51results about How to "Improve output stability" patented technology

This invention discloses an individual motion behavior analysis system based on dynamic modal configuration and rule-guided fusion, belonging to the fields of intelligent perception, sports science, and human-computer interaction. It includes a data input module, a dynamic modal organization and configuration module, a multimodal feature modeling and fusion module, an indicator-level evaluation and behavior state prediction module, a statistical rule-guided analysis and result correction module, and a structured result generation and output module. These modules work together to form a complete closed loop for motion behavior analysis technology. This invention aims to address the problems in existing motion behavior analysis technologies, such as limited dimensionality of single-modal information, rigid multimodal fusion methods, insufficient stability and interpretability of analysis results, and strong dependence on large-scale manually labeled data.

A control circuit for a power converter includes a first sensing circuit, a reference voltage generating circuit, an error amplifier circuit and a pulse width modulation circuit. The first sensing circuit is coupled to a first output circuit to provide a first current sensing signal. The reference voltage generating circuit is coupled to the first sensing circuit and provides a reference voltage according to the first current sensing signal. The error amplifier circuit is coupled to the reference voltage generating circuit to receive the reference voltage and an output feedback voltage of the power converter to provide an error amplifier signal. The pulse width modulation circuit is coupled between the error amplifier circuit and the first output circuit to receive the error amplifier signal and provide a control signal to control the first output circuit. The reference voltage generating circuit also receives the error amplifier signal and adjusts the reference voltage generated according to the error amplifier signal. The control circuit for the power converter can effectively improve the output stability without affecting the output voltage under long time load.

This invention provides a microgap suspension device and a microgap inverted suspension heat treatment process, comprising the following steps: adding p-aminobenzenesulfonic acid to a solvent, mixing thoroughly, and then adding carbon nanotubes to obtain a carbon nanotube dispersion; stirring the carbon nanotube dispersion under heating conditions to obtain a p-aminobenzenesulfonic acid-modified carbon nanotube dispersion; coating a substrate surface with a lead bromide solution, then coating the lead bromide surface with a cesium bromide solution, then coating the cesium bromide surface with the p-aminobenzenesulfonic acid-modified carbon nanotube dispersion; and finally performing heat treatment through a microgap suspension device to form a photoactive layer film. The microgap inverted suspension heat treatment process of this invention, through controllable height suspension heat treatment, ensures that the film is almost unconstrained by the rigidity of the substrate during thermal expansion and cooling contraction, thus making it suitable for flexible substrates as well. It also avoids direct contact damage to flexible substrates during high-temperature heat treatment.

This invention discloses a sensor for detecting ice, water, and dryness states, comprising a substrate, a first metal line, a second metal line, a third metal line disposed on the surface of the substrate, a variable capacitor, an interdigitated capacitor, and a planar inductor. One end of the variable capacitor is connected to one end of the planar inductor via the first metal line, and the other end of the planar inductor is connected to one end of the interdigitated capacitor via the second metal line. The other end of the interdigitated capacitor is connected to the other end of the variable capacitor via the third metal line. The capacitance value of the interdigitated capacitor changes with the different dielectric constants of the materials between the interdigitated fingers. Since the relative dielectric constants of ice, water, and dryness change at different frequencies, the capacitance value of the interdigitated capacitor also changes. Therefore, this sensor for detecting ice, water, and dryness states has advantages such as simple structure, high recognition rate, small size, and process compatibility.

The invention relates to the technical field of machine learning, and discloses a medical field large language model training method and system based on knowledge graph enhancement, and the method comprises the steps: obtaining a medical corpus and a knowledge graph; performing fine-grained semantic analysis to generate an entity sequence; reasoning and extracting a multi-hop sub-graph by a multi-dimensional association path; embedding the graph into the dynamic injection text based on the dynamic weight to generate an enhanced training sample; parsing the codes into a standardized instruction set; and performing iterative optimization to obtain a target model. Through adaptive knowledge fusion and standardized instruction coding, the training efficiency and professional reasoning ability of the medical big language model are improved.

The invention provides a display panel and a display device. A pixel circuit of the display panel comprises a driving transistor, a data writing module and a threshold grabbing module. The first end of the data writing module is electrically connected with a data signal line, and the second end of the data writing module is electrically connected with the first end of the driving transistor; the first end of the threshold grabbing module is electrically connected with the second end of the driving transistor, and the second end of the threshold grabbing module is electrically connected with the first grid electrode of the driving transistor; a second grid electrode of the driving transistor is electrically connected with the grid voltage control line, and a film layer where the second grid electrode of the driving transistor is located is located between the active layer and the substrate. Wherein in the data writing stage, the time-varying signal transmitted by the grid voltage control line is a high-level signal; in the light emitting stage, the time-varying signal transmitted by the grid voltage control line is a low-level signal. According to the scheme, the display abnormity phenomenon can be weakened or eliminated from the two aspects of improving the data signal writing efficiency and enhancing the output stability of the light-emitting driving current.

A lead-free packaged all-quartz resonant pressure sensor comprises a 3J21 alloy shell cover, the 3J21 alloy shell cover is connected with a 3J21 alloy shell in a sealed mode, and a quartz island film pressure conversion unit, a quartz resonant layer element and a quartz base component which are connected through a vacuum glass frit sealing technology are arranged in an inner cavity formed by the 3J21 alloy shell cover and the 3J21 alloy shell. Glass frit dielectric layers are arranged between the quartz island film pressure conversion unit and the quartz resonance layer element as well as between the quartz resonance layer element and the quartz base part; three quartz layers and two glass frit dielectric layers are connected to form an all-quartz pressure chip; the all-quartz pressure chip, the 3J21 alloy shell and the gold-plated lead column are connected through sintering of glass silver paste, and a chip electrode signal is led out through the gold-plated lead column; according to the invention, the long-term stability of the sensor is significantly improved, the mechanical strength and reliability of electrical connection are enhanced, real-time and homologous sensing of temperature signals is realized, and the compensation precision of the sensor in a full temperature range is effectively improved.

The application discloses a power assisting curve self-adaptive generation method and system based on riding habit learning, relates to the field of intelligent control of electric power assisting bicycles, and comprises the following steps: establishing a riding habit library based on historical riding data of target riders, and storing a pedaling force habit center and a pedaling frequency habit center representing long-term pedaling behavior characteristics of the riders; collecting multi-source real-time sensing data such as pedal torque, pedaling frequency, vehicle driving state and battery state in real time; constructing a comfort envelope model through the pedaling force habit center and the pedaling frequency habit center, and calculating a comfort coefficient; generating comprehensive control indexes in combination with the comfort coefficient, the vehicle driving state and the battery state data, dynamically adjusting the shape parameters of a basic power assisting function, generating a self-adaptive power assisting curve matched with a current riding scene, and then determining a motor power assisting torque instruction and outputting the motor power assisting torque instruction to a motor controller. In this way, the individualized coordination and coupling of man-machine torque aiming at natural pedaling force performance of the riding user habits are realized.

The application discloses a smoothing maintenance method, system, device and storage medium for an extremely low bit quantization language model, and belongs to the technical field of artificial intelligence and natural language processing. The method comprises the following steps: obtaining a large language model to be quantized; performing hierarchical quantization processing on the model; constructing a joint optimization target; and finally generating a quantization model. The application aims at the defects that the existing extremely low bit quantization technology only optimizes the forward fitting error and ignores the model smoothing and gradient propagation characteristics, constructs a joint optimization target containing the forward fitting error term and the gradient maintenance term in the quantization process, ensures the output consistency of the quantization model and the original model through the forward fitting error term, and restricts the input gradient consistency of the model before and after quantization through the gradient maintenance term, so that the smoothing of the model is maintained.

This invention discloses a directional adaptive wave energy generator and its power generation method. In this invention, raft one and raft two oscillate back and forth with the waves. During oscillation, a magnetic coupling transmission device and a one-way bearing drive the drive shaft of the triboelectric nano-power generation mechanism located in the sealed chamber to rotate unidirectionally. The drive shaft drives the FEP friction layer in the rotor of each power generation component to contact and rub against the MF foam friction layer in the stator, generating triboelectric charge. Furthermore, the relative positions between the FEP friction layer and the copper electrodes A and B in the stator periodically change, creating an alternating potential difference between copper electrodes A and B. Electrons flow back and forth between copper electrodes A and B, forming an alternating current, thus achieving power generation. Raft two is equipped with a tail fin, enabling the device to automatically adjust its attitude according to the wave or current direction. This invention can convert low-frequency wave energy into a stable current output for power generation and exhibits strong adaptability to multi-directional waves and complex sea conditions.

This invention relates to the field of switching power supply technology, specifically providing an isolated power supply circuit aimed at solving the problems of low integration, poor output consistency, and high cost and energy consumption in existing multi-channel isolated power supplies. To this end, the invention comprises: a power control and drive module, an isolation module, a rectification, filtering, and voltage regulation module, an anti-interference module, and a load module. The power control and drive module receives power, generates a drive voltage signal, and outputs the drive voltage signal to the isolation module, while simultaneously outputting a ground potential voltage signal to the anti-interference module. The isolation module achieves voltage signal isolation output through internal electromagnetic coupling, outputting the isolated voltage signal to the rectification and filtering module. The anti-interference module outputs the filtered, interference-free ground potential voltage signal to the rectification and filtering module. The rectification and filtering module performs rectification and filtering processes, outputting a stable voltage signal to the load module. This invention features high integration and high output consistency.

The invention relates to a high-power green laser synthesis system based on a rotary frequency doubling crystal and multiple beams of infrared pulse laser, and belongs to the technical field of laser. The system comprises a multi-beam pulse infrared laser device, a frequency doubling crystal module, an annular rotating mechanism, a beam combining device, a continuous infrared laser device, a laser frequency regulation and control device and a rotating driving control system. The four beams of pulse infrared laser sources are symmetrically arranged along the circumferential edge, and periodically staggered output is realized through frequency regulation and control; the multiple frequency doubling crystals are distributed along the rotating circular ring at equal intervals and are sequentially aligned with laser pulses under driving of the rotating mechanism, and efficient frequency doubling is achieved. The center continuous infrared laser generates stable green light through frequency doubling of the fixed crystal. Finally, the green light generated by the multiple frequency doubling crystals and the central green light are integrated into a single light beam through a beam combining device to be output. According to the system, the power upper limit and the stability of green laser are remarkably improved through energy sharing of polycrystals, rotary dynamic matching and annular cooling, and the problems that existing green laser is limited in power and serious in heat effect are solved.

The application discloses a kind of plaque three-dimensional parameter atlas reconstruction evaluation methods based on coronary center line constraint, comprising: obtaining and preprocessing coronary CCTA volume data;Using the continuous action deep reinforcement learning model based on Actor-Critic architecture to extract center line and radius value, generate tubular region of interest;Orthogonal section resampling is carried out, section sequence is constructed, and double-coordinate joint representation is formed;Input VM-UNet segmentation network, obtain the segmentation result of lumen, vessel wall and plaque component;Introduce pixel-level contrast learning strategy in the training of segmentation network;From the segmentation result, plaque three-dimensional parameter atlas is reconstructed by back projection;Construct plaque component relationship graph;Input uncertainty perception multimodal fusion network, obtain risk result and corresponding confidence;When confidence result exceeds preset threshold, resample and reconstruct plaque component relationship graph, obtain corrected result.

The application discloses a flood peak prediction method and system, the method comprising: obtaining topographic elevation data of a target area and target rainfall pattern information; inputting the topographic elevation data and the target rainfall pattern information into a pre-trained flood peak prediction model to obtain a flood peak prediction result, the prediction result comprising a maximum water depth spatial distribution of the target area; wherein the flood peak prediction model comprises a generation network constructed based on a conditional generative adversarial network, the generation network introducing a topographic attention mechanism and a multi-scale rainfall encoding mechanism in the process of generating the maximum water depth spatial distribution, and the topographic feature being a feature representation obtained by feature extraction on the topographic elevation data. The application can directly output the maximum water depth spatial distribution in the prediction stage, reduce the dependence on numerical simulation iteration, shorten the result acquisition period, and enhance the adaptability and output stability under different rainfall pattern scenarios.

The invention discloses a material flow self-adaptive control system in a condiment continuous proportioning process, and relates to the technical field of condiment proportioning control, and the material flow self-adaptive control system comprises a main conveying unit used for continuously conveying a liquid basic condiment; the at least one auxiliary material feeding unit is connected with the output pipeline of the main conveying unit in a converging manner and is used for continuously adding solid seasoning auxiliary materials into the liquid basic seasoning; the temperature detection unit is used for detecting the temperature of the liquid basic seasoning in real time; the controller is in signal connection with the main conveying unit, the auxiliary material feeding unit, the pressure detection unit, the flow detection unit, the auxiliary material flow detection device and the temperature detection unit; according to the invention, the rheological state of the main body material is sensed on line through coupling analysis of the pipeline pressure difference and the pump power signal, and according to dynamic feedforward compensation auxiliary material ratio setting, compound control is formed, so that the influence of natural fluctuation of the main body material characteristics on the mixing uniformity can be effectively and adaptively counteracted.

This invention discloses a hybrid high-precision power source control method and device integrating dynamic error cancellation technology, relating to the field of power metering and testing technology. The method first constructs an asymmetric heterogeneous hybrid power stage topology composed of a high-power IGBT inverter main stage and a high-frequency broadband linear auxiliary stage connected in parallel. Then, it utilizes an FPGA high-speed acquisition link to separate the comprehensive error signal, including dead-zone distortion and switching noise, in real time, and introduces an iterative learning control algorithm to establish a dead-zone inverse model reflecting the device characteristics. Subsequently, based on model prediction, a leading inverse cancellation command is generated to drive the linear auxiliary stage to inject a precision compensation current, performing nanosecond-level dynamic shaping and ripple cancellation on the main stage output. Simultaneously, by monitoring the load current amplitude, full-range adaptive operating condition switching is performed, locking the main stage under weak current conditions to avoid nonlinear interference. This invention, through master-slave collaborative control and dynamic error cancellation, effectively overcomes the bottleneck of traditional power sources' difficulty in simultaneously achieving high energy efficiency and high signal-to-noise ratio, completely eliminating zero-crossing distortion, ensuring output stability and high waveform fidelity across the entire range from microamps to hundreds of amps, and providing ideal standard signal support for power metering traceability.

The invention discloses a road state detection method based on zero sample open set learning, and adopts the following innovative scheme: a spatial semantic fusion module is provided, and semantic information can be directly embedded into a network model to enhance the target recognition capability; according to the module, the recognition performance of the model in different scene contexts is remarkably improved, so that accurate road surface state detection can be realized even in an extremely complex environment; according to the process, customized fine adjustment is carried out only for a small part of key parameters in the network, model weight updating can be completed under the extremely low data requirement, the model forgetting phenomenon is effectively relieved, new knowledge can be continuously absorbed, and the adaptability and deployment efficiency of the method are remarkably improved. The bounding box and the pixel-level mask are synchronously output, the mask module takes the bounding box as input, prediction is executed in a limited area, and robustness and output stability in a complex environment are effectively enhanced.

This invention relates to the field of geomagnetic vector measurement technology and proposes an in-situ rapid calibration method for a cubic magnetic gradient tensor measurement array. This invention utilizes a triaxial Helmholtz coil system to generate a constant magnetic field with random orientation and actively compensates for environmental magnetic drift. A hierarchical calibration algorithm is used to sequentially estimate sensor-level inherent errors and array-level alignment errors, and an evaluation framework based on tensor invariants is established to quantitatively analyze the calibration effect. Specifically, this includes: constructing a constant modulus random magnetic field, solving the hierarchical calibration algorithm, and evaluation and verification based on tensor invariants. This invention eliminates the need for mechanical array rotation, achieving in-situ rapid calibration in a static array state. It significantly suppresses pseudo-gradients introduced by channel inconsistencies, improving the stability of magnetic gradient tensor measurement by 3 to 5 orders of magnitude. It is applicable to magnetic anomaly detection and target localization scenarios in fields such as geological exploration, underwater tracking, and space magnetic surveying.

The application discloses a wide working field linear-rotary multi-port generator based on sea conditions and belongs to the technical field of magnetic gears and generators. The application is composed of a field modulation magnetic screw rod and a controllable multi-port generator. The field modulation magnetic screw rod converts low-speed linear motion of waves into high-speed rotary motion, thereby improving the power density of the direct-drive wave generator. By arranging a power winding and a control winding, the axial flux generator and the magnetic gear magnetic circuit are coupled, the air gap magnetic density is improved, the power density of the generator and the instantaneous power transmission capacity of the magnetic gear are increased, and the maximum power transmission capacity of the axial flux generator is improved through electric excitation, so that the wave range that can be captured in the working area of the device is effectively widened.

The embodiment of the invention provides a time sequence signal fault prediction method and device, equipment and a storage medium, and relates to the technical field of anomaly prediction. The method comprises the following steps: generating a simulation vibration signal based on the vibration response of a gear, and inputting the simulation vibration signal into a gear fault prediction model for signal prediction to obtain a predicted vibration signal; physical information loss is calculated based on predicted displacement, predicted speed and predicted acceleration in the predicted vibration signals, predicted error loss is calculated based on the simulated vibration signals and the predicted vibration signals, a total loss value is calculated according to the predicted error loss and the physical information loss, parameter adjustment is carried out, and a trained gear fault prediction model is obtained. And inputting a to-be-predicted signal into the trained gear fault prediction model for signal prediction to obtain a reference vibration signal, and obtaining a time sequence abnormal result based on the reference vibration signal and the to-be-predicted signal. The prediction process fits the actual characteristics of the signal and the physical law of gear vibration, and the recognition accuracy and stability are improved.

The utility model discloses a light intensity adjusting device and a laser optical system, and relates to the technical field of optics. The light intensity adjusting device comprises a beam expanding and shaping assembly and a polarization adjusting assembly which are sequentially arranged along a light path. The beam expanding and shaping assembly comprises a shaft distance adjusting piece, a first optical element and a second optical element, wherein the first optical element and the second optical element are arranged on the two opposite sides of the shaft distance adjusting piece respectively. The shaft distance adjusting piece drives the first optical element to move in the axial direction to get close to or get away from the second optical element. The polarization adjusting assembly comprises a rotation driving part, a wave plate and a polaroid, the wave plate is arranged between the polaroid and the second optical element, and the rotation driving part can drive the wave plate to rotate with the optical axis as the rotation axis; the first optical element, the second optical element, the wave plate and the polarizer are arranged along the same optical axis. The light intensity adjusting device is compact in structure, light intensity can be accurately adjusted, and the output stability and uniformity of laser are improved.

The application discloses an energy storage device based on reverse electrodialysis technology, and designs the structure of the energy storage device, so that the transmission ions can be transmitted in the first structure and the second structure to form an internal loop; meanwhile, due to the existence of oxidizing substances and reducing substances, when the oxidizing substances undergo reduction reactions and the reducing substances undergo oxidation reactions, an oxidation-reduction potential difference can be formed; since the oxidation reactions and the reduction reactions continuously proceed, the oxidation-reduction potential difference can be continuously provided, and when an electric signal is outputted based on at least the oxidation-reduction potential difference, the electric signal can be continuously outputted at a high level, so that the energy storage device has a high output power density and good output stability.

This invention discloses a high-power, broadband, tunable long-wavelength mid-infrared femtosecond laser generation device and method. The laser beam is split into a first beam and a second beam. The first beam pumps a first YAG crystal to generate a first supercontinuum, from which a signal light of a predetermined wavelength is selected. The signal light undergoes at least two cascaded optical parametric amplification stages to generate a high-power mid-infrared pump light. The second beam pumps a second YAG crystal to generate a second supercontinuum covering the long-wavelength infrared band as a seed light. This seed light is then incident on a BGGSe crystal for optical parametric amplification. Using a 1.5μm high-power femtosecond laser as the pump source, combined with the BGGSe nonlinear crystal for optical parametric amplification, the wide transmission range and high nonlinear coefficient of the BGGSe crystal are effectively utilized. Phase matching in the 6–18μm band can be achieved under 1.5μm pumping conditions, thereby obtaining a broadband, high-power, femtosecond-level laser output covering the long-wavelength mid-infrared band.

The invention discloses an end-side palm vein representation attack detection method and device based on key point guidance, and the method comprises the steps: inputting a palm image into a palm detection and key point positioning network model, obtaining a palm target frame, four key point coordinates at a palm finger gap valley point, and corresponding confidence, and completing the judgment of the validity of an input sample, if the samples are invalid samples, outputting marks and ending, and if the samples are valid samples, calculating palm posture correction parameters through key point coordinates, and performing rotation correction on the palm image; on the corrected image, whether key points for constructing the ROI conform to preset legality constraints or not is judged, if yes, the ROI is determined and cut according to the geometric relation of the key points, an ROI normalized image is obtained after zooming, and if not, the image is generated through an ROI rollback strategy; and inputting the ROI normalized image into the end-side palm vein representation attack detection network model to obtain an attack probability of the sample, comparing the attack probability with a preset threshold, outputting a judgment result, and synchronously outputting the attack probability and the preset threshold.

The application discloses a universal pressure sensor, which comprises a pressure seat provided with a measuring hole at the bottom end, and a shell fixedly connected with the top edge of the pressure seat at the bottom end, wherein the top end of the pressure seat is provided with a limiting groove, the opening of the shell is downwardly arranged, the top end of the pressure seat and the inner wall of the shell enclose a cavity, the top end of the pressure seat is fixedly provided with a binding plate and a strain gauge, the binding plate is provided with mounting holes and limiting holes, the strain gauge is arranged in the mounting hole, a metal wire is arranged between the strain gauge and the binding plate, a compensation plate is arranged above the binding plate and fixedly connected with the binding plate through a support, the lower end of the support is provided with a limiting block and a containing cavity, the limiting block is inserted into the limiting groove through the limiting hole, the opening of the containing cavity is downwardly arranged to contain the strain gauge and the metal wire, and a plurality of vertical wires are fixedly connected with the compensation plate. The application improves the overload capacity and the sealing performance of the pressure sensor and expands the range of the pressure sensor.

The utility model discloses a large-scale articulated bus road energy harvesting system, relates to the technical field of energy collection, and aims to solve the technical problem that a road energy harvesting system using vehicle vibration energy is urgently required to be developed in the prior art to generate power and supply power to road sensing equipment. The system comprises an input module, a conversion module and a power generation module which are connected in sequence. The input module comprises a fixing frame connected with a guide mechanism, the guide mechanism makes contact with a vehicle and receives vibration of the vehicle, an elastic energy storage device arranged between the fixing frame and the guide mechanism is arranged on the guide mechanism, and the guide mechanism is connected with the conversion module through a first connecting device. The conversion module comprises a rotating mechanism, the rotating mechanism converts linear motion of the guide mechanism into rotating motion, and the rotating mechanism is connected with a power generation module which performs electromagnetic induction power generation through rotation. The input module, the conversion module and the like are adopted, mechanical energy of vehicle vibration is converted into electric energy, energy is effectively converted and utilized, and energy conversion and utilization efficiency is improved.

This invention belongs to the technical field of silicon-based piezoresistive high-pressure sensors, and specifically relates to a dual-island beam membrane high-pressure sensor chip and its fabrication method. The chip features a dual-island beam on a sensitive membrane; a set of piezoresistive strips are arranged on the upper surface of the dual-island beams, positioned along two center lines of the chip body; the dual-island beams include an island and an island, with their ends connected to the device layer, insulating layer, and support layer via a beam body; the upper surfaces of the islands are flush with the upper surface of the beam body; this significantly enhances the lateral and longitudinal stress difference in the stress concentration area, making the stress concentration effect more pronounced, improving the sensor's sensitivity, and ensuring reliable measurement results under various operating conditions.

This invention relates to a temporal knowledge graph reasoning method based on causal reinforcement and counterfactual intervention. It inputs a historical temporal knowledge graph into a trained temporal knowledge graph reasoning model to output fact predictions corresponding to future timestamps. The temporal knowledge graph reasoning model includes a multi-view construction module, a causal reinforcement heterogeneous dual-stream encoder, an orthogonal feature decoupling module, a counterfactual hybrid intervention module, and a task decoder. The multi-view construction module generates a base view and a perturbed view for the same time snapshot. The causal reinforcement heterogeneous dual-stream encoder obtains a latent representation that considers both local structural information and global semantic information. The orthogonal feature decoupling module separates stable causal representations from environment-sensitive representations. The counterfactual hybrid intervention module constructs a counterfactual hybrid representation. The task decoder performs fact scoring and prediction. This invention improves the accuracy, stability, and generalization ability of temporal knowledge graph reasoning.

This utility model discloses a multi-output flyback power supply. By setting at least one current compensation module, the first end of the current compensation module is connected to the secondary winding, and the second end of the current compensation module is connected to the power output terminal corresponding to the output circuit connected to the secondary winding. The current compensation module can output a corresponding compensation current to the power output terminal according to the current of the output circuit. When the load connected to the output circuit changes, the current compensation module can provide compensation current to the power output terminal connected to the output circuit, reducing the voltage change of the secondary winding connected to the output circuit, thereby reducing the mutual influence of voltage between different secondary windings, optimizing the cross-regulation rate, and improving output stability.