Power management method, device and equipment for cloud mobile phone and medium

Abstract

The invention relates to the technical field of computers, in particular to a power management method and device for a cloud mobile phone, equipment and a medium. The current load data of the computing device is obtained by detecting the operation behavior of the user for the cloud mobile phone in real time, and the limitation of a traditional static power management strategy is broken through. Due to the fact that the user operation can be captured in real time, high-load game operation and low-load text browsing can be perceived in the first time; and meanwhile, the system operation state is comprehensively mastered by combining equipment load data. On the basis, the two types of data are analyzed, the association between the operation behavior and the load change is accurately identified, and then the power supply operation parameters of the computing equipment are dynamically adjusted. Compared with a traditional method, the scheme has the advantages that dynamic perception and response to user behaviors and load changes are achieved, power source resources can be flexibly and accurately distributed according to actual requirements, resource supply can be guaranteed during high-performance tasks, a low-power-consumption mode can be switched during low loads, and the power source utilization efficiency is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a power management method, device, equipment and medium for a cloud phone. Background Art

[0002] With the rapid development of cloud computing and edge computing technologies, cloud phones are becoming mainstream computing devices due to their convenience and flexibility. However, due to their frequent deployment in low-power devices or edge computing environments, they face bottlenecks in battery capacity and performance. Therefore, achieving efficient energy management and extending battery life have become urgent technical challenges.

[0003] Traditional power management methods often employ static strategies, lacking the ability to dynamically perceive and respond to real-time user behavior and system load changes, making it impossible to accurately and flexibly allocate power resources. For example, when processing high-performance tasks, power resources are overly concentrated, resulting in energy waste. However, in low-load tasks, the device maintains a high-power operating mode, failing to fully utilize the energy-saving advantages of low-power mode. These shortcomings result in insufficient battery life for cloud phones, causing users to frequently experience battery anxiety, which seriously impacts the device's user experience and application promotion. Summary of the Invention

[0004] In view of this, an embodiment of the present invention provides a power management method, device, equipment and medium for a cloud phone to solve the problem that traditional power management methods mostly adopt static strategies, lack dynamic perception and response capabilities to users' real-time operating behaviors and system load changes, and cannot achieve accurate and flexible allocation of power resources.

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

[0006] Detect the user's operation behaviors triggered on the cloud phone in the current time period;

[0007] Obtaining first load data of a computing device deployed by the cloud phone in the current time period;

[0008] The operating behavior and the first load data are analyzed, an operating parameter affecting the power supply of the computing device is adjusted, and the computing device is controlled to operate according to the operating parameter.

[0009] Furthermore, analyzing the operation behavior and the first load data and adjusting an operating parameter affecting the power supply of the computing device includes:

[0010] extracting key behavioral features from the operating behavior;

[0011] predicting second load data of the computing device in a future time period based on the key behavior feature and the first load data;

[0012] predicting a power consumption trend of the computing device in a future time period based on the key behavior characteristics and the first load data;

[0013] Determining an adjustment indicator affecting the power supply of the computing device by using the second load data and the power consumption trend;

[0014] The operating parameters of the computing device are adjusted according to the adjustment index.

[0015] Furthermore, predicting second load data of the computing device in a future time period based on the key behavior feature and the first load data includes:

[0016] Extracting key indicators from the first load data;

[0017] Constructing a time series data set using the key indicators and the key behavioral features;

[0018] Performing a correlation analysis on the key indicators and key behavior features in the time series data set according to the time series to obtain a behavior-load causal relationship model, wherein the behavior-load causal relationship model is used to characterize the influence weight and time lag effect of different operating behaviors on future load changes;

[0019] Output second load data of the computing device in a future time period according to the obtained behavior-load causal relationship model.

[0020] Furthermore, predicting the power consumption trend of the computing device in a future time period based on the key behavior characteristics and the first load data includes:

[0021] Normalizing the key behavior features and the first load data, and generating a fused feature vector by a feature fusion method;

[0022] Inputting the fused feature vector into a power consumption trend prediction model to obtain a spatiotemporal dependency relationship between the key behavior feature and the first load data;

[0023] The spatiotemporal dependency is separated through an attention mechanism to obtain decoupled power consumption related features, and the power consumption trend in the future time period is predicted based on the power consumption related features.

[0024] Furthermore, the determining, by utilizing the second load data and the power consumption trend, an adjustment indicator affecting the power supply of the computing device includes:

[0025] Time-aligning the second load data with the power consumption trend, analyzing the correlation between the two, and obtaining a correlation analysis result between the load and power consumption;

[0026] Constructing a power supply impact weight matrix based on the correlation analysis result, wherein the power supply impact weight matrix is ​​used to reflect the contribution of each load component to power consumption;

[0027] Designing a multi-objective optimization function according to the power management objectives, taking the power impact weight matrix as a constraint condition, and obtaining an optimization objective function with constraint conditions;

[0028] Solve the optimization objective function with constraints through the optimization algorithm to obtain the adjustment index that affects the power supply of the computing device

[0029] Furthermore, after controlling the computing device to operate according to the operating parameters, the method further includes:

[0030] Collect key operating data of the power supply in real time and obtain historical failure modes of the power supply;

[0031] extracting correlation features associated with the historical failure mode from the key operating data;

[0032] Inputting the associated features into a fault prediction model to output an early warning signal of power supply failure risk;

[0033] A corresponding power supply warning operation is performed based on the warning signal.

[0034] Furthermore, inputting the associated features into a fault prediction model and outputting a warning signal of power supply failure risk includes:

[0035] Constructing a feature vector of the multidimensional data based on the associated features, and fine-tuning the parameters of the fault prediction model based on the feature vector to adapt it to the characteristics of the power supply;

[0036] Performing forward calculation on the characteristic vector according to the adjusted fault prediction model to obtain a fault risk probability value;

[0037] The fault risk probability value is compared with the preset threshold value, and a warning signal of power supply failure risk is output when the threshold value is exceeded.

[0038] In a second aspect, an embodiment of the present invention provides a power management device for a cloud phone, the device comprising:

[0039] The detection module is used to detect the user's operation behavior triggered on the cloud phone in the current time period;

[0040] An acquisition module, configured to acquire first load data of the computing device deployed by the cloud phone in the current time period;

[0041] An analysis module is used to analyze the operation behavior and the first load data, adjust the operating parameters that affect the power supply of the computing device, and control the computing device to operate according to the operating parameters.

[0042] In a third aspect, an embodiment of the present invention provides a computer 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.

[0043] 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.

[0044] The embodiment of the present application breaks through the limitations of traditional static power management strategies by detecting the user's operating behavior on the cloud phone in real time and obtaining the current load data of the computing device. Due to the ability to capture user operations in real time, whether it is high-load game operation or low-load text browsing, it can be perceived at the first time; at the same time, combined with the device load data, the system operation status can be fully grasped. On this basis, the two types of data are analyzed to accurately identify the relationship between operating behavior and load changes, and then dynamically adjust the power operation parameters of the computing device. Compared with traditional methods, this solution realizes dynamic perception and response to user behavior and load changes, and can flexibly and accurately allocate power resources according to actual needs. It can not only ensure resource supply during high-performance tasks, but also switch to low-power mode under low load, effectively improving power utilization efficiency, extending the battery life of cloud phones, and significantly improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] 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.

[0046] Figure 1 is a flowchart of a power management method for a cloud phone according to some embodiments of the present invention;

[0047] Figure 2 is a flowchart of another power management method for a cloud phone according to some embodiments of the present invention;

[0048] Figure 3 This is a structural block diagram of a power management device for a cloud phone according to an embodiment of the present invention;

[0049] Figure 4 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] 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.

[0051] According to an embodiment of the present invention, a power management method, apparatus, device and medium for a cloud phone 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.

[0052] In this embodiment, a power management method for a cloud phone is provided. Figure 1 is a flow chart of a power management method for a cloud phone according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0053] Step S101: Detect the user's operation behavior triggered on the cloud phone within the current time period.

[0054] In an embodiment of the present application, a behavior monitoring module is embedded in the cloud phone system, and system hook functions, event monitoring APIs and other technologies are used to capture in real time the operation events generated by the user's interaction with the cloud phone. For touch behaviors such as clicks and slides, quantification is performed by recording information such as screen coordinates, trigger time, and operation duration; for behaviors such as application startup and data input, relevant data is obtained through system log parsing and process monitoring. At the same time, in order to avoid data redundancy, a time window (such as 1 second / time) is set for data sampling, and finally a user behavior data set containing dimensions such as operation type, trigger time, and operation frequency is formed.

[0055] Step S102: Obtain first load data of the computing device deployed by the cloud phone in the current time period.

[0056] In an embodiment of the present application, core load indicators such as CPU utilization, memory occupancy, disk I / O throughput, and network bandwidth usage are collected periodically (e.g., once per second) through the system management interface of the computing device (e.g., procfs and sysfs file systems under Linux, and WMI services under Windows). For special hardware such as GPU and FPGA, the performance query interface of the corresponding driver is called to obtain its load data. To ensure data accuracy, the collected raw data is denoised (e.g., removing abnormal spikes) and structured stored by timestamp to generate a real-time data sequence of the device load status in the current time period.

[0057] Step S103 : analyzing the operation behavior and the first load data, adjusting the operation parameters that affect the power supply of the computing device, and controlling the computing device to operate according to the operation parameters.

[0058] In an embodiment of the present application, analyzing the operation behavior and the first load data and adjusting the operating parameters affecting the power supply of the computing device include the following steps A1-A5:

[0059] Step A1: extract key behavior features from the operation behavior.

[0060] Step A2: predicting second load data of the computing device in a future time period based on the key behavior characteristics and the first load data.

[0061] Specifically, predicting the second load data of the computing device in a future time period based on the key behavior characteristics and the first load data includes: extracting key indicators in the first load data; constructing a time series data set using the key indicators and key behavior characteristics; performing correlation analysis on the key indicators and key behavior characteristics in the time series data set according to the time series to obtain a behavior-load causal relationship model, wherein the behavior-load causal relationship model is used to characterize the influence weights and time lag effects of different operating behaviors on future load changes; and outputting the second load data of the computing device in the future time period according to the obtained behavior-load causal relationship model.

[0062] System monitoring tools are used to collect hardware status data of computing devices in real time, extracting basic metrics such as CPU utilization (broken down to the user-mode / kernel-mode ratio), memory usage, disk I / O throughput, and network bandwidth occupancy. Furthermore, based on the characteristics of cloud phone services, additional key metrics such as GPU rendering frame rate, virtual CPU context switching frequency, and video memory bandwidth are extracted. The collected raw data is filtered with a sliding window (5-minute window size) to remove instantaneous outliers, and data is sampled at 10-second intervals, ultimately forming a standardized indicator vector containing 12 dimensions.

[0063] The extracted key metrics are aligned with user behavior features (such as click frequency, swipe speed, and app switching) by timestamp to construct a three-dimensional tensor data structure: [timestep, number of samples, feature dimension]. Data is segmented using a fixed time window (30 minutes), with each window containing 180 timesteps (one sample every 10 seconds). Behavioral features are one-hot encoded (such as operation type) or normalized (such as swipe distance), and temporal features (such as weekday / weekend, daytime / nighttime) are added to enhance temporal regularity. The resulting training dataset has a shape of [1000, 180, 20], containing 1000 time window samples.

[0064] The Granger causality test is used to determine the causal relationship between operational behavior and load indicators. The maximum lag order is set to 12 (i.e., 2 minutes), and significant causal relationship pairs with a p-value < 0.05 are screened (such as "application startup event → CPU load increase"). A time series prediction model is constructed using a gated recurrent unit (GRU). The input layer dimension is 20 (including 12 load indicators and 8 behavioral features), the hidden layer is set to 3 layers with 64 neurons each, and the output layer predicts the load changes in the next 30 minutes. The attention mechanism is introduced during the training process to automatically learn the weight of the impact of different operational behaviors on each load indicator. Finally, the time lag effect is quantified through the model parameter matrix (such as a GPU load peak 3 seconds after the user clicks).

[0065] The behavioral feature vector at the current moment is input into the trained GRU model, and a load change sequence for the next 30 minutes is generated through rolling prediction. Monte Carlo simulation is performed on the prediction results (1000 samples), and the 95% confidence interval of the load value at each time point is calculated. The probability distribution prediction containing the mean, standard deviation, and upper and lower bounds is output. Anomaly detection is performed on key indicators (such as CPU peak load). If the predicted value exceeds 120% of the historical threshold, an early warning is triggered and it is marked as a high-risk period. The final generated second load data is output in JSON format, containing multi-dimensional load prediction values ​​and their confidence levels for 180 time steps.

[0066] Step A3: predicting the power consumption trend of the computing device in a future time period based on the key behavior characteristics and the first load data.

[0067] Specifically, the power consumption trend of the computing device in a future time period is predicted based on the key behavior characteristics and the first load data, including: normalizing the key behavior characteristics and the first load data, and generating a fused feature vector through a feature fusion method; inputting the fused feature vector into a power consumption trend prediction model to obtain the spatiotemporal dependency relationship between the key behavior characteristics and the first load data; separating the spatiotemporal dependency relationship through an attention mechanism to obtain decoupled power consumption related features, and predicting the power consumption trend in the future time period based on the power consumption related features.

[0068] First, the key behavioral features extracted from the operational behavior and the first load data of the computing device in the current time period are normalized to eliminate dimensional differences between the different features and bring them into the same numerical range. Next, feature fusion methods (such as feature concatenation, weighted summation, or feature mapping using a fully connected layer) are used to fuse the normalized key behavioral features and the first load data into a unified feature vector. This fused feature vector contains comprehensive information about the operational behavior and the current load, providing a more comprehensive data foundation for subsequent analysis.

[0069] The fused feature vector is fed into a pre-built power consumption trend prediction model (such as an LSTM- or GRU-based time series prediction model). The model uses internal gating mechanisms and memory cells to capture the dynamic changes in key behavioral features and initial load data at different time steps, as well as the spatiotemporal dependencies between them. This dependency reflects how operational behavior and load changes jointly affect power consumption, providing a key basis for predicting future power consumption trends.

[0070] The attention mechanism is used to separate the spatiotemporal dependencies learned by the model. It automatically focuses on the feature dimensions or time steps that have the greatest impact on power consumption, thereby generating decoupled power consumption-related features. These features focus more on information directly related to power consumption, removing redundancy and noise.

[0071] Finally, based on the decoupled power consumption related features, regression analysis, classification methods or other prediction algorithms are used to predict the power consumption trend in future time periods, such as increase, decrease or stability, to provide decision support for power management.

[0072] Step A4: Determine an adjustment index affecting the power supply of the computing device by using the second load data and the power consumption trend.

[0073] In an embodiment of the present application, the second load data and the power consumption trend are used to determine an adjustment index that affects the power supply of the computing device, including: time-aligning the second load data with the power consumption trend, analyzing the correlation between the two, and obtaining a correlation analysis result between the load and power consumption; constructing a power impact weight matrix based on the correlation analysis result, wherein the power impact weight matrix is ​​used to reflect the contribution of each load component to the power consumption; designing a multi-objective optimization function according to the power management goal, using the power impact weight matrix as a constraint condition, and obtaining an optimization objective function with constraints; solving the optimization objective function with constraints through an optimization algorithm to obtain an adjustment index that affects the power supply of the computing device.

[0074] First, the predicted second load data is aligned with the power consumption trend forecast results in the time dimension to ensure that the two are compared within the same time window. Next, statistical analysis methods (such as Pearson correlation coefficient, mutual information, etc.) or visualization methods (such as line chart comparison) are used to analyze the correlation between load changes and power consumption trends, identifying the synchronous changes in power consumption when the load increases or decreases. This results in a correlation analysis between load and power consumption, providing a basis for subsequent weight allocation.

[0075] Based on the correlation analysis results, determine the impact of different load components (such as CPU load, memory load, and network load) on power consumption. Using quantitative methods (such as regression coefficients and feature importance scores) or qualitative assessments, assign a weight to each load component and construct a power impact weight matrix. This matrix clearly reflects the contribution of each load component to power consumption, with higher weights indicating a greater impact.

[0076] A multi-objective optimization function is designed, combining power management objectives (such as minimizing energy consumption and balancing performance and power consumption). This function typically incorporates multiple optimization objectives (such as minimizing power consumption and maximizing computing performance), and multi-objective processing is performed using methods such as weighted summation and the Pareto front. Furthermore, the power impact weight matrix is ​​introduced as a constraint into the optimization function to ensure that the impact of each load component on power consumption is fully considered during the optimization process, resulting in a constrained optimization objective function.

[0077] Optimization algorithms (such as genetic algorithms, particle swarm optimization, and linear programming) are used to solve the constrained optimization objective function. While satisfying the constraints, the optimization algorithm seeks a combination of adjustment indicators that achieves the optimal (or satisfactory) solution for the multi-objective optimization function. These adjustment indicators may include CPU frequency, voltage regulation parameters, and task scheduling strategies, and are used to adjust the operating parameters of the computing device to achieve power management goals. The final output of these adjustment indicators is the optimization result that affects the computing device's power supply.

[0078] Step A5: Adjust the operating parameters of the computing device according to the adjustment index.

[0079] Specifically, the specific implementation process of adjusting the operating parameters of the computing device according to the adjustment indicators is: after obtaining the adjustment indicators obtained based on the power management target optimization (such as CPU frequency, voltage regulation parameters, memory allocation strategy, task priority setting, etc.), these adjustment indicators are converted into specific control instructions through the device management system or control interface, and sent to the hardware controller or operating system level of the computing device, which dynamically adjusts the CPU clock frequency, power supply voltage, memory usage threshold, task scheduling strategy and other operating parameters to ensure that the computing device meets business needs while operating according to the optimized parameters to reduce power consumption or improve energy efficiency. The whole process involves the collaborative work of multiple links such as parameter analysis, instruction generation, execution and effect feedback.

[0080] The embodiment of the present application breaks through the limitations of traditional static power management strategies by detecting the user's operating behavior on the cloud phone in real time and obtaining the current load data of the computing device. Due to the ability to capture user operations in real time, whether it is high-load game operation or low-load text browsing, it can be perceived at the first time; at the same time, combined with the device load data, the system operation status can be fully grasped. On this basis, the two types of data are analyzed to accurately identify the relationship between operating behavior and load changes, and then dynamically adjust the power operation parameters of the computing device. Compared with traditional methods, this solution realizes dynamic perception and response to user behavior and load changes, and can flexibly and accurately allocate power resources according to actual needs. It can not only ensure resource supply during high-performance tasks, but also switch to low-power mode under low load, effectively improving power utilization efficiency, extending the battery life of cloud phones, and significantly improving user experience.

[0081] In an embodiment of the present application, after controlling the computing device to operate according to the operating parameters, the method further includes:

[0082] Step S201 : collecting key operating data of the power supply in real time and obtaining historical failure modes of the power supply.

[0083] In an embodiment of the present application, by deploying sensors and monitoring equipment on the power module and its related components, key operating data of the power supply are collected in real time, including but not limited to parameters such as output voltage, current, temperature, power factor, etc., and at the same time, past failure modes and corresponding data of the power supply are retrieved from a database or historical records to form a historical failure mode set, providing a data basis for subsequent feature extraction and fault prediction.

[0084] Step S202 : extracting correlation features associated with historical failure modes from key operating data.

[0085] In an embodiment of the present application, data analysis algorithms and pattern recognition technologies are used to process key operating data collected in real time, and data features that are significantly correlated with historical fault patterns in terms of time series, numerical changes, statistical characteristics, etc. are screened out. These features may include voltage fluctuation frequency, temperature rise rate, abnormal power factor changes, etc., which serve as input basis for subsequent fault prediction.

[0086] Step S203: input the associated features into the fault prediction model and output a warning signal of the power supply failure risk.

[0087] In an embodiment of the present application, the associated features are input into a fault prediction model, and a warning signal of the power supply failure risk is output, including: constructing a feature vector of multidimensional data based on the associated features, and fine-tuning the parameters of the fault prediction model based on the feature vector to adapt it to the characteristics of the power supply; forward calculating the feature vector based on the adjusted fault prediction model to obtain a fault risk probability value; comparing the fault risk probability value with a preset threshold value, and outputting a warning signal of the power supply failure risk when the threshold value is exceeded.

[0088] First, the extracted correlated features (such as voltage fluctuation amplitude and temperature change rate) are organized according to chronological order or correlation to construct a multidimensional feature vector, where each dimension corresponds to an eigenvalue, to comprehensively describe the operating status of the power supply. This feature vector is then used to fine-tune the parameters of a pre-trained fault prediction model. Through optimization algorithms such as gradient descent and backpropagation, the weights and bias parameters in the model are adjusted to enable the model to more accurately capture the characteristics of the current power supply system (such as hardware differences and operating environment), thereby improving prediction accuracy. The fine-tuning process is typically performed on a partially labeled validation dataset to evaluate the performance of the model after adaptation.

[0089] Using a finely tuned fault prediction model, the constructed multidimensional feature vector is used as input to perform the model's forward propagation calculations. The model processes the feature vector layer by layer through its internal multi-layer neural network structure (such as fully connected layers and convolutional layers), extracting high-level feature representations and ultimately outputting a failure risk probability value. This probability value indicates the likelihood of a power supply failure given the feature vector and typically ranges from 0 to 1, with values ​​closer to 1 indicating a higher failure risk.

[0090] The failure risk probability value output by the model is compared with a preset threshold. This threshold is set based on power system safety requirements, historical failure data, and operational experience to distinguish between normal conditions and potential failures. When the failure risk probability value exceeds the preset threshold, the power supply is deemed to have a high failure risk. The system automatically triggers the early warning mechanism and issues a warning signal indicating the power supply failure risk. This warning signal can include alarm information, fault type information, and recommended maintenance measures, allowing operations and maintenance personnel to take timely action to prevent or mitigate the failure.

[0091] Step S204: executing a corresponding power warning operation based on the warning signal.

[0092] In an embodiment of the present application, based on the warning signal output by the fault prediction model, the system automatically triggers or manually initiates corresponding power warning operations. These operations may include but are not limited to sending alarm notifications, recording fault logs, starting backup power supplies, adjusting load distribution, performing power maintenance or replacement, etc., aiming to promptly prevent or reduce the impact of power failures on system operations and ensure the stability and reliability of the system.

[0093] The embodiment of the present application can fully grasp the power supply operation status and failure patterns by collecting key power supply operation data in real time and obtaining historical failure modes, breaking the limitations of traditional post-maintenance. Features associated with historical failure modes are extracted from key operation data to accurately locate failure-related factors and avoid interference from invalid information. By inputting these associated features into the fault prediction model, the advantages of data and models can be used to identify potential power supply failure risks in advance and output early warning signals. Compared with traditional passive operation and maintenance, this solution realizes active closed-loop management from operation data to fault prediction. It can not only dynamically perceive power supply anomalies, but also take measures in advance through prediction to avoid sudden failures that cause equipment downtime or safety hazards, significantly improve power supply operation reliability and maintenance efficiency, and reduce operation and maintenance costs and potential losses.

[0094] In this embodiment, a power management device for a cloud phone is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and 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, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0095] This embodiment provides a power management device for a cloud phone, such as Figure 3 As shown, including:

[0096] Detection module 301, used to detect the user's operation behavior triggered on the cloud phone in the current time period;

[0097] An acquisition module 302 is configured to acquire first load data of a computing device deployed by the cloud phone in a current time period;

[0098] The analysis module 303 is configured to analyze the operation behavior and the first load data, adjust the operation parameters that affect the power supply of the computing device, and control the computing device to operate according to the operation parameters.

[0099] In an embodiment of the present application, the analysis module 303 is used to extract key behavioral features from the operating behavior; predict the second load data of the computing device in the future time period based on the key behavioral features and the first load data; predict the power consumption trend of the computing device in the future time period based on the key behavioral features and the first load data; use the second load data and the power consumption trend to determine the adjustment index that affects the power supply of the computing device; and adjust the operating parameters of the computing device according to the adjustment index.

[0100] In an embodiment of the present application, the analysis module 303 is used to extract key indicators from the first load data; construct a time series data set using the key indicators and key behavioral features; perform correlation analysis on the key indicators and key behavioral features in the time series data set according to the time series to obtain a behavior-load causal relationship model, wherein the behavior-load causal relationship model is used to characterize the influence weights and time lag effects of different operating behaviors on future load changes; and output the second load data of the computing device in the future time period based on the obtained behavior-load causal relationship model.

[0101] In an embodiment of the present application, the analysis module 303 is used to normalize the key behavior features and the first load data, and generate a fused feature vector through a feature fusion method; the fused feature vector is input into the power consumption trend prediction model to obtain the spatiotemporal dependency relationship between the key behavior features and the first load data; the spatiotemporal dependency relationship is separated through the attention mechanism to obtain decoupled power consumption related features, and the power consumption trend in the future time period is predicted based on the power consumption related features.

[0102] In an embodiment of the present application, the analysis module 303 is used to time-align the second load data with the power consumption trend, analyze the correlation between the two, and obtain the correlation analysis results of the load and power consumption; construct a power impact weight matrix based on the correlation analysis results, wherein the power impact weight matrix is ​​used to reflect the contribution of each load component to the power consumption; design a multi-objective optimization function according to the power management goal, use the power impact weight matrix as a constraint condition, and obtain an optimization objective function with constraints; solve the optimization objective function with constraints through an optimization algorithm to obtain an adjustment index that affects the power supply of the computing device

[0103] In an embodiment of the present application, the device also includes: a fault warning module, which is used to collect key operating data of the power supply in real time and obtain historical failure modes of the power supply; extract associated features associated with historical failure modes from the key operating data; input the associated features into a fault prediction model, and output a warning signal of the power supply failure risk; and execute corresponding power supply warning operations based on the warning signal.

[0104] In an embodiment of the present application, a fault warning module is used to construct a feature vector of multidimensional data based on associated features, and fine-tune the parameters of a fault prediction model based on the feature vector to adapt it to the characteristics of the power supply; forward calculation is performed on the feature vector based on the adjusted fault prediction model to obtain a fault risk probability value; the fault risk probability value is compared with a preset threshold value, and a warning signal of power supply failure risk is output when the threshold value is exceeded.

[0105] See also Figure 4 , Figure 4 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 4 As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. 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 the instructions executed in the computer device, including instructions stored in the memory or on the memory to display the graphical information of the 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. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system).

[0106] 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.

[0107] 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.

[0108] 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 a computer device for displaying 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 computer 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.

[0109] 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.

[0110] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0111] 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.

[0112] 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 power management method for a cloud phone, characterized in that: The method comprises: Detect the user's operation behaviors triggered on the cloud phone in the current time period; Obtaining first load data of a computing device deployed by the cloud phone in the current time period; The operating behavior and the first load data are analyzed, an operating parameter affecting the power supply of the computing device is adjusted, and the computing device is controlled to operate according to the operating parameter.

2. The method according to claim 1, characterized in that The analyzing the operation behavior and the first load data and adjusting an operating parameter affecting the power supply of the computing device includes: extracting key behavioral features from the operating behavior; predicting second load data of the computing device in a future time period based on the key behavior feature and the first load data; predicting a power consumption trend of the computing device in a future time period based on the key behavior characteristics and the first load data; Determining an adjustment indicator affecting the power supply of the computing device by using the second load data and the power consumption trend; The operating parameters of the computing device are adjusted according to the adjustment index.

3. The method according to claim 3, characterized in that The predicting, based on the key behavior feature and the first load data, second load data of the computing device in a future time period includes: Extracting key indicators from the first load data; Constructing a time series data set using the key indicators and the key behavioral features; Performing a correlation analysis on the key indicators and key behavior features in the time series data set according to the time series to obtain a behavior-load causal relationship model, wherein the behavior-load causal relationship model is used to characterize the influence weight and time lag effect of different operating behaviors on future load changes; Output second load data of the computing device in a future time period according to the obtained behavior-load causal relationship model.

4. The method according to claim 3, characterized in that The predicting the power consumption trend of the computing device in a future time period based on the key behavior feature and the first load data includes: Normalizing the key behavior features and the first load data, and generating a fused feature vector by a feature fusion method; Inputting the fused feature vector into a power consumption trend prediction model to obtain a spatiotemporal dependency relationship between the key behavior feature and the first load data; The spatiotemporal dependency is separated through an attention mechanism to obtain decoupled power consumption related features, and the power consumption trend in the future time period is predicted based on the power consumption related features.

5. The method according to claim 3, characterized in that Determining an adjustment indicator affecting the power supply of the computing device by using the second load data and the power consumption trend includes: Time-aligning the second load data with the power consumption trend, analyzing the correlation between the two, and obtaining a correlation analysis result between the load and power consumption; Constructing a power supply impact weight matrix based on the correlation analysis result, wherein the power supply impact weight matrix is ​​used to reflect the contribution of each load component to power consumption; Designing a multi-objective optimization function according to the power management objectives, taking the power impact weight matrix as a constraint condition, and obtaining an optimization objective function with constraint conditions; An optimization objective function with constraints is solved by an optimization algorithm to obtain an adjustment index that affects the power supply of the computing device.

6. The method according to claim 1, characterized in that After controlling the computing device to operate according to the operating parameters, the method further includes: Collect key operating data of the power supply in real time and obtain historical failure modes of the power supply; extracting correlation features associated with the historical failure mode from the key operating data; Inputting the associated features into a fault prediction model to output an early warning signal of power supply failure risk; A corresponding power supply warning operation is performed based on the warning signal.

7. The method according to claim 1, characterized in that Inputting the associated features into a fault prediction model and outputting a warning signal of a power supply failure risk includes: Constructing a feature vector of the multidimensional data based on the associated features, and fine-tuning the parameters of the fault prediction model based on the feature vector to adapt it to the characteristics of the power supply; Performing forward calculation on the characteristic vector according to the adjusted fault prediction model to obtain a fault risk probability value; The fault risk probability value is compared with the preset threshold value, and a warning signal of power supply failure risk is output when the threshold value is exceeded.

8. A power management device for a cloud phone, characterized in that: The device comprises: The detection module is used to detect the user's operation behavior triggered on the cloud phone in the current time period; An acquisition module, configured to acquire first load data of the computing device deployed by the cloud phone in the current time period; An analysis module is used to analyze the operation behavior and the first load data, adjust the operating parameters that affect the power supply of the computing device, and control the computing device to operate according to the operating parameters.

9. A computer 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