Power management system and device based on artificial intelligence and voltage stabilization protection

Abstract

The invention relates to the technical field of intelligent electric power, in particular to a power management system and device based on artificial intelligence and voltage stabilization protection, and the system comprises a switching power supply module, an explosion-proof protection module and a multi-scale intelligent detection module. Multiple protection mechanisms are designed, the anti-interference capability is improved through electromagnetic filtering, voltage stabilization protection and current limiting protection are designed to ensure that the output voltage is stable, the circuit is rapidly cut off when the output voltage is abnormal, the detection unit feeds back the output voltage in real time, the control unit achieves closed-loop adjustment, the working state of the transformer is dynamically adjusted, and the voltage stabilization precision is improved. Each unit supports independent optimization, and the adaptive capacity is high; according to the signal anomaly detection method designed by the invention, synchronous detection is realized through global-local joint modeling, a mask proportion is dynamically adjusted through signal entropy to focus a key signal segment, the robustness to complex working conditions is improved, a time ordinal number and a channel space relationship are fused, and multi-sensor data relevance is enhanced; and the capability of identifying abnormal current and voltage is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent power, and particularly relates to a power management system and device based on artificial intelligence and voltage stabilization protection. BACKGROUND

[0002] The power management system refers to a combination of hardware and software for efficient distribution, conversion, monitoring and protection of power supply of electronic equipment, and the purpose is to ensure stable operation of the system, but the existing power management system lacks effective and rapid response and circuit protection mechanism, and has poor recognition ability for abnormal current and voltage data. SUMMARY

[0003] In view of the above situation, in order to overcome the defects of the prior art, the application provides a power management system and device based on artificial intelligence and voltage stabilization protection, in view of the problem that the existing power management system lacks effective and rapid response and circuit protection mechanism, the application designs multiple protection mechanisms, suppresses surges, overvoltages and electromagnetic interference at the input end through an electromagnetic filter unit, improves the anti-interference ability, designs a voltage stabilization protection unit and a current limiting protection unit to work cooperatively, ensures stable output voltage, and quickly cuts off the circuit in an abnormal situation to prevent damage to the rear-end load, uses a detection unit to feed back the output voltage in real time, combines a control unit to realize closed-loop regulation, dynamically adjusts the working state of the transformer, improves the voltage stabilization precision, and in addition, the units of the application have clear division of labor and can be optimized independently, and the adaptability to different input environments is enhanced; in view of the problem that the prior art has poor recognition ability for abnormal current and voltage data, the application adopts a cross-modal encoding anomaly detection method based on multi-scale dynamic perception, realizes synchronous detection of slow drift and transient pulse through global-local joint modeling, focuses on key signal segments by dynamically adjusting the mask ratio based on signal entropy, improves the robustness to complex working conditions, fuses the time sequence and channel space relationship, enhances the correlation of multi-sensor data, and finally improves the recognition ability for abnormal current and voltage.

[0004] The application provides a power management system based on artificial intelligence and voltage stabilization protection, which comprises a switching power supply module, an explosion-proof protection module and a multi-scale intelligent detection module.

[0005] The switching power supply module comprises an electromagnetic filter unit, a first rectifier filter unit, a conversion unit, a second rectifier filter unit, a detection unit and a control unit.

[0006] The explosion-proof protection module comprises a filter protection unit, a voltage stabilization protection unit and a current limiting protection unit.

[0007] The multi-scale intelligent detection module adopts a cross-modal encoding anomaly detection method based on multi-scale dynamic perception, and performs data acquisition and anomaly detection alarm on the multi-channel current and voltage data of the switching power supply.

[0008] The electromagnetic filter unit comprises, in sequence, a non-return fuse, a voltage-dependent resistor, a filter capacitor and a coupling inductor, the non-return fuse is connected in series at the input end of the power supply, the voltage-dependent resistor and the filter capacitor are connected in parallel at the output end of the non-return fuse, and the coupling inductor is connected in series at the output end of the filter capacitor.

[0009] The first rectifier filter unit comprises a rectifier bridge and a first filter capacitor, the input end of the rectifier bridge is connected to the output end of the coupling inductor, the first filter capacitor is connected in parallel at the output end of the rectifier bridge, and the rectifier bridge is composed of four rectifier diodes in a full-bridge mode.

[0010] The conversion unit comprises a transformer, the primary winding of the transformer is connected to the output end of the initial rectifier filter unit, and is used for converting the input voltage into a target value voltage.

[0011] The second rectifier filter unit comprises a rectifier diode, a second filter capacitor and a load resistor, the rectifier diode is connected in series at the output end of the conversion unit, and the second filter capacitor and the load resistor are connected in parallel at the output end of the rectifier diode.

[0012] The detection unit comprises a voltage stabilizing tube and a detection resistor, the cathode of the voltage stabilizing tube is connected to the output end of the second rectifier filter unit, the anode is connected to the ground through the detection resistor, and is used for feeding back the target value voltage to the control unit.

[0013] The control unit comprises a switching power supply control chip and an optoelectronic coupler, the input end of the optoelectronic coupler is connected to the output end of the detection unit, the output end is connected to the feedback end of the switching power supply control chip, and the output end of the switching power supply control chip is connected to the control end of the conversion unit.

[0014] The filter protection unit comprises a resistance-capacitance fuse device, which is connected to the output end of the second rectifier filter unit, and is used for filtering and primary protection of the target value voltage, and the resistance-capacitance fuse device comprises a fuse and a filter capacitor connected in series.

[0015] The voltage stabilizing protection unit comprises a silicon controlled rectifier and a voltage stabilizing diode, the cathode of the voltage stabilizing diode is connected to the output end of the second rectifier filter unit, the anode is connected to the control electrode of the silicon controlled rectifier, when the target value voltage is out of limit, the silicon controlled rectifier is turned on to make the resistance-capacitance fuse of the filter protection unit fuse.

[0016] The current limiting protection unit is integrated in the control unit, and is used for realizing overcurrent and overload protection of the power supply.

[0017] Further, in the multi-scale intelligent detection module, a cross-modal encoding anomaly detection method based on multi-scale dynamic perception comprises the following steps:

[0018] Step S1: Data preprocessing, normalize the current and voltage time series signals to obtain the original time series signals, where the signal dimension is M×N, M represents the number of channels, and N is the number of sampling points;

[0019] Step S2: Data segmentation processing, dividing the original time-series signal into non-overlapping segments along the time axis. The length of the time sequence signal is Q, and the length of each non-overlapping segment is Q / T;

[0020] Step S3: Multi-scale region construction, extracting subsequences from the original time-series signal using a sliding window and constructing local regions. ;

[0021] Step S4: Global mask for non-overlapping segments Random mask [ ] paragraph, in which A dynamic parameter that affects the length of the mask segment;

[0022] Step S5: Local masking, for local areas Random mask [ ] paragraph, in which ;

[0023] Step S6: Dynamic mask adjustment, adjusting dynamic parameters according to signal entropy. : ; ;

[0024] In the formula, The probability of the amplitude histogram within the segment. Represents signal entropy. These correspond to their maximum and minimum values, respectively. This represents the dynamically adjusted parameters;

[0025] Step S7: Multimodal feature encoding based on spatiotemporal embedding. Multimodal feature encoding based on spatiotemporal embedding is performed on the non-masked segments (i.e., non-overlapping segments) and local regions to obtain multimodal spatiotemporal codes. This specifically includes the following steps:

[0026] Step S71: Projection of non-masked segments. Projection operations are performed on non-overlapping segments and local regions respectively to obtain global projection and local projection.

[0027] Step S72: Position encoding, performing position encoding embedding based on the sine function for both global and local projections, wherein the position encoding based on the sine function is as follows: ;

[0028] In the formula, a length of a non-overlapping segment or a local region;

[0029] Step S73: Spatial encoding, respectively performing spatial encoding embedding based on channel embedding matrix for global projection and local projection, wherein the spatial encoding based on channel embedding matrix is as follows: ;

[0030] In the formula, representing one-hot encoding, representing a channel m of a non-overlapping segment and a local region, representing an embedding matrix of the channel m;

[0031] Step S74: Multimodal embedding, performing time embedding and spatial embedding of the multimodal for global projection and local projection, and encoding by using a multi-layer residual connection-based lightweight transformer to obtain multimodal spatiotemporal encoding: z 0 =[ g t + e t time + e m space ; l ω + e ω time + e m space ] ; ;

[0032] In the formula, and represent global projection and local projection respectively, representing multimodal spatiotemporal encoding, representing the transformer encoding of the layer, representing layer normalization, representing multi-head self-attention mechanism processing, representing feedforward neural network processing;

[0033] Step S8: Signal reconstruction, adding a mask segment label to the multimodal spatiotemporal encoding, and outputting through a back projection layer after single-layer transformer decoding to obtain a reconstructed signal;

[0034] Step S9: Abnormality detection, calculating an abnormality score through a sliding window, and issuing an abnormality alarm when the abnormality score exceeds a preset threshold, wherein the abnormality score is calculated as follows: ;

[0035] In the formula, a set of mask segments, a value representing the mth channel of the original time series signal, the pth sampling point, and the kth segment, a value representing the corresponding position of the reconstructed signal, a standard deviation representing the mth channel signal, representing gradient calculation, a standard deviation representing the mth channel gradient value, representing a preset weight factor.

[0036] The application provides a power management device based on artificial intelligence and voltage stabilization protection, comprising a switching power supply circuit and an explosion-proof protection circuit.

[0037] The switching power supply circuit comprises an electromagnetic filtering part, a first rectifying filtering part, a conversion part, a second rectifying filtering part, a detection part and a control part.

[0038] The explosion-proof protection circuit comprises a filtering protection part, a voltage stabilization protection part and a current limiting protection part.

[0039] The application has the following beneficial results by adopting the above scheme:

[0040] (1) In view of the problem that the existing power management system lacks effective and rapid response and circuit protection mechanism, the application designs multiple protection mechanisms, suppresses the surge, overvoltage and electromagnetic interference at the input end through the electromagnetic filtering unit, improves the anti-interference ability, designs the voltage stabilization protection unit and the current limiting protection unit to work cooperatively, ensures the stability of the output voltage, quickly cuts off the circuit in an abnormal situation to prevent damage to the rear-end load, uses the detection unit to feed back the output voltage in real time, combines the control unit to realize closed-loop regulation, dynamically adjusts the working state of the transformer, improves the voltage stabilization precision, and in addition, the units of the application have clear division of labor and can be optimized separately, thereby enhancing the adaptability to different input environments.

[0041] (2) In view of the problem that the existing technology has poor recognition ability for abnormal current and voltage data, the application adopts a cross-modal encoding anomaly detection method based on multi-scale dynamic perception, realizes synchronous detection of slow drift and transient pulse through global-local joint modeling, adjusts the mask proportion based on signal entropy to focus on key signal segments, improves the robustness to complex working conditions, fuses the time sequence and channel space relationship, enhances the correlation of multi-sensor data, and finally improves the recognition ability for abnormal current and voltage. DETAILED DESCRIPTION

[0042] Figure 1 A module diagram of the power management system based on artificial intelligence and voltage stabilization protection is provided.

[0043] Figure 2A flowchart of a cross-modal encoding anomaly detection method based on multi-scale dynamic perception.

[0044] The accompanying drawings are included to provide a further understanding of the application, and are incorporated in and constitute a part of the specification, illustrate embodiments of the application, and explain the application without limiting the application. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0046] Embodiment one, refer to Figure 1 The application provides a power management system based on artificial intelligence and voltage stabilization protection, which comprises a switching power supply module, an explosion-proof protection module and a multi-scale intelligent detection module.

[0047] The switching power supply module comprises an electromagnetic filter unit, a first rectifier filter unit, a conversion unit, a second rectifier filter unit, a detection unit and a control unit.

[0048] The explosion-proof protection module comprises a filter protection unit, a voltage stabilization protection unit and a current limiting protection unit.

[0049] The multi-scale intelligent detection module adopts a cross-modal encoding anomaly detection method based on multi-scale dynamic perception to perform data acquisition and anomaly detection alarm on multi-channel current and voltage data of the switching power supply.

[0050] Embodiment two, this embodiment is based on the above-mentioned embodiment, in the switching power supply module:

[0051] The electromagnetic filter unit comprises a non-return fuse, a voltage-dependent resistor, a filter capacitor and a coupling inductor connected in sequence, the non-return fuse is connected in series at the input end of the power supply, the voltage-dependent resistor and the filter capacitor are connected in parallel at the rear stage of the non-return fuse, and the coupling inductor is connected in series at the rear stage of the filter capacitor.

[0052] The first rectifier filter unit comprises a rectifier bridge and a first filter capacitor, the input end of the rectifier bridge is connected to the output end of the coupling inductor, the first filter capacitor is connected in parallel at the output end of the rectifier bridge, and the rectifier bridge is composed of four rectifier diodes connected in a full-bridge mode.

[0053] The conversion unit comprises a transformer, the primary winding of the transformer is connected to the output end of the initial rectifier filter unit, and is used to convert the input voltage into a target value voltage.

[0054] The second rectification filter unit comprises a rectification diode, a second filter capacitor and a load resistor, the rectification diode is connected in series at the output end of the conversion unit, and the second filter capacitor and the load resistor are connected in parallel at the output end of the rectification diode;

[0055] The detection unit comprises a stabilizing tube and a detection resistor, the cathode of the stabilizing tube is connected to the output end of the second rectification filter unit, the anode is connected to the ground through the detection resistor, and the stabilizing tube is used for feeding back the target value voltage to the control unit;

[0056] The control unit comprises a switching power supply control chip and an optoelectronic coupler, the input end of the optoelectronic coupler is connected to the output end of the detection unit, the output end is connected to the feedback end of the switching power supply control chip, and the output end of the switching power supply control chip is connected to the control end of the conversion unit.

[0057] In the explosion-proof protection module in the third embodiment based on the above-mentioned embodiments,

[0058] The filter protection unit comprises a resistance-capacitance fuse device connected to the output end of the second rectification filter unit and used for filtering and primary protection of the target value voltage, and the resistance-capacitance fuse device comprises a fuse and a filter capacitor connected in series;

[0059] The voltage stabilizing protection unit comprises a thyristor and a voltage stabilizing diode, the cathode of the voltage stabilizing diode is connected to the output end of the second rectification filter unit, the anode is connected to the control electrode of the thyristor, and when the target value voltage is out of limit, the thyristor is turned on to make the resistance-capacitance fuse of the filter protection unit fuse;

[0060] The current limiting protection unit is integrated in the control unit and used for realizing overcurrent and overload protection of the power supply.

[0061] In the fourth embodiment based on the above-mentioned embodiments, in the multi-scale intelligent detection module, a cross-modal encoding anomaly detection method based on multi-scale dynamic perception specifically comprises the following steps:

[0062] Step S1: data preprocessing, the time sequence signal of current and voltage is normalized to obtain an original time sequence signal, wherein the dimension of the signal is MxN, M represents the number of channels, and N is the number of sampling points;

[0063] Step S2: data segmentation processing, the original time sequence signal is divided into non-overlapping segments along the time axis , wherein the length of the time sequence signal is Q, and the length of each non-overlapping segment is Q / T;

[0064] Step S3: multi-scale region construction, a subsequence is extracted from the original time sequence signal by a sliding window, and a local region is constructed , wherein This represents the window size, i.e., the length of the local area. Represents the starting position of the subsequence;

[0065] Step S4: Global mask for non-overlapping segments Random mask [ ] paragraph, in which A dynamic parameter that affects the length of the mask segment;

[0066] Step S5: Local masking, for local areas Random mask [ ] paragraph, in which ;

[0067] Step S6: Dynamic mask adjustment, adjusting dynamic parameters according to signal entropy. : ; ;

[0068] In the formula, The probability of the amplitude histogram within the segment. Represents signal entropy. These correspond to their maximum and minimum values, respectively. This represents the dynamically adjusted parameters;

[0069] Step S7: Multimodal feature encoding based on spatiotemporal embedding is performed on the non-masked segment (i.e., non-overlapping segment) and local region to obtain multimodal spatiotemporal encoding;

[0070] Step S8: Signal reconstruction. A mask segment marker is added to the multimodal spatiotemporal coding, and the signal is output through a back projection layer after single-layer transformer decoding to obtain the reconstructed signal.

[0071] Step S9: Anomaly detection. An anomaly score is calculated using a sliding window. When the anomaly score exceeds a preset threshold, an anomaly alarm is issued. The anomaly score is calculated as follows: ;

[0072] In the formula, The set representing the mask segments, This represents the value of the m-th channel, p-th sampling point, and k-th segment of the original time-series signal. The value represents the location of the reconstructed signal. This represents the standard deviation of the m-th channel signal. Represents gradient calculation. This represents the standard deviation of the gradient value of the m-th channel. This represents the preset weighting factor.

[0073] Example 5, this example is based on the above examples, step S7 specifically includes the following steps:

[0074] Step S71: Projection of non-masked segments. Projection operations are performed on non-overlapping segments and local regions respectively to obtain global projection and local projection.

[0075] Step S72: Position encoding, performing position encoding embedding based on the sine function for both global and local projections, wherein the position encoding based on the sine function is as follows: ;

[0076] In the formula, Represents the length of non-overlapping segments or local regions;

[0077] Step S73: Spatial encoding, performing spatial encoding embedding based on the channel embedding matrix for both global and local projections, wherein the spatial encoding based on the channel embedding matrix is ​​as follows: ;

[0078] In the formula, Represents one-hot encoding. Channel m represents the non-overlapping segment and the local region. The embedding matrix representing channel m;

[0079] Step S74: Multimodal embedding, performing multimodal temporal and spatial embedding for global and local projections, and using a multi-layer lightweight transformer based on residual connections to obtain multimodal spatiotemporal coding: z 0 =[ g t + e t time + e m space ; l ω + e ω time + e m space ] ; ;

[0080] In the formula, and These represent global projection and local projection, respectively. Represents multimodal spatiotemporal coding. Representing the Transformer encoding, Representative level normalization, Represents multi-head self-attention mechanism processing, This represents feedforward neural network processing.

[0081] Example 6: Application of the present invention in the power protection circuit of a battery-powered gas meter:

[0082] Scenario: Wireless remote gas meter powered by lithium-ion batteries

[0083] (1) Electromagnetic filter unit

[0084] Non-returning fuse (PPTC): Connected in series with the positive terminal of the battery to prevent short circuits or overcurrent from damaging the battery;

[0085] TVS diodes (replacement for varistors): connected in parallel at the battery input to suppress static electricity or surge damage;

[0086] Filter capacitors: filter out high-frequency noise and improve power supply purity;

[0087] (2) Conversion unit

[0088] DC-DC step-down chip: Reduces the voltage of a 3.6V lithium-ion battery to 2.5V for use by MCU and RF modules;

[0089] (3) Detection unit

[0090] Zener diode + voltage divider resistor: Detects the output voltage; if it exceeds 3.6V (which may indicate a DC-DC fault), it triggers protection.

[0091] (4) Voltage stabilization and protection unit

[0092] Thyristor: When an overvoltage is detected, the thyristor conducts, causing the PPTC to melt and cut off the power supply;

[0093] (5) Current limiting protection unit

[0094] Integrated into DC-DC chip: Built-in overcurrent protection to prevent damage to the circuit from sudden current surges when the RF module transmits.

[0095] Example 7: Application of the present invention in the power protection circuit of a mains-powered gas meter:

[0096] Scenario: IoT gas meter powered by AC mains (220VAC) converted to 5VDC

[0097] (1) Electromagnetic filter unit

[0098] Varistor (MOV, 14D471K): Absorbs power grid surges (lightning strikes).

[0099] X capacitor + common mode inductor: Filter out high-frequency interference from the power grid;

[0100] (2) First-stage rectifier and filter unit

[0101] Rectifier bridge: Converts 220VAC to high-voltage DC;

[0102] Electrolytic capacitor (100μF / 400V): Voltage after smoothing rectification;

[0103] (3) Conversion unit

[0104] Flyback switching power supply: Converts high-voltage DC to 5V / 1A to power the gas meter's main control and communication modules;

[0105] (4) Voltage stabilization and protection unit

[0106] SCR + Zener diode: If the output voltage exceeds 6V (e.g., the feedback optocoupler fails), the SCR will conduct and blow the fuse (250V / 1A).

[0107] (5) Current limiting protection unit

[0108] Fuse resistor: limits short-circuit current and protects the switching transistor.

[0109] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0110] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0111] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A power management system based on artificial intelligence and voltage regulation protection, characterized in that: Includes a switching power supply module, an explosion-proof protection module, and a multi-scale intelligent detection module; The switching power supply module includes an electromagnetic filter unit, a first rectifier filter unit, a conversion unit, a second rectifier filter unit, a detection unit, and a control unit; The explosion-proof protection module includes a filtering protection unit, a voltage stabilizing protection unit, and a current limiting protection unit; The multi-scale intelligent detection module employs a cross-modal coding anomaly detection method based on multi-scale dynamic perception to collect data and detect anomalies in the multi-channel current and voltage data of the switching power supply.

2. The power management system based on artificial intelligence and voltage regulation protection according to claim 1, characterized in that: In the switching power supply module: The electromagnetic filtering unit includes a non-returnable fuse, a varistor, a filter capacitor, and a coupling inductor connected in sequence. The non-returnable fuse is connected in series at the power input terminal, the varistor and the filter capacitor are connected in parallel after the non-returnable fuse, and the coupling inductor is connected in series after the filter capacitor. The first rectifier and filter unit includes a rectifier bridge and a first filter capacitor. The input terminal of the rectifier bridge is connected to the output terminal of the coupling inductor. The first filter capacitor is connected in parallel to the output terminal of the rectifier bridge. The rectifier bridge is composed of four rectifier diodes connected in a full-bridge manner. The conversion unit includes a transformer, the primary winding of which is connected to the output terminal of the initial rectifier and filter unit, and is used to convert the input voltage into the target voltage. The second rectifier and filter unit includes a rectifier diode, a second filter capacitor, and a load resistor. The rectifier diode is connected in series at the output terminal of the conversion unit, and the second filter capacitor and the load resistor are connected in parallel at the output terminal of the rectifier diode. The detection unit includes a Zener diode and a detection resistor. The cathode of the Zener diode is connected to the output terminal of the second rectifier and filter unit, and the anode is grounded through the detection resistor, which is used to feed back the target voltage value to the control unit. The control unit includes a switching power supply control chip and an optocoupler. The input end of the optocoupler is connected to the output end of the detection unit, and the output end is connected to the feedback end of the switching power supply control chip. The output end of the switching power supply control chip is connected to the control end of the conversion unit.

3. The power management system based on artificial intelligence and voltage regulation protection according to claim 1, characterized in that: In the explosion-proof protection module: The filtering protection unit includes a resistor-capacitor fuse device connected to the output terminal of the second rectifier filter unit, used for filtering the target voltage and providing primary protection. The resistor-capacitor fuse device includes a fuse and a filter capacitor connected in series. The voltage regulation protection unit includes a thyristor and a Zener diode. The cathode of the Zener diode is connected to the output terminal of the second rectifier filter unit, and the anode is connected to the control electrode of the thyristor. When the target voltage exceeds the limit, the thyristor conducts, causing the RC fuse of the filter protection unit to blow. The current limiting protection unit is integrated into the control unit and is used to realize overcurrent and overload protection of the power supply.

4. The power management system based on artificial intelligence and voltage regulation protection according to claim 1, characterized in that: In the multi-scale intelligent detection module, a cross-modal coding anomaly detection method based on multi-scale dynamic perception specifically includes the following steps: Step S1: Normalize the current and voltage timing signals to obtain the original timing signals, where the signal dimension is M×N; Step S2: Divide the original time-series signal into non-overlapping segments along the time axis. ; Step S3: Extract subsequences from the original time-series signal using a sliding window and construct local regions. ; Step S4: For non-overlapping segments Random mask [ ] paragraph, in which A dynamic parameter that affects the length of the mask segment; Step S5: For local areas Random mask [ ] paragraph, in which ; Step S6: Adjust dynamic parameters based on signal entropy ; Step S7: Perform multimodal feature encoding based on spatiotemporal embedding on the non-masked segment (i.e., non-overlapping segment) and local region to obtain multimodal spatiotemporal encoding; Step S8: Add mask segment markers to the multimodal spatiotemporal coding, and output it through the back projection layer after single-layer transformer decoding to obtain the reconstructed signal; Step S9: Anomaly detection. An anomaly score is calculated using a sliding window. When the anomaly score exceeds a preset threshold, an anomaly alarm is issued. The anomaly score is calculated as follows: ； In the formula, The set representing the mask segments, This represents the value of the m-th channel, p-th sampling point, and k-th segment of the original time-series signal. The value represents the location of the reconstructed signal. This represents the standard deviation of the m-th channel signal. Represents gradient calculation. This represents the standard deviation of the gradient value of the m-th channel. This represents the preset weighting factor.

5. The present invention provides a power management device based on artificial intelligence and voltage regulation protection, characterized in that: Including switching power supply circuits and explosion-proof protection circuits; The switching power supply circuit includes an electromagnetic filter section, a first rectifier filter section, a conversion section, a second rectifier filter section, a detection section, and a control section. The explosion-proof protection circuit includes a filtering protection section, a voltage regulation protection section, and a current limiting protection section.

Images