Intelligent power supply circuit system and method

Abstract

The present application relates to the field of intelligent power supply technology, and specifically discloses an intelligent power supply circuit system and method, which first obtains the power supply voltage, collects the load status information of each load at the same time, and introduces data processing technology based on deep learning to semantically represent and fine-grained cross-modal fusion of the power supply voltage and the load status information of each load, so as to achieve a fine-grained understanding of the association between the power supply voltage and the load status, and then intelligently recommends a suitable voltage conversion mode on this basis. In this way, the power supply voltage can be dynamically adjusted to meet the specific needs of different loads, thereby further improving the power supply efficiency of the power supply circuit and the endurance of the smart terminal.

Description

Technical Field

[0001] The present application relates to the field of intelligent power supply technology, and more specifically, to an intelligent power supply circuit system and method. Background Art

[0002] As the functions of smart terminals (such as smartphones, tablets, etc.) become increasingly complex and diverse, the power requirements of their internal components have also become more diversified. At present, smart terminals are often equipped with power supply circuits, such as power management units (PMUs), which are connected to the battery and load in the smart terminal respectively, and the load is powered by regulating the output voltage of the battery. However, the traditional power supply circuit design mainly focuses on the step-down operation of the higher voltage provided by the battery, and when the supply voltage is within a certain range, using a buck converter for voltage regulation will reduce the overall conversion efficiency.

[0003] In this regard, the invention patent with publication number CN114498872A proposes a power supply circuit, which includes a control module, a first voltage conversion module and at least one second voltage conversion module. The control module determines the working mode of the first voltage conversion module, such as pass-through mode and buck mode, by comparing the supply voltage with a preset threshold voltage, so that it outputs a stable supply voltage to the second voltage conversion module for further voltage conversion to power each load, thereby improving the power supply efficiency and increasing the battery life of the smart terminal.

[0004] However, the prior art mainly relies on a fixed threshold voltage for mode switching, while ignoring the specific state of each load and its special requirements for power supply characteristics. This may lead to non-optimal operation mode selection and fail to dynamically adapt to actual demand changes in different application scenarios, thereby affecting the overall system performance.

[0005] Therefore, a control method and system for an optimized intelligent power supply circuit are desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an intelligent power supply circuit system and method, which first obtains the power supply voltage, collects the load status information of each load at the same time, and introduces data processing technology based on deep learning to semantically represent the power supply voltage and the load status information of each load and perform fine-grained cross-modal fusion, so as to achieve a fine-grained understanding of the association between the power supply voltage and the load status, and intelligently recommend appropriate voltage conversion modes on this basis. In this way, the power supply voltage can be dynamically adjusted to meet the specific needs of different loads, thereby further improving the power supply efficiency of the power supply circuit and the endurance of the smart terminal.

[0007] Accordingly, according to one aspect of the present application, there is provided an intelligent power supply circuit system, comprising:

[0008] A control module, a first voltage conversion module and at least one second voltage conversion module;

[0009] The control module is connected to the first voltage conversion module and is used to control the first voltage conversion module to be in a pass-through mode or a buck mode according to an input voltage signal, so as to output a first output voltage according to the supply voltage;

[0010] The at least one second voltage conversion module is used to perform voltage conversion on the first output voltage to output a second output voltage for supplying a load connected thereto.

[0011] In the above-mentioned intelligent power supply circuit system, the control module includes: a voltage information acquisition unit, which is used to acquire the power supply voltage; a load state information acquisition unit, which is used to collect the load state information of each load connected to each second voltage conversion module to obtain a set of load state information; a power supply voltage information encoding unit, which is used to perform one-hot encoding on the power supply voltage to obtain a supply voltage one-hot encoding vector; a load information semantic encoding unit, which is used to perform semantic embedding encoding on each load state information in the set of load state information to obtain a set of load state information semantic embedding encoding vectors; a cross-domain aggregation analysis unit, which is used to perform cross-domain aggregation analysis based on a priori clustering centers on the set of the supply voltage one-hot encoding vector and the load state information semantic embedding encoding vector to obtain a supply voltage-load state fine-grained interactive aggregation encoding vector; a mode switching control unit, which is used to generate a mode switching control instruction based on the supply voltage-load state fine-grained interactive aggregation encoding vector.

[0012] According to another aspect of the present application, a control method for an intelligent power supply circuit is provided, comprising:

[0013] Get the supply voltage;

[0014] Collecting load status information of each load connected to each second voltage conversion module to obtain a set of load status information;

[0015] One-hot encoding the supply voltage to obtain a supply voltage one-hot encoding vector;

[0016] Performing semantic embedding coding on each piece of load state information in the set of load state information to obtain a set of load state information semantic embedding coding vectors;

[0017] Performing a cross-domain aggregation analysis based on a priori clustering centers on the set of the power supply voltage one-hot encoding vector and the load state information semantic embedding encoding vector to obtain a power supply voltage-load state fine-grained interactive aggregation encoding vector;

[0018] Based on the power supply voltage-load state fine-grained interactive aggregation coding vector, a mode switching control instruction is generated.

[0019] Compared with the prior art, the intelligent power supply circuit system and method provided by the present application first obtains the power supply voltage, collects the load status information of each load at the same time, and introduces data processing technology based on deep learning to semantically represent and fine-grained cross-modal fusion of the power supply voltage and the load status information of each load, so as to achieve a fine-grained understanding of the association between the power supply voltage and the load status, and then intelligently recommends a suitable voltage conversion mode on this basis. In this way, the power supply voltage can be dynamically adjusted to meet the specific needs of different loads, thereby further improving the power supply efficiency of the power supply circuit and the endurance of the smart terminal. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1 4 is a block diagram of a control module in a smart power supply circuit system according to an embodiment of the present application.

[0022] Figure 2 Schematic diagram of data flow of a control module in a smart power supply circuit system according to an embodiment of the present application.

[0023] Figure 3 It is a block diagram of a cross-domain aggregation analysis unit in a smart power supply circuit system according to an embodiment of the present application.

[0024] Figure 4 Flow chart of a control method of an intelligent power supply circuit according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0026] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0027] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0028] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0029] As mentioned in the background technology above, patent CN114498872A proposes a power supply circuit, which includes a control module, a first voltage conversion module and at least one second voltage conversion module; wherein the control module is connected to the first voltage conversion module, and is used to control the first voltage conversion module to be in a pass-through mode or a buck mode according to an input voltage signal, so as to output a first output voltage according to the power supply voltage; the at least one second voltage conversion module is used to perform voltage conversion on the first output voltage, so as to output a second output voltage for supplying a load connected thereto.

[0030] However, in the prior art, mode switching mainly relies on the execution of a preset threshold voltage, without fully considering the specific state of each load and its specific requirements for power supply properties, which may lead to non-optimal operation mode selection and cannot flexibly respond to changes in actual needs in different usage scenarios, which may have an adverse effect on the overall performance of the system. In response to this technical problem, the present application proposes an optimized intelligent power supply circuit system, which first obtains the power supply voltage, collects the load status information of each load, and introduces data processing technology based on deep learning to semantically represent and fine-grained cross-modal fusion of the power supply voltage and the load status information of each load, so as to achieve a fine-grained understanding of the association between the power supply voltage and the load status, and intelligently recommend appropriate voltage conversion modes on this basis. In this way, the power supply voltage can be dynamically adjusted to meet the specific needs of different loads, thereby further improving the power supply efficiency of the power supply circuit and the endurance of the smart terminal.

[0031] Figure 14 is a block diagram of a control module in a smart power supply circuit system according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of a control module in a smart power supply circuit system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the control module 100 includes: a voltage information acquisition unit 110, used to acquire the supply voltage; a load state information acquisition unit 120, used to collect the load state information of each load connected to each of the second voltage conversion modules to obtain a set of load state information; a supply voltage information encoding unit 130, used to perform one-hot encoding on the supply voltage to obtain a supply voltage one-hot encoding vector; a load information semantic encoding unit 140, used to perform semantic embedding encoding on each load state information in the set of load state information to obtain a set of load state information semantic embedding encoding vectors; a cross-domain aggregation analysis unit 150, used to perform cross-domain aggregation analysis based on a priori clustering centers on the set of the supply voltage one-hot encoding vector and the load state information semantic embedding encoding vector to obtain a supply voltage-load state fine-grained interactive aggregation encoding vector; a mode switching control unit 160, used to generate a mode switching control instruction based on the supply voltage-load state fine-grained interactive aggregation encoding vector.

[0032] In the above-mentioned intelligent power supply circuit system, the voltage information acquisition unit 110 is used to obtain the supply voltage. In a specific example of the present application, the voltage information acquisition unit 110 is used to: use an analog-to-digital converter to sample the input voltage signal to extract the supply voltage value at the current time point in real time to obtain the supply voltage. It should be understood that the supply voltage is the energy source of the entire power supply circuit, and its stability and size directly affect the subsequent power supply strategy. In one embodiment of the present application, the input voltage signal is sampled using an analog-to-digital converter to extract the supply voltage value at the current time point in real time to obtain the supply voltage. For example, in the power management module of the smart terminal, a voltage detection circuit based on resistor voltage division is integrated, and the battery output voltage is divided, and the divided signal is input to the analog input pin of the microcontroller. After the internal analog-to-digital conversion module (ADC), the analog voltage signal is converted into a digital quantity for subsequent processing and analysis. In this way, the supply voltage value can be obtained in real time with high accuracy and stability, providing reliable basic data for subsequent control decisions. Of course, in practical applications, the supply voltage can also be obtained by other means, such as directly sampling from the output end of the battery, or using a dedicated voltage detection chip. The specific method can be determined based on actual needs and cost considerations.

[0033] Specifically, in order to effectively obtain the power supply voltage information, it is first necessary to select a suitable voltage sensor and plan its layout reasonably. Considering the limited internal space and complex wiring of smart terminals, priority should be given to application-specific integrated circuits (ICs) that are small in size, easy to integrate, and have low power consumption when selecting voltage sensing elements. These ICs usually have built-in functional blocks such as amplifiers and filters, and can directly output digital signals to the control unit, simplifying the design of peripheral circuits. For example, a shunt resistor can be used as a basic sensing element, connected in series in the power supply line through a small resistance resistor, and Ohm's law can be used to calculate the voltage drop across the two ends to obtain current and voltage information; or a Hall effect sensor can be used for non-contact measurement. Although it is more expensive and sensitive to environmental magnetic fields, it may be the preferred solution in certain specific application scenarios.

[0034] After selecting the appropriate voltage sensor, the next step is to build a stable and reliable data acquisition circuit. The data acquisition circuit is responsible for converting the analog signal from the sensor into a form suitable for reading by a microcontroller or other processing unit. A typical data acquisition circuit may include a preamplifier to amplify the small amplitude signal output by the sensor and improve the signal-to-noise ratio; a filter to eliminate high-frequency noise and other unnecessary frequency components to ensure the purity of the signal before input to the ADC; an analog-to-digital converter (ADC) to convert the pre-processed analog signal into a digital signal for subsequent digital processing; and a voltage regulator to provide a stable reference voltage to the ADC to ensure the consistency and accuracy of the conversion results. In particular, in order to support the simultaneous sampling requirements of multiple channels (such as monitoring multiple load points at the same time), you can consider using a multiplexer (MUX) to switch different input channels to the same set of data acquisition links in turn, which not only saves hardware resources, but also helps to synchronously obtain voltage information at each point.

[0035] Considering that the power management system in the smart terminal requires a rapid response to changes in the power supply voltage, measures must be taken to ensure that the data acquisition process is sufficiently real-time. Optimizing the sampling frequency is one of the important means to achieve this. According to the needs of the application scenario, a reasonable sampling rate should be set to meet the ability to capture rapid changes while avoiding the waste of resources caused by oversampling. Generally speaking, for battery-powered devices, a sampling frequency of dozens to hundreds of times per second is sufficient. In addition, an interrupt-driven mechanism can be introduced. When it is detected that the power supply voltage exceeds the preset safety threshold, the interrupt notification control module is immediately triggered for emergency processing. This method can respond to abnormal situations in a timely manner without affecting the normal operation of the program, thereby ensuring the stability and security of the system.

[0036] In addition to hardware preparation, the software platform also plays an indispensable role. It is very necessary to write dedicated firmware code to manage and coordinate the workflow of all the above hardware components. This part of the code should include initialization configuration settings for various peripheral parameters, such as ADC resolution, MUX channel order, etc.; periodic task scheduling executes voltage sampling tasks at predetermined time intervals and stores the results in the memory buffer for further analysis; event processing logic defines how to respond to various possible interrupt events, such as overvoltage protection, undervoltage alarm, etc.; and communication interface protocol If you need to exchange information with other subsystems, you must implement the corresponding communication protocol, such as I2C, SPI, etc.

[0037] Specifically, the initialization configuration phase involves configuring the ADC sampling rate, resolution, and MUX channel selection sequence to ensure that each sampling accurately reflects the current power supply voltage status. Periodic task scheduling regularly triggers voltage sampling tasks through the scheduling service provided by the timer or operating system to maintain continuous monitoring of the power supply voltage. For any situation that exceeds the set threshold, the event processing logic will respond quickly and perform preset operations, such as issuing an alarm or adjusting the power management strategy. The communication interface protocol part is to ensure that the power supply voltage information can be correctly transmitted to other related modules or external systems so that they can make corresponding adjustments based on the latest power supply conditions.

[0038] In the above-mentioned intelligent power supply circuit system, the load status information acquisition unit 120 is used to collect the load status information of each load connected to each second voltage conversion module to obtain a set of load status information. It should be understood that different types of loads may have different working states (such as standby, activation, full load, etc.) and working characteristics (such as working current, working voltage, power consumption, etc.) at different time points, which directly affect the required voltage level. Therefore, the present application further deploys corresponding sensors or interfaces for each load connected to the second voltage conversion module to capture its status information, so as to fully understand the actual needs of each load, thereby providing a basic basis for the subsequent voltage conversion strategy.

[0039] In order to accurately obtain the load status information of each load connected to each of the second voltage conversion modules, it is first necessary to clarify the specific content of the load status information to be collected. Load status information usually includes aspects such as current consumption, operating temperature, operating mode, power demand priority, and real-time performance indicators. Current consumption reflects the current activity level of each load; operating temperature is used to prevent overheating risks and take cooling measures; changes in operating modes will significantly affect load characteristics; power demand priority determines which loads should be given priority in power support when power is limited; and delay-sensitive applications need to monitor their real-time performance to ensure that power supply does not become a performance bottleneck.

[0040] For the above-mentioned types of load status information, it is necessary to select a suitable data collection method. For example, the direct measurement method is to use physical sensors such as current sensors and temperature sensors to be directly installed near the load to read the corresponding physical values ​​in real time. This method is simple and intuitive, but it may increase the complexity of wiring. An indirect estimation method can also be used to infer unknown load state parameters based on known load characteristic curves or models combined with other available information (such as input power). In addition, if the load itself has an operating system-level API interface, these interfaces can be accessed programmatically to directly obtain the required status information. For independent subsystems that support standard communication protocols (such as I2C, SPI, UART, etc.), it is also possible to communicate with them by sending specific commands and receive feedback status reports. In this way, not only can the refinement of power management be improved, but also the actual demand changes in different application scenarios can be better met, thereby improving the performance of the entire system.

[0041] In order to efficiently collect all necessary load status information, hardware and software must work closely together. A unified data bus can be built to establish one or more high-speed data transmission paths inside the intelligent terminal to connect all load nodes to the central control unit, which not only simplifies the wiring structure but also improves the speed and reliability of data exchange. For locations far away from the central controller, miniaturized data acquisition nodes can be deployed. The nodes are close to their respective loads, responsible for preliminary data preprocessing and local caching, and then upload the results to the main control end regularly or on demand. Considering the large number of loads and their different update frequencies, concurrency mechanisms should be considered in the design of software architecture to ensure that each load can be paid attention to in time without causing information lag. Middleware is introduced as a bridge between software and hardware. The middleware can shield the underlying hardware differences and provide a consistent service interface to the upper-level application. At the same time, it is responsible for functions such as data format conversion and anomaly detection to reduce the burden on the main program.

[0042] In actual operation, it is equally important to ensure the quality and integrity of the collected data. Since some data may be distorted or lost due to environmental factors, hardware failures, etc., it is necessary to implement a series of data verification and quality assurance measures. Set up redundant backups for key data so that even if there is a problem with one sampling, you can still get the correct information from other sources. Set a reasonable error range. When the collected data exceeds this range, trigger an alarm or re-collect to ensure the accuracy of the data. Regularly calibrate sensors and other measuring equipment to maintain long-term accuracy and reliability. Finally, use data analysis tools to perform statistical analysis on the collected data, identify potential problem trends, and optimize the collection strategy accordingly.

[0043] In the above-mentioned intelligent power supply circuit system, the supply voltage information encoding unit 130 is used to perform one-hot encoding on the supply voltage to obtain a one-hot encoding vector of the supply voltage. It should be understood that one-hot encoding is a technique for converting categorical variables into binary vectors. In the technical solution of the present application, the supply voltage is a discrete physical variable. When the original data is directly processed using a deep learning algorithm, the model may mistakenly believe that there is a certain linear relationship between different supply voltage values ​​(for example, voltage 3V is twice voltage 1.5V), but in fact this proportional relationship is not necessarily applicable to the association between voltage and load demand. Therefore, the present application performs one-hot encoding on the supply voltage to map each possible value interval of the supply voltage to a binary vector of fixed length to obtain a one-hot encoding vector of the supply voltage. Among them, only one position in the one-hot encoding vector of the supply voltage is 1, and the rest of the positions are 0, so as to accurately represent different supply voltage values, thereby eliminating potential misleading assumptions caused by the size of the numerical value.

[0044] In the above-mentioned intelligent power supply circuit system, the load information semantic encoding unit 140 is used to perform semantic embedding encoding on each load state information in the set of load state information to obtain a set of load state information semantic embedding encoding vectors. In a specific example of the present application, the load information semantic encoding unit 140 is used to: use a semantic embedding encoder based on the ALBERT model to perform semantic embedding encoding on each load state information in the set of load state information to obtain a set of load state information semantic embedding encoding vectors. It should be understood that since the load state information contains a variety of load working state information and characteristic descriptions, in order to comprehensively consider the overall characteristic description of each load so as to accurately understand its power supply requirements, the present application further uses the ALBERT model to perform semantic embedding encoding on each load state information. Those of ordinary skill in the art should know that the ALBERT model is a pre-trained language representation model based on the Transformer architecture, which can capture the complex semantic relationship between each vocabulary by learning a large amount of text data. In the technical solution of the present application, the ALBERT model is used to perform semantic embedding encoding on each load state information, and the load state information can be converted from the original text or numerical form into a vector representation with rich semantic information to obtain a set of load state information semantic embedding encoding vectors. Here, each load state information semantic embedding encoding vector in the set corresponds to a state description of a load, reflecting the actual needs and characteristics of the load. In addition, the ALBERT model maps each load state information into the same semantic space, so that load features with similar needs and characteristics are close to each other in the semantic space, retaining the semantic connection and contextual information between each load state, and providing a more refined decision-making basis for subsequent mode switching control.

[0045] In the above-mentioned intelligent power supply circuit system, the cross-domain aggregation analysis unit 150 is used to perform cross-domain aggregation analysis based on a priori clustering centers on the set of the power supply voltage one-hot encoding vector and the load status information semantic embedding encoding vector to obtain a power supply voltage-load status fine-grained interactive aggregation encoding vector. It should be understood that the present application takes into account that the power supply voltage and the load status information come from different physical domains and data modalities, respectively, and directly fusing the features of the two may cause information loss or misunderstanding due to modality mismatch. Therefore, the present application proposes a cross-domain feature fine-grained aggregation analysis method, which integrates the supply voltage information and the load status information by mining the pseudo-clustering centers between cross-modal data as anchors to achieve deeper information extraction and comprehensive characterization. Among them, Figure 3 FIG. 1 is a block diagram of a cross-domain aggregation analysis unit in a smart power supply circuit system according to an embodiment of the present application. Figure 3As shown, the cross-domain aggregation analysis unit 150 includes: a cluster center calculation subunit 151, which is used to calculate the cluster center between the set of the load state information semantic embedding coding vectors and the power supply voltage one-hot coding vector to obtain the power supply voltage-load state prior cluster center coding vector; a fine-grained aggregation subunit 152, which is used to input the supply voltage-load state prior cluster center coding vector and the set of the load state information semantic embedding coding vector into a cluster center-based cross-domain fine-grained aggregation descriptor to obtain the power supply voltage-load state fine-grained interactive aggregation coding vector.

[0046] Specifically, in a specific example of the present application, the cluster center calculation subunit 151 is used to: input the set of the load state information semantic embedding coding vector into the modal kernel feature extraction network to obtain the load state information kernel semantic feature coding vector. It should be understood that in order to achieve cross-domain aggregation interaction between the supply voltage one-hot coding vector and the set of the load state information semantic embedding coding vector, it is necessary to unify the data dimensions of the two, which can be expressed as follows:

[0047]

[0048] in, represents the modal kernel feature extraction network, represents a set of semantically embedded coding vectors of the load status information, , , , and respectively represent the first, second, and third in the set of semantic embedding coding vectors of the load status information , and The semantic embedding encoding vector of load status information, Indicates the A semantic difference factor relative to the set of semantic embedding encoding vectors of the load state information, is the number of the supply voltage-load state priori cluster center encoding vectors, represents the one-norm of a vector, represents the exponential function with base e, A kernel semantic feature encoding vector representing the load status information.

[0049] That is, the present application takes into account that using traditional pooling methods to perform feature aggregation and dimensionality reduction on the set of load state information semantic embedded coding vectors will cause indiscriminate information loss. Based on this, in the technical solution of the present application, a modal kernel feature extraction network is introduced, which can not only effectively perform feature dimensionality reduction on the above-mentioned coding vector set, but also refine more critical and representative semantic features, thereby generating a load state information kernel semantic feature coding vector, thereby reflecting the core characteristics of the load state and providing high-quality data input for subsequent fusion analysis.

[0050] In a specific example of the present application, the cluster center calculation subunit 151 is further used to: determine the supply voltage-load state prior cluster center encoding vector based on the supply voltage one-hot encoding vector and the load state information kernel semantic feature encoding vector, which is expressed as:

[0051]

[0052] in, represents the one-hot encoding vector of the supply voltage, represents the weight matrix, Indicates cascade, represents the bias vector, Represents the supply voltage-load state priori clustering center encoding vector.

[0053] That is, the present application further uses a neural network model to perform cross-domain association learning on the power supply voltage unique-hot encoding vector and the load state information kernel semantic feature encoding vector after dimensional uniform processing, and identifies the implicit relationship between the power supply voltage and the load state. In this way, the cluster center focusing on the core associated information of the power supply voltage information and the load state information can be predicted and set in the high-dimensional feature space, and the power supply voltage-load state prior cluster center encoding vector is generated. In this way, clustering analysis of load state information in a unified space helps to better understand the relationship between the power supply voltage and the load state, and reveals the commonalities and characteristics of different load states when associated with the supply voltage.

[0054] Specifically, in a specific example of the present application, the fine-grained aggregation subunit 152 is used to: calculate the Poincare distance of each load state information semantic embedding encoding vector in the set of the load state information semantic embedding encoding vector relative to the supply voltage-load state prior clustering center encoding vector to obtain a set of load state information semantic clustering distances, which is expressed by the formula:

[0055]

[0056] in, Indicates the and stated The semantic clustering distance of load status information between represents the inverse hyperbolic cosine function;

[0057] The set of load state information semantic clustering distances is input into a sparse encoder based on a gating mechanism to obtain a set of load state information gating clustering weight coefficients, which is expressed as follows:

[0058]

[0059] in, represents the gating threshold, Indicates the The load status information gated clustering weight coefficient;

[0060] Calculate the positional difference of each load state information semantic embedding coding vector in the set of load state information semantic embedding coding vectors relative to the power supply voltage-load state prior clustering center coding vector to obtain a set of load state information semantic clustering differential feature vectors; based on the set of load state information gated clustering weight coefficients, perform weighted aggregation on the set of load state information semantic clustering differential feature vectors to obtain the power supply voltage-load state fine-grained interactive aggregation coding vector, which is expressed as follows:

[0061]

[0062] in, express Middle The eigenvalues ​​at the positions, express Middle The eigenvalues ​​at the positions, A fine-grained interaction aggregation coding vector representing the supply voltage-load state.

[0063] That is, the power supply voltage-load state prior cluster center encoding vector is used as a reference benchmark for fine-grained interaction, and a fine-grained aggregation analysis is performed on the set of load state information semantic embedding encoding vectors, so as to comprehensively reflect the state information of multiple loads and generate a power supply voltage-load state fine-grained interaction aggregation encoding vector. It should be understood that the power supply voltage-load state prior cluster center encoding vector is constructed in the cross-domain feature space based on the feature association between the power supply voltage and the load state. Using it as an anchor for fine-grained interaction can guide the load state features of each load to be more finely aggregated according to the potential association between the power supply voltage and the load state, thereby more accurately characterizing the complex interactive relationship between the power supply voltage and the load state, while avoiding feature confusion and information loss that may be caused by directly fusing different modal data, providing strong support for subsequent intelligent voltage regulation and load management.

[0064] In the above-mentioned intelligent power supply circuit system, the mode switching control unit 160 is used to generate a mode switching control instruction based on the power supply voltage-load state fine-grained interactive aggregation coding vector. In a specific example of the present application, the mode switching control unit 160 is used to: input the power supply voltage-load state fine-grained interactive aggregation coding vector into a classifier-based mode switching logic to obtain the mode switching control instruction, and the mode switching control instruction includes switching to a pass-through mode or switching to a buck mode. In the technical solution of the present application, the classifier is based on a neural network architecture, using its multi-layer neuron structure and nonlinear activation function, and performs feature learning on the power supply voltage-load state fine-grained interactive aggregation coding vector, and combines the decision boundary learned during the training process to achieve accurate judgment of the mode switching requirements under the current power supply voltage and load state combination, and determine the most appropriate voltage conversion mode under the current situation. For example, when the power supply voltage is high and the load state is relatively stable, the classifier may output a control instruction to switch to the buck mode to reduce unnecessary energy consumption and heat; when the power supply voltage is low or the load state fluctuates greatly, the classifier may output a control instruction to switch to the pass-through mode to ensure the stability and reliability of the power supply. In this way, the voltage conversion mode can be intelligently adjusted according to the real-time power supply voltage and load state information to achieve efficient energy utilization and stable operation of the equipment.

[0065] In a specific example of the present application, the mode switching control unit 160 is used to: use the fully connected layer of the mode switching logic to fully connect the power supply voltage-load state fine-grained interactive aggregation coding vector to obtain a power supply voltage-load state fine-grained interactive aggregation fully connected coding vector; input the power supply voltage-load state fine-grained interactive aggregation fully connected coding vector into the Softmax classification function of the mode switching logic to obtain the probability value of the power supply voltage-load state fine-grained interactive aggregation coding vector belonging to each type label, wherein the type label includes a pass-through mode and a buck mode; and determine the type label corresponding to the largest of the probability values ​​as the mode switching control instruction. Specifically, the mode switching logic first uses a fully connected layer to process the power supply voltage-load state fine-grained interactive aggregation coding vector to obtain a more expressive power supply voltage-load state fine-grained interactive aggregation fully connected coding vector. Then, the power supply voltage-load state fine-grained interaction aggregated fully connected encoding vector is input to the Softmax classification function of the mode switching logic, and the Softmax classification function evaluates and outputs the probability value of the power supply voltage-load state fine-grained interaction aggregated encoding vector belonging to a predefined type label (such as pass-through mode or buck mode) by applying probability distribution calculation. Finally, according to the probability distribution generated by the Softmax classification function, the type label with a higher probability value is selected as the mode switching control instruction output.

[0066] In a preferred example of the present application, inputting the supply voltage-load state fine-grained interaction aggregated coding vector into a classifier-based mode switching logic to obtain a mode switching control instruction includes:

[0067] First, determine the eigenvalue mean corresponding to the power supply voltage-load state fine-grained interactive aggregation coding vector and the standard deviation of the eigenvalues ;

[0068] Secondly, the power supply voltage-load state fine-grained interactive aggregate coding vector and the point-subtraction vector of the eigenvalue mean and the eigenvalue standard deviation are multiplied to obtain a first power supply voltage-load state fine-grained interactive aggregate coding fairness target vector, which is expressed as:

[0069]

[0070] in, Represents the fine-grained interaction aggregation encoding vector of supply voltage and load state, represents the first supply voltage-load state fine-grained interaction aggregation encoding fairness target vector, Indicates point reduction, represents dot product;

[0071] Next, the point subtraction vector of the power supply voltage-load state fine-grained interactive aggregate coding vector and the eigenvalue standard deviation is multiplied by the eigenvalue mean to obtain a second power supply voltage-load state fine-grained interactive aggregate coding fairness target vector, which is expressed as:

[0072]

[0073] in, Representing a second supply voltage-load state fine-grained interactive aggregate encoding fairness target vector;

[0074] Then, after multiplying the bit-by-bit reciprocal of the second power supply voltage-load state fine-grained interactive aggregate coding fairness target vector by the first power supply voltage-load state fine-grained interactive aggregate coding fairness target vector, take the bit-by-bit logarithm with base 2 to obtain the power supply voltage-load state fine-grained interactive aggregate coding information correction vector, which is expressed by the formula:

[0075]

[0076] in, represents the bit-wise reciprocal of a vector, A correction vector representing the supply voltage-load state fine-grained interaction aggregation encoding information;

[0077] Next, the eigenvalue mean Divide by the standard deviation of the eigenvalue The square root of the quotient is multiplied by the weight hyperparameter, and then the optimized power supply voltage-load state fine-grained interactive aggregation coding vector is obtained by the power supply voltage-load state fine-grained interactive aggregation coding information correction vector point, which is expressed as follows:

[0078]

[0079] in, represents the weight hyperparameter, Indicates point addition, Representing the optimized supply voltage-load state fine-grained interaction aggregate encoding vector;

[0080] Finally, the optimized power supply voltage-load state fine-grained interaction aggregate coding vector is input into the classifier-based mode switching logic to obtain the mode switching control instruction.

[0081] Here, since the power supply voltage one-hot encoding vector and the load state information semantic aggregation encoding feature map respectively represent the one-hot encoding features of the power supply voltage and the semantic aggregation features of the full sample domain of the load state information, when performing cross-domain fine-grained aggregation analysis based on modal prior, the semantic feature population attributes corresponding to different modal data will have differences in cross-domain fine-grained aggregation fairness based on the modal prior level, thereby affecting the inclusiveness of the aggregation feature distribution of the power supply voltage-load state fine-grained interactive aggregation encoding vector, and reducing the accuracy of the mode switching control instructions obtained by inputting the classifier-based mode switching logic.

[0082] Therefore, taking into account the attribute level fairness differences of the data population corresponding to the sequence fusion features of the power supply voltage-load state fine-grained interactive aggregation coding vector, in order to improve the aggregation inclusiveness under the feature distribution diversity of the power supply voltage-load state fine-grained interactive aggregation coding vector, the crossover probability value constraint based on the power supply voltage-load state fine-grained interactive aggregation coding vector is used as the interactive fairness target representation to correct the group feature information interactive propagation of the power supply voltage-load state fine-grained interactive aggregation coding vector, and the unified statistical feature response interaction based on the power supply voltage-load state fine-grained interactive aggregation coding vector is used as the feature distribution multi-level fairness target bias to achieve a robust distribution fairness unified representation of the power supply voltage-load state fine-grained interactive aggregation coding vector, form a fair collaboration paradigm under the feature distribution framework of the power supply voltage-load state fine-grained interactive aggregation coding vector, and improve the accuracy of the mode switching control instructions obtained by its input classifier-based mode switching logic.

[0083] After receiving the mode switching control instruction, the central control unit (usually a microcontroller or a dedicated power management integrated circuit) needs to deeply analyze this instruction. The analysis process aims to understand the specific meaning of the instruction and the operation content it indicates, such as switching the first voltage conversion module to pass-through mode or buck mode, or adjusting the output voltage level of the second voltage conversion module. For complex multi-level conversion architectures, it may be necessary to further refine the operating parameters of each sub-module, such as setting specific output current limits, frequency modulation strategies, etc. In addition, considering the special requirements in different application scenarios, the control instruction may also contain additional information such as priority settings and safety protection measures, all of which are factors that must be considered during the analysis process.

[0084] After completing the instruction parsing, the next step is to convert the parsing results into actual operable electrical signals and send them to the corresponding execution components through the preset data transmission path. The data transmission path here can be a directly connected hardwire interface or a serial bus based on a communication protocol (such as I2C, SPI, UART, etc.). The choice of which method depends on the overall architecture of the system and the physical distance between the components. For close-range and high-real-time scenarios, the hardwire interface provides the fastest and most stable solution; for distributed complex systems, it is more suitable to use a serial bus to achieve flexible and efficient communication. Regardless of the form, it is crucial to ensure signal integrity and anti-interference capabilities. To this end, auxiliary equipment such as filters and isolators can be added at the hardware level, and verification codes, retransmission mechanisms and other means can be used at the software level to enhance the reliability of data transmission.

[0085] When the execution component receives the parsed and packaged control signal, it will respond immediately according to the internal logic. Taking the first voltage conversion module as an example, if it receives an instruction to switch to the pass-through mode, the module will stop its original step-down conversion work and enter an almost lossless state, so that the input voltage can be directly transmitted to the downstream second voltage conversion module. Similarly, for the second voltage conversion module, it may adjust its output characteristics according to the new configuration parameters, such as changing the output voltage value to meet the needs of a specific load. It is worth noting that since different voltage conversion modules may have different reaction times, this should be fully considered when designing the control system to ensure that the modules are coordinated and consistent to avoid overall performance degradation due to differences in response speed.

[0086] In summary, the intelligent power supply circuit system based on the embodiment of the present application is explained, which first obtains the power supply voltage, collects the load status information of each load at the same time, and introduces the data processing technology based on deep learning to semantically represent the power supply voltage and the load status information of each load and perform fine-grained cross-modal fusion, so as to achieve a fine-grained understanding of the association between the power supply voltage and the load status, and then intelligently recommends the appropriate voltage conversion mode on this basis. In this way, the power supply voltage can be dynamically adjusted to meet the specific needs of different loads, thereby further improving the power supply efficiency of the power supply circuit and the endurance of the smart terminal.

[0087] Furthermore, a control method for an intelligent power supply circuit is also provided.

[0088] Figure 4 FIG. 1 is a flow chart of a control method of an intelligent power supply circuit according to an embodiment of the present application. Figure 4As shown, the control method of the intelligent power supply circuit includes the steps of: S1, obtaining the supply voltage; S2, collecting the load status information of each load connected to each second voltage conversion module to obtain a set of load status information; S3, performing one-hot encoding on the supply voltage to obtain a supply voltage one-hot encoding vector; S4, performing semantic embedding encoding on each load status information in the set of load status information to obtain a set of load status information semantic embedding encoding vectors; S5, performing cross-domain aggregation analysis based on a priori clustering centers on the supply voltage one-hot encoding vector and the set of load status information semantic embedding encoding vectors to obtain a supply voltage-load status fine-grained interactive aggregation encoding vector; S6, generating a mode switching control instruction based on the supply voltage-load status fine-grained interactive aggregation encoding vector.

[0089] Here, those skilled in the art can understand that the specific operations of each step in the control method of the above-mentioned intelligent power supply circuit have been referred to above. Figures 1 to 3 The intelligent power supply circuit system has been described in detail, and therefore, its repeated description will be omitted.

[0090] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0091] In the above embodiments, the description of each embodiment has its own emphasis. For the parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0092] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0093] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0094] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. An intelligent power supply circuit system, comprising a control module, a first voltage conversion module and at least one second voltage conversion module; wherein: The control module is connected to the first voltage conversion module, and is used to control the first voltage conversion module to be in a pass-through mode or a step-down mode according to an input voltage signal, so as to output a first output voltage according to a power supply voltage; the at least one second voltage conversion module is used to perform voltage conversion on the first output voltage, so as to output a second output voltage for supplying a load connected thereto, wherein the control module comprises: A voltage information acquisition unit, used to acquire the supply voltage; A load status information acquisition unit, used to collect the load status information of each load connected to each of the second voltage conversion modules to obtain a set of load status information; A supply voltage information encoding unit, configured to perform one-hot encoding on the supply voltage to obtain a one-hot encoding vector of the supply voltage; A load information semantic encoding unit, used for performing semantic embedding encoding on each load state information in the set of load state information to obtain a set of load state information semantic embedding encoding vectors; A cross-domain aggregation analysis unit, used for performing a cross-domain aggregation analysis based on a priori clustering centers on the set of the power supply voltage one-hot encoding vector and the load state information semantic embedding encoding vector to obtain a power supply voltage-load state fine-grained interactive aggregation encoding vector; A mode switching control unit, configured to generate a mode switching control instruction based on the supply voltage-load state fine-grained interactive aggregation coding vector, wherein the mode switching control instruction includes switching to a pass-through mode or switching to a buck mode; The cross-domain aggregation analysis unit includes: A cluster center calculation subunit, used to calculate the cluster center between the set of the load state information semantic embedding encoding vectors and the supply voltage one-hot encoding vector to obtain a supply voltage-load state prior cluster center encoding vector; The fine-grained aggregation subunit is used to input the set of the power supply voltage-load state prior clustering center encoding vector and the load state information semantic embedding encoding vector into the cluster center-based cross-domain fine-grained aggregation descriptor to obtain the power supply voltage-load state fine-grained interactive aggregation encoding vector.

2. The intelligent power supply circuit system according to claim 1, characterized in that: The voltage information acquisition unit is used to: The input voltage signal is sampled using an analog-to-digital converter to extract the supply voltage value at the current time point in real time to obtain the supply voltage.

3. The intelligent power supply circuit system according to claim 2, characterized in that: The load information semantic encoding unit is used to: Use a semantic embedding encoder based on the ALBERT model to perform semantic embedding encoding on each load state information in the set of load state information to obtain a set of semantic embedding encoding vectors of the load state information.

4. The intelligent power supply circuit system according to claim 3, characterized in that: The cluster center calculation subunit is used for: Inputting the set of load state information semantic embedding coding vectors into a modal kernel feature extraction network to obtain a load state information kernel semantic feature coding vector; Based on the supply voltage one-hot encoding vector and the load state information kernel semantic feature encoding vector, the supply voltage-load state prior clustering center encoding vector is determined.

5. The intelligent power supply circuit system according to claim 4, characterized in that: The fine-grained aggregation subunit is used to: Calculating the Poincare distance of each load state information semantic embedding coding vector in the set of load state information semantic embedding coding vectors relative to the supply voltage-load state priori clustering center coding vector to obtain a set of load state information semantic clustering distances; Inputting the set of load state information semantic clustering distances into a sparse encoder based on a gating mechanism to obtain a set of load state information gating clustering weight coefficients; Calculating the positional difference of each load state information semantic embedding coding vector in the set of load state information semantic embedding coding vectors relative to the supply voltage-load state prior clustering center coding vector to obtain a set of load state information semantic clustering differential feature vectors; Based on the set of load state information gated clustering weight coefficients, weighted aggregation is performed on the set of load state information semantic clustering differential feature vectors to obtain the supply voltage-load state fine-grained interactive aggregation coding vector.

6. The intelligent power supply circuit system according to claim 5, characterized in that: The mode switching control unit is used to: The supply voltage-load state fine-grained interaction aggregate coding vector is input into the classifier-based mode switching logic to obtain the mode switching control instruction.

7. The intelligent power supply circuit system according to claim 6, characterized in that: The mode switching control unit is used to: Using the fully connected layer of the mode switching logic to fully connect the power supply voltage-load state fine-grained interactive aggregation coding vector to obtain a power supply voltage-load state fine-grained interactive aggregation fully connected coding vector; Input the power supply voltage-load state fine-grained interactive aggregated fully connected coding vector into the Softmax classification function of the mode switching logic to obtain the probability value of the power supply voltage-load state fine-grained interactive aggregated coding vector belonging to each type label, wherein the type label includes a pass-through mode and a buck mode; The type label corresponding to the largest one of the probability values ​​is determined as the mode switching control instruction.

8. A control method for an intelligent power supply circuit, used in the intelligent power supply circuit system according to claim 1, characterized in that: include: Get the supply voltage; Collecting load status information of each load connected to each second voltage conversion module to obtain a set of load status information; One-hot encoding the supply voltage to obtain a supply voltage one-hot encoding vector; Performing semantic embedding coding on each piece of load state information in the set of load state information to obtain a set of load state information semantic embedding coding vectors; Performing a cross-domain aggregation analysis based on a priori clustering centers on the set of the power supply voltage one-hot encoding vector and the load state information semantic embedding encoding vector to obtain a power supply voltage-load state fine-grained interactive aggregation encoding vector; Based on the power supply voltage-load state fine-grained interactive aggregation coding vector, a mode switching control instruction is generated.

Images