A multi-cell charging power supply abnormality detection method and system based on current analysis

Abstract

This invention relates to the field of power management and anomaly detection technology, specifically to a method and system for detecting anomalies in multi-port charging power supplies based on current analysis. It is applied to a charging system containing multiple load power supply buses and several branch switching transistors. The method includes: extracting and sampling AC ripple current from the total current signal of the multiple load power supply buses and each branch; performing delay mapping on the AC ripple current to construct a multi-dimensional phase space matrix and calculate a Poincaré cross-section scatter plot; calculating the current topological information entropy based on the scatter distribution probability and comparing it with the baseline information entropy to generate an information entropy residual; comparing the information entropy residual with a safety manifold boundary threshold range, determining a charging anomaly when it exceeds the threshold, outputting a hardware interrupt signal to control the state of the corresponding branch switching transistor, and generating a diagnostic report containing physical feature parameters extracted based on the multi-dimensional phase space matrix. This invention enables early identification of anomalies in multi-port charging systems and rapid hardware-level response.

Description

Technical Field

[0001] This invention relates to the field of power management and anomaly detection technology, specifically to a method and system for detecting anomalies in multi-mobile phone charging power supplies based on current analysis. Background Technology

[0002] With the rapid popularization of multi-terminal centralized charging equipment in hospitals, offices and public duty environments, multi-phone charging power supplies need to be compatible with multiple terminal types, charging protocols and frequent plugging and unplugging conditions. Under the condition of multi-circuit parallel power supply, the coupling disturbance between branches is more complex. How to identify charging abnormalities in a timely manner without relying on terminal communication data has become an important issue in the safety management of multi-port charging equipment. Traditional multi-phone charging anomaly detection currently relies mainly on the following methods: DC current threshold monitoring, temperature detection, and charging protocol status judgment; However, DC current threshold monitoring, temperature detection, and charging protocol status judgment all have certain drawbacks. For example, DC current threshold monitoring is usually only triggered when the abnormality is already obvious, making it difficult to detect early abnormalities such as interface oxidation, cable aging, or micro-short circuits inside the battery in a timely manner; temperature detection has a response lag problem and is easily detected only after local overheating has formed. Charging protocol status judgment is easily affected by factors such as differences in terminal compatibility, the closed nature of proprietary protocols, and inconsistencies in actual access devices, making it difficult to meet the needs for early warning and rapid blocking in scenarios with multiple terminals connected together. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis, thereby solving the following technical problems: This avoids the difficulty of timely identification of early anomalies such as interface oxidation, cable aging, or internal micro-short circuits in batteries by traditional monitoring methods based on DC current thresholds. Moreover, it can achieve early identification of anomalies in multi-port charging systems and rapid hardware-level response based solely on electrical variable measurements without relying on terminal protocol data. This enables more refined safety classification and handling and explainable fault tracing.

[0004] The objective of this invention can be achieved through the following technical solutions: A method for detecting power supply anomalies in multi-mobile phone chargers based on current analysis, applied to a charging system containing multiple load power supply buses and several branch switching transistors, includes the following steps: The total current signal of multiple load power supply buses and each branch is obtained by a preset high-frequency current extraction circuit, and the corresponding AC ripple current is extracted and sampled from each total current signal by the preset high-frequency current extraction circuit. The phase space reconstruction module is used to perform delay mapping on each AC ripple current to construct the corresponding multidimensional phase space matrix, and the corresponding Poincaré cross-section scatter plot is calculated based on the respective multidimensional phase space matrix. The topological invariant calculator calculates the current topological information entropy for each path based on the scatter distribution probability of the scatter plots of each Poincaré section, and compares the current topological information entropy with the preset baseline information entropy to generate their respective information entropy residuals. The information entropy residual is compared with a preset safe manifold boundary threshold range using a hardware closed-loop blocking module. When the information entropy residual exceeds the safe manifold boundary threshold range, a charging abnormality is determined, a hardware interrupt signal is output to control the state of the corresponding branch switch tube, and a diagnostic report containing physical feature parameters extracted based on the multidimensional phase space matrix is ​​generated.

[0005] Furthermore, the step of acquiring the total current signal of multiple load power supply buses and their branches through a preset high-frequency current extraction circuit, and then stripping and sampling the AC ripple current from the total current signal through the preset high-frequency current extraction circuit, includes: The total current signal of the multi-load power supply bus and each branch is obtained by a current transformer. The total current signal is filtered using a hardware high-pass filter to remove the DC component and extract the high-frequency AC signal. The cutoff frequency of the hardware high-pass filter is set to a preset high-frequency threshold. The high-frequency AC signal is synchronously sampled by a high-speed analog-to-digital converter to generate the AC ripple current.

[0006] Furthermore, the AC ripple current is delayed and mapped using a phase space reconstruction module to construct a multi-dimensional phase space matrix, including: Obtain the preset delay time parameter; Based on the current sampling value of the AC ripple current, a preset number of delayed sampling values ​​that match the phase space embedding dimension are extracted in combination with the delay time parameter. The current sample value and the delayed sample value are combined to generate the multidimensional phase space matrix.

[0007] Furthermore, calculating the Poincaré section scatter plot based on the multidimensional phase space matrix includes: Define a cross-sectional hyperplane that matches the dimensions of the multidimensional phase space matrix; Calculate the continuous intersection points between the phase space trajectory represented by the multidimensional phase space matrix and the cross-sectional hyperplane; The coordinate data of all the continuous intersection points are summarized to generate the Poincaré section scatter plot.

[0008] Furthermore, based on the scatter distribution probability of the Poincaré section scatter plot, the current topological information entropy is calculated using a topological invariant calculator, and the current topological information entropy is compared with a preset baseline information entropy to generate an information entropy residual, including: The Poincaré section scatter plot is divided into a preset number of sub-regions based on a preset spatial division rule; Calculate the probability distribution of scattered points within each sub-region; The negative summation of the distribution probability of each sub-region and its logarithmic product is taken as the current topological information entropy; Calculate the absolute value of the difference between the current topological information entropy and the baseline information entropy, and use the absolute value of the difference as the information entropy residual.

[0009] Furthermore, a hardware closed-loop blocking module is used to compare the information entropy residual with a preset safe manifold boundary threshold range; when the information entropy residual exceeds the safe manifold boundary threshold range, a charging abnormality is determined, and a hardware interrupt signal is output to control the state of the corresponding branch switch, including: When the information entropy residual is greater than the maximum upper limit of the safe manifold boundary threshold interval, a topology breaking anomaly is determined to have occurred, and the hardware interrupt signal is output to the branch switch corresponding to the charging anomaly to generate a cut-off control command. When the information entropy residual is less than the minimum lower limit of the safety manifold boundary threshold interval, the system is determined to be in an over-coupling abnormal state, an early warning signal is output, and a state flag is generated to keep the branch switch tube conducting. When the information entropy residual is greater than or equal to the minimum lower limit and less than or equal to the maximum upper limit, the system is determined to be in a stable resonance state, a stable state data tag is generated, and no interruption signal is output.

[0010] Furthermore, a diagnostic report is generated that includes physical feature parameters extracted based on the multidimensional phase space matrix, including: When a charging anomaly is detected, the current Lyapunov index is obtained by calculating the exponential divergence rate of the phase space trajectory represented by the multidimensional phase space matrix during the evolution process, and compared with the preset Lyapunov baseline index to obtain the Lyapunov index offset. The Lyapunov exponent offset is used as the physical characteristic parameter; Match the preset device topology feature library to identify the type of load device that is malfunctioning; The diagnostic report is generated by combining the load device type with the physical characteristic parameters.

[0011] Furthermore, by matching the preset device topology feature library, the types of load devices that have experienced anomalies are identified, including: Extract the geometric boundary feature data of the Poincaré section scatter plot; The geometric boundary feature data is compared with the topology templates of various power management chips pre-stored in the device topology feature library for similarity. The device topology feature library is constructed based on pure physical current features. When the highest similarity is greater than or equal to the preset similarity threshold, the corresponding load device type is determined based on the topology template with the highest similarity; when the highest similarity is less than the preset similarity threshold, the load device type is marked as an unknown type.

[0012] A multi-mobile phone charging power supply anomaly detection system based on current analysis is applied to a charging system containing multiple load power supply buses and several branch switching transistors, comprising: A high-frequency current extraction circuit is used to acquire the total current signal of multiple load power supply buses and each branch, and to strip and sample the AC ripple current from the total current signal. The phase space reconstruction module is used to perform delay mapping processing on the AC ripple current, construct a multi-dimensional phase space matrix, and calculate the Poincaré section scatter plot based on the multi-dimensional phase space matrix. The topological invariant calculator is used to calculate the current topological information entropy based on the scatter distribution probability of the Poincaré section scatter plot, and compare the current topological information entropy with the preset baseline information entropy to generate information entropy residuals; The hardware closed-loop blocking module is used to compare the information entropy residual with a preset safe manifold boundary threshold range; when the information entropy residual exceeds the safe manifold boundary threshold range, it is determined that a charging abnormality has occurred, outputs a hardware interrupt signal to control the state of the corresponding branch switch tube, and generates a diagnostic report containing physical feature parameters extracted based on the multidimensional phase space matrix.

[0013] The beneficial effects of this invention are: 1. This invention removes AC ripple and constructs a multi-dimensional phase space matrix by pre-setting a high-frequency extraction circuit, transforming a one-dimensional waveform into a measurable topological state quantity. This overcomes the hysteresis of traditional DC monitoring and can effectively extract early weak anomalies such as interface oxidation or cable aging before overcurrent occurs, achieving timely protection based on hardware closed loop. 2. This invention calculates the Poincaré scatter probability and information entropy residual based on pure physical current characteristics, without relying on or reading terminal communication message data throughout the process; this design effectively avoids the defects of traditional detection that are easily affected by differences in terminal compatibility and the closed nature of proprietary protocols, and significantly enhances the anti-tampering and generalization capabilities in complex terminal mixed-connection environments. 3. This invention utilizes a hardware closed-loop blocking module to compare the information entropy residual with the safety manifold boundary threshold range, and implements hierarchical control for different degrees of evolution deviation; it cuts off branch switching tubes for topology violations, provides early warning for over-coupling outputs and keeps them on, optimizes the single fixed threshold, and takes into account both system safety and multi-port power supply continuity. 4. After the present invention determines an anomaly, the system extracts physical parameters such as the Lyapunov exponent offset and combines them with the scatter plot geometric boundary feature comparison template to automatically identify the load type; thereby generating a diagnostic report containing accurate physical features and priority investigation directions, transforming simple binary anomaly alarms into interpretable professional diagnoses, and thus outputting fault feature data with multi-dimensional physical characteristics. Attached Figure Description

[0014] The invention will now be further described with reference to the accompanying drawings.

[0015] Figure 1 A flowchart illustrating a method for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis, provided in an embodiment of this application; Figure 2 This is a schematic diagram of a multi-mobile phone charging power supply anomaly detection system based on current analysis provided in an embodiment of this application. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 A method for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis is applied to a charging system containing multiple load power supply buses and several branch switching transistors. The method includes the following steps: acquiring the total current signal of the multiple load power supply buses and each branch through a preset high-frequency current extraction circuit; extracting and sampling the corresponding AC ripple current from each of the total current signals using the same preset high-frequency current extraction circuit; performing delay mapping processing on each of the AC ripple currents using a phase space reconstruction module to construct a corresponding multi-dimensional phase space matrix; and calculating the corresponding Poincaré scatter plot based on the respective multi-dimensional phase space matrix. The topological invariant calculator calculates the current topological information entropy for each path based on the scatter distribution probability of the scatter plots of each Poincaré section, and compares the current topological information entropy with the preset baseline information entropy to generate their respective information entropy residuals. When the information entropy residual exceeds the safety manifold boundary threshold range, a charging abnormality is determined, a hardware interrupt signal is output to control the state of the corresponding branch switch tube, and a diagnostic report containing physical feature parameters extracted based on the multidimensional phase space matrix is ​​generated.

[0018] This embodiment provides a multi-mobile phone charging power supply anomaly detection mechanism based on current analysis; specifically, the mechanism is deployed in a multi-port charging cabinet in the hospital emergency command center, which simultaneously powers multiple medical and nursing work mobile phones, mobile nursing terminals and backup communication terminals. In this scenario, with mixed terminal brands, different fast charging protocols, and frequent plugging and unplugging, traditional monitoring methods based on DC current thresholds are unable to identify early anomalies such as interface oxidation, cable aging, or micro-short circuits inside the battery in a timely manner. Therefore, this embodiment measures and determines the dynamic topological characteristics of high-frequency AC ripple to quickly block abnormal branches. The details are as follows: Current acquisition channels are arranged on the multi-load power supply bus and each branch to obtain the total current signal of each channel; the total current signal here includes both DC component and high-frequency ripple component. In this embodiment, the AC ripple sequence obtained by high-frequency extraction of the link is sent into the main topology analysis process, while the DC component is mainly used for idle port identification, protection range configuration and result verification, and does not directly participate in the phase space matrix construction; the AC ripple part is stripped out by the high-frequency current extraction circuit, and a one-dimensional ripple sequence arranged continuously in time is formed by the synchronous sampling unit; To facilitate understanding, a simplified numerical example can be used: Suppose that the ripple sampling values ​​obtained by a certain branch at six consecutive sampling times are 0.12, 0.18, 0.11, -0.03, -0.10, and 0.05, respectively, and the unit can be the normalized value in amperes; these sampling values ​​on their own are not enough to intuitively reflect the anomaly, but after subsequent delay mapping, they can form trajectory points that reflect the coupling state of the system; During the phase space reconstruction stage, the processing unit extracts the current sampled value and several delay terms according to a preset delay time. If a three-dimensional reconstruction is taken, the above sequence can be reconstructed into several row vectors. For example, the first row consists of 0.12, 0.18, and 0.11, the second row consists of 0.18, 0.11, and -0.03, and the third row consists of 0.11, -0.03, and -0.10. The multiple row vectors are superimposed to obtain a multi-dimensional phase space matrix. This matrix can be regarded as the trajectory of the ripple current in a high-dimensional state space. Unlike conventional spectrum analysis, the focus here is not on the strength of a single frequency point, but on whether the trajectory shape is stable under the combined action of multiple coupled power sources. A cross-sectional hyperplane is set on the high-dimensional trajectory, and the continuous intersections formed when the trajectory crosses the cross-section are calculated. The results are then summarized to obtain a Poincaré cross-sectional scatter plot. If the system is stable, the scatter points are usually concentrated within the finite boundary. If a device parasitic parameter changes at a certain charging port, the scatter points will expand, split, or become discrete. To quantify this change, the topological invariant calculator spatially divides the scatter plot, counts the probability of scatter points in each sub-region, and calculates the current topological information entropy accordingly. Then, the current value is compared with the baseline information entropy pre-calibrated under normal operating conditions to obtain the information entropy residual. The hardware closed-loop blocking module compares the residual with the safe manifold boundary threshold range; if the residual remains within the range, it is considered that the current multi-terminal coupling is in a stable state; if the residual exceeds the range, it is considered that at least one branch has experienced a topology breaking anomaly or a contraction coupling anomaly. In the case of upper boundary overflow, the control logic immediately outputs a hardware interrupt signal to drive the switch of the corresponding branch to turn off, preventing the abnormality from evolving into overheating, ablation or a larger-scale bus disturbance; in the case of lower boundary overflow, in this embodiment, it is preferable to output a warning first and keep it on, while combining the scattered point shrinkage characteristics for continuous verification, so as to avoid misjudging a simple low-fluctuation steady state as an abnormality. Meanwhile, the system extracts diagnostic physical feature parameters from the constructed multidimensional phase space matrix and generates a traceable report; furthermore, to make the lower bound judgment conditions clearer, in the specific implementation, the early warning process for contraction coupling anomaly is only entered when the information entropy residual is lower than the minimum lower limit value and the Poincaré section scatter plot simultaneously satisfies the contraction feature. If only low residuals occur without any of the following: decrease in scatter point boundary area, shortening of principal axis length, or abnormal increase in intersection concentration, then the window is treated as a stable fluctuation. In terms of handling abnormal operating conditions, if there are insufficient effective sampling points in a certain sampling window, for example, due to hot-plugging causing the window length to be incomplete, the current round will not enter the topology determination, but will be marked as needing to be supplemented and recalculated after one window; if the synchronous sampling clock detects that the drift exceeds the limit, the data in this round will be directly discarded to avoid misjudging the clock deviation as a topology abnormality. If multiple branches exceed the threshold simultaneously, the branch with the largest residual deviation will be cut off first, and the remaining branches will be put into high-frequency verification mode to avoid cascading accidental cuts caused by bus cascade disturbances; if a window only shows a lower boundary without evidence of scatter point contraction, only an observation mark will be generated and no abnormal report upgrade will be triggered. For example, in the 8-port charging cabinet of the emergency command center, port 3 is connected to a mobile nursing terminal that is used frequently for a long time, and port 7 is connected to a duty mobile phone with a newly replaced battery. During system operation, although the DC charging current of both ports did not exceed the rated value, the high-frequency ripple phase space trajectory corresponding to port 3 began to show boundary expansion. The calculated information entropy residual was 0.42, while the safe range was set to 0.05 to 0.25. Since 0.42 has exceeded the upper limit, the closed-loop blocking module cuts off the MOSFET at port 3 before the DC overcurrent occurs and records the broken coupling topology at port 3 in the diagnostic report, which is suspected to be due to interface contact degradation or abnormal input capacitor. The purpose of this step is to convert the one-dimensional current waveform, which is difficult to interpret directly under the condition of multiple devices aliasing, into a measurable topological state quantity. This allows for early identification of anomalies in multi-port charging systems and rapid hardware-level response based solely on electrical variable measurements, without relying on terminal protocol data.

[0019] In this embodiment of the invention, the step of obtaining the total current signal of the multiple load power supply buses and each branch through a preset high-frequency current extraction circuit, and stripping and sampling the AC ripple current from the total current signal through the preset high-frequency current extraction circuit, includes: obtaining the total current signal of the multiple load power supply buses and each branch through a current transformer. The total current signal is filtered using a hardware high-pass filter to remove the DC component and extract the high-frequency AC signal. The cutoff frequency of the hardware high-pass filter is set to a preset high-frequency threshold. The high-frequency AC signal is synchronously sampled by a high-speed analog-to-digital converter to generate the AC ripple current. This embodiment provides a hardware extraction step for AC ripple current; specifically, in the multi-port charging cabinet of the aforementioned hospital emergency command center, relying solely on software to read charging protocol messages is easily affected by terminal compatibility, proprietary protocol closure, and cable contact fluctuations. Therefore, this embodiment uses a pure current measurement link to complete ripple acquisition. The details are as follows: A current acquisition link is set at the input end of each load power supply bus and each branch to sense the total current signal. Since the amplitude of the DC component in the total current is greater than the preset component threshold, and the abnormal precursors used to characterize the power supply coupling state usually fall in the preset high frequency band, a hardware high-pass filter is introduced to suppress the low frequency and DC components. The cutoff frequency can be preset to 100kHz, or it can be set between 80kHz and 300kHz depending on the range of fast charging protocols supported by the charging cabinet. The reason for this setting is that most fast charging switching power supplies and their higher harmonic components are located near this frequency band, which can retain ripple characteristics with diagnostic value, or retain ripple characteristics used for abnormal diagnosis. It should be noted that in the DC charging scenario of this embodiment, the DC magnitude of the total current and the high-frequency ripple are collected in a division of labor: the main channel used for topology determination is the AC coupling channel, which is preferably handled by the current transformer for high-frequency component sensing; while the DC amplitude in the total current can be provided by the shunt resistor or Hall sensor to provide auxiliary calibration values ​​for range selection, idle port identification and alignment with the protection link. In other words, in this embodiment, what is actually sent to the subsequent phase space reconstruction is the high-frequency component obtained after AC coupling, so as to avoid the principle deviation caused by directly using the current transformer for DC component measurement; the aforementioned acquisition of the total current signal through the current transformer is specifically manifested in this embodiment as the acquisition of the high-frequency change part in the total current and the alignment of it with the DC auxiliary measurement result on the timestamp, forming an equivalent representation of the total current state of the same branch. Furthermore, the total current signal, in terms of hardware implementation, is not limited to a single sensor fully covering all DC and AC components, but rather refers to a unified representation of the current state within the same branch and the same time window. Among these, the current transformer outputs a high-frequency change, a high-pass filter further suppresses low-frequency drift on this high-frequency change, a high-speed analog-to-digital converter synchronously samples the filtered signal, and the DC auxiliary channel is only used to provide background operating condition labels and protection verification data, and is not used as a direct input for the Poincaré section calculation. For ease of understanding, suppose the total current of a branch at a certain moment can be simplified as a DC current of 1.8A superimposed with a high-frequency fluctuation of 0.06A; where 1.8A can be given by the shunt or Hall channel as the range reference, and the high-frequency fluctuation of 0.06A is given by the current transformer and the subsequent high-pass link; after high-pass filtering, the DC current of 1.8A is filtered out, and only the AC part with an amplitude of approximately 0.06A is retained. The high-speed analog-to-digital converter samples synchronously at a sampling rate of, for example, 10MSps. If the bus and all eight branches need to be compared at the same time, synchronous sampling can avoid error propagation due to phase shift. In this case, each sampling window outputs a set of time-aligned ripple arrays, for example, the bus is [0.03, 0.05, 0.01, -0.02], branch 3 is [0.01, 0.04, 0.00, -0.01], and branch 7 is [0.02, 0.03, -0.02, -0.03]. These arrays will be used as inputs for subsequent reconstruction. In the previous embodiment, if phase space analysis is performed directly on the original total current, the DC bias will dominate, causing the phase space trajectory to mainly reflect the size of the charging load rather than the high-frequency coupling state, thereby masking early fault characteristics. Therefore, this embodiment uses a link of high-frequency sensing channel + high-pass filter + high-speed synchronous sampling to separate the low-amplitude AC ripple from the high-amplitude DC reference, providing usable input for subsequent topology analysis. In terms of handling abnormal operating conditions, if the signal of a branch is close to zero due to the disconnection of the terminal, the current transformer output may be lower than the effective range lower limit. At this time, the branch is marked as an idle port and will not participate in this round of coupling analysis. If the amplitude of the signal after high-pass filtering exceeds the full scale of the analog-to-digital converter, the pre-stage attenuation channel is triggered and resampled to prevent truncated distortion from being mistaken for discrete outliers. If a sample is lost in a certain channel during synchronous sampling, all cross-branch comparisons involving that timestamp in this round will be suspended, and only the channels without lost samples will be monitored locally. If the timestamps of the DC auxiliary measurement and the high-frequency channel are not aligned, the high-frequency channel window will be the main one, and the DC value will only be used for the background label of the window and will not be directly involved in the topology calculation. For example, during the emergency room night shift, 8 ports are charging different brands of terminals simultaneously; due to slight oxidation of the deteriorated cable plug connected to port 7, the ripple peak value obtained after high-pass filtering of its branch is higher than that of the adjacent port, and the phase jitter is obvious. Once the high-speed analog-to-digital converter captures this feature, it provides a complete input for subsequent phase space reconstruction, while the traditional solution that only monitors DC current still shows normal charging at this stage; Furthermore, to maintain consistency in terminology throughout the text, in the subsequent embodiments of this specification, the total current signal refers to the total current state of the same branch or bus within the same time window, which can be characterized in hardware by the DC auxiliary measurement value and the high-frequency change portion obtained by the current transformer. Among them, the data that actually enters the high-pass filtering, phase space reconstruction and Poincaré section calculation is always the AC ripple current; accordingly, the high-frequency threshold and the cutoff frequency of the hardware high-pass filter are the same set value in this paper, both of which are used to limit the lower boundary of the frequency band retained by the high-frequency extraction channel, and no other threshold is referred to, so as to avoid ambiguity for the input objects of subsequent modules. The purpose of this step is to obtain high-frequency AC ripple corresponding to the actual power coupling state through a clear hardware link, thereby improving the detectability of abnormal precursors from the source and laying a data foundation for subsequent topology determination.

[0020] In this embodiment of the invention, the phase space reconstruction module is used to perform delay mapping processing on the AC ripple current to construct a multi-dimensional phase space matrix, including: obtaining a preset delay time parameter; using the current sampling value of the AC ripple current as a reference, and combining the delay time parameter to extract a preset number of delayed time sampling values ​​that match the phase space embedding dimension; and combining the current sampling value and the delayed time sampling values ​​to generate the multi-dimensional phase space matrix.

[0021] This embodiment provides a phase space reconstruction step; specifically, the ripple sequence obtained after high-frequency extraction is still a one-dimensional time series. If the threshold is directly determined on the time axis, the coupling behavior between multiple ports is still difficult to show. Therefore, this embodiment uses delay mapping to reconstruct the one-dimensional ripple sequence into a multi-dimensional phase space matrix. The details are as follows: The system first obtains the preset delay time parameter; this parameter can be represented by an integer multiple of the sampling point interval. For example, when the sampling period is 0.1 microseconds, the delay amount can be 2 sampling points. Based on the current sampling value, the system extracts the sampling values ​​at several delay moments backward or forward according to the delay amount and combines them. If three-dimensional reconstruction is used, any row can be written as a combination of the current value, the value delayed once, and the value delayed twice. Taking the ripple sequence [0.12, 0.18, 0.11, -0.03, -0.10, 0.05] in the previous embodiment as an example, when the delay is 1 sampling point, the matrix can be obtained: the first row is [0.12, 0.18, 0.11], the second row is [0.18, 0.11, -0.03], the third row is [0.11, -0.03, -0.10], and the fourth row is [-0.03, -0.10, 0.05]. Each row generated in this way represents the state vector of the system at a local moment. When the number of system ports increases and the coupling relationship becomes more complex, simply observing the ripple amplitude may not be able to distinguish between two situations: high load but stable and normal load but with micro-jitter anomalies. After introducing delay mapping, the trajectory shapes of these two situations in the matrix are usually different: the former, although the amplitude is large, has a concentrated and continuous trajectory; the latter, even if the amplitude changes little, may exhibit cross-dimensional jumps and boundary distortions. Therefore, subsequent calculations can be based on the state evolution structure, rather than being limited to instantaneous numerical values. Regarding the fault tolerance control mechanism, if the sampling window length is less than the minimum number of points required for reconstruction, such as three consecutive valid points required for 3D reconstruction, then the window is only used for caching and does not participate in matrix generation; if the delay parameter exceeds the set upper limit of the parameter, causing the correlation between adjacent vectors to be lower than the preset correlation threshold, the system can automatically reduce the delay amount based on the average mutual information result of the calibration stage. If switching charging protocols at the port causes a sudden change in the sampling statistical characteristics within a short period of time, a transition flag will be enabled within a few windows after the protocol switch to avoid mistaking the natural switching process as an anomaly. For example, in the aforementioned charging cabinet, the No. 3 port nursing terminal experiences slight switching jitter due to the aging of the input capacitor; its original ripple sequence shows only sporadic fluctuations in the time domain, but in the reconstructed three-dimensional matrix, the adjacent state vectors no longer follow the stable ring band distribution, but show a tendency to escape outward; this trend will be directly reflected in the cross-sectional scatter plot. The purpose of this step is to recover the one-dimensional ripple signal into a multi-dimensional structure that can reflect the nonlinear dynamic state, so as to capture weak and continuous anomalous evolution in subsequent processing.

[0022] In this embodiment of the invention, calculating the Poincaré cross-sectional scatter plot based on the multidimensional phase space matrix includes: setting a cross-sectional hyperplane that matches the dimensions of the multidimensional phase space matrix; calculating the continuous intersection points between the phase space trajectory represented by the multidimensional phase space matrix and the cross-sectional hyperplane; and summarizing the coordinate data of all the continuous intersection points to generate the Poincaré cross-sectional scatter plot.

[0023] This embodiment provides a Poincaré section scatter plot generation step; specifically, after a multidimensional phase space matrix has been formed, directly interpreting all trajectory points one by one would exceed the preset calculation threshold and be detrimental to rapid hardware implementation. Therefore, this embodiment compresses the continuous trajectory into a set of intersection points by setting a cross-sectional hyperplane in order to extract the core structure of state evolution. The details are as follows: For the three-dimensional reconstruction case, a fixed cross section can be set, such as the plane corresponding to a certain reference value in the second dimension; the system checks whether two adjacent state vectors cross the plane segment by segment; if they cross, the intersection point of the line segment and the plane is interpolated and the coordinates of the intersection point in the other dimensions are recorded. Taking specific system operation data as an example: if two adjacent state vectors are [0.10, 0.15, 0.08] and [0.02, -0.05, -0.01] respectively, and the cross section is defined as having a second dimension equal to 0, then it can be determined that the trajectory segment has crossed, and the coordinates of the intersection point can be estimated proportionally; after accumulating multiple intersection points, a scatter plot can be obtained; Although the previous scheme can reconstruct the trajectory, if all trajectory points are used directly for probability statistics, the dense sampling of the trajectory in time will cause a large number of adjacent redundant points, making the system more sensitive to the sampling rate and window length. Therefore, this embodiment uses cross-sectional sampling to retain only the intersection points that best represent the dynamic repeating structure. Under normal circumstances, scatter plots usually form clusters with clear boundaries; when the branch switch status is abnormally disturbed, the intersection points will show discrete outward divergence, the cluster centers will separate into multiple clusters or exhibit an asymmetric divergent shape. Regarding the fault tolerance control mechanism, if the number of intersection points in a certain window is too small, for example, less than the preset 20, then the window is considered insufficient to form reliable scattered point statistics, and only the trajectory cache is retained without entering the entropy calculation; if all the intersection points are concentrated in a very small range, resulting in most sub-regions being empty during subsequent division, then the system automatically enables a finer coordinate resolution. If the number of intersection points increases abnormally, first check whether it is caused by sampling jitter or improper section selection, and switch to the backup section plane for verification. For example, during continuous operation of the emergency charging cabinet, the non-original fast charging cable used by the on-duty mobile phone at Gate 7 caused current ripple phase drift. Although the reconstructed trajectory still shows a closed trend, the intersection points formed on the set cross section gradually stretch from the original single elliptical cluster to a distribution with a tail; this change is more easily captured by subsequent probability statistics and topological entropy models at the scatter plot level. The purpose of this step is to extract key structural features from high-dimensional trajectories so that abnormal states can be stably observed in the form of geometrical distribution changes.

[0024] In this embodiment of the invention, a topological invariant calculator is used to calculate the current topological information entropy based on the scatter distribution probability of the Poincaré cross-section scatter plot, and the current topological information entropy is compared with the baseline information entropy to generate an information entropy residual. This includes: dividing the Poincaré cross-section scatter plot into a preset number of sub-regions based on a preset spatial partitioning rule; and calculating the distribution probability of scatter points within each sub-region. The negative summation of the product of the distribution probability of each sub-region and its logarithm is taken as the current topological information entropy; the absolute value of the difference between the current topological information entropy and the baseline information entropy is calculated, and the absolute value of the difference is taken as the information entropy residual.

[0025] This embodiment provides a step for calculating topological information entropy. Specifically, although scatter plots can intuitively show clustering, stretching, or discretization phenomena, in automated monitoring systems, geometric changes still need to be converted into numerical quantities that can be compared with thresholds. Therefore, this embodiment quantifies the probability of scatter distribution and generates information entropy residuals. The system divides the scatter plot into several sub-regions according to a preset spatial partitioning rule. For example, a two-dimensional scatter plot can be divided into four 2×2 sub-regions, denoted as R1, R2, R3, and R4 respectively. Assuming that the total number of intersection points of a certain window is 100, of which R1 has 40, R2 has 30, R3 has 20, and R4 has 10, the corresponding probabilities are 0.4, 0.3, 0.2, and 0.1. The current topological information entropy is obtained by summing the negative products of each probability and its logarithm. If the probability distribution corresponding to the normal baseline window is more concentrated, such as 0.7, 0.2, 0.1, and 0, then its baseline entropy is lower. Once the current distribution becomes more dispersed, the entropy value will increase. The system takes the absolute value of the difference between the current entropy value and the baseline entropy value as the information entropy residual. In the previous layer implementation, scatter plots can already reveal geometric changes, but manual observation relies on experience and is not suitable for continuous online operation. After introducing topological information entropy, the system can transform whether the scatter points have become scattered and whether the boundaries have become disordered into numbers of a uniform scale, thus being compatible with the comparison needs of different terminal models and different charging powers. In addition, using residual form instead of using the current entropy value alone is beneficial to eliminating the bias caused by the difference in the original scatter point density of different branches. Regarding the fault tolerance control mechanism, if the number of scattered points in a certain sub-region is zero, the probability of that sub-region is treated as zero, and the logarithmic term can be skipped during calculation to avoid numerical overflow; if all scattered points are concentrated in a single region, the current entropy will approach zero, and at this time the system will further check whether a pseudo-steady state is caused by sampling freeze. If the residuals of multiple consecutive windows are close to the upper limit but have not exceeded the limit, the system enters the early warning buffer state, requiring the next window to be reviewed before hardware action is taken, in order to reduce misjudgments caused by occasional noise in a single window. For example, in the aforementioned emergency charging cabinet, when the nursing terminal at port 3 is in the early stage of potential failure, its scatter plot, which originally mainly falls on R1 and R2, gradually spreads to R3 and R4; the system calculates that the current entropy is 1.28, the baseline entropy is 0.74, and the residual is 0.54; although the DC current is still in the normal charging range of 1.6A at this time, the topological residual clearly reflects that the coupling relationship has been disturbed. The purpose of this step is to compress the changes in high-dimensional dynamic structures into a single-valued index that is easy to determine the threshold, thereby achieving a rapid, unified, and repeatable quantitative evaluation of complex charging states.

[0026] In this embodiment of the invention, a hardware closed-loop blocking module is used to compare the information entropy residual with a preset safe manifold boundary threshold range; when the information entropy residual exceeds the safe manifold boundary threshold range, a charging abnormality is determined to have occurred, and a hardware interrupt signal is output to control the state of the corresponding branch switch, including: when the information entropy residual is greater than the maximum upper limit of the safe manifold boundary threshold range, a topology breaking abnormality is determined to have occurred, and the hardware interrupt signal is output to the branch switch corresponding to the charging abnormality to generate a cut-off control command; When the information entropy residual is less than the minimum lower limit of the safe manifold boundary threshold interval, the system is determined to be in an over-coupling abnormal state, an early warning signal is output, and a state flag is generated to keep the branch switch tube conducting; when the information entropy residual is greater than or equal to the minimum lower limit and less than or equal to the maximum upper limit, the system is determined to be in a stable resonance state, a stable state data tag is generated, and no interruption signal is output.

[0027] This embodiment provides a closed-loop determination and blocking step based on a safe manifold boundary; specifically, the residual magnitude alone is not enough to distinguish between dangerous anomalies and observable but non-disconnectable state shifts, so this embodiment divides the residual landing point into three intervals and provides different hardware handling methods; The details are as follows: The system presets a safety manifold boundary threshold range, with a minimum lower limit of 0.05 and a maximum upper limit of 0.25. If the information entropy residual is between 0.05 and 0.25, it indicates that although there are certain natural fluctuations in the phase space topology, the overall system is still in a stable resonance state. The system only generates stable state data tags and does not output interruption signals. If the residual is greater than 0.25, for example, reaching 0.42, it indicates that the scatter topology has obviously deviated from the normal manifold boundary, which can be identified as a topology breaking anomaly. At this time, the hardware closed-loop blocking module directly outputs the interrupt signal to the switch driver of the abnormal branch to perform cut-off control. If the residual is less than 0.05, for example, only 0.01, it does not necessarily mean that it is more stable. On the contrary, it may mean that there is excessive coupling between multiple ports and abnormal scatter shrinkage. The system outputs a warning signal but keeps the conduction state to prompt the operation and maintenance personnel to check whether there is a certain type of abnormal synchronization frequency locking phenomenon. It should be further explained that since the information entropy residual uses the absolute value of the difference between the current entropy value and the baseline entropy value in the preceding steps, the small residual itself only indicates that the current entropy value is close to the baseline entropy value, and does not constitute an abnormal conclusion on its own. Therefore, in this embodiment, the excessive coupling anomaly less than the lower limit is triggered by a combination criterion: on the one hand, the residual falls below the lower limit, and on the other hand, the Poincaré section scatter plot must also meet the shrinkage characteristics, such as the shortening of the principal axis length of the scatter plot, the decrease of the boundary area, or the abnormal increase of the intersection concentration in multiple consecutive windows. Only when low residuals and significant contraction occur simultaneously can it be interpreted as an over-coupling anomaly; if only the residuals are small and the scatter geometry does not show abnormal contraction, it is still treated as a stable fluctuation; to avoid unclear judgment criteria, in specific implementation, the minimum lower limit value can be understood as the entry threshold for contraction anomaly warning, that is, after the residuals are lower than this value, the system starts the contraction feature review process. When the verification is successful, an early warning signal is output and a state flag indicating that the circuit remains on is generated. When the verification is unsuccessful, the stable state data label is retained and no abnormal upgrade is performed. The reason for this is that some normal high-consistency charging conditions may also produce small residuals. If no shrinkage feature constraint is added, it is easy to misjudge the normal steady state as an abnormality. The previous layer can identify residual deviations, but in actual engineering, the direction of deviation is also significant. If a single threshold cutoff strategy is simply adopted, which determines an anomaly as the absolute value of the residual exceeds a preset upper limit, it may fail to identify some excessively contracted anomalies, or even mistake a stable high-load condition for a dangerous condition. Therefore, this embodiment introduces upper and lower boundaries to jointly define the safe manifold interval, and corresponds to the three actions of cutoff, warning and hold, respectively, thereby improving the precision of system decision-making. In terms of handling abnormal operating conditions, if two consecutive windows are one above the upper limit and the other below the lower limit, it indicates that the system is in a rapid transition phase. At this time, the sampling clock is locked and a high-priority retest is performed first, and then it is decided whether to cut off. If the residual is exactly equal to the boundary value, it is processed within the interval according to the safety policy to avoid repeated interruption caused by boundary jitter. If the abnormal branch location result is not unique, then first take the downgrade measures of bus current limiting and reducing the duty cycle of all branches, and then determine the target port by retesting branch by branch; if the low residual state only lasts for a single window and the scattered point shrinkage index is not continuously met, then only record one early warning candidate event and do not immediately generate an over-coupling abnormal conclusion. For example, during the peak of emergency duty, the residual information entropy of the No. 3 terminal rose to 0.42, and the system immediately determined it to be a topology failure and cut off the No. 3 terminal. Meanwhile, the residual of port 7 dropped to 0.03 during the same period, and the area of ​​the scattered point boundary in three consecutive windows shrank to less than 60% of the baseline. The system did not cut off the power directly, but kept the power supply to port 7 and issued an over-coupling warning, indicating that the port may have compatibility abnormalities or frequency-locked abnormal ripples, and that the cable and terminal adaptation status needed to be manually checked. Furthermore, to avoid ambiguity in understanding between exceeding the safety manifold boundary threshold range and the three handling actions in this embodiment, the term "exceeding the limit" in this specification uniformly refers to the residual falling on either side outside the range, that is, it includes both upper limit exceeding the maximum upper limit value and lower limit less than the minimum lower limit value; wherein, upper limit corresponds to hardware cut-off anomaly, and lower limit corresponds to warning anomaly that maintains conduction. Accordingly, the term "branch switch" is a unified term in this embodiment and subsequent system embodiments, and is preferably implemented by a MOS transistor. Therefore, the statements in the text such as "the MOS transistor driving terminal cuts off the corresponding port MOS transistor" are all specific hardware implementations that control the state of the corresponding branch switch, and do not separately indicate different execution objects. The purpose of this step is to enable the system to move away from coarse control using a single overcurrent logic and instead implement graded handling based on the dynamic boundary position, thereby balancing safety, continuous power supply capability, and false alarm suppression capability.

[0028] In this embodiment of the invention, generating a diagnostic report containing physical feature parameters extracted based on the multidimensional phase space matrix includes: when a charging anomaly is determined to occur, obtaining the current Lyapunov index by calculating the exponential divergence rate of the phase space trajectory represented by the multidimensional phase space matrix during the evolution process, and comparing it with a preset Lyapunov baseline index to obtain the Lyapunov index offset; and using the Lyapunov index offset as the physical feature parameter. The system matches a preset device topology feature library to identify the type of load device that is malfunctioning; and combines the load device type with the physical feature parameters to generate the diagnostic report.

[0029] This embodiment provides a step for generating an anomaly diagnosis report. Specifically, after completing the hardware disconnection, if only an anomaly conclusion is given, maintenance personnel will still find it difficult to determine whether the problem is with the terminal body, cable, interface or power supply branch. Therefore, this embodiment further extracts physical feature parameters from the phase space trajectory and generates an interpretable diagnosis report by combining the device type identification results. The details are as follows: After confirming the anomaly, the system calculates the exponential divergence rate of the multidimensional phase space trajectory during the evolution process to obtain the current Lyapunov exponent; this exponent can be understood as the sensitivity of adjacent state trajectories to initial small perturbations; for ease of explanation, it is assumed that the baseline exponent is 0.08, the current anomaly window is calculated to be 0.21, and the offset is 0.13. If the offset is greater than the preset offset threshold, it usually indicates that the system has entered a more obvious unstable or chaotic state; if the offset is less than or equal to the preset offset threshold but the residual has exceeded the limit, it may indicate that the anomaly is mainly manifested as structural boundary offset rather than overall divergence; the system uses this offset as one of the physical characteristic parameters in the report. The previous layer implementation can already perform the cut-off action, but in high-reliability scenarios such as hospitals, simply cutting off without giving a reason will result in a lack of feature data support for fault location and will not be conducive to subsequent traceability. Therefore, this embodiment continues to extract features from the abnormal window after cutting off and matches them with the device topology feature library to identify the corresponding load device type. Ultimately, the report can include the port number, time of the anomaly, device type, Lyapunov exponent offset, residual value, and suggested direction of investigation. Regarding the fault tolerance control mechanism, if the sampling window is too short when an anomaly occurs, resulting in insufficient stability of the index calculation, the report will retain the marker that the index has not reached the confidence length, but will still output other usable features; if the feature library matching results are not unique, the top two candidate types and their similarity will be listed at the same time to avoid making overly certain judgments. If the abnormality disappears after a retest following disconnection, a transient abnormality marker will be added to the report to distinguish between persistent faults and short-term contact jitter. For example, after the No. 3 port of the emergency charging cabinet is disconnected, the system calculates the current Lyapunov index to be 0.21, which is 0.13 off from the baseline of 0.08; at the same time, it identifies that the port is connected to a mobile nursing terminal rather than a regular mobile phone; the final report can state: No. 3 port, mobile nursing terminal, topology information entropy residual 0.42, Lyapunov index offset 0.13, and recommends prioritizing the inspection of interface oxidation and input filter capacitor degradation; for on-duty engineers, this report can be directly used for troubleshooting. The purpose of this step is to transform abnormal results from a binary judgment of whether an error has occurred into an interpretable diagnostic output that includes information such as the type of device, the type of physical offset, and the priority direction for investigation.

[0030] In this embodiment of the invention, matching a preset device topology feature library to identify the type of load device that has malfunctioned includes: extracting geometric boundary feature data from the Poincaré scatter plot; comparing the geometric boundary feature data with the topology map templates of various power management chips pre-stored in the device topology feature library, wherein the device topology feature library is constructed based on pure physical current features; when the highest similarity is greater than or equal to a preset similarity threshold, determining the corresponding load device type based on the topology map template with the highest similarity; when the highest similarity is less than the preset similarity threshold, marking the load device type as an unknown type.

[0031] This embodiment provides a device type identification step based on pure physical current characteristics; specifically, in anomaly diagnosis, if the device type is inferred solely from port registration information, there may be discrepancies between the registered device and the actual connected device, especially when multiple people share a charging cabinet; Therefore, this embodiment directly extracts the device fingerprint from the geometric boundary of the Poincaré section scatter plot and matches it with a preset template library; The details are as follows: The system extracts geometric boundary feature data from the scatter plot, such as principal axis length, boundary closure, discrete outlier ratio, center offset, and boundary curvature change trend; for intuitive explanation, we can assume that the feature vector extracted from a certain abnormal window is [major axis 0.82, minor axis 0.46, discrete ratio 0.12, center offset 0.08]; The device topology feature library contains templates corresponding to various types of power management chips. For example, template T1 corresponds to a certain type of Apple terminal power management chip, template T2 corresponds to a certain type of Android fast charging chip, and template T3 corresponds to a dedicated power management structure for mobile nursing handheld terminals. The system calculates the similarity between the current feature and each template one by one. For example, if the similarity with T1 is 0.61, with T2 is 0.74, and with T3 is 0.89, then T3 with the highest similarity is determined as the matching result, and the type of load device is determined accordingly. Specifically, the method for extracting geometric boundary feature data includes: using the convex hull algorithm to calculate the boundary contour of the Poincaré section scatter plot, and calculating the boundary closure based on the ratio of the contour area to the area of ​​the inscribed ellipse; the similarity comparison specifically uses the cosine similarity algorithm to calculate the similarity score between the current geometric boundary feature vector and each template feature vector; In the previous layer implementation, although diagnostic reports can be output, the reports are still limited in their relevance if the origin of the device is unknown. After introducing the geometric boundary feature library, the system does not need to read the device identifier through USB communication, nor does it rely on terminal-reported information. It can complete the type identification based solely on the current measurement link. This is particularly suitable for scenarios that are privacy-sensitive, have closed protocols, or where terminal information has been tampered with. Regarding the fault tolerance control mechanism, if the highest similarity is lower than a preset threshold, such as below 0.75, the system will mark the device as an unknown type and add the current feature to the sample pool to be reviewed. If the difference between the highest similarity and the second highest similarity is too small, for example, less than 0.03, the report will mark it as type uncertain, indicating that it may be a mixed feature introduced by similar chip platforms or adapters; if the scatter plot is affected by sudden plugging and unplugging and the boundaries are incomplete, type identification will be suspended and only the basic anomaly report will be retained. For example, in the aforementioned case of port 3 anomaly, the similarity between the scatter point boundary extracted by the system and the mobile nursing terminal template was 0.89, which was higher than the 0.74 similarity with the ordinary Android phone template. Therefore, the report clearly stated that the target device was a mobile nursing terminal. If another port is connected to an unregistered private mobile phone, although the system does not read the phone information, it can still determine that it is closer to a certain type of consumer-grade fast-charging mobile phone through pure physical ripple topology features, thereby assisting managers in verifying the actual access object. The purpose of this step is to form a device-level physical fingerprint through pure electrical variable measurement, realize the integrated output of anomaly location and device identification, and improve the interpretability and anti-tampering capability of the system in complex multi-terminal environments.

[0032] Please see Figure 2 A multi-mobile phone charging power supply anomaly detection system based on current analysis is applied to a charging system containing multiple load power supply buses and several branch switching transistors. The system includes: a high-frequency current extraction circuit, used to acquire the total current signal of the multiple load power supply buses and each branch, and to strip and sample the AC ripple current from the total current signal. The phase space reconstruction module is used to perform delay mapping processing on the AC ripple current, construct a multi-dimensional phase space matrix, and calculate the Poincaré section scatter plot based on the multi-dimensional phase space matrix. The topological invariant calculator is used to calculate the current topological information entropy based on the scatter distribution probability of the Poincaré section scatter plot, and compare the current topological information entropy with the preset baseline information entropy to generate information entropy residuals; The hardware closed-loop blocking module is used to compare the information entropy residual with a preset safe manifold boundary threshold range; when the information entropy residual exceeds the safe manifold boundary threshold range, it is determined that a charging abnormality has occurred, outputs a hardware interrupt signal to control the state of the corresponding branch switch tube, and generates a diagnostic report containing physical feature parameters extracted based on the multidimensional phase space matrix.

[0033] This embodiment provides a multi-mobile phone charging power supply anomaly detection system based on current analysis; specifically, the system can be integrated into the main control board of an 8-port or 16-port charging cabinet in a hospital emergency command center, or it can be used as an independent monitoring board attached to existing charging equipment. The system is based on a hardware measurement link, and is equipped with programmable logic devices and control execution circuits to complete the entire process from AC ripple acquisition, phase space reconstruction, topology entropy calculation to hardware closed-loop blocking and diagnostic output. The high-frequency current extraction circuit may include current transformers, front-end isolation networks, hardware high-pass filters, and high-speed analog-to-digital converters arranged on the bus and each branch; the phase space reconstruction module may be implemented by FPGA, DSP, or processor with high-speed parallel capabilities, and its internal components include at least a delay buffer unit, a vector splicing unit, and a cross-section intersection calculation unit. The topology invariant calculator may include a spatial partitioning unit, a scatter point counting unit, a probability calculation unit, and an information entropy calculation unit; the hardware closed-loop blocking module may include a threshold comparator, an abnormal branch mapping table, a MOS transistor driving circuit, and an alarm output interface; the diagnostic report unit may be implemented by an embedded controller to store the exponential offset, device type, timestamp, port number, and action result; The following describes the collaborative working process of each module in conjunction with the specific processing flow; assuming that after port 3 outputs the ripple array in a certain sampling window, the high-frequency current extraction circuit sends it to the reconstruction module, the reconstruction module generates a three-dimensional state matrix and calculates the intersection points of the cross sections; the topological invariant calculator obtains the current information entropy of 1.28, the baseline of 0.74, and the residual of 0.54; The hardware closed-loop blocking module determines that the upper limit of 0.25 is exceeded and then outputs a shutdown pulse to the MOS drive pin of port 3; the diagnostic report unit continues to calculate the exponential offset of 0.13 based on the trajectory divergence rate of the window, and identifies the device as a mobile nursing terminal in combination with the feature library, and finally forms a complete event entry on the local screen or in the background record. To avoid system fragility caused by overly tight coupling between modules, this embodiment preferably aligns data between modules using timestamps and port numbers; this way, even if a module is temporarily reset, the pipeline can be restored based on the unprocessed data in the buffer; if the high-frequency current extraction circuit fails, the system can degrade to a backup mode that only monitors DC overcurrent. If the phase space reconstruction module malfunctions, the blocking module will not perform topology-based disconnection, but will only retain traditional overcurrent protection; if the diagnostic report unit fails, it will not affect the disconnection action itself, and the relevant raw data can be written to the circular buffer first, and a report will be generated after recovery. For example, during continuous 24-hour operation in the emergency department, the multi-port charging cabinet primarily serves medical staff's work terminals during the day, while simultaneously connecting to staff's personal mobile phones and backup communication devices at night. The system can reliably identify ripple coupling boundaries formed by various terminals under high concurrency during the day, and directly disconnects the device at night if a topology breach is detected at port 3, leaving a traceable record of the device type and physical characteristics in the background. Maintenance personnel can directly replace the cable and interface module of port 3 the next day based on the record, without having to reproduce the fault one by one. Furthermore, in order to maintain consistency in terminology and hardware objects between the system embodiments and the aforementioned method embodiments, the branch switch transistors are all uniformly named in this specification, and their preferred device form is a MOS transistor. Therefore, the MOS transistor drive circuit in the system-level description outputs a turn-off pulse to the MOS drive pin of port 3, which corresponds to controlling the state of the corresponding branch switch transistor. The diagnostic report unit corresponds to the implementation carrier in the method that generates a diagnostic report containing physical feature parameters extracted based on the multidimensional phase space matrix, without changing the aforementioned functional division; It should also be noted that the total current signal in the system is still understood as the total current state characterization under the same time window according to the aforementioned method embodiment. The AC ripple current output by the high-frequency current extraction circuit is used for the main link of topology analysis, while the DC measurement branch only undertakes the functions of range calibration, idle port identification and protection verification. The purpose of this system is to solidify the aforementioned methods into a measurement and control device with a clearly defined hardware structure, thereby enabling real-time detection, graded handling, and interpretable traceability of early anomalies in multi-port charging environments.

[0034] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis, characterized in that, The method is applied to a charging system that includes multiple load power supply buses and several branch switching transistors, and includes the following steps: The total current signal of multiple load power supply buses and each branch is obtained by a preset high-frequency current extraction circuit, and the corresponding AC ripple current is extracted and sampled from each total current signal by the preset high-frequency current extraction circuit. The phase space reconstruction module is used to perform delay mapping on each AC ripple current to construct the corresponding multidimensional phase space matrix, and the corresponding Poincaré cross-section scatter plot is calculated based on the respective multidimensional phase space matrix. The topological invariant calculator calculates the current topological information entropy for each path based on the scatter distribution probability of the scatter plots of each Poincaré section, and compares the current topological information entropy with the preset baseline information entropy to generate their respective information entropy residuals. The information entropy residual is compared with a preset safe manifold boundary threshold range using a hardware closed-loop blocking module. When the information entropy residual exceeds the safe manifold boundary threshold range, a charging abnormality is determined, a hardware interrupt signal is output to control the state of the corresponding branch switch tube, and a diagnostic report containing physical feature parameters extracted based on the multidimensional phase space matrix is ​​generated.

2. The method for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis according to claim 1, characterized in that, The step of acquiring the total current signal of multiple load power supply buses and their branches through a preset high-frequency current extraction circuit, and then extracting and sampling the AC ripple current from the total current signal through the same circuit, includes: The total current signal of the multi-load power supply bus and each branch is obtained by a current transformer. The total current signal is filtered using a hardware high-pass filter to remove the DC component and extract the high-frequency AC signal. The cutoff frequency of the hardware high-pass filter is set to a preset high-frequency threshold. The high-frequency AC signal is synchronously sampled by a high-speed analog-to-digital converter to generate the AC ripple current.

3. The method for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis according to claim 1, characterized in that, The phase space reconstruction module performs delay mapping processing on the AC ripple current to construct a multi-dimensional phase space matrix, including: Obtain the preset delay time parameter; Based on the current sampling value of the AC ripple current, a preset number of delayed sampling values ​​that match the phase space embedding dimension are extracted in combination with the delay time parameter. The current sample value and the delayed sample value are combined to generate the multidimensional phase space matrix.

4. The method for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis according to claim 1, characterized in that, The calculation of the Poincaré section scatter plot based on the multidimensional phase space matrix includes: Define a cross-sectional hyperplane that matches the dimensions of the multidimensional phase space matrix; Calculate the continuous intersection points between the phase space trajectory represented by the multidimensional phase space matrix and the cross-sectional hyperplane; The coordinate data of all the continuous intersection points are summarized to generate the Poincaré section scatter plot.

5. The method for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis according to claim 1, characterized in that, The step of calculating the current topological information entropy using a topological invariant calculator based on the scatter distribution probability of the Poincaré cross-section scatter plot, and comparing the current topological information entropy with a preset baseline information entropy to generate an information entropy residual, includes: The Poincaré section scatter plot is divided into a preset number of sub-regions based on a preset spatial division rule; Calculate the probability distribution of scattered points within each sub-region; The negative summation of the distribution probability of each sub-region and its logarithmic product is taken as the current topological information entropy; Calculate the absolute value of the difference between the current topological information entropy and the baseline information entropy, and use the absolute value of the difference as the information entropy residual.

6. The method for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis according to claim 1, characterized in that, The method utilizes a hardware closed-loop blocking module to compare the information entropy residual with a preset safe manifold boundary threshold range; when the information entropy residual exceeds the safe manifold boundary threshold range, a charging abnormality is determined, and a hardware interrupt signal is output to control the state of the corresponding branch switch, including: When the information entropy residual is greater than the maximum upper limit of the safe manifold boundary threshold interval, a topology breaking anomaly is determined to have occurred, and the hardware interrupt signal is output to the branch switch corresponding to the charging anomaly to generate a cut-off control command. When the information entropy residual is less than the minimum lower limit of the safety manifold boundary threshold interval, the system is determined to be in an over-coupling abnormal state, an early warning signal is output, and a state flag is generated to keep the branch switch tube conducting. When the information entropy residual is greater than or equal to the minimum lower limit and less than or equal to the maximum upper limit, the system is determined to be in a stable resonance state, a stable state data tag is generated, and no interruption signal is output.

7. The method for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis according to claim 1, characterized in that, The generation of a diagnostic report containing physical feature parameters extracted based on the multidimensional phase space matrix includes: When a charging anomaly is detected, the current Lyapunov index is obtained by calculating the exponential divergence rate of the phase space trajectory represented by the multidimensional phase space matrix during the evolution process, and compared with the preset Lyapunov baseline index to obtain the Lyapunov index offset. The Lyapunov exponent offset is used as the physical characteristic parameter; Match the preset device topology feature library to identify the type of load device that is malfunctioning; The diagnostic report is generated by combining the load device type with the physical characteristic parameters.

8. The method for detecting abnormalities in multi-mobile phone charging power supplies based on current analysis according to claim 7, characterized in that, The matching of the preset device topology feature library identifies the types of load devices that have experienced anomalies, including: Extract the geometric boundary feature data of the Poincaré section scatter plot; The geometric boundary feature data is compared with the topology templates of various power management chips pre-stored in the device topology feature library for similarity. The device topology feature library is constructed based on pure physical current features. When the highest similarity is greater than or equal to the preset similarity threshold, the corresponding load device type is determined based on the topology template with the highest similarity; when the highest similarity is less than the preset similarity threshold, the load device type is marked as an unknown type.

9. A multi-mobile phone charging power supply anomaly detection system based on current analysis, applied to a charging system including multiple load power supply buses and several branch switching transistors, characterized in that, include: A high-frequency current extraction circuit is used to acquire the total current signal of multiple load power supply buses and each branch, and to strip and sample the AC ripple current from the total current signal. The phase space reconstruction module is used to perform delay mapping processing on the AC ripple current, construct a multi-dimensional phase space matrix, and calculate the Poincaré section scatter plot based on the multi-dimensional phase space matrix. The topological invariant calculator is used to calculate the current topological information entropy based on the scatter distribution probability of the Poincaré section scatter plot, and compare the current topological information entropy with the preset baseline information entropy to generate information entropy residuals; The hardware closed-loop blocking module is used to compare the information entropy residual with a preset safe manifold boundary threshold range; when the information entropy residual exceeds the safe manifold boundary threshold range, it is determined that a charging abnormality has occurred, outputs a hardware interrupt signal to control the state of the corresponding branch switch tube, and generates a diagnostic report containing physical feature parameters extracted based on the multidimensional phase space matrix.

Images