CN-121984313-A - PMIC load self-adaptive data processing power supply scheduling method and system

CN121984313ACN 121984313 ACN121984313 ACN 121984313ACN-121984313-A

Abstract

The invention provides a PMIC load self-adaptive data processing power supply scheduling method and a system, wherein the method comprises the steps of monitoring load electrical parameters in real time and collecting historical task data, generating a composite working state identifier by combining current power characteristics and future power requirements, matching an optimal power supply strategy according to the composite working state identifier, generating a self-adaptive scheduling instruction, dynamically adjusting PMIC output voltage, current and loop compensation parameters, and realizing efficient and stable load power supply. The invention can improve the efficiency, stability and adaptability of the power supply system, solve the problems of power supply lag, low efficiency, inflexible strategy matching and the like in the prior art, and provide a reliable power management solution for high-performance operation of electronic equipment.

Inventors

LIU KAI
HUANG PENG

Assignees

上海格州微电子技术有限公司

Dates

Publication Date: 20260505
Application Date: 20260130

Claims (10)

1. The PMIC load self-adaptive data processing power supply scheduling method is characterized by comprising the following steps of: s1, monitoring the electrical parameters of at least one load unit in real time during operation, and collecting historical task data of the load; S2, dynamically calculating the current power characteristic parameters of the load through a pre-constructed first calculation model based on the electric parameters during operation; S3, predicting short-term future power demand of the load through a pre-constructed second calculation model according to the historical task data; s4, fusing the current power characteristic parameter and the short-term future power demand to generate a composite working state identifier of the load; s5, based on the composite working state identification, matching an optimal power supply strategy from a pre-configured power supply strategy library, and generating a corresponding self-adaptive power supply scheduling instruction; And S6, adjusting the power supply output of the PMIC to the at least one load unit according to the self-adaptive power supply scheduling instruction.
2. The method of claim 1, wherein the real-time monitoring of the runtime electrical parameters in step S1 includes high frequency synchronous acquisition by current and voltage sensors, obtaining instantaneous current and voltage values of the load, and generating a sequence of current-voltage sample data pairs, the acquisition history task data including acquisition of task queues to be performed and their corresponding task attribute metadata for the load within a predetermined time window in the future from a system task scheduler.
3. The method according to claim 2, wherein in step S2, the dynamically calculating the current power characteristic parameter of the load by the pre-constructed first calculation model includes: S21, calculating to obtain an instantaneous power sequence based on the current-voltage sampling data pair sequence; s22, performing time domain analysis on the instantaneous power sequence, and extracting a power ripple characteristic value; s23, carrying out frequency domain analysis on the instantaneous power sequence, and extracting the power spectrum density of a specific frequency band; And S24, calculating to obtain a current dynamic stability index representing load stability based on the power ripple characteristic value and the power spectrum density.
4. The method of claim 3, wherein in step S22, the extracting the power ripple characteristic value is implemented by calculating a standard deviation of the instantaneous power sequence within a sliding time window, as shown in the following formula: Wherein, the Which represents the characteristic value of the power ripple, The size of the sliding window is indicated, Representing the in-window first The value of the respective instantaneous power value, Representing the average of all instantaneous power values within a window, And starting index for the window.
5. A method according to claim 3, wherein in step S24, the current dynamic stability index is represented by the formula: Wherein, the Which represents the characteristic value of the power ripple, Represents the power spectral density integral values within a preset critical frequency band, And For the pre-defined normalized weight coefficients, In order to prevent and eliminate the zero constant, Representing the current dynamic stability index.
6. The method of claim 2, wherein in step S3, the predicting short-term future power demand of the load by the pre-constructed second calculation model comprises: S31, analyzing task attribute metadata in the task queue to be executed, and inquiring a task power model library based on task types and calculation complexity to obtain a basic power predicted value; And S32, dynamically correcting the base power predicted value by adopting a time sequence prediction algorithm based on the historical power consumption data of the load, and generating the short-term future power demand.
7. The method of claim 6, wherein in step S32, the dynamic correction using the time series prediction algorithm specifically uses the following formula: Wherein, the Representing a short-term future power demand, The base power prediction value is represented as such, Representing the actual average power observed over similar historical task periods, Representing the expected power values during the similar historical task periods, Is a learning rate factor.
8. The method of claim 1, wherein in step S4, the generating the composite operating state identification of the load comprises: s41, comparing a current dynamic stability index in the current power characteristic parameter with a preset stability threshold interval to determine a stability grade; S42, comparing the short-term future power demand with the current average power, and calculating a power demand change gradient; s43, combining the stability grade and the power demand change gradient into a multidimensional feature vector serving as the composite working state identifier.
9. The method of claim 1, wherein in step S5, the matching the optimal power supply policy from the preconfigured power supply policy library and generating the adaptive power supply scheduling instruction comprises: S51, calculating Euclidean distance between the composite working state identifier and a state vector corresponding to each strategy in the power supply strategy library; s52, selecting a strategy with the minimum Euclidean distance from the composite working state identifier as the optimal power supply strategy, wherein the optimal power supply strategy at least comprises target output voltage, maximum allowable current and loop compensation parameters; S53, generating an adaptive power supply scheduling instruction containing the parameters based on the target output voltage, the maximum allowable current and the loop compensation parameters; in step S6, the adjusting the power supply output of the PMIC according to the adaptive power supply scheduling instruction includes: S61, inputting target output voltage, maximum allowable current and loop compensation parameters in the adaptive power supply scheduling instruction to a digital pulse width modulation controller; S62, generating corresponding driving signals according to the target output voltage and loop compensation parameters through the digital pulse width modulation controller; And S63, controlling the on-duty ratio and the switching frequency of a power switch tube in the PMIC by using the driving signal, so as to accurately adjust the power supply output voltage and the current of the at least one load unit.
10. A PMIC load adaptive data processing power supply scheduling system, comprising: The parameter monitoring module is used for monitoring the electrical parameters of at least one load unit in real time during operation and collecting historical task data of the load; the current power analysis module is used for dynamically calculating the current power characteristic parameters of the load through a pre-constructed first calculation model based on the electric parameters during operation; the future demand prediction module is used for predicting the short-term future power demand of the load through a pre-constructed second calculation model according to the historical task data; the state fusion judging module is used for fusing the current power characteristic parameter and the short-term future power demand to generate a composite working state identifier of the load; The self-adaptive strategy generation module is used for matching an optimal power supply strategy from a pre-configured power supply strategy library based on the composite working state identification and generating a corresponding self-adaptive power supply scheduling instruction; And the power supply output execution module is used for adjusting the power supply output of the PMIC to the at least one load unit according to the self-adaptive power supply scheduling instruction.

Description

PMIC load self-adaptive data processing power supply scheduling method and system Technical Field The invention relates to the technical field of power supply data processing, in particular to a PMIC load self-adaptive data processing power supply scheduling method and system. Background In the context of rapid development of electronic devices today, a power management integrated circuit PMIC serves as a core power supply unit for the device, and plays an important role in providing stable and efficient power supply to various load units. The conventional PMIC power supply scheduling method mainly relies on fixed power supply parameter settings, and these parameters are preset according to the maximum power requirement of the load in the design stage, so as to ensure that the power supply requirement of the load can be met in any case. However, this approach has significant limitations. On the one hand, the actual power demand of the load can be dynamically changed due to factors such as task types, calculation complexity and the like in the running process, and the fixed power supply parameters cannot be adapted to the changes in real time, so that the power supply efficiency is low, and the energy waste is serious. On the other hand, the traditional power supply scheduling method lacks the capability of predicting the future power demand of the load and cannot adjust the power supply strategy in advance, so that the condition of unstable power supply easily occurs when the power demand of the load changes sharply, and the performance and the reliability of the equipment are affected. While there are some attempts in the prior art to dynamically adjust the power supply parameters by monitoring the real-time power of the load, these methods mostly only focus on the instantaneous power of the load, ignoring the dynamic characteristics of the load and the predictions of future power demands. For example, some methods calculate instantaneous power by simple current and voltage sampling and adjust the power output accordingly, but such adjustments tend to lag the actual demands of the load and do not effectively address the problems of unstable power and inefficiency. In addition, some methods attempt to predict future power demands by using historical task data of the load, but these methods generally lack deep analysis of load dynamic characteristics, have limited prediction accuracy, and cannot provide accurate basis for optimizing a power supply strategy. Therefore, the prior art has at least the following problems or defects that firstly, the dynamic power characteristics of a load cannot be accurately monitored and analyzed in real time, so that power supply adjustment lags behind load demands, secondly, the accurate prediction capability of future power demands of the load is lacking, the power supply strategy cannot be optimized in advance, thirdly, the existing power supply strategy matching method is not flexible enough, and the power supply parameters cannot be dynamically adjusted according to the complex working state of the load. These problems seriously affect the efficiency and stability of the PMIC power supply system, limiting further improvement of the performance of the electronic device. Disclosure of Invention The invention provides a PMIC load self-adaptive data processing power supply scheduling method and a system. In a first aspect of the present invention, there is provided a PMIC load adaptive data processing power supply scheduling method, including: s1, monitoring the electrical parameters of at least one load unit in real time during operation, and collecting historical task data of the load; S2, dynamically calculating the current power characteristic parameters of the load through a pre-constructed first calculation model based on the electric parameters during operation; S3, predicting short-term future power demand of the load through a pre-constructed second calculation model according to the historical task data; s4, fusing the current power characteristic parameter and the short-term future power demand to generate a composite working state identifier of the load; s5, based on the composite working state identification, matching an optimal power supply strategy from a pre-configured power supply strategy library, and generating a corresponding self-adaptive power supply scheduling instruction; And S6, adjusting the power supply output of the PMIC to the at least one load unit according to the self-adaptive power supply scheduling instruction. Further, in step S1, the real-time monitoring of the electric parameters during operation includes high-frequency synchronous acquisition through a current sensor and a voltage sensor, obtaining of an instantaneous current value and an instantaneous voltage value of a load, and generation of a current-voltage sampling data pair sequence, and the acquisition of historical task data includes acquisition of a task queue to be executed of