Dynamic power management method and system based on low-power meter circuit

Abstract

This invention discloses a dynamic power management method and system based on low-power meter circuits, relating to the field of power measurement technology. The method includes: power prediction training based on the module hardware characteristics of the target low-power meter circuit and historical operating datasets of the meter circuit; constructing a multi-level power supply mode switching strategy according to the meter circuit power consumption predictor, and logically connecting it with a MOSFET power switching circuit to generate a power supply mode switching channel; deploying a monitoring sensor network to collect multi-module operating datasets in real time; and performing power demand prediction and dynamic power switching management on the multi-module operating datasets. This invention solves the technical problems of rigid power management modes, lack of accurate prediction capability for dynamic power consumption, and low efficiency of multi-module collaborative energy consumption management in existing technologies. It achieves the technical effects of improving the flexibility and response speed of power path switching, the overall energy efficiency ratio of the circuit, and the intelligent level of dynamic power management.

Description

Technical Field

[0001] This invention relates to the field of power measurement technology, specifically to a dynamic power management method and system based on low-power meter circuits. Background Technology

[0002] As the core terminal for collecting electricity consumption information, smart meters have seen explosive growth in deployment numbers. They are typically installed in complex environments with difficult wiring and rely on battery power or energy harvesting technology to operate, placing extremely high demands on the low-power design of the circuitry. Existing power management in electricity meter circuits typically employs fixed power supply modes or switching strategies based on simple threshold judgments. For example, different power levels are preset depending on whether the meter is in metering, communication, or standby mode. However, this approach has several limitations. On the one hand, fixed power supply modes cannot adapt to the dynamically changing power consumption demands of the meter in different application scenarios, easily leading to energy waste due to insufficient power supply or insufficient power during peak demand. On the other hand, threshold-based switching strategies often react slowly and cannot accurately predict instantaneous power consumption changes, resulting in low power conversion efficiency and high overall circuit energy consumption. Furthermore, the hardware characteristics and operating states of various functional modules integrated in low-power meter circuits, such as microcontrollers, metering chips, wireless communication modules, and real-time clocks, vary, and their power consumption characteristics during joint operation are complex, making it difficult to coordinate the dynamic energy consumption of multiple modules in both time and functional dimensions. At the same time, existing power switching mechanisms lack flexible logic combinations with high-precision, low-on-resistance MOSFET switching circuits, resulting in insufficient precision and response speed in power path management.

[0003] Therefore, current technologies suffer from problems such as rigid power management modes, lack of accurate prediction of dynamic power consumption, and low efficiency in multi-module collaborative energy management. Summary of the Invention

[0004] This application provides a dynamic power management method and system based on low-power meter circuits, which solves the technical problems of rigid power management mode, lack of accurate prediction capability of dynamic power consumption and low efficiency of multi-module collaborative energy consumption management in the prior art. It achieves the technical effect of improving the flexibility and response speed of power path switching, the overall energy efficiency ratio of the circuit, and the intelligent level of dynamic power management.

[0005] This application provides a dynamic power management method based on a low-power meter circuit. The method includes: training a power consumption prediction function based on the module hardware characteristics of the target low-power meter circuit and the historical operation dataset of the meter circuit to construct a power consumption predictor for the meter circuit; constructing a multi-level power supply mode switching strategy based on the power consumption predictor for the meter circuit; logically connecting the target low-power meter circuit with a MOSFET power switching circuit based on the multi-level power supply mode switching strategy to generate a power supply mode switching channel; deploying a monitoring sensor network within the target low-power meter circuit to collect multi-module operation datasets in real time through the monitoring sensor network; and performing power demand prediction and dynamic power switching management on the multi-module operation dataset based on the power consumption predictor for the meter circuit and the power supply mode switching channel.

[0006] In a possible implementation, constructing a power consumption predictor for the electricity meter circuit includes: standardizing the historical operating dataset of the electricity meter circuit according to the electricity meter data application standard to obtain an available operating dataset of the electricity meter circuit; performing feature extraction and operating state clustering on the available operating dataset of the electricity meter circuit to construct multiple operating state data clusters of the electricity meter circuit; performing operating power consumption correlation analysis based on the module hardware characteristics of the target low-power electricity meter circuit to construct a classification rule for the operating state-power consumption level of the electricity meter circuit; and performing power consumption prediction training on the multiple operating state data clusters of the electricity meter circuit according to the classification rule for the operating state-power consumption level of the electricity meter circuit to construct the power consumption predictor for the electricity meter circuit.

[0007] In a possible implementation, multiple electricity meter circuit operation status data clusters are constructed, including: extracting and standardizing multidimensional features from the available electricity meter circuit operation dataset to obtain a multidimensional operation feature dataset of electricity meter circuits; plotting the intra-cluster sum of squares curves of the multidimensional operation feature dataset of electricity meter circuits under different K values, filtering the inflection points of the intra-cluster sum of squares curves to determine the target clustering K value; and using the target clustering K value to perform operation status clustering on the multidimensional operation feature dataset of electricity meter circuits to construct multiple electricity meter circuit operation status data clusters.

[0008] In a possible implementation, the rule for classifying the operating state and power consumption level of the meter circuit is constructed, including: performing module-level power consumption modeling based on the module hardware characteristics of the target low-power meter circuit to generate a power consumption model set for the meter circuit module; using the power consumption model set for the meter circuit module to perform power consumption simulation analysis on the central features of each cluster in the multiple meter circuit operating state data clusters to obtain a circuit module operating state-power consumption mapping table; and performing total power consumption calculation and level classification preset on the circuit module operating state-power consumption mapping table to construct the rule for classifying the operating state and power consumption level of the meter circuit.

[0009] In a possible implementation, a power consumption prediction predictor is constructed by training the power consumption prediction of the multiple power meter circuit operating state data clusters according to the power consumption level classification rule of the power meter circuit operating state. This includes: classifying and identifying the operating data in the multiple power meter circuit operating state data clusters according to the power consumption level classification rule to obtain a power consumption level sample set of the power meter circuit operating state; training the power consumption prediction of the power consumption level sample set of the power meter circuit operating state using a deep neural network to obtain an initial circuit power consumption predictor; and performing performance testing, evaluation, and optimization on the initial circuit power consumption predictor to construct the power consumption predictor of the power meter circuit.

[0010] In a possible implementation, a multi-level power supply mode switching strategy is constructed, including: obtaining power consumption level information of the meter circuit based on the power consumption predictor of the meter circuit; configuring the power supply mode for each power consumption level in the power consumption level information of the meter circuit to obtain a multi-power consumption level matching power supply mode; designing power switching trigger conditions; and constructing a multi-level power supply mode switching strategy based on the multi-power consumption level matching power supply mode and the power switching trigger conditions.

[0011] In a possible implementation, generating a power supply mode switching channel includes: constructing a multi-level power switching channel architecture according to the multi-level power supply mode switching strategy; logically connecting the target low-power meter circuit and the MOSFET power switching circuit according to the multi-level power switching channel architecture to construct a multi-level power supply branch channel; and combining the multi-level power supply branch channels in parallel topology to generate a power supply mode switching channel.

[0012] In a possible implementation, the dynamic power management method based on the low-power meter circuit further includes: performing response test analysis on the power supply mode switching channel, setting a switching dead time, and optimizing the switching of the power supply mode switching channel based on the switching dead time.

[0013] In a possible implementation, power demand prediction and dynamic power switching management of the multi-module operation dataset are performed based on the power meter circuit power consumption predictor and the power supply mode switching channel, including: predicting power demand of the multi-module operation dataset based on the power meter circuit power consumption predictor to obtain the power circuit power consumption demand level; and using the power supply mode switching channel to perform dynamic power switching management of the power circuit power consumption demand level.

[0014] This application also provides a dynamic power management system based on a low-power meter circuit. The system includes: a power prediction training unit, used to perform power prediction training based on the module hardware characteristics of the target low-power meter circuit and the historical operation dataset of the meter circuit, and construct a power prediction predictor for the meter circuit; a switching strategy generation unit, used to construct a multi-level power supply mode switching strategy based on the power prediction predictor of the meter circuit, and logically connect the target low-power meter circuit and the MOSFET power switching circuit based on the multi-level power supply mode switching strategy to generate a power supply mode switching channel; an operation data acquisition unit, used to deploy a monitoring sensor network in the target low-power meter circuit, and collect multi-module operation datasets in real time through the monitoring sensor network; and a power demand prediction management unit, used to perform power demand prediction and dynamic power switching management on the multi-module operation dataset based on the power prediction predictor of the meter circuit and the power supply mode switching channel.

[0015] This application proposes a dynamic power management method and system based on low-power meter circuits. The method involves training power prediction based on the module hardware characteristics of the target low-power meter circuit and historical operating data. A multi-level power supply mode switching strategy is constructed based on the meter circuit power consumption predictor and logically connected to a MOSFET power switching circuit to generate a power supply mode switching channel. A monitoring sensor network is deployed to collect multi-module operating data in real time. Power demand prediction and dynamic power switching management are then performed on the multi-module operating data. This addresses the technical problems of rigid power management modes, lack of accurate prediction of dynamic power consumption, and low efficiency of multi-module collaborative energy management in existing technologies. The method achieves improved flexibility and response speed of power path switching, overall circuit energy efficiency ratio, and intelligent level of dynamic power management. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a schematic flowchart of a dynamic power management method based on a low-power meter circuit, provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of a dynamic power management system based on a low-power meter circuit, provided as an embodiment of this application.

[0019] Figure labeling: Power consumption prediction training unit 10, switching strategy generation unit 20, running data acquisition unit 30, power consumption demand prediction management unit 40. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0021] This application provides a dynamic power management method based on a low-power meter circuit, such as... Figure 1 As shown, the method includes: Step S100: Based on the module hardware characteristics of the target low-power meter circuit and the historical operation dataset of the meter circuit, power consumption prediction training is performed to construct a power consumption predictor for the meter circuit.

[0022] Step S100 further includes: standardizing the historical operating dataset of the meter circuit according to the meter data application standard to obtain an available meter circuit operating dataset; performing feature extraction and operating state clustering on the available meter circuit operating dataset to construct multiple meter circuit operating state data clusters; performing operating power consumption correlation analysis based on the module hardware characteristics of the target low-power meter circuit to construct meter circuit operating state-power consumption level classification rules; and performing power consumption prediction training on the multiple meter circuit operating state data clusters according to the meter circuit operating state-power consumption level classification rules to construct a meter circuit power consumption predictor.

[0023] Preferably, the standards for electricity meter data application may include data formats, database table structures, or communication protocols specified by the electricity meter industry, such as DL / T645 and IEC. 62056. Based on this, the historical operating dataset of the target low-power meter circuit, such as voltage, current, power, temperature, and event records, is standardized. This standardization includes at least data cleaning, format unification, and missing value handling. For example, outliers, duplicates, or data points with acquisition errors are removed. Data from different sources and time granularities are converted into a unified unit of measurement and time frequency. Fields missing during data acquisition are interpolated or filled to obtain a usable meter circuit operating dataset. Then, feature extraction is performed on the usable meter circuit operating dataset. This involves identifying key features that characterize the circuit's operating mode from the original time-series data, such as calculating the current fluctuation variance, average load rate over a certain period, harmonic content, and the ratio of active to reactive power. Unsupervised K-Means clustering is then used to group data points with similar feature vectors into the same category, forming multiple meter circuit operating state data clusters. Each cluster represents a typical circuit operating state, such as a stable light-load metering state, a frequent wireless communication state, or a centralized data processing state.

[0024] Preferably, the module hardware characteristics refer to the hardware parameters of specific components in the circuit, such as the operating current of the microcontroller unit (MCU), the transmission power of the wireless module, and the sampling rate of the metering chip. Based on the hardware parameters of the target low-power meter circuit, an operational power consumption correlation analysis is performed. Specifically, for each identified operational state data cluster, the analysis identifies the hardware modules that can be activated in that state, the operating frequency, or the corresponding voltage, and calculates the total power consumption value corresponding to that state. This establishes a correspondence between operational states and power consumption. Based on the calculated power consumption range, different operational states are classified into different power consumption levels, thus determining the meter circuit's operational state-power consumption. The power consumption is classified into different levels, such as level 1 representing power consumption <1mW, level 2 representing 1mW~5mW, etc. Then, power consumption prediction training is performed on multiple power meter circuit operating status data clusters according to the power consumption level classification rules. Specifically, the operating status data clusters are used as input samples, and the power consumption level is used as the output label. Deep neural networks are used for supervised training to learn the mapping relationship between specific operating status feature combinations and corresponding power consumption levels. After training, a power consumption predictor is obtained. When new real-time operating data is input, the predictor can output the power consumption level that the current circuit may be in based on the learned mapping relationship.

[0025] Furthermore, step S100 also includes: extracting and standardizing multidimensional features from the available meter circuit operation dataset to obtain a multidimensional operation feature dataset of the meter circuit; plotting the intra-cluster sum of squares curves of the multidimensional operation feature dataset of the meter circuit under different K values; filtering inflection points on the intra-cluster sum of squares curves to determine the target clustering K value; and using the target clustering K value to perform operation state clustering on the multidimensional operation feature dataset of the meter circuit to construct multiple meter circuit operation state data clusters.

[0026] Preferably, the available meter circuit operation dataset is mathematically transformed to generate new feature columns. Multiple statistical indicators that can characterize the circuit operation state from different dimensions are calculated and derived. For example, the sliding window average of the current is calculated to represent the load trend, the variance of the voltage is calculated to represent the fluctuation level, the power factor is calculated to represent the load characteristics, the timestamps of data collection are extracted to separate peak and off-peak periods, and the extracted feature values ​​of different dimensions are normalized by Z-score or scaled by Min-Max to map to the same numerical range to eliminate the influence of dimensions. This generates a multi-dimensional operation feature dataset of the meter circuit, where rows represent operation records at specific times and columns represent dimensionless statistical features. Then, different clustering numbers K are tried one by one, such as from 1 to 10. For each K value, a clustering algorithm is run and the sum of squared errors of that clustering result is calculated. The K value is then compared with the corresponding error... The sum of squared differences is used as the coordinate points to connect the curves, forming the intra-cluster sum of squares curves of the multi-dimensional operating feature dataset of the electricity meter circuit under different K values. Then, the inflection point of the intra-cluster sum of squares curve is identified and screened, that is, the turning point where the curve changes from a rapid drop to a gentle slide. The K value corresponding to the inflection point is determined as the target clustering K value, which can ensure classification accuracy without overfitting. Then, using the target clustering K value as the input parameter, an unsupervised clustering algorithm is run to perform operating state clustering on the multi-dimensional operating feature dataset of the electricity meter circuit. That is, according to the Euclidean distance between feature vectors, each record in the multi-dimensional operating feature dataset is divided into the nearest category. Finally, multiple electricity meter circuit operating state data clusters are generated. Each electricity meter circuit operating state data cluster represents a circuit operating mode with stable electrical characteristics, such as "light load steady-state metering mode", "carrier communication transient mode", "battery float charging mode", etc.

[0027] Furthermore, step S100 also includes: performing module-level power consumption modeling based on the module hardware characteristics of the target low-power meter circuit to generate a power consumption model set for the meter circuit module; using the power consumption model set for the meter circuit module to perform power consumption simulation analysis on the central features of each cluster in the multiple meter circuit operating status data clusters to obtain a circuit module operating status-power consumption mapping table; and performing total power consumption calculation and level classification preset on the circuit module operating status-power consumption mapping table to construct a meter circuit operating status-power consumption level classification rule.

[0028] Preferably, the module hardware characteristics refer to the electrical parameters of each hardware module under specific operating conditions, such as the unit frequency operating current of the microcontroller in active mode, the leakage current in deep sleep mode, the RF power amplification efficiency of the wireless communication module in the transmitting state, the sampling channel power consumption of the metering chip, and the conversion efficiency curve of the power management unit. For each independent hardware functional module of the target low-power meter circuit, based on its datasheet parameters or measured electrical behavior, a power consumption calculation model describing the module under different operating conditions is generated, forming a power consumption model set for the meter circuit module. Each power consumption model can independently calculate the instantaneous power consumption of the corresponding hardware module based on the input operating parameters. The central feature of each cluster in multiple electricity meter circuit operation status data clusters is used as input conditions. The central feature of each data cluster represents the average electrical characteristics of that type of operation status, such as average current, average voltage, power factor range, and communication frequency. For each cluster, its central feature is parsed into the working parameters of each hardware module, and substituted into the electricity meter circuit module power consumption model set for power consumption simulation analysis. The power consumption of each hardware module in this state is calculated, and then a circuit module operation status-power consumption mapping table is generated. The rows of the table correspond to different operation statuses, the columns correspond to different hardware modules, and the cells are filled with the calculated power consumption value of the module in that state. Then, for each operating state, the power consumption values ​​of all hardware modules in that row are summed to obtain the estimated total power consumption of the circuit in that state. Then, based on the designed power supply capacity or energy-saving target, several power consumption threshold ranges are set, such as ultra-low power consumption range 0~1mW, low power consumption range 1~5mW, medium power consumption range 5~20mW, and high performance range >20mW. The range in which the estimated total power consumption of the circuit falls is marked as the power consumption level corresponding to that state. Finally, the operating state-power consumption level classification rule of the meter circuit is determined to clarify the corresponding total power consumption level when the circuit exhibits a certain operating state characteristic.

[0029] Furthermore, step S100 also includes: classifying and identifying the operating data in the multiple meter circuit operating status data clusters according to the meter circuit operating status-power consumption level classification rules to obtain a meter circuit operating status power consumption level sample set; using a deep neural network to perform power consumption prediction training on the meter circuit operating status power consumption level sample set to obtain an initial circuit power consumption predictor; and performing performance testing, evaluation, and optimization on the initial circuit power consumption predictor to construct a meter circuit power consumption predictor.

[0030] Preferably, the power consumption level classification rule of the electricity meter circuit operation status is used as a label generator to classify and label the operation data in multiple electricity meter circuit operation status data clusters according to their power consumption levels. Specifically, a corresponding power consumption level label is added to each data record within each data cluster based on its power consumption level. This transforms the unsupervised clustering results into a supervised training dataset, resulting in a power consumption level sample set for the electricity meter circuit operation status, containing the input operation data feature vector and the output power consumption level label. A prediction model is then constructed based on a deep neural network. The power consumption level sample set of the electricity meter circuit operation status is used to train the prediction model for power consumption prediction, enabling it to learn the nonlinear mapping relationship from input features to power consumption levels. Specifically, the power consumption level sample set of the electricity meter circuit operation status... This set is divided into training, validation, and test sets. The feature data of the training set is input into the prediction model for layer-by-layer computation, outputting the probability distribution of predicted power consumption levels. The loss function is used to calculate the error between the probability distribution output by the network and the true label. Based on the loss value, the gradient is calculated through the Adam optimizer, and the weights and biases of each layer in the network are updated. The iteration continues until the performance of the prediction model on the validation set no longer improves or reaches the preset number of training epochs, thus obtaining the initial circuit power consumption predictor. The performance of the trained model is evaluated using the test set that was not used in the training, and the model is fine-tuned based on the evaluation results. For example, hyperparameter adjustment, network structure adjustment, data augmentation, or regularization techniques are used to prevent overfitting, resulting in a deployable meter circuit power consumption predictor for real-time power consumption demand prediction.

[0031] Preferably, a multilayer perceptron is used for power consumption prediction training. The number of neurons in the input layer is equal to the dimension of the input feature vector, such as the total number of features including average current, current variance, voltage, power factor, temperature, and communication flags, which may be 10 to 50. The hidden layers consist of 2 to 4 fully connected layers, with the number of neurons decreasing layer by layer, for example, [128, 64, 32] or [256, 128, 64]. The ReLU activation function is used to alleviate gradient vanishing. The number of neurons in the output layer is equal to the number of power consumption levels, for example, 4 power consumption levels. The Softmax activation function is used to convert the network output into a probability distribution, representing the probability that the input sample belongs to each power consumption level. The cross-entropy loss function is used as the loss function. The initial learning rate is set to 0.001 or 0.0001, the batch size is 32, 64, 128, or 256, and the training epochs are 50 to 200. An early stopping strategy is used to prevent overfitting. The performance evaluation metrics include accuracy, precision, and recall.

[0032] Step S200: Based on the power consumption predictor of the meter circuit, a multi-level power supply mode switching strategy is constructed. Based on the multi-level power supply mode switching strategy, the target low-power meter circuit and the MOSFET power switching circuit are logically connected to generate a power supply mode switching channel.

[0033] Step S200 further includes: obtaining power consumption level information of the meter circuit according to the power consumption predictor of the meter circuit; configuring the power supply mode for each power consumption level in the power consumption level information of the meter circuit to obtain a multi-power consumption level matching power supply mode; designing power switching trigger conditions; and constructing a multi-level power supply mode switching strategy based on the multi-power consumption level matching power supply mode and the power switching trigger conditions.

[0034] Preferably, real-time acquisition of circuit operation characteristic data is input into the meter circuit power consumption predictor for forward calculation, identifying the operating state of the meter circuit, and outputting the power consumption level label of the circuit at the current or future time, determining the power consumption level information of the meter circuit, such as Level 1 ultra-low power consumption, Level 2 low power consumption, Level 3 medium power consumption, and Level 4 high-performance power consumption, representing the overall power supply capability level required by the current circuit; the power supply mode refers to the specific electrical parameter configuration included in each power supply mode, such as voltage setting value, maximum output current capability, power source selection, and clock frequency limit, thereby establishing a mapping table between power consumption level and power supply mode. For example, if the predicted level is Level 1, it is configured as "ultra-low leakage current LDO mode and clock gating", if the predicted level is Level 2, it is configured as "high-efficiency DC-DC mode and peripheral power supply", and if the predicted level is Level 3, it is configured as "high-performance DC-DC mode and full-speed clock", finally determining the power supply mode matching multiple power consumption levels.

[0035] Preferably, the power switching trigger conditions are designed, including but not limited to: predictive triggering (the power predictor outputs a new level and remains stable for several consecutive cycles); threshold triggering (detecting a specific proportion where the load current exceeds the current mode's upper limit); and event triggering (external interrupt wake-up, timed task startup, or communication request arrival). Then, the multi-power level matching power supply mode and power switching trigger conditions are combined to construct a multi-level power supply mode switching strategy. This includes generating a complete logic control unit, receiving power level information, determining whether the trigger conditions are met, logically connecting the target low-power meter circuit to the MOSFET power switching circuit, and finally determining the power supply mode switching channel. The input is the control signal for the multi-level power supply mode switching strategy, and the output is a stable power supply to the meter circuit, while also including switching timing control to ensure voltage stability and no overshoot when changing the power path. At any given time, at most one branch channel's MOSFET is in the on state, while other channels are in the off state, thus ensuring that the meter circuit can only draw power from the selected power mode.

[0036] Furthermore, step S200 also includes: constructing a multi-level power switching channel architecture according to the multi-level power supply mode switching strategy; logically connecting the target low-power meter circuit and the MOSFET power switching circuit according to the multi-level power switching channel architecture to construct a multi-level power supply branch channel; and combining the multi-level power supply branch channels in parallel topology to generate a power supply mode switching channel.

[0037] Preferably, based on the number of power consumption levels and corresponding power requirements in the multi-level power supply mode switching strategy, multiple parallel power supply branches are designed, each corresponding to a power supply mode. The power source for each branch is determined; for example, a high-efficiency buck converter is used for the high-power mode, a low-dropout linear regulator for the medium-power mode, and battery pass-through for the lowest-power mode. The source and logic level of the control signals required for each MOSFET switch are then determined, thereby defining the multi-level power switching channel architecture. The target low-power meter circuit is then logically connected to the MOSFET power switching circuit, including connecting the output terminals of different power supplies to the input terminals of the corresponding MOSFET switches, and transferring the power output of the meter circuit... The input terminals are connected to the output terminals of all MOSFET switches to form a common connection point. The control signal line is connected to the gate of each MOSFET switch to define a multi-level power supply branch channel. This ensures that the turn-on and turn-off of each MOSFET are controlled by the output signal of the switching strategy. Each branch channel corresponds to a power consumption level. When the power consumption level is activated, the corresponding MOSFET turns on, and the power mode configured for that power consumption level is applied to the meter circuit. Finally, the multi-level power supply branch channels are combined in parallel topology, that is, the input terminals are independent of each other and the output terminals are short-circuited to generate a power supply mode switching channel. This allows the meter circuit to be switched to a preset power supply mode that best suits the current power consumption requirements in real time and with low loss, according to the control signal.

[0038] Furthermore, step S200 also includes performing response test analysis on the power supply mode switching channel, setting a switching dead time, and optimizing the switching of the power supply mode switching channel based on the switching dead time.

[0039] Preferably, the dynamic electrical response characteristics of the power supply mode switching channel in actual operation are evaluated by experimental simulation measurement. Specifically, this includes measuring the time difference from the issuance of the control signal to the actual turn-on or turn-off of the MOSFET, measuring the lowest and highest points of the power supply voltage of the meter circuit at the moment of switching, measuring the instantaneous large current of the input power supply during the switching process, and measuring the total time required to completely disconnect from one power supply mode to another. This determines whether the current switching channel has the risk of power failure reset due to excessively long "disconnect-then-connect" time or the risk of power short circuit due to "connect-then-disconnect", and whether the voltage fluctuation amplitude exceeds the allowable range of the circuit. Based on the response test analysis results, a very short time interval is set between turning off the MOSFET of the current power supply branch and turning on the MOSFET of the next power supply branch. This time is called the switching dead time, which must be greater than the MOSFET turn-off delay time to ensure that the current branch has been completely cut off. This time must also be less than the power supply maintenance time of the meter circuit to prevent the circuit from resetting due to excessively low voltage. Then, the power supply mode switching channel is optimized based on the switching dead time. Specifically, in the power management logic of the FPGA and microcontroller, the timing control of turning off the current channel, waiting for the dead time, and then turning on the target channel is programmed to ensure that the dead time is adjustable to adapt to different load conditions or the characteristics of different MOSFETs, thereby improving the safety and stability of the meter circuit operation.

[0040] Step S300: Deploy a monitoring sensor network within the target low-power meter circuit, and collect multi-module operation datasets in real time through the monitoring sensor network.

[0041] Preferably, a monitoring sensor network is deployed according to the functional module division and power consumption analysis requirements of the meter circuit. Specifically, current sensors are connected in series on the power supply branches of core modules such as the microcontroller, wireless communication module, and metering chip to collect the real-time current consumption of each module; voltage sensors are connected to the power input terminals of each module to monitor the stability and fluctuation of the power supply voltage; temperature sensors are mounted or integrated near heat-generating components such as power amplifier chips and power management chips to monitor the operating temperature; and status sensors are connected to each functional module through digital interfaces to read the module's operating status flags, such as the sleep / wake-up status of the microcontroller and the transmit / receive status of the wireless module. The monitoring sensor network collects multi-module operating datasets in real time, where each module's operating data includes at least the module identifier, timestamp, electrical parameters such as current and voltage, the module's current temperature value, and its operating mode, reflecting the actual energy consumption of the meter circuit under different times and tasks.

[0042] Step S400: Based on the power consumption predictor of the meter circuit and the power supply mode switching channel, perform power demand prediction and dynamic power switching management on the multi-module operation dataset.

[0043] Step S400 further includes: predicting the power demand of the multi-module operation dataset based on the power consumption predictor of the meter circuit to obtain the power consumption demand level of the power circuit; and using the power supply mode switching channel to perform dynamic power switching management of the power consumption demand level of the power circuit.

[0044] Preferably, the real-time collected multi-module operation dataset is subjected to feature extraction and standardization to form a real-time feature vector. This vector is then input into the meter circuit power consumption predictor for power demand prediction. The meter circuit power consumption predictor performs forward computation, calculating the input features and weight parameters learned during the training phase, and outputs power consumption level labels or probability distributions belonging to each level. This determines the power consumption demand level of the power circuit, quantifying the power supply capacity required to meet the normal operation needs of the meter circuit under the current operating state, while avoiding energy waste. Then, the power consumption demand level is used as a control command to execute the actual power path switching operation through the power supply mode switching channel, realizing dynamic adjustment of the power supply to the meter circuit. Specifically, the power consumption demand level is received, and the target power supply corresponding to that level is determined according to the multi-power consumption level matching power supply mode mapping table. The system checks whether the current power switching trigger conditions are met. If the trigger conditions are met, a control signal is sent to the MOSFET power switching circuit according to the switching sequence. Based on the switching strategy and the set dead time, the system precisely controls the on and off of the MOSFETs on the corresponding branch channel. The MOSFET power switching circuit responds to the control signal, turns off the current power supply branch, and turns on the target power supply branch, switching the power input of the meter circuit from the current mode to a new mode that matches the predicted power consumption demand level. This process is repeated. As the operating status of the meter circuit changes, the monitoring sensor network updates data in real time, the power consumption predictor continuously outputs new power consumption demand levels, and the power supply mode switching channel dynamically adjusts the power supply mode to achieve real-time matching between power supply capacity and load demand. This improves the flexibility and response speed of power path switching and minimizes energy consumption while ensuring performance.

[0045] In the above text, refer to Figure 1 A dynamic power management method based on a low-power meter circuit according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2 A dynamic power management system based on a low-power meter circuit according to an embodiment of the present invention is described.

[0046] The dynamic power management system based on a low-power meter circuit according to embodiments of the present invention addresses the technical problems in the prior art, such as rigid power management modes, lack of accurate prediction of dynamic power consumption, and low efficiency of multi-module collaborative energy management. It achieves the technical effects of improving the flexibility and response speed of power path switching, the overall energy efficiency ratio of the circuit, and the intelligent level of dynamic power management. Figure 2 As shown, the dynamic power management system based on low-power meter circuit includes: a power consumption prediction training unit 10, a switching strategy generation unit 20, a running data acquisition unit 30, and a power consumption demand prediction management unit 40.

[0047] The power consumption prediction training unit 10 is used to perform power consumption prediction training based on the module hardware characteristics of the target low-power meter circuit and the historical operation dataset of the meter circuit, and to construct a power consumption predictor for the meter circuit. The switching strategy generation unit 20 is used to construct a multi-level power supply mode switching strategy based on the power consumption predictor of the meter circuit, and to logically connect the target low-power meter circuit and the MOSFET power switching circuit based on the multi-level power supply mode switching strategy to generate a power supply mode switching channel. The operation data acquisition unit 30 is used to deploy a monitoring sensor network in the target low-power meter circuit, and to collect multi-module operation datasets in real time through the monitoring sensor network. The power consumption demand prediction management unit 40 is used to perform power consumption demand prediction and dynamic power switching management on the multi-module operation dataset based on the power consumption predictor of the meter circuit and the power supply mode switching channel.

[0048] The specific configuration of the power consumption prediction training unit 10 will be described in detail below. The power consumption prediction training unit 10 further includes: standardizing the historical operating dataset of the meter circuit according to the meter data application standard to obtain an available meter circuit operating dataset; performing feature extraction and operating state clustering on the available meter circuit operating dataset to construct multiple meter circuit operating state data clusters; performing operating power consumption correlation analysis based on the module hardware characteristics of the target low-power meter circuit to construct meter circuit operating state-power consumption level classification rules; and performing power consumption prediction training on the multiple meter circuit operating state data clusters according to the meter circuit operating state-power consumption level classification rules to construct a meter circuit power consumption predictor.

[0049] The specific configuration of the power consumption prediction training unit 10 will be described in detail below. The power consumption prediction training unit 10 further includes: extracting and standardizing multi-dimensional features from the available meter circuit operation dataset to obtain a multi-dimensional operation feature dataset of the meter circuit; plotting the intra-cluster sum of squares curves of the multi-dimensional operation feature dataset of the meter circuit under different K values; filtering inflection points on the intra-cluster sum of squares curves to determine the target clustering K value; and using the target clustering K value to perform operation state clustering on the multi-dimensional operation feature dataset of the meter circuit to construct multiple meter circuit operation state data clusters.

[0050] The specific configuration of the power prediction training unit 10 will be described in detail below. The power prediction training unit 10 further includes: performing module-level power consumption modeling based on the module hardware characteristics of the target low-power meter circuit to generate a power consumption model set for the meter circuit module; using the power consumption model set to perform power consumption simulation analysis on the central features of each cluster in the multiple meter circuit operating state data clusters to obtain a circuit module operating state-power consumption mapping table; and performing total power consumption calculation and level classification preset on the circuit module operating state-power consumption mapping table to construct a meter circuit operating state-power consumption level classification rule.

[0051] The specific configuration of the power consumption prediction training unit 10 will be described in detail below. The power consumption prediction training unit 10 further includes: classifying and identifying the operating data in the multiple meter circuit operating state data clusters according to the meter circuit operating state-power consumption level classification rules, thereby obtaining a meter circuit operating state power consumption level sample set; using a deep neural network to perform power consumption prediction training on the meter circuit operating state power consumption level sample set to obtain an initial circuit power consumption predictor; and performing performance testing, evaluation, and optimization on the initial circuit power consumption predictor to construct a meter circuit power consumption predictor.

[0052] The specific configuration of the switching strategy generation unit 20 will be described in detail below. The switching strategy generation unit 20 further includes: obtaining power consumption level information of the meter circuit based on the meter circuit power consumption predictor; configuring the power supply mode for each power consumption level in the meter circuit power consumption level information to obtain a multi-power consumption level matching power supply mode; designing power switching trigger conditions; and constructing a multi-level power supply mode switching strategy based on the multi-power consumption level matching power supply mode and the power switching trigger conditions.

[0053] The specific configuration of the switching strategy generation unit 20 will be described in detail below. The switching strategy generation unit 20 further includes: constructing a multi-level power switching channel architecture according to the multi-level power supply mode switching strategy; logically connecting the target low-power meter circuit and the MOSFET power switching circuit according to the multi-level power switching channel architecture to construct a multi-level power supply branch channel; and combining the multi-level power supply branch channels in parallel topology to generate a power supply mode switching channel.

[0054] The specific configuration of the switching strategy generation unit 20 will be described in detail below. The switching strategy generation unit 20 further includes: performing response test analysis on the power supply mode switching channel, setting a switching dead time, and optimizing the switching of the power supply mode switching channel based on the switching dead time.

[0055] The specific configuration of the power consumption demand prediction management unit 40 will be described in detail below. The power consumption demand prediction management unit 40 further includes: predicting the power consumption demand of the multi-module operation dataset based on the power consumption predictor of the meter circuit to obtain the power consumption demand level of the power circuit; and using the power supply mode switching channel to perform dynamic power switching management of the power consumption demand level of the power circuit.

[0056] The dynamic power management system based on low-power meter circuit provided in the embodiments of the present invention can execute the dynamic power management method based on low-power meter circuit provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method execution.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A dynamic power management method based on a low-power meter circuit, characterized in that, The method includes: Power consumption prediction training is conducted based on the module hardware characteristics of the target low-power meter circuit and the historical operation dataset of the meter circuit to construct a power consumption predictor for the meter circuit. Based on the power consumption predictor of the meter circuit, a multi-level power supply mode switching strategy is constructed. Based on the multi-level power supply mode switching strategy, the target low-power meter circuit and the MOSFET power switching circuit are logically connected to generate a power supply mode switching channel. A monitoring sensor network is deployed within the target low-power meter circuit, and the monitoring sensor network is used to collect multi-module operation datasets in real time. Based on the power consumption predictor of the meter circuit and the power supply mode switching channel, power consumption demand prediction and dynamic power switching management are performed on the multi-module operation dataset.

2. The dynamic power management method based on a low-power meter circuit as described in claim 1, characterized in that, Constructing a power consumption predictor for an electricity meter circuit, including: The historical operation dataset of the electricity meter circuit is standardized according to the electricity meter data application standard to obtain the usable electricity meter circuit operation dataset. Feature extraction and operation status clustering are performed on the available meter circuit operation dataset to construct multiple meter circuit operation status data clusters; Based on the module hardware characteristics of the target low-power meter circuit, an operational power consumption correlation analysis is performed to construct a meter circuit operation state-power consumption level classification rule. According to the power consumption level classification rule of the power meter circuit operation status, the power consumption prediction training is performed on the multiple power meter circuit operation status data clusters to construct a power consumption predictor for the power meter circuit.

3. The dynamic power management method based on a low-power meter circuit as described in claim 2, characterized in that, Construct multiple data clusters of electricity meter circuit operation status, including: Multidimensional feature extraction and standardization are performed on the available meter circuit operation dataset to obtain a multidimensional operation feature dataset of meter circuits. Plot the intra-cluster sum of squares curves of the multi-dimensional operating feature dataset of the meter circuit under different K values, filter the inflection points of the intra-cluster sum of squares curves, and determine the target clustering K value; The target clustering K value is used to perform operational state clustering on the multidimensional operational feature dataset of the electricity meter circuit, thereby constructing multiple operational state data clusters of the electricity meter circuit.

4. The dynamic power management method based on a low-power meter circuit as described in claim 2, characterized in that, Establish rules for classifying the operating status and power consumption levels of electricity meter circuits, including: Based on the module hardware characteristics of the target low-power meter circuit, module-level power consumption modeling is performed to generate a power consumption model set for the meter circuit module. The power consumption model set of the meter circuit module is used to perform power consumption simulation analysis on the central features of each cluster in the multiple meter circuit operation status data clusters to obtain a circuit module operation status-power consumption mapping table. The total power consumption is calculated and the level classification is preset for the circuit module operation status-power consumption mapping table, and the meter circuit operation status-power consumption level classification rules are constructed.

5. The dynamic power management method based on a low-power meter circuit as described in claim 2, characterized in that, According to the aforementioned electricity meter circuit operating state-power consumption level classification rule, power consumption prediction training is performed on the multiple electricity meter circuit operating state data clusters to construct an electricity meter circuit power consumption predictor, including: According to the power consumption level classification rule of the power meter circuit operation status, the operation data in the multiple power meter circuit operation status data clusters are classified and identified by power consumption level to obtain the power consumption level sample set of the power meter circuit operation status. A deep neural network is used to train a power consumption prediction system on a sample set of power consumption levels of the meter circuit operating status to obtain an initial circuit power consumption predictor. The initial circuit power consumption predictor was tested, evaluated, and optimized to construct a power consumption predictor for the electricity meter circuit.

6. The dynamic power management method based on a low-power meter circuit as described in claim 1, characterized in that, Construct a multi-level power supply mode switching strategy, including: Based on the power consumption predictor of the meter circuit, obtain the power consumption level information of the meter circuit; Configure the power supply mode for each power consumption level in the power consumption level information of the meter circuit to obtain a multi-power consumption level matching power supply mode. Design power switching trigger conditions, and construct a multi-level power supply mode switching strategy based on the multi-power level matching power supply mode and the power switching trigger conditions.

7. The dynamic power management method based on a low-power meter circuit as described in claim 1, characterized in that, Generate a power supply mode switching channel, including: Based on the multi-level power supply mode switching strategy, a multi-level power switching channel architecture is constructed. The target low-power meter circuit and the MOSFET power switching circuit are logically connected according to the multi-level power switching channel architecture to construct a multi-level power supply branch channel. The multi-level power supply branch channels are combined in parallel topology to generate a power supply mode switching channel.

8. The dynamic power management method based on a low-power meter circuit as described in claim 7, characterized in that, The method further includes: The power supply mode switching channel is subjected to response test analysis, a switching dead time is set, and the switching of the power supply mode switching channel is optimized based on the switching dead time.

9. The dynamic power management method based on a low-power meter circuit as described in claim 1, characterized in that, Based on the power consumption predictor of the meter circuit and the power supply mode switching channel, power demand prediction and dynamic power switching management are performed on the multi-module operation dataset, including: Based on the power consumption predictor of the meter circuit, the power consumption demand of the multi-module operation dataset is predicted to obtain the power consumption demand level of the power circuit. The power supply mode switching channel is used to dynamically manage the power consumption requirements of the power circuit.

10. A dynamic power management system based on a low-power meter circuit, characterized in that, The system is used to implement the dynamic power management method based on a low-power meter circuit as described in any one of claims 1 to 9, and the system comprises: The power consumption prediction training unit is used to perform power consumption prediction training based on the module hardware characteristics of the target low-power meter circuit and the historical operation dataset of the meter circuit, and to build a power consumption predictor for the meter circuit. The switching strategy generation unit is used to construct a multi-level power supply mode switching strategy based on the power consumption predictor of the meter circuit, and to logically connect the target low-power meter circuit and the MOSFET power switching circuit based on the multi-level power supply mode switching strategy to generate a power supply mode switching channel. The data acquisition unit is used to deploy a monitoring sensor network in the target low-power meter circuit and collect multi-module operation data in real time through the monitoring sensor network. The power consumption demand prediction and management unit is used to perform power consumption demand prediction and dynamic power switching management on the multi-module operation dataset based on the power consumption predictor of the meter circuit and the power supply mode switching channel.

Images