power consumption state vector in a buffer according to the time sequence to obtain a multi-mode power consumption state vector sequence, and perform fusion encoding on the multi-mode power consumption state vector sequence with vehicle motion state information and network coverage quality information to obtain a scene representation vector; The power consumption prediction module is used to input the multi-mode power consumption state vector sequence into a shared encoder to obtain a historical time-series encoding vector, perform a concatenation transformation on the historical time-series encoding vector and the scene representation vector to obtain a joint condition vector, input the joint condition vector into a dual-branch decoder to generate a service demand prediction sequence and a power consumption trend prediction sequence, and input the service demand prediction sequence and the power consumption trend prediction sequence into a multi-objective optimization solver to perform Pareto optimization to obtain the optimal solution set. The dynamic adjustment module is used to make a decision selection based on the current vehicle operating mode and power status to obtain a target decision scheme from the optimal solution set, decode the target decision scheme to generate a configuration instruction sequence for each communication module, and send the configuration instruction sequence to each communication module to perform power consumption control.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the intelligent power consumption management method for vehicle-mounted connected terminals based on multi-mode fusion as described in any one of claims 1 to 7.

Images