A method for energy management of a self-powered temperature difference vibration sensor based on capsule networks

Abstract

This invention relates to the field of sensor energy management technology, specifically to a self-powered temperature difference and vibration sensor energy management method based on capsule networks. The invention collects environmental parameters such as vibration, temperature difference, humidity, airflow, and resonant frequency through a sensor network and performs filtering and normalization preprocessing. A selective attention capsule module is constructed to dynamically allocate weights to each parameter, achieving feature fusion. A dynamic time memory factor is introduced to model the correlation between historical and current features, enhancing temporal awareness. The synergistic relationships between parameters are mined through an environmental feature interaction matrix to optimize feature representation. Combined with a power generation mode scoring mechanism, multiple energy management strategies are evaluated, and the optimal mode is selected based on softmax, improving energy harvesting and utilization efficiency.

Description

Technical Field

[0001] This invention relates to the field of sensor energy management technology, and specifically to an energy management method for a self-powered temperature difference vibration sensor based on a capsule network. Background Technology

[0002] With the rapid development of IoT and wireless sensor network technologies, an increasing number of sensor nodes are being deployed in various environments for real-time monitoring and data acquisition. However, since most traditional sensors rely on external power sources or batteries, their energy replenishment and maintenance costs are high, and they are prone to insufficient power supply and limited lifespan in harsh environments. Therefore, self-powered sensors have gradually become a research hotspot, with their core being the achievement of long-term, stable self-powering through environmental energy harvesting.

[0003] Currently, temperature difference energy and vibration energy are two common environmental energy harvesting methods that have attracted widespread attention. Temperature difference sensors utilize the temperature gradient in the environment for energy conversion, while vibration sensors drive the energy harvesting unit through vibration signals; both can achieve self-powered operation. However, in practical applications, the time-varying and diverse nature of environmental parameters presents the following technical challenges to energy harvesting:

[0004] Various parameters in the environment exhibit significant temporal fluctuations. Traditional single-signal processing methods struggle to take into account multi-dimensional environmental information, resulting in low energy harvesting efficiency and delayed response of energy management strategies. Existing technologies often lack the collaborative analysis and fusion processing of multi-dimensional environmental parameters such as temperature difference and vibration, making it difficult to establish a complete environmental characteristic model, thus affecting the determination of the optimal energy harvesting and management mode.

[0005] In practical applications, different power generation modes differ in terms of energy conversion efficiency and environmental adaptability. Traditional static or single optimization strategies cannot adjust energy management schemes in real time and cannot meet energy demands under conditions of instantaneous environmental changes. Summary of the Invention

[0006] To address the problems of existing technologies, this invention proposes an energy management method for self-powered temperature difference vibration sensors based on capsule networks. This scheme utilizes an environmental sensor network to collect various parameters such as vibration signals, temperature gradients, humidity, airflow disturbances, and resonant frequencies. It constructs a temporal correlation between historical and current data through signal filtering, normalization, and time-series segmentation techniques. Simultaneously, it introduces selective attention capsule modules and dynamic time memory factors to dynamically adjust the importance of various environmental parameters, achieving deep fusion of environmental characteristics and joint temporal modeling. Using interaction matrices and capsule routing mechanisms, it performs adaptive evaluation and multi-objective optimization of various power generation modes, ultimately determining the optimal energy harvesting and management strategy. This method significantly improves the energy management system's responsiveness to environmental changes and the accuracy of decision-making, thereby solving the problems of unstable energy harvesting, insufficient information fusion, and inflexible management mode selection inherent in traditional self-powered sensors.

[0007] This invention is achieved through the following technical solution:

[0008] A method for energy management of a self-powered temperature difference vibration sensor based on capsule networks includes the following steps:

[0009] S1. Collect environmental parameters through an environmental sensor network, perform signal filtering and normalization preprocessing on the collected environmental parameters, and establish a time-series correlation between historical data and current data using time series segmentation technology.

[0010] S2. Construct a selective attention capsule module, dynamically adjust the feature importance of each environmental parameter using learnable parameters, and normalize the weights of each environmental parameter using the softmax function to obtain a weighted fused environmental feature capsule vector.

[0011] S3. In the prediction capsule module, a dynamic time memory factor based on environment is introduced. According to the changing state of environmental features, a dynamic time memory factor corresponding to each environmental parameter is established, and an environmental correlation state update mechanism is adopted to realize the temporal joint modeling of historical state and current input features.

[0012] S4. By establishing an environmental parameter interaction matrix through feature fusion capsules, the interaction relationship between various environmental features is calculated, and the fused feature representation is obtained, further enhancing the collaborative expression capability of environmental information.

[0013] S5. Based on the power generation mode scoring mechanism, an adaptive assessment of different energy management modes is conducted. The probability values ​​of different modes are calculated using the softmax function to determine the optimal energy harvesting and management strategy.

[0014] Furthermore, the environmental parameters are categorized into five types, including vibration signal, temperature gradient, humidity, airflow disturbance, and resonant frequency.

[0015] Furthermore, step S1 specifically includes the following sub-steps:

[0016] S101. Collect raw data of five types of environmental parameters: vibration, temperature gradient, humidity, airflow disturbance and resonant frequency, to provide multi-dimensional environmental information input for subsequent analysis;

[0017] S102. Construct historical feature sequences for each type of parameter, extract time-series segments by defining a fixed window length, capture short-term trends in environmental parameters, and provide a data foundation for fluctuation feature calculation;

[0018] S103. Calculate the fluctuation index of each type of parameter, quantify the data fluctuation amplitude within the historical window, and use it for subsequent dynamic filtering parameter adjustment.

[0019] S104. Dynamically adjust the filtering factor according to the fluctuation index, perform adaptive filtering on the current data, suppress high-frequency noise and retain effective signals, and balance the weight of real-time data and historical filtering results.

[0020] S105. Normalize the filtered parameters to eliminate the differences in the dimensions of data from different sensors, ensure that multi-source data are fused at the same scale, and avoid a single parameter dominating model decision-making.

[0021] S106. The normalized multi-class parameters are concatenated into a multi-dimensional feature vector, integrating multi-dimensional environmental information to provide a unified input for subsequent feature fusion and attention weight allocation.

[0022] S107. Using time series segmentation technology, multi-dimensional features within the historical window are correlated in chronological order to form time-series joint feature fragments, capturing the dynamic evolution of environmental states.

[0023] Furthermore, step S2 specifically includes the following sub-steps:

[0024] S201. The importance weights of each environmental parameter are dynamically adjusted through learnable parameters, so that the model can automatically focus on key environmental features based on the current input.

[0025] S202. Use the Softmax function to normalize the learnable parameters and generate probabilistic feature weights to avoid excessive weight for a single parameter.

[0026] S203. Weighted fusion of multiple normalized features highlights the role of important parameters, suppresses the interference of minor parameters, and generates a more representative fusion feature vector.

[0027] S204. By using nonlinear transformation, the fused features are mapped to capsule vectors, which enhances the abstract expressive power of the features, preserves the spatial relationship between parameters, and provides high-order feature input for time series modeling.

[0028] Furthermore, step S3 specifically includes the following sub-steps:

[0029] S301. Generate a dynamic time memory factor based on the current environmental parameters and environmental feature capsule vectors, and use the Sigmoid function to control the memory update intensity so that the model can adapt to the difference between sudden environmental changes and stable states.

[0030] S302. Define the forget gate and the input gate, and jointly control the retention ratio of historical memory and the fusion ratio of new features by using environmental feature capsule vectors and fluctuation indicators to achieve time-dependent dynamic adjustment.

[0031] S303. Update the memory state of each type of parameter through a gating mechanism, and capture long-term environmental evolution patterns by combining historical memory with current input.

[0032] S304. The memory states of the five types of parameters are concatenated and mapped to contextual features to extract global temporal dependencies across parameters, thereby enhancing the model's understanding of interactions in complex environments.

[0033] S305: Integrate contextual features with the current capsule vector, balance the contributions of historical and current information through trainable coefficients, improve training stability through layer standardization, and output temporal joint features.

[0034] Furthermore, step S4 specifically includes the following sub-steps:

[0035] S401. Calculate the interaction matrix between environmental parameters, and measure the influence strength between parameters based on the cosine similarity of environmental feature capsule vectors and temporal joint features to reveal implicit association patterns.

[0036] S402. Dynamically weight the interaction matrix according to the fluctuation index to enhance the interaction weight of high fluctuation parameters, suppress noise interference, and make the interaction relationship conform to the actual environmental changes.

[0037] S403. By fusing temporal joint features and environmental feature capsule vectors through a weighted interaction matrix, information from different dimensions is spliced ​​together to generate fused features that include the synergistic effect of parameters.

[0038] S404. Perform nonlinear transformation and standardization on the fused features to generate enhanced collaborative features, retaining high-order interaction information and providing discriminative input for pattern evaluation.

[0039] Furthermore, S5 specifically includes the following sub-steps:

[0040] S501. Construct a power generation mode score tensor based on collaborative features and time-series joint features, calculate mode adaptability through parameter interaction weights and learnable score weights, and quantify the degree of multi-dimensional matching.

[0041] S502. The capsule routing mechanism is used to iteratively optimize the mode score. Local features are aggregated through dynamic routing weights, and the mode confidence is represented by the vector magnitude.

[0042] S503. Perform multi-objective correction on the model score, introduce a time-series smoothing term to suppress abrupt changes, and a volatility penalty term to reduce the risk of model switching when the environment is unstable, thus balancing short-term and long-term benefits.

[0043] S504. The corrected score is converted into a probability distribution through the Softmax function, which intuitively reflects the applicability probability of each power generation mode and supports probability-based decision optimization.

[0044] S505. Select the highest probability mode that satisfies the constraints of capsule strength, synergistic concentration and historical stability to ensure the optimality of the strategy in terms of reliability, environmental adaptability and temporal continuity.

[0045] The beneficial effects of this invention are:

[0046] This invention proposes a self-powered temperature difference vibration sensor energy management method based on capsule networks, effectively solving the problems of weak multi-dimensional environment modeling capabilities and lagging energy strategy response in existing technologies. By constructing a dynamic time memory factor and introducing an environment-related state update mechanism, the system can dynamically capture the temporal evolution characteristics of various environmental parameters, improving the modeling accuracy for non-stationary conditions and compensating for the deficiency of insufficient utilization of historical states in traditional methods. Simultaneously, by utilizing a selective attention capsule module to achieve feature weighted fusion of multi-source environmental parameters, the system adaptively adjusts the influence of different environmental factors, overcoming the significant bottleneck of fixed feature weights and inability to dynamically adjust them in previous methods. Furthermore, by constructing collaborative features and temporal joint features, this invention achieves deep interactive fusion of multimodal information, effectively improving the system's overall perception capability of coupled environmental changes under complex operating conditions. In addition, by constructing a power generation mode score tensor and an optimal strategy selection mechanism, the system can select the optimal energy management scheme based on the current environmental state and historical evolution trends, solving the problems of slow strategy response and poor adaptability. The overall scheme has clear modules, controllable mechanisms, and high feasibility, demonstrating significant practical application value in improving the intelligent level of energy efficiency management in sensor systems. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1This is a flowchart of an energy management method for a self-powered temperature difference vibration sensor based on capsule networks proposed in this invention;

[0049] Figure 2 This is a schematic diagram of a terminal device for a self-powered temperature difference vibration sensor energy management method based on capsule networks proposed in this invention.

[0050] Figure 3 This is a schematic diagram of a readable storage medium for a self-powered temperature difference vibration sensor energy management method based on capsule networks proposed in this invention.

[0051] In the diagram, 200 is the terminal device, 210 is the memory, 211 is the RAM, 212 is the cache memory, 213 is the ROM, 214 is the program / utility, 215 is the program module, 220 is the processor, 230 is the bus, 240 is the external device, 250 is the I / O interface, 260 is the network adapter, and 300 is the program product. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0053] Example 1

[0054] A method for energy management of a self-powered temperature difference vibration sensor based on capsule networks includes the following steps:

[0055] S1. Five types of environmental parameters, namely vibration signal, temperature gradient, humidity, airflow disturbance and resonant frequency, are collected through an environmental sensor network. The collected environmental parameters are preprocessed by signal filtering and normalization. Time series segmentation technology is used to establish the time series correlation between historical data and current data.

[0056] S2. Construct a selective attention capsule module, dynamically adjust the feature importance of each environmental parameter using learnable parameters, and normalize the weights of each environmental parameter using the softmax function to obtain a weighted fused environmental feature capsule vector.

[0057] S3. In the prediction capsule module, a dynamic time memory factor based on environment is introduced. According to the changing state of environmental features, a dynamic time memory factor corresponding to each environmental parameter is established, and an environmental correlation state update mechanism is adopted to realize the temporal joint modeling of historical state and current input features.

[0058] S4. By establishing an environmental parameter interaction matrix through feature fusion capsules, the interaction relationship between various environmental features is calculated, and the fused feature representation is obtained, further enhancing the collaborative expression capability of environmental information.

[0059] S5. Based on the power generation mode scoring mechanism, an adaptive assessment of different energy management modes is conducted. The probability values ​​of different modes are calculated using the softmax function to determine the optimal energy harvesting and management strategy.

[0060] Furthermore, step S1 specifically includes the following sub-steps:

[0061] S101. Collect raw data of five types of environmental parameters: vibration signal, temperature gradient, humidity, airflow disturbance, and resonant frequency, and record them as follows: , , , , ;

[0062] S102. Construct historical feature sequences for each type of parameter, and define the historical window length as follows: ,get: ,in ;

[0063] S103. For each type of parameter's historical sequence, calculate its volatility index to describe the magnitude of data change. The specific calculation formula is as follows: ;in Standard deviation measures the fluctuation of data within a historical window. The mean is used as a standardized volatility indicator. This represents the historical sequence of the i-th type of parameter, with a window length of k;

[0064] S104. Based on the fluctuation index, the filtering modulation factor is dynamically adjusted to perform dynamic filtering on the current values ​​of each parameter. The filtered data is as follows: ;

[0065] in Represents the filter modulation factor, by calculate, Indicates the current parameter value. This represents the filtered value at the previous time step;

[0066] S105. Based on the filtered data, normalize each parameter to ensure data scale consistency. The calculation formula is as follows: ,in and The first The historical minimum and maximum values ​​of the class parameter;

[0067] S106. Concatenate all normalized parameters to construct the multidimensional feature vector for the current time step:

[0068] ;

[0069] S107. Employ time series segmentation techniques to correlate historical data with current features in a temporal sequence, forming a joint temporal feature segment: .

[0070] Furthermore, step S2 specifically includes the following sub-steps:

[0071] S201, Based on learnable parameters The characteristic importance of each environmental parameter is dynamically adjusted using the following formula: ,in A function for obtaining the learnable feature importance parameters. , Represents a multidimensional feature vector;

[0072] S202. The learnable parameters of each environmental parameter obtained in S201 are normalized using the softmax function to obtain the feature weights of each environmental parameter. The formula is: ;

[0073] S203, Utilizing the characteristic weights of each environmental parameter For the corresponding normalized environmental characteristics The environmental characteristics of the rings are calculated by weighted fusion. The formula is: ,in This represents the feature weights after Softmax normalization;

[0074] S204, The weighted and fused environmental feature results Input nonlinear transformation function The environmental feature capsule vector is obtained. The formula is: ,in Used to enhance feature representation capabilities, it is a non-linear activation or transformation function.

[0075] Furthermore, step S3 specifically includes the following sub-steps:

[0076] S301, Based on normalized environmental parameters and environmental feature capsule vector Generate dynamic time memory factors for each type of environmental parameter: ,in For the sigmoid function, and Each is for the first The learnable weight matrix and bias terms of the class parameters have dimensions that are automatically matched based on the input features;

[0077] S302, Using environmental feature capsule vectors and volatility indicators Define the forget gate Input gate The formula is: , ,in These are learnable parameters whose dimensions match the dimensions of the input features;

[0078] S303. Update the historical memory state of each type of parameter through the forget gate and the input gate. The formula is:

[0079] ,in This represents element-wise multiplication. This represents the memory state from the previous time step, with the initial state set as a zero vector;

[0080] S304. Memorize the state of the five types of environmental parameters. , , , , The concatenated features generate contextual features, as shown in the formula:

[0081]

[0082] in This represents a vector concatenation operation. and Indicates learnable parameters;

[0083] S305, Fusion of Contextual Features With the current capsule vector Output the joint temporal features, the formula is:

[0084]

[0085] in This represents the trainable fusion coefficient, with an initial value set to 0.5. Standardized operations at the presentation layer.

[0086] Furthermore, step S4 specifically includes the following sub-steps:

[0087] S401, Capsule Vector Based on Environmental Characteristics Joint features of time series Calculate the interaction matrix of five types of environmental parameters The formula is:

[0088]

[0089] in Represents the environmental feature capsule subvector of the m-th class of parameters. , The temporal joint eigenvector representing the nth type of parameters , Represents the cosine similarity function;

[0090] S402, Based on the volatility index The interaction matrix is ​​dynamically weighted using the following formula:

[0091] in This represents the volatility index of the m-th type of parameter. This represents the fluctuation index of the p-th type of parameter;

[0092] S403. The joint temporal features and feature capsule vectors are fused using a collaborative weight matrix, as shown in the following formula:

[0093]

[0094] in This represents a vector concatenation operation;

[0095] S404. Input the fused features into a nonlinear transformation function to generate enhanced collaborative features, as shown in the formula:

[0096]

[0097] in , This represents the learnable parameters.

[0098] Furthermore, S5 specifically includes the following sub-steps:

[0099] S501, Based on collaborative characteristics Joint features of time series The power generation mode score tensor is constructed as follows:

[0100]

[0101] in This represents the collaborative weight, i.e., the interaction strength between environmental parameters m and n. , This represents the learnable scoring weights of the k-th power generation mode for the m-th class of collaborative features and the n-th class of temporal joint features;

[0102] S502, Iteratively optimize the power generation mode scoring capsule through the capsule routing mechanism. To enhance the geometric interpretability of pattern decisions, initialize route weights using the following formula:

[0103]

[0104]

[0105]

[0106] in The initial routing coupling weights for the k-th power generation mode are represented and are uniformly initialized before the capsule routing iteration. The tensor representing the power generation mode score All elements of the k-th pattern, This represents the intermediate score capsule for the k-th power generation mode in the r-th routing iteration, currently in an inactive state. Representing collaborative features Dimensions Representing temporal joint features Dimensions Let represent the interaction score of the k-th mode, the i-th dimension collaborative feature, and the j-th dimension temporal joint feature in the score tensor. Let represent the capsule vector activated by nonlinear compression after the r-th routing iteration;

[0107] Update route weights using the following formula:

[0108]

[0109] in ; This represents the routing weight of the k-th mode after the r-th iteration. The index variable represents the summation index, indicating the first index. Various power generation modes;

[0110] The final power generation mode score capsule is Electric model, its model length Representational pattern confidence;

[0111] S503. Perform multi-objective optimization correction on the power generation mode score to generate the final score. The formula is as follows:

[0112]

[0113] in Indicates the base score. Represents the time-series smoothing term. This indicates a fluctuation penalty term. This represents the collaborative weight distribution of the k-th mode. This indicates the degree of concentration of the weight distribution. The time-domain gradient of the volatility index quantifies the intensity of environmental abrupt changes. and Indicates the trainable balance coefficient;

[0114] S504. Calculate the probability distribution of power generation modes based on Softmax parameters, using the following formula:

[0115] ;

[0116] S505. Select the mode with the highest probability that satisfies the constraints:

[0117] .

[0118] Example 2

[0119] This embodiment is a specific implementation of a self-powered temperature difference vibration sensor energy management method based on capsule networks, proposed in Embodiment 1.

[0120] In wind turbine generators, real-time health monitoring of gearboxes and bearings is crucial for preventing mechanical failures. Since wind turbines are often deployed in remote areas with towers exceeding 100 meters in height, traditional wired power sensors suffer from difficulties in wiring and high maintenance costs. This embodiment employs a self-powered sensor network powered by a temperature difference-vibration hybrid system, deployed at key points on the gearbox surface and bearing housing. The challenges in this application scenario include: sudden wind speed changes causing drastic fluctuations in vibration energy (5-200Hz frequency band); changes in gearbox load causing nonlinear changes in the temperature gradient (ΔT=5-50℃); resonant frequency shifts reflecting the degree of mechanical damage; and humidity changes (30-95%RH) affecting the Seebeck coefficient of thermoelectric materials, requiring joint analysis; high instantaneous power of vibration energy but poor stability (peak-to-average power ratio >10); and strong persistence of temperature difference energy but slow response (thermal relaxation time >30 seconds).

[0121] This embodiment performs multi-parameter synchronous acquisition and fluctuation quantization through steps S101-S103.

[0122] The sensor synchronously acquires the effective value of vibration at a frequency of 10Hz. Temperature gradient ,humidity airflow pressure and resonant frequency shift Build a 60-second historical window of data Calculate the fluctuation index of each parameter:

[0123]

[0124] When premature wear occurs in the gearbox, the vibration signal... The increase from 0.18 to 0.45 indicates an increase in the random shock component.

[0125] Steps S104-S105 involve adaptive filtering and normalization.

[0126] The filter factor for the vibration signal is designed to be: ,when hour, Adaptive processing is employed to retain high-frequency impact components and conversely enhance smoothness.

[0127] The normalized features are:

[0128]

[0129] in , This data was obtained from historical statistics.

[0130] Steps S106-S107 perform time-series correlation modeling

[0131] Constructing the current time vector by splicing five-dimensional normalized features And organize a 60-second time sequence. This forms a 60×5 input matrix.

[0132] Steps S201-202 involve dynamic weight allocation.

[0133] Through learnable parameter matrix Generate importance weights for each parameter, and normalize them using Softmax:

[0134]

[0135] In actual operation, when the temperature gradient During a sudden increase, The value increased from 0.25 to 0.41, highlighting the potential for thermoelectric conversion.

[0136] Steps S203-204 generate environmental feature capsule vectors.

[0137] The weighted fusion features are:

[0138]

[0139] A 32-dimensional capsule vector is output through nonlinear transformation. .

[0140] Step S301: Dynamic generation of memory factors.

[0141] Construct memory factors for vibration parameters:

[0142] When continuous monitoring shows an increase in vibration energy The value increased from 0.32 to 0.78, reinforcing the memory of the oscillation trend.

[0143] Steps S302-S304 update the gate status.

[0144] Define the forget gate and the input gate:

[0145]

[0146] When the resonance frequency is abnormal hour, If the value drops below 0.2, old memories will be forcibly forgotten. Memory status will be updated as follows: .

[0147] Step S305: Perform temporal joint feature output.

[0148] Integrating historical memory with current characteristics:

[0149] .

[0150] Steps S401-S402 involve constructing the interaction matrix.

[0151] Calculate the combined weights of vibration and airflow:

[0152]

[0153] When the wind speed suddenly increases, it causes hour, The value increased from 0.15 to 0.28, indicating an enhanced contribution of aerodynamic excitation to vibration energy.

[0154] Steps S403-S404 perform collaborative feature enhancement

[0155] Integrating interaction weights and features:

[0156] 64-dimensional cooperative features are generated through nonlinear transformation. To capture multi-parameter coupling effects.

[0157] Steps S501-S502 perform capsule routing optimization.

[0158] Initialization power generation mode score capsule (vibration priority / temperature difference priority / hybrid mode):

[0159]

[0160] The coupling weight is updated through three routing iterations. Final Mode Capsule Length of the module Confidence of the representation pattern.

[0161] Steps S503-505: Perform multi-objective decision-making.

[0162] Introducing a time series smoothing term and a fluctuation penalty term:

[0163]

[0164] When the environment changes At that time, the penalty term reduces the score of the high-entropy mode. The final choice satisfies:

[0165] Capsule strength Cooperative entropy Historical stability Power generation mode .

[0166] Example 3

[0167] refer to Figure 2 Based on Example 1, this example proposes a terminal device for a self-powered temperature difference vibration sensor energy management method based on capsule network. The terminal device 200 includes at least one memory 210, at least one processor 220, and a bus 230 connecting different platform systems.

[0168] The memory 210 may include a readable medium in the form of volatile memory, such as RAM 211 and / or cache memory 212, and may further include ROM 213.

[0169] The memory 210 also stores a computer program that can be executed by the processor 220, causing the processor 220 to execute any of the above-described energy management methods for a self-powered temperature difference vibration sensor based on a capsule network in this application embodiment. The specific implementation and technical effects achieved are consistent with those described in the embodiments of the above applications, and some details will not be repeated here. The memory 210 may also include a program / utility 214 having a set (at least one) of program modules 215. Such program modules include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment.

[0170] Accordingly, processor 220 can execute the aforementioned computer program, as well as executable program / utility 214.

[0171] Bus 230 can represent one or more of several types of bus structures, including a memory bus or memory controller, peripheral bus, graphics acceleration port, processor, or a local bus using any of the various bus structures.

[0172] Terminal device 200 can also communicate with one or more external devices 240, such as keyboards, pointing devices, Bluetooth devices, etc., and with one or more devices capable of interacting with it, and / or with any device that enables it to communicate with one or more other computing devices (e.g., routers, modems, etc.). This communication can be performed via I / O interface 250. Furthermore, terminal device 200 can communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via network adapter 260. Network adapter 260 can communicate with other modules of terminal device 200 via bus 230. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with terminal device 200, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0173] Example 4

[0174] refer to Figure 3 This embodiment proposes a readable storage medium for a self-powered temperature difference vibration sensor energy management method based on capsule networks. The computer-readable storage medium stores instructions that, when executed by a processor, implement any of the above-mentioned self-powered temperature difference vibration sensor energy management methods based on capsule networks. The specific implementation method and the technical effects achieved are consistent with those described in the above-mentioned application embodiments, and some details will not be repeated.

[0175] Figure 3The present embodiment illustrates a program product 300 for implementing the above-described applications. This product may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product 300 of the present invention is not limited thereto. In this embodiment, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device. The program product 300 may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0176] Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof. Program code for performing operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on a user computing device, partially on a user device, as a standalone software package, partially on a user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing devices can be connected to user computing devices via any type of network, including local area networks (LANs) or wide area networks (WANs), or they can be connected to external computing devices (e.g., via the Internet using an Internet service provider).

[0177] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A self-powered temperature difference vibration sensor energy management method based on a capsule network, characterized by, Includes the following steps: S1. Environmental parameters are collected through an environmental sensor network. The collected environmental parameters are preprocessed by signal filtering and normalization. Time series segmentation technology is used to establish a time series correlation between historical data and current data. The environmental parameters are of five types, including vibration signal, temperature gradient, humidity, airflow disturbance and resonant frequency. S1 specifically includes the following sub-steps: S101. Collect raw data of five types of environmental parameters: vibration, temperature gradient, humidity, airflow disturbance and resonant frequency, to provide multi-dimensional environmental information input for subsequent analysis; S102. Construct historical feature sequences for each type of parameter, extract time-series segments by defining a fixed window length, capture short-term trends in environmental parameters, and provide a data foundation for fluctuation feature calculation; S103. Calculate the fluctuation index for each type of parameter to quantify the data fluctuation amplitude within the historical window, which is used for subsequent dynamic filtering parameter adjustment; the fluctuation index is used to describe the data change amplitude, and the specific calculation formula is as follows: ;in Standard deviation measures the magnitude of fluctuation in data within a historical window. The mean is used as a standardized volatility indicator. This represents the historical sequence of the i-th type of parameter, with a window length of k; S104. Dynamically adjust the filtering factor according to the fluctuation index, perform adaptive filtering on the current data, suppress high-frequency noise and retain effective signals, and balance the weight of real-time data and historical filtering results. S105. Normalize the filtered parameters to eliminate the differences in the dimensions of data from different sensors, ensure that multi-source data are fused at the same scale, and avoid a single parameter dominating model decision-making. S106. The normalized multi-class parameters are concatenated into a multi-dimensional feature vector, integrating multi-dimensional environmental information to provide a unified input for subsequent feature fusion and attention weight allocation. S107. Using time series segmentation technology, multi-dimensional features within the historical window are associated in chronological order to form time-series joint feature segments, capturing the dynamic evolution of environmental states. S2. Construct a selective attention capsule module, dynamically adjust the feature importance of each environmental parameter using learnable parameters, and normalize the weights of each environmental parameter using the softmax function to obtain a weighted fused environmental feature capsule vector. S3. In the prediction capsule module, a dynamic time memory factor based on environment is introduced. According to the changing state of environmental features, a dynamic time memory factor corresponding to each environmental parameter is established, and an environmental correlation state update mechanism is adopted to realize the temporal joint modeling of historical state and current input features. S3 specifically includes the following sub-steps: S301. Generate a dynamic time memory factor based on the current environmental parameters and environmental feature capsule vectors, and use the Sigmoid function to control the memory update intensity so that the model can adapt to the difference between sudden environmental changes and stable states. S302. Define the forget gate and the input gate, and jointly control the retention ratio of historical memory and the fusion ratio of new features by using environmental feature capsule vectors and fluctuation indicators to achieve time-dependent dynamic adjustment. S303. Update the memory state of each type of parameter through a gating mechanism, and capture long-term environmental evolution patterns by combining historical memory with current input. S304. The memory states of the five types of parameters are concatenated and mapped to contextual features to extract global temporal dependencies across parameters, thereby enhancing the model's understanding of interactions in complex environments. S305: Fuse contextual features with the current capsule vector, balance the contributions of historical and current information through trainable coefficients, improve training stability through layer standardization, and output temporal joint features; S4. Based on the environmental feature capsule vector and the temporal joint feature, establish an environmental parameter interaction matrix, calculate the interaction relationship between each environmental feature, obtain the fused feature representation, and further enhance the collaborative expression capability of environmental information. S4 specifically includes the following sub-steps: S401. Calculate the interaction matrix between environmental parameters, and measure the influence strength between parameters based on the cosine similarity of environmental feature capsule vectors and temporal joint features to reveal implicit association patterns. S402. Dynamically weight the interaction matrix according to the fluctuation index to enhance the interaction weight of high fluctuation parameters, suppress noise interference, and make the interaction relationship conform to the actual environmental changes. S403. By fusing temporal joint features and environmental feature capsule vectors through a weighted interaction matrix, information from different dimensions is spliced ​​together to generate fused features that include the synergistic effect of parameters. S404. Perform nonlinear transformation and standardization on the fused features to generate enhanced collaborative features, retaining high-order interaction information and providing discriminative input for pattern evaluation; S5. Based on the power generation mode scoring mechanism, an adaptive assessment of different energy management modes is conducted. The probability values ​​of different modes are calculated using the softmax function to determine the optimal energy harvesting and management strategy.

2. The energy management method for a self-powered temperature difference vibration sensor based on a capsule network according to claim 1, characterized in that, S2 specifically includes the following sub-steps: S201. The importance weights of each environmental parameter are dynamically adjusted through learnable parameters, so that the model can automatically focus on key environmental features based on the current input. S202. Use the Softmax function to normalize the learnable parameters and generate probabilistic feature weights to avoid excessive weight for a single parameter. S203. Weighted fusion of multiple normalized features highlights the role of important parameters, suppresses the interference of minor parameters, and generates a more representative fusion feature vector. S204. By using nonlinear transformation, the fused features are mapped to capsule vectors, which enhances the abstract expressive power of the features, preserves the spatial relationship between parameters, and provides high-order feature input for time series modeling.

3. The energy management method for a self-powered temperature difference vibration sensor based on a capsule network according to claim 1, characterized in that, S5 specifically includes the following sub-steps: S501. Construct a power generation mode score tensor based on collaborative features and time-series joint features, calculate mode adaptability through parameter interaction weights and learnable score weights, and quantify the degree of multi-dimensional matching. S502. The capsule routing mechanism is used to iteratively optimize the mode score. The interaction scores are weighted and summed using dynamic routing weights to obtain the intermediate score capsule vector of the k-th power generation mode. Subsequently, the intermediate score capsule vector is compressed using a nonlinear compression function. Activation is performed to obtain the capsule vector of the k-th power generation mode after the r-th routing iteration. The formula is as follows: ； ； ； in The initial routing coupling weights for the k-th power generation mode are represented and are uniformly initialized before the capsule routing iteration. The tensor representing the power generation mode score All elements of the k-th pattern, This represents the intermediate score capsule vector for the k-th power generation mode in the r-th routing iteration. Representing collaborative features Dimensions Representing temporal joint features Dimensions Let represent the interaction score of the k-th mode, the i-th dimension collaborative feature, and the j-th dimension temporal joint feature in the score tensor. Let represent the capsule vector activated by nonlinear compression after the r-th routing iteration; Update route weights using the following formula: ； in ; This represents the routing weight of the k-th mode after the r-th iteration. The index variable represents the summation index, and the index represents the first index. Various power generation modes; The final power generation mode score capsule is Its mold length Representational pattern confidence; S503. Perform multi-objective correction on the model score, introduce a time-series smoothing term to suppress abrupt changes, and a volatility penalty term to reduce the risk of model switching when the environment is unstable, thus balancing short-term and long-term benefits. S504. The corrected score is converted into a probability distribution through the Softmax function, which intuitively reflects the applicability probability of each power generation mode and supports probability-based decision optimization. S505. Select the highest probability mode that satisfies the constraints of capsule strength, synergistic concentration and historical stability to ensure the optimality of the strategy in terms of reliability, environmental adaptability and temporal continuity. Select the highest probability pattern that satisfies the constraints: ； in The preset capsule strength constraint threshold; The preset threshold for collaborative concentration constraint; The preset historical stability constraint threshold; Let be the collaborative weight distribution for the k-th mode; The preset historical time window length.

Images