Power adapter charging protocol identification and compatibility self-learning optimization method

Abstract

The invention relates to a power adapter charging protocol identification and compatibility self-learning optimization method. After the adapter establishes physical connection with a target terminal, a voltage signal output by the target terminal is collected, a handshake data sequence is formed, feature extraction is performed on the handshake data sequence, and a first protocol feature vector is generated. And performing similarity matching on the feature vectors in a protocol feature library to determine a historical protocol category and extract a corresponding historical charging parameter template. And during voltage and current step-by-step adjustment, collecting target terminal load impedance change data, extracting impedance spectrum features, and fusing the impedance spectrum features with the first protocol feature vector to form an enhanced protocol feature vector. And based on the enhancement protocol feature vector and the historical parameter template, predicting negotiation parameters supported by the target terminal, generating a predicted negotiation parameter set, and applying the predicted negotiation parameter set to a charging negotiation process, thereby realizing stable charging output with the target terminal. According to the method, the compatibility and the self-adaptive capability of the power adapter to various fast charging protocols can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power adapters, and in particular to a method for self-learning and optimizing charging protocol identification and compatibility of a power adapter. Background Art

[0002] In the prior art, after the power adapter establishes a connection with the terminal device, it usually negotiates parameters through a charging protocol to achieve fast charging. Common charging protocols include USB Power Delivery (USB-PD), Qualcomm Quick Charge (QC), MediaTek Pump Express, Huawei SuperCharge (SCP), and some manufacturer-defined private protocols. During the handshake phase, these protocols exchange information such as voltage requests, current capabilities, power negotiation levels, protocol versions, etc. through physical connection lines such as CC lines, D+ / D- lines, or specific data channels. The adapter identifies the protocol type supported by the terminal based on these communication signals, and then outputs the corresponding charging voltage and current according to the parameters specified by the protocol. In the existing technical solutions, the adapter is usually pre-installed with fixed identification logic or hardware control circuits for various standard protocols. Some adapters are integrated with control chips that support multi-protocol identification to adapt to the standard features of mainstream fast charging protocols.

[0003] However, as terminal manufacturers continue to introduce new charging technologies and customized protocols, protocol standards are becoming increasingly complex and diverse. Existing adapters are often unable to accurately identify terminal characteristics when faced with new protocols or protocol variants that are not included in the preset protocol library. In some scenarios, the adapter may misidentify the protocol category, fail to complete effective charging negotiation, or even fall back to low-power charging mode, affecting charging efficiency and terminal compatibility experience. In addition, existing adapters mainly rely on static rule matching and lack the ability to dynamically learn terminal behavior characteristics based on real-time interactive data and continuously optimize recognition logic, resulting in obvious limitations in the adapter's recognition of new protocols and long-term compatibility evolution.

[0004] In view of this, there is an urgent need to provide a new charging protocol identification and compatibility self-learning optimization solution. Summary of the Invention

[0005] This application provides a power adapter charging protocol identification and compatibility self-learning optimization method to improve the compatibility and adaptability of the power adapter to multiple fast charging protocols.

[0006] This application provides a method for self-learning and optimizing charging protocol identification and compatibility of a power adapter, including:

[0007] After the adapter establishes a physical connection with the target terminal, the voltage signal, current signal and communication protocol signal output by the target terminal during the handshake phase are collected to form a handshake data sequence reflecting the initial connection status;

[0008] Performing feature extraction on the handshake data sequence, extracting multidimensional features including handshake signal segments, voltage step amplitudes, current change slopes, and communication command fields according to preset feature mapping rules, and generating a first protocol feature vector;

[0009] Using the first protocol feature vector, a similarity match is performed in a protocol feature library preset by the adapter to determine a historical protocol category with the highest similarity to the first protocol feature vector, and a historical charging parameter template corresponding to the historical protocol category is retrieved from the protocol feature library, the historical charging parameter template including voltage, current, and communication parameter setting ranges associated with the historical protocol category;

[0010] During the gradual adjustment of the output voltage and current of the adapter, the load impedance change data of the target terminal is continuously collected, and the impedance spectrum characteristics are extracted based on the time series information of the load impedance change; and the impedance spectrum characteristics are fused with the first protocol feature vector to generate an enhanced protocol feature vector;

[0011] Based on the enhanced protocol feature vector and the historical parameter template, the negotiation parameters that the target terminal may support are predicted, a predicted negotiation parameter set is generated and applied to the charging negotiation process to establish a stable charging output with the target terminal.

[0012] The beneficial effects of the technical solution provided by this application include:

[0013] (1) By collecting and extracting multi-dimensional feature information output by the target terminal during the handshake phase, the terminal protocol features can be fully captured, improving the accuracy and completeness of protocol recognition. (2) By introducing load impedance change data and its spectral characteristics, the expressive power of the protocol feature vector is enhanced, effectively improving the recognition and discrimination capabilities in scenarios with similar or ambiguous protocols. (3) By integrating the enhanced protocol feature vector with the historical charging parameter template, the dynamic prediction of charging negotiation parameters is achieved, improving the success rate of charging negotiation and the stability of charging output. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flowchart of a power adapter charging protocol identification and compatibility self-learning optimization method provided in the first embodiment of the present application. DETAILED DESCRIPTION

[0015] The following description sets forth many specific details to facilitate a thorough understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the scope of the present application. Therefore, the present application is not limited to the specific implementations disclosed below.

[0016] The first embodiment of the present application provides a method for self-learning and optimizing the charging protocol identification and compatibility of a power adapter. Figure 1 , which is a schematic diagram of the first embodiment of this application. Figure 1 A first embodiment of the present application provides a method for self-learning and optimizing charging protocols and compatibility of a power adapter, which is described in detail.

[0017] Step S101: After the adapter establishes a physical connection with the target terminal, the voltage signal, current signal and communication protocol signal output by the target terminal during the handshake phase are collected to form a handshake data sequence reflecting the initial connection state.

[0018] During step S101 of the present invention, the power adapter first establishes an effective connection with the target terminal via a physical connection interface, which may include, but is not limited to, a USB Type-C interface, a Micro-USB interface, or a dedicated charging interface. After detecting that the physical layer connection is established, the adapter initiates the handshake data collection process. The handshake phase is the initialization phase of the charging protocol negotiation. During this phase, the target terminal outputs multiple signal types, including its charging capabilities and negotiation requests, via the communication line. The adapter must fully collect all relevant signals during this phase.

[0019] Voltage signal acquisition involves continuous sampling of voltage changes on the Vbus line in the connection cable. The sampling frequency must be at least five times the communication protocol change frequency to ensure that all possible voltage jumps and waveform characteristics are captured. A sampling frequency range of 100kHz to 1MHz is generally recommended. Current signal acquisition involves connecting a high-precision sampling resistor in series with the adapter output and using a differential amplifier circuit or a high-precision Hall effect current sensor to measure the current change curve in real time. This records the current step response generated during the handshake period, the initial load current, and the current rise characteristics within a short period of time.

[0020] The acquisition of communication protocol signals is configured accordingly based on the type of connected interface. For example, in the USB Type-C protocol, logic level transitions, voltage encoding signals, and serial data communication content are collected via CC1 and CC2 lines, respectively. In the D+ / D- communication system, differential signals are used to capture level changes and data packet information on the positive and negative data lines. To ensure the integrity of communication data, the adapter must have a built-in high-speed digital acquisition unit or protocol parsing circuit capable of real-time capture of protocol information in handshake packets, including handshake commands, requested voltage levels, supported power ranges, protocol version numbers, and power object request packets.

[0021] After collecting the raw data of all the aforementioned voltage, current, and communication protocol signals, they are time-sequenced and integrated according to timestamp order to form a complete handshake data sequence. This handshake data sequence accurately reflects the complete interaction between the adapter and the target terminal during the initial physical connection. Subsequent feature extraction, protocol identification, and parameter prediction are analyzed and processed based on this data sequence.

[0022] Furthermore, after the adapter establishes a physical connection with the target terminal, the voltage signal, current signal and communication protocol signal output by the target terminal during the handshake phase are collected to form a handshake data sequence reflecting the initial connection state, including:

[0023] After the adapter establishes a physical connection with the target terminal, a perturbation voltage sweep with an amplitude not exceeding the standard handshake threshold variation range is applied to the target terminal. At the same time, the voltage signal output by the target terminal during the perturbation process is collected in real time, and the voltage stability interval width and perturbation recovery delay time are extracted to generate voltage perturbation stability characteristic data.

[0024] Determine a load dynamic window based on the voltage perturbation stability characteristic data, perform high-frequency short-window current sampling at the initial stage of load change within the window, collect the current signal of the target terminal during the transient load transition, extract the current climbing slope, impedance oscillation frequency and dynamic damping characteristics, and generate current transient dynamic characteristic data;

[0025] Determining a communication acquisition timing window based on the voltage perturbation stability characteristic data and the current transient dynamic characteristic data, acquiring a communication protocol signal within the window, extracting a standard communication command field and an inter-code drift amount and a logic synchronization offset time between adjacent communication frames, and generating communication synchronization offset characteristic data;

[0026] The voltage perturbation stability characteristic data, the current transient dynamic characteristic data and the communication synchronization offset characteristic data are synchronously integrated through a timestamp alignment mechanism under unified clock control to generate a handshake data sequence reflecting the initial connection state.

[0027] After the adapter establishes a physical connection with the target terminal, it initiates the handshake phase data acquisition process. To generate a handshake data sequence reflecting the initial connection status, the adapter first collects voltage signals. To enhance the characterization of terminal characteristics, during handshake initialization, a perturbation voltage sweep is applied to the target terminal with an amplitude within the standard handshake threshold range. This perturbation voltage can be implemented using slow step adjustments or low-amplitude sinusoidal wave adjustments. The perturbation amplitude is controlled to a level that does not interfere with the execution of the handshake protocol, with a typical perturbation amplitude set between ±100mV and ±200mV.

[0028] During the perturbation application process, the adapter collects the voltage signal output by the target terminal in real time at a high frequency rate. The voltage sampling frequency is generally set at no less than 200kHz to ensure that the voltage response details are fully captured. By analyzing the stability performance of the target terminal under the action of the perturbation voltage, the time delay required for the terminal to recover to a steady state in each perturbation segment and the width of each stable voltage interval are calculated to form voltage perturbation stability characteristic data. The characteristic data includes the perturbation recovery delay time, the length of the steady-state maintenance interval, and the disturbance response smoothness measurement value, which is used to reflect the short-term voltage regulation capability and internal regulation speed of the terminal power control system.

[0029] After obtaining the voltage perturbation stability characteristic data, the system further determines the load dynamic window interval based on this characteristic data. First, the start time of the dynamic window is set based on the perturbation recovery delay time T1, and a buffer zone ΔT is set to avoid the influence of residual jitter at the end of the perturbation response. ΔT can be set to approximately 1ms. Subsequently, the system dynamically determines the load dynamic window length L1 based on the length of the perturbation stability interval and the amplitude of the perturbation's later fluctuations. It is usually set to add 5ms to 10ms after T1+ΔT.

[0030] Within this load dynamic window, the system implements short-window, high-frequency current sampling. The current sampling frequency is set between 500kHz and 1MHz to fully capture transient current behavior during load transitions. Through current rise curve slope analysis, short-period impedance oscillation frequency extraction, and amplitude stability testing, the system extracts the current rise slope, dynamic impedance oscillation frequency, and damping characteristics, generating current transient dynamic characteristic data. This characteristic data is used to characterize the transient response performance of the target terminal's internal power management system under rapid load changes.

[0031] After extracting the current transient dynamic characteristic data, the system uses the voltage perturbation stability characteristic data in conjunction with the current transient dynamic characteristic data to determine the communication acquisition timing window. The system first predicts the dynamic convergence time interval of the terminal power control system based on the impedance oscillation frequency F0 and damping time constant τ0 measured in the current transient response. The system sets the communication acquisition window starting point to a delay of 3τ0 to 5τ0 after the end of the load dynamic window to avoid instabilities in the final stages of current transient oscillations. This allows communication acquisition to occur during a period when power management is generally stable but protocol control is still in the dynamic interaction phase.

[0032] Within this timing window, the system collects communication protocol signals, extracts the communication command field, and further extracts the logical inter-symbol drift and synchronization offset time difference between adjacent communication frames. Communication drift is obtained by accumulating logical timing errors frame by frame. Inter-symbol drift uses clock reference synchronization measurement to calculate the periodic offset error between logical symbols, forming communication synchronization offset signature data.

[0033] Ultimately, the system integrates voltage perturbation stability characteristic data, current transient dynamic characteristic data, and communication synchronization offset characteristic data into a fused characteristic data sequence through a timestamp alignment mechanism controlled by a unified clock. This fused data is corrected for consistency in both the time and data dimensions, forming a handshake data sequence with a unified encoding format. This handshake data sequence fully characterizes the voltage, current, and communication multi-channel behavior of the adapter and target terminal during the initial physical connection phase, serving as an important input for subsequent feature extraction, protocol identification, and parameter prediction analysis.

[0034] Step S102: performing feature extraction on the handshake data sequence, extracting multi-dimensional features including handshake signal segments, voltage step amplitudes, current change slopes, and communication command fields according to preset feature mapping rules, and generating a first protocol feature vector.

[0035] After completing the handshake data sequence acquisition in step S101, proceed to step S102 for feature extraction. The goal of this step is to systematically analyze and quantify the various charging protocol interaction features contained in the handshake data sequence, extract key feature data that facilitates subsequent protocol identification and parameter prediction, and represent them in a unified data structure to ultimately form a first protocol feature vector.

[0036] First, the handshake data sequence undergoes data preprocessing, performing data denoising, time synchronization, and feature windowing for different signal types. Low-frequency noise is suppressed on the voltage and current signals using sliding average and median filters to ensure waveform smoothness and feature stability. Communication protocol signals are decoded and frame synchronized according to the communication standard to identify valid data segments. All signals are uniformly aligned to a sampling timeline with a unified time base, ensuring strict alignment between different data sources during feature extraction.

[0037] After data preprocessing is complete, multi-dimensional feature extraction is performed. To extract handshake signal segments, the system uses pre-set feature window partitioning rules to extract timing segments from the handshake sequence that contain key negotiation processes, such as the handshake start frame, power request frame, power negotiation confirmation frame, and charging mode switch signal segment. Each segment is recorded based on its timing position, duration, and signal encoding characteristics.

[0038] The voltage step amplitude characteristics are calculated by detecting the location of the transition points and the amplitude change in the voltage signal curve. Specifically, by detecting monotonically increasing, decreasing, or stable platform changes in the Vbus voltage curve, the start voltage, end voltage, and transition amplitude of each change are recorded to form a multi-segment step amplitude data set. This data reflects the voltage capability request made by the terminal during the handshake process and the voltage output response characteristics of the adapter.

[0039] The current slope characteristic is obtained by calculating the slope of the current signal at different load-loading moments during the handshake period. Using a sliding window derivative algorithm, the current rate of change is calculated for the current sampling sequence before and after the power request loading moment, obtaining the forward loading slope and the release unloading slope values. This characteristic reflects the terminal's instantaneous load behavior and current stability characteristics under different power requests.

[0040] Communication command field feature extraction: Based on the decoded communication data content, the main negotiation command fields are extracted according to the protocol specification, including but not limited to the protocol version identifier, supported power range, power object index, requested current limit, maximum voltage support range, fast charging feature identifier, reserved extension fields, etc., to form a standardized command field vector. For non-standard protocols, custom command fields can also be extracted through protocol parsing templates.

[0041] After normalization and feature encoding conversion, these various feature data are combined into multi-dimensional numerical vector data to form the first protocol feature vector. This feature vector fully describes the charging protocol behavior characteristics exhibited by the current target terminal during the handshake phase, providing standardized, high-dimensional input data for subsequent protocol matching, impedance fusion, and parameter prediction.

[0042] Furthermore, the feature extraction is performed on the handshake data sequence, and multi-dimensional features including handshake signal segments, voltage step amplitudes, current change slopes, and communication command fields are extracted according to a preset feature mapping rule to generate a first protocol feature vector, including:

[0043] Performing time segmentation on the voltage signal in the handshake data sequence, identifying the continuous voltage stable platform and the step jump point, extracting the voltage starting value, ending value and jump amplitude of each step segment, and forming voltage step amplitude characteristic data;

[0044] Dividing the current signal in the handshake data sequence into a perturbation response window, extracting the perturbation recovery time and the current change slope during the recovery process in a short period after each voltage step jump, and generating current change slope characteristic data and short-term perturbation stability characteristic data in combination with the short-term fluctuation amplitude in the perturbation response window;

[0045] Performing frame-level analysis on the communication protocol signal in the handshake data sequence, extracting the standard communication command field, and simultaneously extracting the sending timing and inter-frame interval corresponding to the communication command field during the handshake process to form timing-annotated communication command feature data;

[0046] Cross-correlating the current change slope characteristic data with the communication command field in the corresponding time period, calculating the response consistency offset between different current climbing behaviors and specific communication command fields, and forming cross-coupling characteristic data;

[0047] The voltage step amplitude characteristic data, current change slope characteristic data, short-term disturbance stability characteristic data, timing annotation communication command characteristic data and cross-coupling characteristic data are subjected to multi-level nested feature encoding, and a high-dimensional nested coding vector is generated by combining the main features and derived features to form the first protocol feature vector.

[0048] After collecting the handshake data sequence, the system performs feature extraction on it to achieve high-precision identification of charging protocol types and optimize compatibility. This method uses multidimensional feature extraction logic based on preset feature mapping rules to fully capture voltage behavior, current dynamics, communication logic, and the interactive response characteristics of the three, ultimately generating a first protocol feature vector that fully characterizes the terminal protocol behavior.

[0049] First, the system performs time-sequential segmentation on the voltage signal in the handshake data sequence. During the charging handshake phase, the voltage signal typically exhibits multiple stable segments and transient transitions. Using a continuity detection algorithm, the system divides the entire voltage sequence into stable segments and transient transitions.

[0050] The detection logic for a stable segment is as follows: if the absolute value of the voltage rate of change falls below a set rate-of-change threshold over multiple consecutive sampling periods, the segment is considered stable. The voltage rate-of-change threshold is set in volts per millisecond and is typically no greater than 10 millivolts per millisecond. For example, at any two consecutive sampling points, if the rate of change (the voltage value at the subsequent sampling point minus the voltage value at the previous sampling point, divided by the sampling interval) falls below the threshold, the voltage is considered to be changing slowly during that period and is considered stable.

[0051] For each stable segment, the system records the segment's start voltage, end voltage, and duration. The start voltage is the voltage at the first sampling point in the segment, and the end voltage is the voltage at the last sampling point. The stable duration is the number of sampling points in the segment multiplied by the sampling interval.

[0052] The logic for detecting a trip event is independent of the stable segment detection. A trip event is defined as a significant voltage jump between two adjacent sampling points. The system calculates the voltage difference between two consecutive sampling points and identifies a trip event if the absolute value of the difference exceeds the trip amplitude threshold. The trip amplitude threshold is measured in volts and is typically set at 50 millivolts.

[0053] For each transition event, the system records the transition amplitude, defined as the voltage value at the next sampling point minus the voltage value at the previous sampling point. The temporal position of the transition event is recorded as the sampling point index where the transition occurred, which is used for subsequent feature association.

[0054] The voltage step amplitude signature data is a sequence of all jump amplitudes, along with the time index of their occurrence, fully reflecting the terminal's voltage level request, capability declaration, and power switching intention during the handshake. Capability advertisement is a common concept at the protocol level. In all fast-charging protocols (such as USB PD, PPS, QC, SCP, VOOC, etc.), the charger typically announces to the terminal its supported voltage levels, current limits, power limits, and other capabilities at the beginning of the handshake.

[0055] After completing voltage feature extraction, the system proceeds to the perturbation response window partitioning and dynamic feature extraction of the current signal. Because current behavior typically lags behind voltage transitions, the system defines a perturbation response window based on each transition event. The perturbation response window begins at the moment of the transition event, and its length is determined by the terminal hardware control system's adjustment capabilities, typically ranging from 5 to 20 milliseconds.

[0056] Within each perturbation response window, the system first extracts the current perturbation recovery time. This is defined as the time it takes for the current to recover from a rapid change after a jump and eventually enter a new stable segment. The stable segment is determined when the current value enters a 5% tolerance range above or below the target steady-state current value and remains stable within this range for a minimum period of time.

[0057] The system then calculates the current change slope characteristic. This characteristic is approximated by calculating the first-order derivative of the current. Taking each sampling point within the perturbation window as the center, take a sampling point before and after, calculate their current difference, and then divide it by their time interval to obtain the current change rate near that time point. This method eliminates the influence of single-point instantaneous fluctuations and forms a smoother instantaneous slope sequence. The system selects the maximum value of the change rate within the entire perturbation window as the maximum current change slope characteristic of the jump event. This characteristic directly reflects the rapid adjustment capability and regulation efficiency of the internal current control loop of the target terminal after power switching.

[0058] In addition to extracting the current slope, the system also extracts short-term perturbation stability characteristics. This characteristic is obtained by calculating the RMS offset of the current signal within the perturbation window. The RMS offset is equal to the square root of the sum of the squares of the differences between the current values ​​at all sampling points within the window and the average current value within that window. The RMS offset reflects small residual oscillations during short-term dynamic adjustments and can be used to assess the residual oscillation suppression capability of the terminal's internal filter circuit.

[0059] After completing current feature extraction, the system proceeds to process communication protocol signal features. The communication signals during the handshake period typically appear as a sequence of data frames defined by the standard protocol. The system uses a frame format parsing algorithm to identify each complete data frame and extract the protocol command fields according to the protocol standard. Typical command fields include voltage request objects, power capability objects, protocol version identifiers, and role switch requests.

[0060] While extracting each command field, the system also records the time the command frame was sent and the time interval between it and the previous frame. This time information forms the communication timing characteristics, which can reflect the response speed of the terminal protocol control logic during the handshake interaction, the density of data exchange, and the stability of the protocol state machine.

[0061] After extracting voltage, current, and communication features, the system performs a cross-modal feature cross-correlation operation. This operation is used to explore the synchronous and coordinated relationship between electrical transient behavior and communication control logic. The system uses each jump event as the analysis benchmark, locates the nearest communication command frame within the corresponding perturbation current window, and records the time difference between the frame transmission time and the time when the maximum current slope occurs, which is called the response time offset. The system further records the command field type corresponding to the frame and the maximum current change slope of the jump event, forming a cross-correlation feature group that includes the command type, response time offset, and current change rate.

[0062] All extracted single-dimensional and cross-coupling features are ultimately fed into a multi-level nested encoding process. Taking each transition event as the primary unit, its voltage step amplitude characteristics, current change slope characteristics, short-term disturbance stability characteristics, timing-annotated communication command characteristics, and cross-coupling characteristics are organized into a high-dimensional nested feature vector. The entire sequence of nested feature vectors constitutes a complete first protocol feature vector, which is used in the subsequent protocol recognition model.

[0063] To eliminate the influence of different dimensions, the system preprocesses all features by normalization. Voltage features are normalized to the rated maximum voltage, current features to the rated maximum current, and time features to the system's unified clock unit. The command field is encoded as a fixed numerical index using the standard protocol command table. The fully normalized first protocol feature vector provides high-quality input data for subsequent model training and compatibility self-learning optimization.

[0064] Through the above-mentioned multi-dimensional feature modeling method, the present invention realizes the deep feature expression of the handshake data sequence in the dimensions of voltage signal stable segment identification, voltage jump event identification, current transient behavior modeling, communication protocol timing feature extraction and cross-modal interaction correlation, fully capturing the dynamic characteristics of the terminal's complex control logic during the handshake, and effectively improving the accuracy of charging protocol identification and the adaptive learning ability of new protocol compatibility.

[0065] Step S103: Use the first protocol feature vector to perform similarity matching in the protocol feature library preset by the adapter, determine the historical protocol category with the highest similarity to the first protocol feature vector, and call the historical charging parameter template corresponding to the historical protocol category from the protocol feature library, wherein the historical charging parameter template includes the voltage, current and communication parameter setting ranges associated with the historical protocol category.

[0066] After completing the generation of the first protocol feature vector in step S102, the process proceeds to step S103 for protocol matching and historical parameter template invocation. This step aims to perform similarity retrieval and classification of the first protocol feature vector within the adapter's internal protocol feature library, determine the closest historical protocol category, and retrieve the associated historical charging parameter template accordingly, which serves as an initial reference for subsequent charging parameter prediction and negotiation.

[0067] First, prepare a protocol signature library. This library is a collection of charging protocol data pre-established in the adapter firmware. Each protocol signature library data item contains a set of historical protocol feature vectors and their corresponding protocol category identifiers, as well as historical charging parameter templates accumulated during actual charging. This protocol signature library can cover various standard protocols such as USB-PD, QC, SCP, and PEP, as well as new protocol feature samples that are automatically expanded during the adapter's long-term learning process.

[0068] After receiving the first protocol feature vector, the system initiates feature vector similarity matching. This matching process uses a multidimensional vector distance metric, such as Euclidean distance, cosine similarity, Mahalanobis distance, or a weighted combination thereof, to calculate a similarity score between the first protocol feature vector and each historical protocol feature vector in the protocol feature library. To ensure accuracy, the feature vectors undergo dimensional standardization, numerical normalization, and feature weight adjustment to prevent dimensional differences from interfering with the matching results.

[0069] After completing the full database match calculation, the system sorts the similarity scores from high to low and selects the one or more historical protocol categories with the highest similarity as the current protocol recognition result. If the system has a similarity threshold condition set, when the highest matching result falls below the set threshold, the scenario can be marked as an unknown protocol or a new protocol sample for subsequent self-learning model expansion.

[0070] After determining the protocol type, the system calls the historical charging parameter template corresponding to the historical protocol type in the protocol feature library. This template is a set of stable charging behavior parameters accumulated over a long period of charging for the corresponding protocol type. It includes charging control parameters such as the typical voltage output range, maximum supported current value, communication negotiation window size, expected handshake success probability, terminal response delay distribution, power change sensitive range, and protection current setting under the protocol type.

[0071] The retrieved historical charging parameter template serves as the basis for predicting subsequent charging negotiation parameters. It is further integrated with the enhanced feature vector to create a customized charging negotiation solution that dynamically adapts to the target terminal. This similarity matching and parameter template invocation significantly improves the system's recognition accuracy and adaptability to changing protocol versions, protocol extension features, and the evolving characteristics of new terminal models.

[0072] Furthermore, performing similarity matching in a protocol feature library preset by the adapter using the first protocol feature vector to determine a historical protocol category having the highest similarity to the first protocol feature vector includes:

[0073] After receiving the first protocol feature vector, the adapter uses a built-in dynamic weight adjustment mechanism to perform weighted adjustment on different feature dimensions in the first protocol feature vector by calling the historical weight coefficients corresponding to the respective dimensions to form a weighted feature vector, wherein the historical weight coefficients are updated based on the stability performance, parameter adaptation success rate, and compatibility feedback of the target terminal in previous charging negotiation processes;

[0074] Based on the weighted feature vector, a similarity determination algorithm including Mahalanobis distance calculation and a confidence penalty factor is adopted to calculate the Mahalanobis distance between the weighted feature vector and the feature vector corresponding to each historical protocol category in the protocol feature library, wherein the covariance matrix used in the Mahalanobis distance calculation process is estimated in real time based on the stability clustering results of the historical protocol categories;

[0075] The Mahalanobis distance and the confidence penalty factor accumulated by each historical protocol category are combined to form a comprehensive similarity score, and the historical protocol category with the best current match is determined based on the comprehensive similarity score sorting;

[0076] When the comprehensive similarity score is lower than the preset matching confidence threshold, the semi-supervised learning label buffer mechanism is activated, and the current first protocol feature vector is used as a potential new protocol sample and entered into the feature buffer pool. Dynamic labels are collected during subsequent charging negotiations and compatibility feedback processes for subsequent self-learning model expansion of the protocol feature library.

[0077] After generating the first protocol feature vector, the system enters a similarity matching process within the protocol feature library to determine the charging protocol category currently used by the target terminal. This matching process not only considers the similarity calculation of the feature vector itself, but also combines historical charging compatibility data, dynamically learned weights, distance measurement, and a confidence penalty mechanism to form a complete and adaptively evolving protocol recognition architecture.

[0078] First, after receiving the first protocol feature vector, the system adjusts the weights of each dimension of the feature vector according to an internal dynamic weight adjustment mechanism. This dynamic weight adjustment mechanism means that during the maintenance of the protocol feature library, the system continuously records and analyzes the stability performance, parameter adaptation success rate, and compatibility feedback records of the terminal during previous charging negotiations. Based on these historical performance indicators, the system dynamically updates the weight contribution of each feature dimension in the similarity calculation.

[0079] Stability performance refers to whether the target terminal can maintain stable output for typical negotiation parameters under a historical protocol category over a long period of time without abnormal events such as interruptions, overcurrent, or downshifting during historical charging. The parameter adaptation success rate refers to whether the terminal frequently adjusts, corrects, or renegotiates the negotiation parameters predicted and applied by the system. Compatibility feedback comes from the compatibility score label obtained by the self-learning module during dynamic monitoring of the charging process, including indicators such as current interruption records, voltage drift records, load anomaly feedback, and termination due to user manual intervention.

[0080] During the dynamic weight adjustment process, the system will design a separate historical weight coefficient for each dimensional feature. Taking the voltage step amplitude feature as an example, if it is found in the historical charging process that a certain terminal model has a very regular and distinctive voltage jump behavior characteristic, the system will increase the weight coefficient of the voltage step feature, and vice versa. The same is true for the current change slope feature and the short-term disturbance stability feature. The weights are dynamically assigned based on their effectiveness in distinguishing the protocol category in the long-term data accumulation. Finally, the system forms a complete set of historical weight coefficients, multiplies it with the data of each dimension of the current first protocol feature vector, and forms a weighted feature vector after weighted processing.

[0081] After completing the weighted feature vector calculation, the system enters the core calculation phase of the similarity determination algorithm. Traditional Euclidean distance algorithms suffer from significant feature scale imbalance in high-dimensional spaces, which can easily lead to some high-variance features dominating the matching results. Therefore, this paper employs a composite similarity algorithm that incorporates the Mahalanobis distance and a confidence penalty factor to comprehensively enhance the robustness of similarity calculations and their ability to distinguish between classes.

[0082] The Mahalanobis distance is a distance metric widely used in multidimensional statistical analysis. It measures the actual similarity between two sample vectors even when the feature dimensions are correlated, scaled, and have uneven variances. Specifically, the system maintains a set of feature vectors for all training samples within each historical protocol category in the protocol feature library. For each historical protocol category, the system precomputes the covariance matrix between its feature dimensions and iterates this matrix in real time after each library update.

[0083] When calculating similarity, the system sequentially calculates the Mahalanobis distance between the current weighted feature vector and the center vector of each historical protocol category in the protocol feature library. The core of the Mahalanobis distance calculation lies in simultaneously considering the inverse matrix of the covariance matrix between features, which is used to correct the collaborative change trend between each feature dimension. Assuming that the current weighted feature vector is X, the center vector of the historical protocol category feature is μ, and the covariance matrix is ​​∑, the Mahalanobis distance D calculation logic is: Among them, X-μ is the characteristic difference vector, ∑ -1 is the inverse matrix of the covariance matrix. This distance metric can effectively avoid the problem of unbalanced matching caused by excessive variance of certain dimensional features or linear correlation between features.

[0084] Based on the Mahalanobis distance, the system introduces a confidence penalty factor mechanism to further dynamically adjust the matching results. The confidence penalty factor is derived from the long-term matching stability records accumulated and maintained for each historical protocol category in the protocol feature library. If a historical protocol category has been incorrectly matched multiple times recently, or if compatibility anomalies are frequently reported during the actual charging process, the system automatically increases its confidence penalty weight, reducing its score contribution in the current matching process. Conversely, if a protocol category has successfully and stably completed charging negotiations over a long period of time and maintains a high compatibility label, its penalty factor remains low, giving it higher priority in the matching ranking.

[0085] Ultimately, the system combines the Mahalanobis distance and the confidence penalty factor to form a comprehensive similarity score for the protocol category. The comprehensive similarity score uses the Mahalanobis distance as the primary distance, weighted by the confidence penalty factor to form a complete score sequence. The system sorts the protocols from lowest to highest based on the comprehensive score, with the lowest score being the optimal historical protocol category match determined by the system. The historical charging parameter template corresponding to this protocol category is then called by the system to enter the charging negotiation prediction application process.

[0086] It should be noted that in actual applications, some new terminals may not be fully included in the historical protocol feature library, or terminal behavior may exhibit fuzzy cross-library matching boundaries. In such low-similarity scenarios, the system automatically triggers the semi-supervised learning label buffering mechanism when the overall similarity score falls below the built-in confidence threshold.

[0087] When this mechanism is activated, the system temporarily stores the current first protocol feature vector in a feature buffer pool, marking it as a potential new protocol sample. This buffer pool sample is not immediately included in the formal training set. Instead, it is dynamically supplemented with labels during the subsequent complete charging negotiation process by collecting real-time terminal compatibility feedback data. Feedback data includes, but is not limited to, typical compatibility indicators such as whether parameter negotiation was completed, whether power degradation occurred, whether an interruption occurred, and whether there was long-term current drift or voltage jitter.

[0088] When the potential samples in the buffer pool have accumulated sufficiently stable compatibility feedback labels and stable charging behavior has been confirmed, the system will formally incorporate the new protocol sample into the protocol feature library training set, realizing the dynamic expansion and knowledge update of the self-learning model in the new protocol environment, thereby achieving the continuous evolution of the charging adapter protocol recognition capability.

[0089] Through the collaboration of the above-mentioned dynamic adjustment of feature weights, precise matching of Mahalanobis distance, confidence penalty correction and semi-supervised buffering mechanism, the present invention can continuously maintain high recognition accuracy and rapid adaptation capabilities to new terminals in a complex and diverse fast charging protocol environment.

[0090] Step S104: During the gradual adjustment of the adapter output voltage and current, continuously collect the load impedance change data of the target terminal, extract the impedance spectrum characteristics based on the time series information of the load impedance change; and fuse the impedance spectrum characteristics with the first protocol feature vector to generate an enhanced protocol feature vector.

[0091] After completing the protocol type identification and historical charging parameter template call in step S103, the process proceeds to step S104 to collect impedance change data, extract impedance spectrum features, and perform feature fusion. This step aims to further enrich the charging protocol feature information by introducing the impedance response behavior of the target terminal under different voltage and current conditions, thereby improving the completeness of the protocol characterization and the accuracy of subsequent parameter prediction.

[0092] When the adapter enters the actual charging output adjustment phase, it gradually adjusts the output voltage and current according to the voltage and current output ranges recorded in the historical charging parameter template and the preset charging regulation strategy. Each voltage or current adjustment is accompanied by a dynamic response behavior of the terminal load circuit. To capture this response process, the adapter synchronously collects the instantaneous voltage and current values ​​at the output terminal in real time at each output change, forming an impedance change data sequence.

[0093] Impedance change data is obtained by calculating the voltage-to-current ratio. To improve calculation accuracy, a short-window moving average method is used to smooth the instantaneous voltage and current signals, then their respective averages are calculated before impedance calculation. Each set of impedance data is accompanied by a precise timestamp, forming a complete impedance change time series. The sampling frequency must be less than the characteristic time constant of the terminal load response change. A sampling frequency of 10kHz to 100kHz is generally recommended.

[0094] After acquiring the impedance change time series, the system performs spectrum reconstruction. Spectral reconstruction uses fast Fourier transform (FFT), short-time Fourier transform (STFT), or wavelet transform algorithms to convert the impedance time series data from a time-domain signal into a frequency-domain feature representation. The spectrum features contain the impedance response characteristics of the target terminal at different frequency components, reflecting the dynamic characteristics of the terminal's power management circuitry, battery management system, and internal load switching logic, providing complementary differentiation capabilities for protocol identification.

[0095] The resulting impedance spectrum features are then quantized, encoding key parameters such as spectrum amplitude, phase, primary frequency components, and frequency-domain energy distribution into a unified impedance spectrum feature vector. To ensure feature consistency, all spectral features are uniformly divided into frequency bands, amplitude normalized, and scaled.

[0096] The system then fuses the acquired impedance spectrum feature vector with the first protocol feature vector generated in step S102. This fusion method can employ various algorithms, such as feature concatenation, weighted superposition, or feature coding mapping, to combine the original handshake phase feature information with the charging output dynamic feature information into a unified enhanced protocol feature vector. This enhanced protocol feature vector, which incorporates both static handshake signals and dynamic impedance behavior in terms of data dimensions, more completely reflects the target terminal's protocol behavior pattern in actual charging scenarios, providing a higher-dimensional and more accurate input data foundation for subsequent charging parameter predictions.

[0097] Furthermore, during the stepwise adjustment of the output voltage and current of the adapter, the load impedance change data of the target terminal is continuously collected, and the impedance spectrum features are extracted based on the time series information of the load impedance change; and the impedance spectrum features are fused with the first protocol feature vector to generate an enhanced protocol feature vector, including:

[0098] During the gradual adjustment of the adapter output voltage and current, a dynamic scanning disturbance signal is actively applied to the target terminal. The dynamic scanning disturbance signal includes a periodic, small-amplitude voltage disturbance with controlled amplitude. The dynamic scanning disturbance signal covers multiple preset frequency bands and selectively stimulates the dynamic impedance response behavior of the target terminal's power regulation loop without affecting normal charging negotiation and stability.

[0099] During each dynamic scanning disturbance, the voltage and current synchronous sampling data under the disturbance are collected in real time. The instantaneous impedance value at each moment is calculated based on the voltage and current sampling values ​​at that moment. The instantaneous impedance value is defined as the voltage sampling value at that moment divided by the corresponding current sampling value. All instantaneous impedance values ​​are arranged in chronological order to form an impedance change time series.

[0100] The impedance change time series is input into a dual-window resolution frequency band feature extraction module. First, within the low-frequency window, the impedance amplitude stability range under slow perturbations is analyzed. The low-frequency impedance amplitude characteristics are determined by statistically analyzing the mean and fluctuation amplitude of the impedance values ​​within the low-frequency sampling segment. Simultaneously, the impedance phase lag characteristics within the corresponding low-frequency band are analyzed, and the impedance phase delay phenomenon and the slowly varying load adaptation rate are recorded.

[0101] In the high-frequency window, the dynamic behavior of impedance under rapid disturbances is analyzed. By detecting the rate of change of the oscillation amplitude and the position of the oscillation peak in the high-frequency impedance numerical sequence, the high-frequency resonance peak position, the high-frequency oscillation frequency boundary, and the damping attenuation trend during the high-frequency impedance change process are extracted to characterize the rapid adjustment capability and resonance control characteristics of the terminal power management system.

[0102] After the extraction of low-frequency impedance features and high-frequency impedance features, the two are combined to form a complete impedance spectrum feature vector, and the impedance spectrum feature vector and the first protocol feature vector are fused according to the tensor cascade coding logic. The tensor cascade uniformly maps the feature dimensions, and integrates the impedance features and the original protocol features in a cross-modal high-dimensional expansion to form the final enhanced protocol feature vector for use by subsequent charging protocol identification and compatibility self-learning modules.

[0103] After generating the first protocol feature vector, to further improve protocol recognition accuracy and adaptive compatibility optimization, the system also needs to combine the target terminal's dynamic load impedance response during the handshake to construct a higher-dimensional representation of protocol behavior. This embodiment details the impedance change data collection method, the impedance spectrum feature extraction process, and the enhanced protocol feature vector fusion mechanism.

[0104] After the adapter and the target terminal complete the basic handshake process and enter the stage of gradual adjustment of the output voltage and current, the system starts the impedance data active acquisition module. Unlike the traditional passive observation of terminal load changes, the present invention adopts dynamic scanning disturbance technology to actively apply dynamic disturbance signals to the output end of the target terminal to induce the multi-band response behavior of the terminal power regulation loop, thereby obtaining more distinctive impedance characteristic data. The so-called dynamic scanning disturbance signal refers to the superposition of a slightly periodic disturbance signal of controlled amplitude to the output voltage channel of the adapter without affecting the stability of normal charging negotiation. The amplitude of the disturbance signal is designed to be within ±1% of the rated output voltage of the adapter, and is usually controlled in the range of tens of millivolts to one hundred millivolts to ensure that it will not interfere with the transmission of the protocol logic frame or trigger the terminal security protection mechanism.

[0105] In order to cover the regulation characteristics of the target terminal's internal control system under different time constants, the dynamic scanning disturbance signal covers multiple preset frequency bands in the frequency domain. Generally, the scanning frequency range extends from a few hertz to several thousand hertz. The low-frequency part is used to excite the slowly varying load regulation loop, and the high-frequency part is used to excite the resonant control unit and the fast current loop response characteristics. Within each frequency band, the system can adopt a segmented linear sweep mode, or a pseudo-random multi-frequency excitation sequence to improve frequency coverage efficiency and noise resistance. The disturbance duration of each frequency band is reasonably set according to the sampling window design to ensure the integrity of data acquisition while avoiding excessively long total test time.

[0106] During the dynamic scanning perturbation, the system simultaneously activates the high-precision voltage and current sampling module. This module uses a unified clock reference to control dual-channel synchronous sampling, ensuring perfect time alignment between the voltage and current sampling values ​​at each sampling moment. The sampling frequency is set with a sampling theorem margin based on the maximum scanning frequency, typically no less than ten times the maximum perturbation frequency, to ensure sufficient frequency domain resolution for the impedance calculation results.

[0107] Impedance is calculated by dividing the voltage sample value at each sampling moment by the corresponding current sample value to obtain the instantaneous impedance value at that moment. Specifically, at any sampling moment, the voltage sample value reflects the actual output voltage after the dynamic disturbance signal is superimposed on the current output terminal, and the current sample value reflects the real-time current absorbed by the terminal under the voltage condition. The impedance value is calculated by directly ratioing the two, avoiding the complex calculations that require additional phase shift or power factor measurements, making it suitable for real-time, high-speed processing. All instantaneous impedance values ​​are arranged in sequence according to the sampling time sequence, forming a complete impedance change time series, laying the data foundation for subsequent spectral feature extraction.

[0108] The system inputs the impedance change time series into the dual-window resolution frequency band feature extraction module to extract the dynamic impedance features of the low-frequency and high-frequency bands respectively. First, the low-frequency window data is processed. The low-frequency window covers the data segment collected during the low-frequency excitation stage of the dynamic scanning disturbance, with a typical frequency range of 1 Hz to tens of Hz. In the low-frequency band, the impedance change is mainly affected by the terminal load slow-changing adjustment logic, showing a certain degree of slow amplitude drift and small stable fluctuations. The system first calculates the interval stability characteristics of the impedance amplitude in this window, and obtains the low-frequency impedance mean by calculating the arithmetic mean of the impedance value in the window. The fluctuation amplitude index is obtained by averaging the absolute value of the deviation of all sampling points from the mean. This low-frequency impedance mean reflects the equivalent load level of the terminal under steady-state charging conditions, while the fluctuation amplitude reveals the subtle dynamic stability of its load regulation.

[0109] In addition, in the low-frequency band, the system further calculates the impedance phase lag characteristics. Since the terminal power management system has a certain phase delay when responding to disturbances, the system calculates the phase lag angle by analyzing the delay time between the input disturbance signal and the impedance amplitude fluctuation. The delay time is represented by the offset duration of the two signal zero crossing points, and then converted into a phase angle according to the excitation frequency. The larger the phase lag angle, the slower the response speed of the terminal's internal adjustment system; the smaller the phase lag angle, the more sensitive the load adjustment capability. The system simultaneously records the load slow-changing adaptation rate and models it by tracking the climbing slope trend of the impedance change curve in the low-frequency band. This feature can reveal the terminal's load adaptive balancing capability under slow power disturbances.

[0110] After extracting the low-frequency impedance signature, the system enters the high-frequency window data processing phase. The high-frequency window covers the data segment during the high-frequency excitation phase of the dynamic scanning perturbation, with a typical frequency range extending from hundreds of hertz to several kilohertz. Within the high-frequency band, impedance changes are affected by the terminal fast current loop, resonant compensation, and electromagnetic filtering components, resulting in a far more complex overall response than in the low-frequency band. The system first analyzes the rate of change of the oscillation amplitude in the impedance sequence, identifying the high-frequency response sensitivity by analyzing the trends in the amplitude rise and fall rates within the sliding window.

[0111] At the same time, the system performs oscillation peak position detection in the high-frequency band to determine high-frequency resonant behavior. Specifically, by calculating the changing trends of the first-order derivative and second-order derivative of the impedance amplitude in the frequency domain, the local extreme point is found to locate the resonant peak. The resonant peak frequency reveals the control closed-loop bandwidth range of the terminal under high-speed regulation, and is one of the core indicators for evaluating the stability of fast power jumps. The system continues to extract the damping attenuation trend during the high-frequency impedance change process, and estimates the damping coefficient by tracking the amplitude decreasing slope after the peak, reflecting its high-frequency oscillation suppression capability. The comprehensive characteristics of the high-frequency impedance spectrum effectively characterize the high-speed dynamic behavior and short-period stability of the terminal power regulation module.

[0112] After extracting the low-frequency and high-frequency impedance features, the system combines them in a preset order to form a complete impedance spectrum feature vector. To avoid the problem of feature dimension accumulation caused by simple splicing, the present invention uses tensor concatenation encoding logic to fuse the impedance spectrum feature vector with the first protocol feature vector. Tensor concatenation establishes a cross-modal correspondence between the two feature vectors in a unified feature space, forming a high-dimensional tensor representation through feature dimension alignment, weight scaling, and cross-mapping operations.

[0113] During the tensor concatenation process, mapping weights are first set based on the importance of each feature group. The voltage step amplitude, current change slope, and impedance low-frequency mean are aligned as the first-level master control features. The communication command field and the impedance high-frequency resonance peak are mapped to the protocol dynamic control layer. High- and low-frequency oscillation amplitudes and damping coefficients are used across domains to supplement the short-term dynamic correction layer. Each feature group is normalized and scaled within the tensor space to eliminate dimensional inconsistencies and ensure complete feature representation and structural stability after high-dimensional data fusion.

[0114] The resulting enhanced protocol feature vector retains the protocol control behavior characteristics of the original handshake phase while incorporating the target terminal's complete impedance behavior under multi-band dynamic load response, significantly enriching the input data dimension of the protocol classification model. This enhanced protocol feature vector is directly used by the subsequent charging protocol identification and compatibility self-learning module to improve the accuracy of new protocol recognition, enhance multi-terminal compatibility and adaptability, and support the continuous optimization of the adaptive charging parameter prediction algorithm.

[0115] Step S105: Based on the enhanced protocol feature vector and the historical parameter template, predict the negotiation parameters that the target terminal may support, generate a predicted negotiation parameter set and apply it to the charging negotiation process to establish a stable charging output with the target terminal.

[0116] After completing the generation of the enhanced protocol feature vector in step S104, the process proceeds to step S105 for charging parameter prediction and negotiation. The goal of this step is to derive the charging negotiation parameters most likely supported by the target terminal under a specific charging environment based on the enhanced protocol feature vector and the historical parameter template. These parameters are then applied to the subsequent actual charging negotiation process to ultimately establish a stable charging output state.

[0117] First, the system inputs the enhanced protocol feature vector into the charging parameter prediction model. This prediction model performs multidimensional correlation analysis based on the existing historical parameter template and the current terminal characteristics contained in the enhanced feature vector. The prediction model can use a parameter interval derivation algorithm based on similar sample clustering, a continuous parameter fitting algorithm based on a regression model, or a complex nonlinear feature mapping algorithm based on a multi-layer neural network. Regardless of the type of algorithm used, it must be fully trained before the adapter leaves the factory or during subsequent self-learning training to ensure that it can provide sufficiently reliable charging parameter prediction results when the new enhanced protocol feature vector is input.

[0118] The output of the charging parameter prediction is a set of predicted negotiation parameters, including but not limited to key charging control parameters such as the target terminal's acceptable voltage range, current limit, power object selection, output voltage fine-tuning offset, dynamic power allocation window, communication control timing window, and terminal feedback timeout control value. These parameters are generated not only within the theoretical range allowed by the standard protocol specifications, but also comprehensively consider the target terminal's stable charging capability under actual load changes, impedance response characteristics, and historical behavior patterns.

[0119] After the predicted negotiation parameter set is generated, the system applies the parameter set to the adapter charging control module and initiates the charging negotiation process with the target terminal through the protocol control unit. The adapter encapsulates various power request commands, protocol negotiation frames, and charging capability declaration data packets in accordance with the protocol standard format and sends them to the target terminal via a communication line (such as the CC line or D+ / D- channel). After receiving the negotiation request, the terminal completes the power negotiation confirmation or parameter adjustment feedback according to its internal protocol control logic.

[0120] If the terminal accepts the charging plan defined by the predicted negotiation parameter set, the adapter will officially enter the stable charging output mode and perform closed-loop output regulation and dynamic power control according to the negotiated voltage and current combination; if the terminal feedback partially modifies the parameters or the negotiation fails, the system can also execute an adaptive negotiation fallback strategy based on the data frame information returned by the terminal, iteratively optimize the negotiation parameters within a safe range, and increase the success rate of charging adaptation.

[0121] Through the prediction generation and negotiation application in this step, the adapter can achieve efficient adaptation to multiple standard protocols and variant protocols in actual application scenarios, while improving overall charging stability and compatibility, and accumulating reliable training sample data for subsequent self-learning optimization models.

[0122] A specific example is given below to facilitate understanding by those skilled in the art.

[0123] For example, the system first uses the enhanced protocol feature vector obtained in step S104 as input. Assuming that this vector includes, for example, the protocol version identifier extracted during the handshake phase, a voltage step amplitude sequence of 5V-9V-12V, a rapidly rising current slope, and a PPS support flag in the communication command field, combined with the impedance spectrum characteristics showing that the main frequency component is concentrated in the low-frequency range of 1kHz-3kHz, indicating that the terminal has the ability to quickly adjust to a steady state. Based on this enhanced protocol feature vector, the system retrieves protocol category samples from the historical parameter template library that are highly similar to Qualcomm QC4.0 and USB-PD 3.0.

[0124] In this embodiment, the charging parameter prediction model can be implemented using a support vector regression (SVR) algorithm. This algorithm is trained using a large number of terminal charging samples during the adapter's factory phase, learning the nonlinear mapping relationship between enhanced feature vectors and the optimal negotiation parameter set. After receiving the current input features, the SVR model outputs a predicted negotiation parameter set, which recommends adjustable voltage levels of 12V and 15V, a current limit of 3.0A, a power target request of 45W, a recommended output voltage fine-tuning offset of +150mV, a dynamic power allocation window of ±10%, and an 800ms timeout.

[0125] The adapter then packages a standard power request packet and sends a negotiation request to the target device. If the target device accepts the negotiation conditions, the adapter will then maintain a stable 12V / 3.0A output. If the device responds with a power downshift signal, the system automatically adjusts the output to 9V / 3.0A mode based on fallback logic, ensuring a stable and continuous charging process.

[0126] Furthermore, the method of predicting negotiation parameters that may be supported by the target terminal based on the enhanced protocol feature vector and the historical parameter template, generating a predicted negotiation parameter set and applying it to the charging negotiation process to establish a stable charging output with the target terminal includes:

[0127] receiving the enhanced protocol feature vector and the historical parameter template, performing global mapping using the enhanced protocol feature vector, determining the voltage parameter interval, the current parameter interval, and the communication control parameter interval covered by the corresponding historical parameter template, and forming a preliminary predicted parameter interval;

[0128] The preliminary prediction parameter interval is calculated jointly with the impedance spectrum characteristics and historical compatibility feedback weights in the enhanced protocol feature vector, and the central prediction value of the preliminary prediction parameter interval is corrected to form a predicted negotiation parameter set after the offset adjustment;

[0129] For the predicted negotiation parameter set, dynamically calculate the confidence boundary compensation range of voltage, current, and communication parameters based on the historical compatibility feedback weight and the boundary stability fluctuation record of the terminal's historical protocol category during the charging execution process, and generate a dynamic safety parameter interval containing upper and lower confidence boundaries;

[0130] Applying the dynamic safety parameter interval to the charging negotiation request output, initiating a power request command according to the parameter combination within the dynamic safety parameter interval during the charging negotiation process, and monitoring the negotiation response signal and real-time current and voltage fluctuation behavior returned by the terminal in real time;

[0131] Detect whether there is a negotiation failure signal, current over-limit fluctuation or instantaneous current interruption trend in the negotiation response signal and real-time electrical behavior. If so, perform a second-order downshift adjustment based on the fast feedback fallback logic, revise the current output voltage and current parameters downward by two steady-state power levels, form a second-order downshift parameter set, and update the output charging control instruction in real time to achieve abnormal negotiation fast fallback control.

[0132] After generating the enhanced protocol feature vector, the system enters the charging parameter prediction and negotiation application phase. This phase aims to derive the charging negotiation parameter combinations that the target terminal may support based on the multi-dimensional feature information extracted earlier and historical protocol library data. It also uses dynamic security mechanisms to control parameter application risks, ultimately achieving stable and efficient charging output.

[0133] In this invention's technical approach, charging parameter prediction doesn't rely solely on single parameter calculations from the enhanced protocol feature vector. Instead, it employs a multi-layered inference logic, first performing a global mapping of parameter ranges, followed by refined calibration and dynamic risk compensation. The entire charging parameter prediction process revolves around two input data types: the enhanced protocol feature vector and the historical parameter template.

[0134] First, the system receives the enhanced protocol feature vector and historical parameter template. The enhanced protocol feature vector, generated during the aforementioned protocol feature extraction process, integrates the communication behavior, electrical response behavior, and multi-band impedance dynamic behavior during the handshake phase. It is a high-dimensional feature set that fully characterizes the charging protocol characteristics of the target terminal. The historical parameter template, derived from the protocol feature library, is a pre-archived set of standard charging parameter configurations for each historical protocol category, including voltage levels, current limits, communication protocol timing windows, power object mapping rules, and control strategy boundaries.

[0135] The system uses the enhanced protocol feature vector to perform a global mapping operation, thereby screening out all available parameter intervals that cover the characteristic behavior performance of the current terminal in the historical parameter template. The so-called global mapping refers to determining which voltage, current and communication control parameter intervals are feature compatible with the current terminal by calculating the similarity index between the enhanced protocol feature vector and the parameter dimensions of the historical parameter template. The voltage parameter interval includes the range of steady-state output voltage levels that the target terminal may support, the current parameter interval covers the upper and lower limits of the maximum allowable output current, and the communication control parameter interval covers logical protocol parameters such as the number of handshake retries, negotiation response delay, and safety timeout window settings.

[0136] After the global mapping operation, the system forms a preliminary prediction parameter range, which is a roughly locked set of possible support parameters. This preliminary prediction parameter range is not yet safe for direct application and requires further refinement and optimization.

[0137] The system continues to calculate the preliminary predicted parameter interval together with the impedance spectrum characteristics and historical compatibility feedback weights in the enhanced protocol feature vector, and performs parameter offset adjustment reasoning. The core role of the impedance spectrum characteristics at this stage is to reflect the dynamic adaptability of the terminal's physical load under different power adjustment states. If the impedance spectrum characteristics show a higher frequency resonance peak, a low impedance damping coefficient, and good high-frequency stability, it indicates that the terminal has a strong ability to withstand rapid power switching, and the center value of the predicted voltage and current parameters can be appropriately adjusted upward; on the contrary, if obvious high-frequency impedance oscillation is detected and the residual fluctuation amplitude is large, the system automatically adjusts the predicted center point downward to prevent the predicted parameters from exceeding the terminal's internal control loop margin.

[0138] At the same time, historical compatibility feedback weighting is applied throughout the entire process. This weighting is derived from multiple compatibility feedback annotations recorded during the terminal's historical charging process. This compatibility feedback record integrates multiple compatibility indicators, including negotiation success rate, long-term charging stability, electrical fluctuation event statistics, and current interruption protection records. Based on this historical compatibility feedback weighting, the system applies an adjustment coefficient to the predicted offset amplitude, taking into account both current characteristic behavior and long-term stability performance, to form an offset-adjusted predicted negotiation parameter set that is both adaptable and secure.

[0139] The historical compatibility feedback weight is calculated based on the system's statistical analysis and quantification of key feedback events from previous charging sessions. The system continuously records multiple core compatibility monitoring indicators during each charging negotiation and execution cycle, including but not limited to the charging negotiation completion rate, parameter confirmation times, negotiation failure retries, power downshift frequency, current interruption protection intervention times, instantaneous current peak exceeding the limit, voltage jitter amplitude statistics, and the average duration of the overall charging process.

[0140] For each protocol category, the system assigns a weighted score to each metric based on the accumulated data from multiple charging sessions. For example, a higher charge negotiation completion rate earns a higher score, while a higher number of negotiation failures earns a lower score. A higher number of disconnections earns a heavier penalty, while longer stable charging times earn positive points. Each scoring criterion is optimized through engineering experience and long-term sample distribution, ensuring that the overall evaluation reflects the terminal's comprehensive protocol adaptation stability.

[0141] For example, if a terminal completed parameter confirmation on the first negotiation 93 times out of the last 100 charging cycles, had parameter renegotiation 7 times, and ultimately had current cutoff protection intervene 2 times, with an average charging stabilization time of 40 minutes throughout the entire process, and the average output current fluctuation range remained within 3% of the nominal value, the system would assign high-weighted points to the completion rate based on the scoring model, deduct points for renegotiation and current cutoff events, and receive positive corrections for long-term stable operation, ultimately calculating a comprehensive compatibility score. For example, based on a 100-point model, this terminal might receive an 87 in this round of historical compatibility scoring.

[0142] To facilitate model usage, the system normalizes the overall score proportionally and maps it to a weight coefficient range, forming the historical compatibility feedback weight for the current terminal under that protocol category (for example, a normalized value of 0.87). This weight directly contributes to the calculation of offset correction coefficients, dynamic confidence bound expansion intervals, and fallback logic trigger thresholds during subsequent parameter prediction, ensuring that the prediction model consistently converges toward optimal stability based on long-term learning.

[0143] The system doesn't directly use the predicted negotiation parameter set to initiate actual charging negotiations. Instead, it incorporates dynamic risk compensation logic and builds a confidence boundary model to provide a safety buffer for subsequent negotiation execution. For each voltage, current, and communication control parameter in the predicted negotiation parameter set, the system dynamically calculates the confidence boundary compensation range based on historical compatibility feedback weights and the protocol type's boundary stability fluctuations during charging execution.

[0144] The confidence bound compensation range is determined based on a statistical fluctuation model. The system performs variance analysis on the operating stability of each power level in historical charging data to determine its long-term standard deviation range, then applies a safety factor to form upper and lower confidence limits. The voltage parameter confidence bound reflects the degree of steady-state jitter after different output voltage steps, the current parameter confidence bound reflects the dynamic current peak excursion range during load changes, and the communication parameter confidence bound covers handshake frame response delay, control command delay, and retry frequency upper limit.

[0145] After the dynamic safety parameter interval calculation is complete, the system officially enters the charging negotiation application phase. When initiating a negotiation request, the system generates an initial negotiation power request command based on the confidence interval center weight within the dynamic safety parameter interval. This power request command is encapsulated in real time into a standard protocol format command frame and sent to the target terminal via the communication line.

[0146] During charging negotiation, the system continuously monitors the negotiation response signals and electrical feedback behavior returned by the terminal in real time, forming a complete real-time closed-loop control chain. Negotiation response signals include but are not limited to terminal acceptance confirmation frames, negotiation parameter revision frames, renegotiation rejection frames, and abnormal power interruption status instructions. Real-time electrical feedback behavior monitoring uses the adapter's internal sampling module to detect output current fluctuations, voltage regulation smoothness, and power switching response stability.

[0147] During the charging process, once the system detects negotiation failure feedback or refusal confirmation signal in the negotiation response signal, or real-time electrical monitoring finds abnormal phenomena such as abnormal amplification of output current fluctuation amplitude in a short period of time, voltage jitter exceeding the upper and lower confidence limits, and obvious current interruption trend, the system immediately triggers the fast feedback fallback logic to perform a second-order downshift adjustment.

[0148] The fast feedback fallback logic does not require completely exiting the current charging session. Instead, it quickly downgrades the output power level by two levels in the steady-state gear sequence based on the existing output power level, forming a second-order downshift parameter set. The voltage and current parameters are adjusted simultaneously during the downshift process, and the amplitude of each adjustment is determined according to the protocol standard step gear preset. For example, in the USB-PD protocol, the voltage downshift can adopt the standard voltage object step logic, and the current is synchronously reduced to the lower limit of the corresponding power gear. The second-order downshift control is completed in milliseconds, which greatly improves the ability to quickly suppress abnormal fluctuations, avoids the overcurrent protection from being forcibly triggered or triggering the terminal safety protection intervention, and improves the overall negotiation adaptation stability.

[0149] After applying the second-order downshift parameter set, the system continues to monitor charging stability after the downshift in real time. If terminal behavior stabilizes during the downshift mode, the system maintains charging in the new gear. If residual fluctuations persist after the downshift, the system gradually decreases the downshift level within the safety threshold until the steady-state charging safety zone is reached. The entire rapid feedback fallback logic is fully embedded in the negotiation execution process, operating as a real-time linkage control rather than independent fault handling, ensuring highly flexible protocol compatibility processing.

[0150] In the present invention, the second-order downshift adjustment refers to a power safety fallback control mechanism that is quickly executed during the charging negotiation process when the system detects that the terminal feedback contains a serious compatibility anomaly signal or the electrical parameters suddenly fluctuate beyond the confidence boundary threshold in a short period of time.

[0151] The core logic of the second-order downshift adjustment is: taking the currently executing charging power gear as the base gear, continuously lowering the two power gears according to the standard gear descending sequence supported by the protocol to form a new output control parameter set. For voltage gears, the system follows the optional voltage gear sequence defined in the charging protocol standard. For example, in the USB PowerDelivery protocol, if the current output voltage is 15V and the standard gear sequence is 20V, 15V, 9V, and 5V, the second-order downshift target is the 9V gear. For current gears, the system automatically adjusts according to the maximum safe current value corresponding to the target voltage gear to ensure that the total power output value is always within the safe power range.

[0152] In some configurations that support continuously adjustable voltage levels (e.g., the PPS protocol), the second-step downshift adjustment uses relative voltage offset calculation logic, which means that the current voltage is reduced by two step sizes. For example, if the voltage adjustment step size is 20mV, the second-step downshift will reduce the current output voltage by 40mV. The current adjustment is automatically adjusted downward according to the current-voltage coordinated regulation curve set by the protocol.

[0153] The second-order downshift parameter set represents the combined output control parameters of the new voltage level and current limit calculated for this downshift adjustment. The output takes effect immediately after the downshift, ensuring that the terminal quickly returns to a stable operating range even in the event of a dynamic anomaly, improving the overall safety and stability of the charging process.

[0154] Through the charging parameter prediction and negotiation application logic described in this embodiment, the system implements a complete charging parameter adaptive reasoning and execution system starting from enhanced feature vectors, through global mapping, offset adjustment, confidence boundary compensation, and fast fallback control. Compared with traditional fast charging compatibility algorithms based on single field recognition or fixed interval heuristics, this invention takes into account both protocol behavior modeling and terminal dynamic physical behavior analysis, integrating long-term compatibility feedback data to achieve rapid adaptation capabilities for unknown terminals, semi-standard protocol terminals, and new protocol extension terminals.

[0155] Furthermore, the power adapter charging protocol identification and compatibility self-learning optimization method further includes:

[0156] During the charging negotiation and charging execution process, the negotiation response signal, voltage and current stability, and abnormal current interruption feedback from the target terminal are monitored in real time. Based on the preset compatibility evaluation criteria, the feedback compatibility label is determined for updating the adaptive performance of the self-learning model in subsequent optimization.

[0157] During the charging negotiation and execution process, the adapter monitors the target terminal's feedback in real time during the negotiation and actual charging phases, collecting negotiation response signals returned by the terminal in response to negotiation requests initiated by the adapter. These negotiation response signals include, but are not limited to, standard protocol control frames such as handshake confirmation, power object request confirmation, power adjustment confirmation, error code return, timeout failure, and negotiation rejection. When the adapter detects each negotiation response, it records the corresponding response type, response delay, feedback parameters, and protocol version fields.

[0158] During the actual charging phase, the adapter continuously monitors the output voltage and current stability of the charging circuit. Specifically, the adapter periodically collects instantaneous measurements from the voltage and current sensors and evaluates output fluctuations using a sliding window averaging method and short-term variance calculation. If, under the negotiated parameter settings, the voltage or current outputs exhibit persistent oscillations, frequent jumps, or a tendency to deviate from the negotiated target range, the system is deemed to be unstable.

[0159] The adapter also monitors the charging process for abnormal current loss. An abnormal current loss event occurs when the output current drops below the preset current loss threshold without completing natural unloading, or when the terminal suddenly loses communication during stable charging. The adapter records the time of occurrence, voltage and current status, and contextual negotiation parameters for each abnormal current loss event.

[0160] By comprehensively evaluating negotiation response signals, voltage and current stability, and abnormal disconnections across multiple dimensions, the system generates a feedback compatibility label for the current charging cycle based on pre-set compatibility evaluation criteria. This compatibility evaluation utilizes a weighted scoring model, taking into account metrics such as negotiation success rate, stability score, number of abnormal disconnections, feedback error code weight, and negotiation retry frequency to output a compatibility label. Compatibility labels typically use discrete grading, such as high compatibility, medium compatibility, low compatibility, or failure; continuous numerical scores, such as a 0-1 floating-point compatibility score, can also be used.

[0161] The feedback compatibility label is input into the self-learning model together with the feature data involved in this round of charging process for model training and adaptive performance update. The core goal of the self-learning model is to improve the generalization ability and prediction accuracy of the protocol characteristics of new terminals in the process of protocol identification and parameter prediction. The self-learning model adopts a supervised incremental learning mechanism, and its input includes an enhanced protocol feature vector, a historical charging parameter template, and a feedback compatibility label. The enhanced protocol feature vector is derived from the aforementioned feature fusion step, which fully reflects the protocol characteristics of the current terminal in the handshake phase and the impedance response phase; the historical charging parameter template provides basic charging behavior samples associated with the protocol category; and the feedback compatibility label is used as a supervised training signal to guide the directional optimization of the model parameters.

[0162] The output of the self-learning model is the updated weight set of the charging parameter prediction model. The training process can utilize stochastic gradient descent (SGD) with regularization constraints, the Adam optimization algorithm, or its derivatives. A sliding window sample buffering mechanism can be incorporated to control the model memory window length to avoid overfitting to historical samples. The training loss function is defined by combining the compatibility score residual, prediction bias, stability weight, and parameter convergence speed. After each training session, the self-learning model synchronizes the latest model parameters to the protocol feature library, expanding the knowledge base coverage and providing an optimized model foundation for subsequent terminal adaptation.

[0163] Through the above method, the adapter can continuously absorb the evolution trend of terminal behavior characteristics after each round of charging process, and gradually improve the protocol recognition accuracy, charging negotiation success rate and overall adaptation and compatibility performance.

[0164] A second embodiment of the application provides an electronic device, comprising:

[0165] processor;

[0166] The memory is used to store a program, which, when read and executed by the processor, executes a power adapter charging protocol identification and compatibility self-learning optimization method provided in the first embodiment of the present application.

[0167] The third embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for self-learning optimization of power adapter charging protocol identification and compatibility provided in the first embodiment of the present application is executed.

[0168] Although the present application is disclosed as above with the preferred embodiments, it is not intended to limit the present application. Any person skilled in the art may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.

Claims

1. A method for self-learning and optimizing charging protocol identification and compatibility of a power adapter, characterized in that: include: After the adapter establishes a physical connection with the target terminal, the voltage signal, current signal and communication protocol signal output by the target terminal during the handshake phase are collected to form a handshake data sequence reflecting the initial connection status; Performing feature extraction on the handshake data sequence, extracting multidimensional features including handshake signal segments, voltage step amplitudes, current change slopes, and communication command fields according to preset feature mapping rules, and generating a first protocol feature vector; Using the first protocol feature vector, a similarity match is performed in a protocol feature library preset by the adapter to determine a historical protocol category with the highest similarity to the first protocol feature vector, and a historical charging parameter template corresponding to the historical protocol category is retrieved from the protocol feature library, the historical charging parameter template including voltage, current, and communication parameter setting ranges associated with the historical protocol category; During the gradual adjustment of the output voltage and current of the adapter, the load impedance change data of the target terminal is continuously collected, and the impedance spectrum characteristics are extracted based on the time series information of the load impedance change; and the impedance spectrum characteristics are fused with the first protocol feature vector to generate an enhanced protocol feature vector; Based on the enhanced protocol feature vector and the historical parameter template, the negotiation parameters that the target terminal may support are predicted, a predicted negotiation parameter set is generated and applied to the charging negotiation process to establish a stable charging output with the target terminal.

2. The power adapter charging protocol identification and compatibility self-learning optimization method according to claim 1, characterized in that: Also includes: During the charging negotiation and charging execution process, the negotiation response signal, voltage and current stability, and abnormal current interruption feedback from the target terminal are monitored in real time. Based on the preset compatibility evaluation criteria, the feedback compatibility label is determined for updating the adaptive performance of the self-learning model in subsequent optimization.

3. The power adapter charging protocol identification and compatibility self-learning optimization method according to claim 1, characterized in that: After the adapter establishes a physical connection with the target terminal, collecting the voltage signal, current signal and communication protocol signal output by the target terminal during the handshake phase to form a handshake data sequence reflecting the initial connection state, including: After the adapter establishes a physical connection with the target terminal, a perturbation voltage sweep with an amplitude not exceeding the standard handshake threshold variation range is applied to the target terminal. At the same time, the voltage signal output by the target terminal during the perturbation process is collected in real time, and the voltage stability interval width and perturbation recovery delay time are extracted to generate voltage perturbation stability characteristic data. Determine a load dynamic window based on the voltage perturbation stability characteristic data, perform high-frequency short-window current sampling at the initial stage of load change within the window, collect the current signal of the target terminal during the transient load transition, extract the current climbing slope, impedance oscillation frequency and dynamic damping characteristics, and generate current transient dynamic characteristic data; Determining a communication acquisition timing window based on the voltage perturbation stability characteristic data and the current transient dynamic characteristic data, acquiring a communication protocol signal within the window, extracting a standard communication command field and an inter-code drift amount and a logic synchronization offset time between adjacent communication frames, and generating communication synchronization offset characteristic data; The voltage perturbation stability characteristic data, the current transient dynamic characteristic data and the communication synchronization offset characteristic data are synchronously integrated through a timestamp alignment mechanism under unified clock control to generate a handshake data sequence reflecting the initial connection state.

4. The power adapter charging protocol identification and compatibility self-learning optimization method according to claim 1, characterized in that: The feature extraction of the handshake data sequence is performed, and multi-dimensional features including handshake signal segments, voltage step amplitudes, current change slopes, and communication command fields are extracted according to preset feature mapping rules to generate a first protocol feature vector, including: Performing time segmentation on the voltage signal in the handshake data sequence, identifying the continuous voltage stable platform and the step jump point, extracting the voltage starting value, ending value and jump amplitude of each step segment, and forming voltage step amplitude characteristic data; Dividing the current signal in the handshake data sequence into a perturbation response window, extracting the perturbation recovery time and the current change slope during the recovery process in a short period after each voltage step jump, and generating current change slope characteristic data and short-term perturbation stability characteristic data in combination with the short-term fluctuation amplitude in the perturbation response window; Performing frame-level analysis on the communication protocol signal in the handshake data sequence, extracting the standard communication command field, and simultaneously extracting the sending timing and inter-frame interval corresponding to the communication command field during the handshake process to form timing-annotated communication command feature data; Cross-correlating the current change slope characteristic data with the communication command field in the corresponding time period, calculating the response consistency offset between different current climbing behaviors and specific communication command fields, and forming cross-coupling characteristic data; The voltage step amplitude characteristic data, current change slope characteristic data, short-term disturbance stability characteristic data, timing annotation communication command characteristic data and cross-coupling characteristic data are subjected to multi-level nested feature encoding, and a high-dimensional nested coding vector is generated by combining the main features and derived features to form the first protocol feature vector.

5. The power adapter charging protocol identification and compatibility self-learning optimization method according to claim 1, characterized in that: The performing similarity matching in a protocol feature library preset by the adapter using the first protocol feature vector to determine a historical protocol category having the highest similarity to the first protocol feature vector includes: After receiving the first protocol feature vector, the adapter uses a built-in dynamic weight adjustment mechanism to perform weighted adjustment on different feature dimensions in the first protocol feature vector by calling the historical weight coefficients corresponding to the respective dimensions to form a weighted feature vector, wherein the historical weight coefficients are updated based on the stability performance, parameter adaptation success rate, and compatibility feedback of the target terminal in previous charging negotiation processes; Based on the weighted feature vector, a similarity determination algorithm including Mahalanobis distance calculation and a confidence penalty factor is adopted to calculate the Mahalanobis distance between the weighted feature vector and the feature vector corresponding to each historical protocol category in the protocol feature library, wherein the covariance matrix used in the Mahalanobis distance calculation process is estimated in real time based on the stability clustering results of the historical protocol categories; The Mahalanobis distance and the confidence penalty factor accumulated by each historical protocol category are combined to form a comprehensive similarity score, and the historical protocol category with the best current match is determined based on the comprehensive similarity score sorting; When the comprehensive similarity score is lower than the preset matching confidence threshold, the semi-supervised learning label buffer mechanism is activated, and the current first protocol feature vector is used as a potential new protocol sample and entered into the feature buffer pool. Dynamic labels are collected during subsequent charging negotiations and compatibility feedback processes for subsequent self-learning model expansion of the protocol feature library.

6. The power adapter charging protocol identification and compatibility self-learning optimization method according to claim 1, characterized in that: During the stepwise adjustment of the output voltage and current of the adapter, the load impedance change data of the target terminal is continuously collected, and the impedance spectrum characteristics are extracted based on the time series information of the load impedance change; The impedance spectrum feature is fused with the first protocol feature vector to generate an enhanced protocol feature vector, including: During the gradual adjustment of the adapter output voltage and current, a dynamic scanning disturbance signal is actively applied to the target terminal. The dynamic scanning disturbance signal includes a periodic, small-amplitude voltage disturbance with controlled amplitude. The dynamic scanning disturbance signal covers multiple preset frequency bands and selectively stimulates the dynamic impedance response behavior of the target terminal's power regulation loop without affecting normal charging negotiation and stability. During each dynamic scanning disturbance, the voltage and current synchronous sampling data under the disturbance are collected in real time. The instantaneous impedance value at each moment is calculated based on the voltage and current sampling values ​​at that moment. The instantaneous impedance value is defined as the voltage sampling value at that moment divided by the corresponding current sampling value. All instantaneous impedance values ​​are arranged in chronological order to form an impedance change time series. The impedance change time series is input into a dual-window resolution frequency band feature extraction module. First, within the low-frequency window, the impedance amplitude stability range under slow perturbations is analyzed. The low-frequency impedance amplitude characteristics are determined by statistically analyzing the mean and fluctuation amplitude of the impedance values ​​within the low-frequency sampling segment. Simultaneously, the impedance phase lag characteristics within the corresponding low-frequency band are analyzed, and the impedance phase delay phenomenon and the slowly varying load adaptation rate are recorded. In the high-frequency window, the dynamic behavior of impedance under rapid disturbances is analyzed. By detecting the rate of change of the oscillation amplitude and the position of the oscillation peak in the high-frequency impedance numerical sequence, the high-frequency resonance peak position, the high-frequency oscillation frequency boundary, and the damping attenuation trend during the high-frequency impedance change process are extracted to characterize the rapid adjustment capability and resonance control characteristics of the terminal power management system. After the extraction of low-frequency impedance features and high-frequency impedance features, the two are combined to form a complete impedance spectrum feature vector, and the impedance spectrum feature vector and the first protocol feature vector are fused according to the tensor cascade coding logic, wherein the tensor cascade uniformly maps the feature dimensions, and integrates the impedance features and the original protocol features in a cross-modal high-dimensional expansion to form the final enhanced protocol feature vector.

7. The power adapter charging protocol identification and compatibility self-learning optimization method according to claim 1, characterized in that: The method of predicting negotiation parameters that may be supported by the target terminal based on the enhanced protocol feature vector and the historical parameter template, generating a predicted negotiation parameter set and applying it to the charging negotiation process to establish a stable charging output with the target terminal includes: receiving the enhanced protocol feature vector and the historical parameter template, performing global mapping using the enhanced protocol feature vector, determining the voltage parameter interval, the current parameter interval, and the communication control parameter interval covered by the corresponding historical parameter template, and forming a preliminary predicted parameter interval; The preliminary prediction parameter interval is calculated jointly with the impedance spectrum characteristics and historical compatibility feedback weights in the enhanced protocol feature vector, and the central prediction value of the preliminary prediction parameter interval is corrected to form a predicted negotiation parameter set after the offset adjustment; For the predicted negotiation parameter set, dynamically calculate the confidence boundary compensation range of voltage, current, and communication parameters based on the historical compatibility feedback weight and the boundary stability fluctuation record of the terminal's historical protocol category during the charging execution process, and generate a dynamic safety parameter interval containing upper and lower confidence boundaries; Applying the dynamic safety parameter interval to the charging negotiation request output, initiating a power request command according to the parameter combination within the dynamic safety parameter interval during the charging negotiation process, and monitoring the negotiation response signal and real-time current and voltage fluctuation behavior returned by the terminal in real time; Detect whether there is a negotiation failure signal, current over-limit fluctuation or instantaneous current interruption trend in the negotiation response signal and real-time electrical behavior. If so, perform a second-order downshift adjustment based on the fast feedback fallback logic, revise the current output voltage and current parameters downward by two steady-state power levels, form a second-order downshift parameter set, and update the output charging control instruction in real time to achieve abnormal negotiation fast fallback control.

Images