Power management strategy prediction method and device, electronic equipment and storage medium

Abstract

The invention relates to the field of power supply intelligent management, in particular to a power supply management strategy prediction method and device, electronic equipment and a storage medium. According to the embodiment of the invention, the touch behavior sequence of the user in the current time period is obtained, and the future touch behavior is predicted based on the sequence, so that compared with a traditional touch behavior prediction technology, the method does not depend on simple statistics or preset rules any more, but is based on the real behavior data of the user, and a complex and changeable user behavior mode can be analyzed more accurately. Meanwhile, by analyzing the influence data generated by the touch behavior on the power supply and combining the current performance data of the power supply, a power supply management strategy is generated, and a traditional static power supply management mode lacking dynamic response is changed. And finally, the operation is executed according to the strategy, so that full-link dynamic management from behavior prediction to strategy making and execution is realized, the blank that the power management strategy cannot be dynamically optimized based on user behaviors in the traditional technology is filled, and the power management efficiency and the energy consumption performance of the intelligent equipment are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent power management, and in particular to a prediction method, device, electronic device and storage medium for power management strategies. Background Art

[0002] Currently, traditional power management strategies mostly use static methods, allocating hardware resources based on preset rules. This approach has significant limitations. Due to its lack of awareness and response to real-time user actions, it cannot flexibly adjust power management strategies based on dynamic changes in user touch behavior, resulting in long-term low device energy efficiency.

[0003] For example, when users frequently swipe, drag, and tap the screen, the device's high-performance mode persists excessively, causing unnecessary power consumption. When user activity decreases or is interrupted, the device fails to switch to low-power mode in a timely manner, further wasting battery life. Furthermore, existing touch behavior prediction technologies often rely on simple statistical or rule-based models, making it difficult to accurately capture and effectively analyze complex and changing user behavior patterns. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a power management strategy prediction method, device, electronic device, and storage medium to solve the problem that existing touch behavior prediction technologies are difficult to accurately analyze complex user behaviors.

[0005] In a first aspect, an embodiment of the present invention provides a method for predicting a power management policy, the method comprising:

[0006] Get the sequence of touch actions performed by the user on the smart device during the current time period;

[0007] predicting the touch behavior performed by the user on the smart device in a future time period based on the touch behavior sequence;

[0008] Analyzing data on the impact of the touch behavior on the power supply of the smart device, and generating a power management policy corresponding to the power supply based on the impact data and current performance data of the power supply;

[0009] Perform corresponding power management operations on the power supply according to the power management policy.

[0010] Furthermore, predicting the touch behavior performed by the user on the smart device in a future time period based on the touch behavior sequence includes:

[0011] extracting time variation rules and touch behavior patterns from the touch behavior sequence;

[0012] Inputting the time variation pattern and the touch behavior pattern into a pre-trained prediction model, analyzing the dependency relationship between the time variation pattern and the touch behavior pattern through the prediction model, and predicting the touch behavior performed by the user on the smart device in a future time period based on the dependency relationship;

[0013] Furthermore, analyzing the dependency between the time variation pattern and the touch behavior pattern through the prediction model, and predicting the touch behavior performed by the user on the smart device in a future time period based on the dependency, includes:

[0014] Identifying local features of the touch behavior model at different time scales using convolution kernels of different sizes of the temporal convolutional network in the prediction model to obtain a first feature vector;

[0015] Analyzing the temporal evolution of the touch behavior pattern in the first feature vector using a long short-term memory network in the prediction model to obtain a second feature vector;

[0016] Performing an attention mechanism weighted operation based on the second feature vector to obtain a weighted behavior feature;

[0017] The weighted behavior features are projected into the operation pattern space through the output layer in the prediction model, and the touch behavior performed by the user on the smart device in the future time period is output by combining the time change law and the dependency relationship between the touch behavior pattern.

[0018] Furthermore, the analyzing the data on the impact of the touch behavior on the power supply of the smart device includes:

[0019] Obtaining an energy efficiency characteristic curve of the power supply;

[0020] Calculating the power consumption distribution of the power supply based on the energy efficiency characteristic curve to obtain energy consumption impact data corresponding to the touch behavior;

[0021] Evaluating health impact data of the touch behavior on the power supply based on the energy consumption impact data and the environmental impact data;

[0022] The impact data is calculated based on the energy consumption impact data and the health impact data.

[0023] Furthermore, the calculating the impact data based on the energy consumption impact data and the health impact data includes:

[0024] Calculating the cumulative energy consumption and energy consumption fluctuation coefficient under different operation modes based on the power consumption distribution and time series in the energy consumption impact data to obtain a quantitative feature vector of the energy consumption impact;

[0025] Performing a multi-dimensional correlation analysis based on the energy consumption impact data and the health impact data, constructing an energy consumption-health coupling model, and obtaining a correlation feature vector between touch behavior and power supply performance degradation;

[0026] The influence data is obtained by performing weighted calculation using the quantized feature vector and the associated feature vector.

[0027] Furthermore, generating a power management policy corresponding to the power supply based on the impact data and the current performance data of the power supply includes:

[0028] Performing feature fusion on the impact data and the current performance data of the power supply to construct a power supply status assessment vector;

[0029] Using a preset rule base to judge the power state evaluation vector, a candidate strategy set that meets the safety constraint conditions is obtained;

[0030] The cumulative benefit of each strategy in the candidate strategy set under different scenarios is evaluated, and the strategy with the best cumulative benefit is used as the power management strategy.

[0031] Furthermore, the evaluation of the cumulative benefits of each strategy in the candidate strategy set under different scenarios includes:

[0032] Substitute each strategy in the candidate strategy set into various scenarios in the scenario library to simulate the operation of the power supply in the future. During the simulation of the power supply operation, calculate the energy consumption and health changes of the power supply after the strategy is executed;

[0033] For each scenario, a weighted sum is performed based on the power consumption and the health change to obtain a single scenario benefit;

[0034] Based on the accumulation of the single-scenario benefits of each strategy in all scenarios, the cumulative benefits of each strategy in different scenarios are obtained.

[0035] In a second aspect, an embodiment of the present invention provides a device for predicting a power management strategy, the device comprising:

[0036] An acquisition module, used to acquire a sequence of touch actions performed by a user on a smart device within a current time period;

[0037] A prediction module, configured to predict, based on the touch behavior sequence, the touch behavior that the user will perform on the smart device within a future time period;

[0038] an analysis module, configured to analyze data on the impact of the touch behavior on the power supply of the smart device, and generate a power management policy corresponding to the power supply based on the impact data and current performance data of the power supply;

[0039] An execution module is used to perform corresponding power management operations on the power supply according to the power management policy.

[0040] In a third aspect, an embodiment of the present invention provides an electronic device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, computer instructions being stored in the memory, and the processor executing the method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0041] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method of the first aspect or any corresponding embodiment thereof.

[0042] The embodiment of the present application obtains the touch behavior sequence of the user in the current time period and predicts future touch behavior based on this sequence. Compared with traditional touch behavior prediction technology, it no longer relies on simple statistics or preset rules, but is based on the user's real behavior data, and can more accurately analyze complex and changeable user behavior patterns. At the same time, by analyzing the impact data of touch behavior on the power supply and generating a power management strategy based on the current performance data of the power supply, the traditional static power management method that lacks dynamic response is changed. Finally, operations are executed according to the strategy, realizing full-link dynamic management from behavior prediction to strategy formulation and execution, filling the gap that traditional technology cannot dynamically optimize power management strategies based on user behavior, and effectively improving the power management efficiency and energy consumption performance of smart devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1 is a flowchart of a method for predicting a power management strategy according to some embodiments of the present invention;

[0045] Figure 2 is a flow chart of another XX method according to some embodiments of the present invention;

[0046] Figure 3 is a structural block diagram of a prediction device for a power management strategy according to an embodiment of the present invention;

[0047] Figure 4 FIG. 4 is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0049] According to an embodiment of the present invention, a method, apparatus, electronic device, and storage medium for predicting a power management strategy are provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0050] In this embodiment, a method for predicting a power management strategy is provided. Figure 1 FIG. 1 is a flow chart of a method for predicting a power management strategy according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0051] Step S101: Acquire a sequence of touch actions performed by a user on a smart device within a current time period.

[0052] In an embodiment of the present application, first, user touch data is collected in real time through sensors (such as capacitive touch sensors, pressure sensors, etc.) carried by smart devices, including original signals such as touch position, touch duration, number of clicks, and sliding trajectory. Secondly, the original signal is preprocessed, noise interference is removed by filtering algorithms, and specific touch action types (such as single click, double click, long press, sliding, etc.) are identified through threshold detection, feature extraction and other technologies. Then, the identified touch actions are serialized and stored in timestamp order to form a touch behavior sequence within the current time period.

[0053] Step S102 : predicting the touch actions that the user will perform on the smart device in a future time period based on the touch action sequence.

[0054] In an embodiment of the present application, predicting a user's touch behavior on a smart device in a future time period based on a touch behavior sequence includes:

[0055] Step A1: extracting time variation rules and touch behavior patterns from the touch behavior sequence.

[0056] Specifically, first, the time stamp data in the touch action sequence is analyzed to calculate the time intervals between adjacent touch actions (such as Δt1, Δt2, ..., Δt n ), statistically analyze the distribution characteristics of time intervals (such as mean, variance, and probability density), identify periodic patterns (such as high-frequency operations during fixed time periods each day) or trend changes (such as gradually shortening operation intervals), and form temporal change patterns. Secondly, model the type, combination sequence, and spatial position of touch behaviors, and use sequence pattern mining algorithms (such as PrefixSpan and GSP) to identify recurring subsequences (such as a fixed operation chain of "long press-slide-click"), or use clustering algorithms (such as DBSCAN) to group similar behavior patterns and extract typical touch behavior patterns (such as a sliding combination for quickly adjusting brightness, and a single-click pattern for low-frequency operations at night).

[0057] In step A2, the time variation pattern and the touch behavior pattern are input into a pre-trained prediction model, the dependency relationship between the time variation pattern and the touch behavior pattern is analyzed by the prediction model, and the touch behavior of the user on the smart device in the future time period is predicted based on the dependency relationship.

[0058] Specifically, the dependency relationship between the time change law and the touch behavior pattern is analyzed through the prediction model, and the touch behavior performed by the user on the smart device in the future time period is predicted based on the dependency relationship, including: identifying the local features of the touch behavior model at different time scales through the convolution kernels of different sizes of the temporal convolutional network in the prediction model to obtain a first eigenvector; analyzing the evolution law of the touch behavior pattern in the first eigenvector over time through the long short-term memory network in the prediction model to obtain a second eigenvector; performing a weighted operation of the attention mechanism based on the second eigenvector to obtain a weighted behavior feature; projecting the weighted behavior feature into the operation mode space through the output layer in the prediction model, and combining the dependency relationship between the time change law and the touch behavior pattern to output the touch behavior performed by the user on the smart device in the future time period.

[0059] Secondly, the preprocessed touch behavior sequence (such as time series data composed of time intervals and action type encoding) is converted into an input format suitable for convolution operation (such as a two-dimensional matrix, where rows represent time steps and columns represent feature dimensions).

[0060] Then, convolution kernels of different sizes (such as 3, 5, and 7) are deployed in parallel within a temporal convolutional network (TCN). Each kernel performs a sliding window-style local feature extraction on the input sequence: small kernels capture fine features at short time scales (such as the rapid operation pattern of continuous single clicks), while large kernels capture macro trends at long time scales (such as changes in operation frequency across time periods). After generating feature maps with different receptive fields through convolution operations, the output results of each branch are concatenated and fused to form a first feature vector containing multi-scale local features, which characterizes the pattern details and contextual associations of touch behavior at different time granularities.

[0061] Next, the first feature vector is fed into the hidden layer of a long short-term memory (LSTM) network, using the LSTM unit's gating mechanism (forget gate, input gate, and output gate) to selectively memorize historical information. At each time step, the LSTM unit updates the cell state by calculating the interaction between the current input features and past memories to preserve long-term dependencies (such as a user's frequent charging habits on weekday mornings) while outputting a hidden state vector to capture short-term behavioral changes (such as multiple sudden power mode switches within a single day).

[0062] As the sequence is gradually input, the LSTM network extracts the hierarchical representation of the temporal features layer by layer, and finally integrates the hidden state and cell state of the last time step into the second feature vector, which encodes the evolution trajectory of the touch behavior pattern over time and the long-term and short-term dependencies.

[0063] Next, the second feature vector is fed into the attention mechanism, where the importance weights of the features at each time step are calculated using a query-key-value architecture. The query vector is generated from the current feature state, while the key and value vectors are derived from the hidden states at each historical time step. The similarity score between the query and key is calculated using a dot product or neural network. After normalization using the softmax function, an attention weight matrix is ​​generated, which represents the degree to which future behavior prediction depends on the features at each historical time step.

[0064] The attention weight is multiplied and accumulated by the value vector to obtain the weighted behavioral feature. This feature highlights the key historical patterns that have a greater impact on the prediction results and suppresses irrelevant or noisy information, thereby improving the pertinence of feature representation and prediction accuracy.

[0065] Finally, a fully connected neural network is constructed at the output layer to map the weighted behavioral features from the hidden layer feature space to the operation mode space (e.g., a set of preset power operation categories, including power on, power off, power adjustment, and mode switching). An activation function (such as softmax) is used to calculate the probability distribution of each operation category over the future time period. For example, the probability of "power adjustment" is 0.6, and the probability of "mode switching" is 0.3.

[0066] The extracted temporal variation features are used as auxiliary input and integrated with the weighted behavioral features to strengthen the time constraint on behavioral patterns. Finally, the touch behavior prediction results are output based on the operation category with the highest probability distribution and time prediction (e.g., within the next 10 minutes).

[0067] As an example, Figure 2 As shown,

[0068] Step S103 : analyzing the impact data of the touch behavior on the power supply, and generating a power management policy corresponding to the power supply based on the impact data and the current performance data of the power supply.

[0069] In the embodiment of the present application, analyzing the impact of touch behavior on power supply data includes the following steps B1-B4:

[0070] Step B1: Obtain the energy efficiency characteristic curve of the power supply.

[0071] Specifically, first, test equipment (such as electronic loads and power analyzers) is used to sample the power supply's input power, output power, efficiency, and other parameters in real time under different load conditions (such as light load, rated load, and overload). The sampling frequency must cover the key time nodes of the power supply's dynamic response. Secondly, typical operating scenarios (such as standby mode, full load operation, and periodic start and stop) are simulated in a laboratory environment, and continuous data points of power supply efficiency changes with load in each scenario are recorded.

[0072] The sampled data is then processed using data fitting algorithms (such as polynomial fitting and piecewise linear fitting) to generate energy efficiency characteristic curves (such as efficiency-load ratio curves and loss-output power curves). Standardization is then performed to ensure comparability across different power supply models. Finally, the energy efficiency characteristic curves are stored as parameter tables or mathematical expressions for subsequent energy consumption calculations.

[0073] Step B2: Calculate the power consumption distribution based on the energy efficiency characteristic curve to obtain energy consumption impact data corresponding to the touch behavior.

[0074] Specifically, the user's touch behavior sequence is mapped to the actual working state changes of the power supply (such as single-clicking the power button corresponds to power on / off, and sliding operation corresponds to power adjustment), and the behavior-load-energy consumption mapping relationship is established in combination with the energy efficiency characteristic curve. For example, when the user performs the "adjust power" operation, the corresponding efficiency value is queried from the curve according to the current set power value, and the input power (= output power / efficiency) and the energy consumption accumulation rate are calculated. Secondly, the load state triggered by each touch behavior is segmented and integrated according to the time series, and the power consumption during the duration of each behavior is calculated (such as kilowatt-hours, kWh), and the energy consumption proportion of different behavior types is counted (such as the power on / off operation accounts for 15% of energy consumption, and the power adjustment operation accounts for 60%). Finally, the energy consumption data of each touch behavior is associated with the behavior sequence timestamp to form a three-dimensional impact data matrix of "behavior-time-energy consumption", which is used to analyze the real-time impact of different operations on power supply energy consumption.

[0075] Step B3: Evaluate the impact of the touch behavior on the health of the power supply based on the energy consumption impact data and the environmental impact data.

[0076] Specifically, we first extract key indicators from the energy consumption impact data (such as peak power, cumulative energy consumption, and load fluctuation frequency). Combined with physical aging parameters from the power health model (such as capacitor life and number of semiconductor device thermal cycles), we establish a mapping relationship between energy consumption and device loss. For example, high-frequency start-stop operations increase the number of capacitor charge and discharge cycles, and the equivalent aging time can be calculated using the Arrhenius model.

[0077] Secondly, environmental impact data (such as ambient temperature, humidity, and dust concentration) are introduced as correction factors, and the superimposed effect of environmental stress on power supply health is evaluated through multivariate regression analysis (for example, the same energy consumption operation in a high temperature environment will accelerate device degradation).

[0078] Then, using the hierarchical analysis method or fuzzy comprehensive evaluation method, the energy consumption indicators and environmental indicators are weighted and integrated into a health impact index (such as 0-100 points, where the higher the score, the greater the health loss), and the main sources of loss are located (such as a user frequently adjusting the power, resulting in a converter loss accounting for 70%).

[0079] Finally, a health impact data report is generated, recording information such as the degree of device wear and remaining life prediction corresponding to each touch behavior.

[0080] Step B4: Calculate impact data based on the energy consumption impact data and the health impact data.

[0081] Specifically, the impact data is calculated based on the energy consumption impact data and the health impact data, including the following steps: based on the power consumption distribution and time series in the energy consumption impact data, the cumulative energy consumption and the energy consumption fluctuation coefficient under different operating modes are calculated to obtain the quantitative characteristic vector of the energy consumption impact; based on the energy consumption impact data and the health impact data, a multi-dimensional correlation analysis is performed to construct an energy consumption-health coupling model to obtain the correlation characteristic vector of touch behavior and power supply performance degradation; and the quantitative characteristic vector and the correlation characteristic vector are used to perform weighted calculation to obtain the impact data.

[0082] First, the energy consumption impact data is classified and aggregated according to the operating mode (such as power on, power off, power adjustment, etc.), and the power consumption time series in each mode is integrated to obtain the cumulative energy consumption of the mode (unit: kWh). For example, the power change curve of the "power adjustment" operation each time it is triggered in the past week is counted, and the total energy consumption is accumulated. Secondly, the fluctuation coefficient of power consumption under each operating mode is calculated, and the degree of energy consumption fluctuation is measured by the ratio of the standard deviation to the mean (i.e., the coefficient of variation). The formula is: Fluctuation coefficient = μσ. Among them, σ is the standard deviation of power consumption, and μ is the mean. A high fluctuation coefficient indicates that power adjustment is frequent in this mode (such as the user quickly switches the power gear multiple times).

[0083] Then, the indicators such as cumulative energy consumption, fluctuation coefficient and operating frequency are normalized and combined into a multi-dimensional vector to form a quantitative feature vector of energy consumption impact. For example, [0.8, 0.6, 0.3] corresponds to high cumulative energy consumption, medium fluctuation coefficient and low operating frequency respectively.

[0084] Key variables from the energy consumption impact data (such as cumulative energy consumption and peak load) and degradation indicators from the health impact data (such as the increase in capacitor equivalent series resistance (ESR) and the fluctuation in semiconductor junction temperature) were extracted. Correlations and causal relationships between these variables were analyzed using the Pearson correlation coefficient or Granger causality test. For example, a significant positive correlation was found for every 100 kWh increase in cumulative energy consumption, an increase in capacitor ESR by 0.05 Ω. Next, a coupled model was constructed using a structural equation model (SEM) or random forest algorithm. The inputs were touch behavior characteristics (such as the sequence of operating modes) and energy consumption indicators, and the output was a predicted health degradation value (such as the power supply efficiency drop rate). During model training, parameters were optimized through cross-validation to identify key influencing paths (such as high-frequency, high-energy-consumption fluctuation operation → device thermal fatigue → accelerated efficiency degradation). Finally, the contribution weights of each input variable to health degradation were extracted from the model to form an associated feature vector. For example, [0.4, 0.3, 0.2, 0.1] correspond to the impact weights of cumulative energy consumption, fluctuation coefficient, operating frequency, and ambient temperature, respectively.

[0085] Afterwards, the quantified feature vector of energy consumption impact is dimensionally aligned with the associated feature vector, and the comprehensive impact data is calculated using the weighted summation formula:

[0086] Impact data = λ × ∑ (quantitative feature × energy consumption weight) + (1-λ) × ∑ (quantitative feature × health weight), weight coefficient λ (e.g., λ = 0.6 means that the energy consumption impact accounts for 60%).

[0087] In an embodiment of the present application, generating a power management policy corresponding to a power supply based on the impact data and the current performance data of the power supply includes the following steps C1-C3:

[0088] Step C1: perform feature fusion on the impact data and the current performance data of the power supply to construct a power supply status assessment vector.

[0089] Specifically, current power supply performance data is collected, including real-time operating parameters (such as input voltage / current, output power, efficiency, temperature, and fan speed), device status data (such as capacitor capacity decay rate and MOSFET on-resistance change), and historical maintenance records (such as component replacement time and fault alarm logs). Secondly, impact data (such as the comprehensive energy consumption-health impact value and operating mode correlation features) are temporally and spatially aligned with the performance data, and dimensional differences are eliminated through data normalization (Z-score normalization) or normalization. Then, feature concatenation (such as concatenating the impact data vector with the performance indicator vector end-to-end) or feature cross-linking (such as calculating the product of the impact value and temperature) is used to fuse them, forming a multidimensional vector that includes historical behavioral impacts, real-time operating status, and device health trends, such as [0.7 (impact value), 0.9 (efficiency), 0.8 (temperature), 0.6 (capacitor status)]. Finally, principal component analysis (PCA) or autoencoder (AE) is used to reduce the dimensionality of the high-dimensional vector, retaining key features and suppressing noise, to generate a power supply status assessment vector.

[0090] Step C2: Using a preset rule base to judge the power state evaluation vector, a candidate strategy set that meets the safety constraint conditions is obtained.

[0091] Specifically, a preset rule base is established, containing hard constraints for safe power supply operation (such as output voltage fluctuations must not exceed ±5% and device temperature must not exceed the threshold Tmax) and optimization-oriented rules (such as efficiency must be above 90% in energy-saving mode and the load fluctuation coefficient must be below 0.3 in health-first mode). Rules are stored as logical expressions or decision tree nodes, for example, "IF temperature > 0.8 × Tmax THEN prohibit high-load operation." Next, the power supply status evaluation vector is input into the rule engine, and rule conditions are matched one by one. For numerical indicators, threshold comparison is used to determine whether the rule is triggered (for example, when the temperature component of the evaluation vector is > 0.8, the high-temperature warning rule is triggered). For behavioral characteristics, pattern matching is used to identify risky operations (for example, the "inrush current limit" rule in the rule base corresponds to continuous high-frequency start-stop). Then, all operational strategies that do not violate safety constraints are selected as candidate sets. For example, when the evaluation vector indicates capacitor state degradation, the rule base automatically excludes the "continuous rapid charge and discharge" strategy and retains the "low-frequency stable operation" strategy. Finally, the candidate strategies are initially filtered to eliminate obviously inefficient or redundant options, resulting in a streamlined list of feasible strategies.

[0092] In step C3, the cumulative benefits of each strategy in the candidate strategy set under different scenarios are evaluated, and the strategy with the best cumulative benefit is used as the power management strategy.

[0093] First, define the scenario division dimensions (such as time scenario - daytime / nighttime, load scenario - light load / heavy load, environmental scenario - high temperature / normal temperature), and preset benefit evaluation indicators for each scenario (such as the energy consumption reduction rate as an indicator in the energy-saving scenario, and the device loss reduction as an indicator in the health scenario). Secondly, for each candidate strategy, simulate the strategy execution effect in each scenario through simulation models or historical data: for example, in the "light load + high temperature at night" scenario, simulate the impact of the "automatic frequency reduction strategy" on energy consumption (reduction by 15%) and temperature (reduction by 8°C), and calculate the cumulative benefit value based on the benefit function benefit = 0.6 × energy consumption reduction rate + 0.4 × temperature improvement rate. Then, use multi-attribute decision-making methods (such as TOPSIS and hierarchical analysis method AHP) to comprehensively rank the benefit values ​​across scenarios, considering the robustness of the strategy (such as the benefit fluctuation range in different scenarios) and long-term impact (such as the cumulative effect of delayed device aging). Finally, the strategy with the highest cumulative benefit and that meets the scenario priority (such as the default priority execution of the health protection strategy) is selected as the final power management strategy. For example, in high-temperature and heavy-load scenarios, the "derated operation + forced air cooling" strategy is preferred. Its comprehensive benefit value is 20% higher than other strategies and complies with safety rules.

[0094] Step S104: performing corresponding power management operations on the power supply according to the power management policy.

[0095] In an embodiment of the present application, first, the selected power management policy content is parsed to clarify the operation type and parameter configuration (such as adjusting the output voltage to a specific gear, switching the working mode to energy-saving mode, starting fan heat dissipation threshold adjustment, etc.).

[0096] Then, instructions are sent to the power supply's microcontroller (MCU) or digital signal processor (DSP) through the power supply's internal control bus (such as I2C, SPI) or external interface (such as RS485, USB). The instructions must comply with the preset communication protocol (such as ModbusRTU, CANopen) to ensure accurate transmission of the instructions.

[0097] Then, the power supply hardware module performs specific operations according to the instructions:

[0098] For voltage regulation, the power conversion circuit parameters are adjusted through a digital potentiometer or pulse width modulation (PWM) controller; for mode switching, a relay or solid-state switch is triggered to change the topology; for cooling strategy, the fan speed or start / stop status is adjusted.

[0099] During this process, circuit feedback signals (such as voltage / current sampling values ​​and temperature sensor data) are monitored in real time, and closed-loop control algorithms (such as PID control) are used to ensure that the power supply status remains stable within the target range after the operation. Simultaneously, operation logs are recorded, including the policy name, execution time, parameter changes, and real-time status data, for subsequent energy efficiency analysis and fault tracing. If an anomaly (such as overcurrent or overtemperature) is detected during operation, pre-set safety protection mechanisms (such as soft start and automatic current limiting) are immediately triggered, and an alarm is sent to the system monitoring platform to ensure safe operation of the power supply.

[0100] The embodiment of the present application obtains the touch behavior sequence of the user in the current time period and predicts future touch behavior based on this sequence. Compared with traditional touch behavior prediction technology, it no longer relies on simple statistics or preset rules, but is based on the user's real behavior data, and can more accurately analyze complex and changeable user behavior patterns. At the same time, by analyzing the impact data of touch behavior on the power supply and generating a power management strategy based on the current performance data of the power supply, the traditional static power management method that lacks dynamic response is changed. Finally, operations are executed according to the strategy, realizing full-link dynamic management from behavior prediction to strategy formulation and execution, filling the gap that traditional technology cannot dynamically optimize power management strategies based on user behavior, and effectively improving the power management efficiency and energy consumption performance of smart devices.

[0101] In this embodiment, a power management strategy prediction device is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments. The details that have been described will not be repeated here. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.

[0102] This embodiment provides a prediction device for power management strategy, such as Figure 3 As shown, including:

[0103] The acquisition module 301 is used to acquire a sequence of touch actions performed by a user on a smart device within a current time period;

[0104] Prediction module 302, configured to predict the touch behavior that the user will perform on the smart device in a future time period based on the touch behavior sequence;

[0105] An analysis module 303 is configured to analyze the impact of touch behavior on the power supply, and generate a power management policy corresponding to the power supply based on the impact data and the current performance data of the power supply;

[0106] The execution module 304 is configured to execute corresponding power management operations on the power supply according to the power management policy.

[0107] In the embodiment of the present application, the prediction module 302 is configured to extract a temporal variation pattern and a touch behavior pattern from a touch behavior sequence; input the temporal variation pattern and the touch behavior pattern into a pre-trained prediction model; analyze the dependency relationship between the temporal variation pattern and the touch behavior pattern through the prediction model; and predict the user's touch behavior on the smart device in a future time period based on the dependency relationship;

[0108] In an embodiment of the present application, the prediction module 302 is used to identify the local features of the touch behavior model at different time scales through convolution kernels of different sizes of the temporal convolutional network in the prediction model to obtain a first eigenvector; analyze the evolution of the touch behavior pattern in the first eigenvector over time through the long short-term memory network in the prediction model to obtain a second eigenvector; perform an attention mechanism weighted operation based on the second eigenvector to obtain a weighted behavior feature; project the weighted behavior feature into the operation mode space through the output layer in the prediction model, and output the touch behavior performed by the user on the smart device in the future time period in combination with the time change law and the dependency relationship between the touch behavior pattern.

[0109] In an embodiment of the present application, the analysis module 303 is used to obtain an energy efficiency characteristic curve of a power supply; calculate the power consumption distribution of the power supply based on the energy efficiency characteristic curve to obtain energy consumption impact data corresponding to the touch behavior; evaluate the health impact data of the touch behavior on the power supply based on the energy consumption impact data and the environmental impact data; and calculate the impact data based on the energy consumption impact data and the health impact data.

[0110] In an embodiment of the present application, based on the power consumption distribution and time series in the energy consumption impact data, the cumulative energy consumption and energy consumption fluctuation coefficient under different operating modes are calculated to obtain a quantitative characteristic vector of the energy consumption impact; a multi-dimensional correlation analysis is performed based on the energy consumption impact data and the health impact data, and an energy consumption-health coupling model is constructed to obtain a correlation characteristic vector between touch behavior and power supply performance degradation; and a weighted calculation is performed using the quantitative characteristic vector and the correlation characteristic vector to obtain the impact data.

[0111] In an embodiment of the present application, the analysis module 303 is used to perform feature fusion of the impact data and the current performance data of the power supply to construct a power supply status evaluation vector; use a preset rule base to judge the power supply status evaluation vector to obtain a set of candidate strategies that meet the safety constraints; evaluate the cumulative benefits of each strategy in the candidate strategy set under different scenarios, and use the strategy with the best cumulative benefits as the power management strategy.

[0112] In an embodiment of the present application, the analysis module 303 is used to substitute each strategy in the candidate strategy set into various scenarios in the scenario library, simulate the operation of the power supply in the future, and calculate the power consumption and health changes of the power supply after the strategy is executed during the simulation of the power supply operation; for each scenario, a weighted sum is taken based on the power consumption and health changes to obtain the single-scenario benefit; the single-scenario benefits of each strategy in all scenarios are accumulated to obtain the cumulative benefits of each strategy in different scenarios.

[0113] See also Figure 4 , Figure 4 is a structural diagram of an electronic device provided by an optional embodiment of the present invention, such as Figure 4As shown, the electronic device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the electronic device, including instructions stored in or on the memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple electronic devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system).

[0114] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0115] The memory 20 stores instructions that can be executed by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0116] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created based on the use of an electronic device presented by a small program landing page, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0117] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0118] The electronic device further includes a communication interface 30 for the electronic device to communicate with other devices or a communication network.

[0119] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0120] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for predicting a power management strategy, characterized in that: The method comprises: Get the sequence of touch actions performed by the user on the smart device during the current time period; predicting the touch behavior performed by the user on the smart device in a future time period based on the touch behavior sequence; Analyzing data on the impact of the touch behavior on the power supply of the smart device, and generating a power management policy corresponding to the power supply based on the impact data and current performance data of the power supply; Perform corresponding power management operations on the power supply according to the power management policy.

2. The method according to claim 1, characterized in that The predicting, based on the touch behavior sequence, the touch behavior performed by the user on the smart device in a future time period includes: extracting time variation rules and touch behavior patterns from the touch behavior sequence; The time change law and the touch behavior pattern are input into a pre-trained prediction model, the dependency relationship between the time change law and the touch behavior pattern is analyzed by the prediction model, and the touch behavior performed by the user on the smart device in a future time period is predicted based on the dependency relationship.

3. The method according to claim 2, characterized in that Analyzing the dependency between the time variation rule and the touch behavior pattern through the prediction model, and predicting the touch behavior performed by the user on the smart device in a future time period based on the dependency, includes: Identifying local features of the touch behavior model at different time scales using convolution kernels of different sizes of the temporal convolutional network in the prediction model to obtain a first feature vector; Analyzing the temporal evolution of the touch behavior pattern in the first feature vector using a long short-term memory network in the prediction model to obtain a second feature vector; Performing an attention mechanism weighted operation based on the second feature vector to obtain a weighted behavior feature; The weighted behavior features are projected into the operation pattern space through the output layer in the prediction model, and the touch behavior performed by the user on the smart device in the future time period is output by combining the time change law and the dependency relationship between the touch behavior pattern.

4. The method according to claim 1, wherein The analyzing the data of the impact of the touch behavior on the power supply of the smart device includes: Obtaining an energy efficiency characteristic curve of the power supply; Calculating the power consumption distribution of the power supply based on the energy efficiency characteristic curve to obtain energy consumption impact data corresponding to the touch behavior; Evaluating health impact data of the touch behavior on the power supply based on the energy consumption impact data and the environmental impact data; The impact data is calculated based on the energy consumption impact data and the health impact data.

5. The method according to claim 4, characterized in that The calculating the impact data according to the energy consumption impact data and the health impact data includes: Calculating the cumulative energy consumption and energy consumption fluctuation coefficient under different operation modes based on the power consumption distribution and time series in the energy consumption impact data to obtain a quantitative feature vector of the energy consumption impact; Performing a multi-dimensional correlation analysis based on the energy consumption impact data and the health impact data, constructing an energy consumption-health coupling model, and obtaining a correlation feature vector between touch behavior and power supply performance degradation; The influence data is obtained by performing weighted calculation using the quantized feature vector and the associated feature vector.

6. The method according to claim 1, characterized in that Generating a power management policy corresponding to the power supply based on the impact data and the current performance data of the power supply includes: Performing feature fusion on the impact data and the current performance data of the power supply to construct a power supply status assessment vector; Using a preset rule base to judge the power state evaluation vector, a candidate strategy set that meets the safety constraint conditions is obtained; The cumulative benefit of each strategy in the candidate strategy set under different scenarios is evaluated, and the strategy with the best cumulative benefit is used as the power management strategy.

7. The method according to claim 6, characterized in that The evaluating the cumulative benefits of each strategy in the candidate strategy set in different scenarios includes: Substitute each strategy in the candidate strategy set into various scenarios in the scenario library to simulate the operation of the power supply in the future. During the simulation of the power supply operation, calculate the energy consumption and health changes of the power supply after the strategy is executed; For each scenario, a weighted sum is performed based on the power consumption and the health change to obtain a single scenario benefit; Based on the accumulation of the single-scenario benefits of each strategy in all scenarios, the cumulative benefits of each strategy in different scenarios are obtained.

8. A prediction device for power management strategy, characterized in that: The device comprises: An acquisition module, used to acquire a sequence of touch actions performed by a user on a smart device within a current time period; A prediction module, configured to predict, based on the touch behavior sequence, the touch behavior that the user will perform on the smart device within a future time period; an analysis module, configured to analyze data on the impact of the touch behavior on the power supply of the smart device, and generate a power management policy corresponding to the power supply based on the impact data and current performance data of the power supply; An execution module is used to perform corresponding power management operations on the power supply according to the power management policy.

9. An electronic device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images