23results about How to "Reduce computing burden" patented technology

The invention relates to the technical field of machine learning and software development, in particular to a software development system based on a machine learning algorithm, which comprises an acquisition processing unit, a model training unit, an interaction prediction unit and an optimization feedback unit, and is characterized in that the acquisition processing unit constructs a database, extracts key elements such as variable names and converts the key elements into vectors; the model training unit adopts recurrent neural network hierarchical training and gradually introduces intermediate and complex structure codes and learns logic relation and semantic relation from code snippets related to a basic grammar structure, the interaction prediction unit screens and extracts the code snippets according to code function module labels and historical use frequency, the code snippets are input into a model after vector conversion, and the model is used for prediction. A probability weighting algorithm is used for predicting subsequent fragments, an optimization feedback unit calculates the similarity between predicted fragments and actual fragments, an optimization mechanism is automatically triggered if a threshold value is not reached, parameters are adjusted by utilizing reinforcement learning, and prediction accuracy is used as a key part of a reward function so as to improve the system performance.

The invention provides a training method of a steel surface defect detection model. The training method of the steel surface defect detection model comprises the following steps that S1, the steel surface defect detection model is initialized, the steel surface defect detection model is a YOLO v8 network, and a convolution-to-full connection C2f module in a backbone network of the YOLO v8 network adopts double branches for coding; s2, a training data set is constructed, the training data set comprises images of steel of different defect types and images of steel without defects, and each image further comprises the corresponding defect type as a label; and S3, training the steel surface defect detection model based on the training data set, and updating model parameters based on a preset loss function. Through a mode of double-branch parallel coding and final feature fusion, the feature extraction capability of a backbone network is improved, and the accuracy of a defect detection model is enhanced; and the calculation burden is reduced by introducing the dynamic convolution and the phantom convolution.

The invention relates to the technical field of data processing, in particular to a covariance inverse matrix recursive updating method for on-satellite hyperspectral anomaly detection, which comprises the following steps: acquiring a hyperspectral data stream; carrying out adaptive dimension reduction processing on the data stream and loading an initial background model; extracting local background statistics by adopting a sliding window mechanism, generating a background spectrum dictionary by utilizing online dictionary learning, calculating a reconstruction error between a current pixel and the dictionary and a Mahalanobis distance between the current pixel and a background model, and fusing to generate an abnormal score; performing abnormal confidence coefficient evaluation by combining the spatial context information and the spectral angle matching degree, and updating a spectral mean vector by using an exponential weighted moving average algorithm based on non-abnormal pixel data; and the updated model is injected into the next round of processing to form a recursive chain. According to the method, the high calculation complexity of direct inversion of a covariance matrix is avoided, real-time anomaly detection on a satellite is realized, the downloading amount of original data is remarkably reduced, and the congestion of a satellite-ground communication link is relieved.

The invention provides a power distribution network grounding fault line selection method based on characteristic frequency band optimization, and the method comprises the following steps: obtaining a transient zero sequence signal of each line, the transient zero sequence signal comprising a normalized transient zero sequence current and a transient zero sequence voltage derivative; decomposing the transient zero-sequence signal by adopting a wavelet packet transformation method; a historical data set is constructed, and frequency bands capable of effectively representing fault features in the historical data set are extracted in an off-line self-adaptive mode through a genetic algorithm; collecting data after the fault, and analyzing and extracting a characteristic frequency band in the fault data by utilizing wavelet packet energy entropy; performing intersection extraction processing on the characteristic frequency bands extracted by the genetic algorithm and the wavelet packet energy entropy to obtain a final optimized characteristic frequency band result; and for a real-time fault, carrying out fault research and judgment by utilizing the transient zero-sequence signal in the optimal characteristic frequency band. According to the invention, waveform data collected by all detection nodes does not need to be uploaded to a power distribution master station, and the position of the single-phase earth fault can be researched and judged on site.

The invention discloses a camera exposure time control method under high-dynamic background light intensity change, and aims to solve the problem of camera overexposure caused by environment light intensity change in optical measurement. The method comprises the following steps: building a test system, and setting a camera frame rate; setting initial exposure time, and shooting an image of the measured object to obtain an optimal gray scale; adjusting parameters of the signal processing circuit; and starting the to-be-tested equipment, and obtaining the exposure time when the ambient light intensity rapidly changes. According to the invention, an open-loop control method is applied, the ambient light intensity is captured in real time through the photodiode, and the ambient light intensity is converted into the camera exposure time inversely proportional to the ambient light intensity through the signal processing circuit, so that rapid adjustment of the camera exposure time is realized. Compared with a fixed exposure time measuring method and a traditional method based on an image gray value, the method has the advantages of being high in response speed, small in calculation burden, real-time in feedback and the like, and is particularly suitable for an application scene where the light intensity changes rapidly in a high-temperature environment.

The application relates to the technical field of multi-modal data processing, and discloses an AI-based cross-medium artistic design collaborative creation system, which comprises the following modules: an initialization and anchor point module that extracts features and clusters to generate a static mapping matrix; a residual extraction and evaluation module that calculates the feature difference before and after source medium modification and outputs a scalar energy value; a manifold cutting and routing module that performs spatial cutting according to the scalar energy value or starts a self-attention mechanism to generate disturbed subspace data with a transition calculation domain; a damping iteration synchronization module that generates a dynamic damping matrix in combination with a physical topological density scalar field, iteratively optimizes the feature parameters, and outputs convergent features; and a decoding fusion module that recombines and decodes the convergent features and the frozen static data and outputs a target medium physical file. The application realizes accurate quantification of local modification and prevents topological distortion during synchronous conversion of multi-modal data.

This invention patent provides a sound coordination processing method and apparatus for binaural hearing aids. The binaural hearing aid has two audio acquisition devices located on either side of the head, and the two audio acquisition devices are communicatively connected. The method includes: the two audio acquisition devices synchronously acquiring sounds from the surrounding environment and adding timestamps to the sounds; parsing the sounds into multiple sound signals with different audio characteristics, the sound signals excluding noise signals; adaptively adjusting the sound signals according to the order in which they are acquired; merging the adaptively adjusted sound signals into an output signal; and recording the sound signal characteristics acquired by one hearing aid within a certain time domain so that when the other hearing aid acquires the sound signal with the same characteristics, there is no need for secondary filtering of that sound signal.

The application discloses a kind of based on anti-amplitude and rate saturation paddle drive unmanned vehicle formation control method, belong to unmanned vehicle formation control technical field.The existing paddle drive unmanned platform executor is aimed at solving the problem of control failure and system oscillation caused by amplitude and rate saturation constraint in physical level, including: paddle drive unmanned vehicle dynamics model is established integrated executor dynamic characteristics;Based on the dynamics model, the control constraint model for paddle drive unmanned vehicle is constructed;Design anti-saturation formation cooperative control law under combined connected topology.The method offsets saturation deviation by auxiliary compensation system, and combines neural network to estimate system disturbance online.The amplitude and rate saturation problem of paddle drive unmanned vehicle under the vertical plane working condition is solved, the formation stability is maintained under the topology switching environment, and the control accuracy and safety of the system are improved.

The invention relates to the technical field of power supplies, in particular to an output voltage calibration method and a power supply. Comprises: obtaining a target voltage; based on the output calibration coefficient and the target voltage, the current PWM duty ratio is calculated through the master controller according to a PWM duty ratio calibration formula; adjusting the output voltage based on the current PWM duty ratio through a hardware control loop; sampling the adjusted output voltage through a sampling module to obtain a sampling voltage; judging whether the error between the sampling voltage and the target voltage is within an allowable error range or not through the main controller; if yes, the calibration is successful; if not, updating the output calibration coefficient according to an output calibration coefficient updating formula, and then recalculating the current PWM duty ratio; by introducing the output calibration coefficient and dynamically updating the output calibration coefficient, the inherent error of an actual electronic element is accurately offset, so that the output voltage error is stabilized in an allowable range, the requirement of a high-precision voltage output scene is met, special working conditions do not need to be identified, and the software development difficulty is reduced.

This application relates to the field of water treatment technology, and discloses a control method, water treatment device, and medium for a water treatment apparatus. The method includes: in response to a venting command, a controller controls the venting valve to open and controls the water pump to start; after the water pump starts, a negative pressure is generated inside the hot tank; a one-way air inlet valve is configured to automatically open when its positive pressure value is greater than a first threshold, and a one-way air outlet valve is configured to automatically open when its positive pressure value is greater than a second threshold; during the venting process, the controller keeps the venting valve open and the water pump running until a preset venting termination condition is met. This application utilizes the automatic pressure response characteristics of a mechanical one-way valve to achieve adaptive adjustment of negative pressure during the venting process without the need for a pressure sensor and an electrically controlled air inlet valve, simplifying the control logic, reducing hardware costs, and improving venting efficiency and system reliability.

The invention relates to an equipment management system based on data acquisition, which is applied to the field of control systems, and comprises an equipment management processing unit and equipment ends, and adopts cooperation of a multi-layer equipment data acquisition unit, a lower-level equipment verification unit and each equipment end, so that on one hand, data acquisition steps are effectively simplified; the complexity of data collection and the single-time transmission data volume are reduced, the processing difficulty and operation burden of a subsequent equipment management processing unit on the collected data are effectively reduced, the operation efficiency of the equipment management system is improved, meanwhile, the accuracy of equipment management can be effectively improved, on the other hand, the accuracy of the collected data can be effectively improved, and the user experience is improved. According to the method, the collected data are verified by utilizing the mutual relationship of the equipment ends, so that the accuracy and effectiveness of the collected data are effectively ensured, and the economical efficiency of the application of the equipment management system can be effectively promoted while the construction cost of the equipment management system is reduced.

The application discloses an image cropping method, an image acquisition device and a storage medium, relates to the technical field of image processing, and discloses an image cropping method, which comprises the following steps: determining a face existence detection result of a current image photographed by an image acquisition device based on a face detection head of a CNN convolutional neural network; determining predicted cropping point coordinates based on a cropping point positioning head parallel to the face detection head in the CNN convolutional neural network; and if the face existence detection result is that there is a face, cropping the current image based on the predicted cropping point coordinates. Through the architecture of the parallel work of the double-task head of the CNN convolutional neural network after image shooting, face existence detection and positioning of cropping coordinates are synchronously completed, the image can be automatically cropped without user operation, the image does not need to be screened one by one, the image processing process is simplified, and the image processing efficiency is improved.

The utility model discloses a kind of operation fault self-diagnosis devices of distributed energy storage system. The main body of device is equipped with rotatable photovoltaic module, and lower end is provided with battery compartment. Photovoltaic module is powered for battery compartment, and battery compartment is powered for detection unit. Battery compartment is connected with main body through the through hole of distributed energy storage system upper cover, so that the device is stably installed on energy storage system. When installing, install battery compartment and connect each component in the top hole of energy storage container;When debugging and operating, adjust the angle of photovoltaic module and collect data through internal detection unit, data processing module analyzes operating state, and triggers alarm when abnormal. The device can monitor the operating state of battery pack in energy storage system in real time, so as to discover fault in time, widely applicable, effectively improve the stability and safety of energy storage system operation.

This invention belongs to the field of spectrum analysis technology and relates to a low-resource adaptive spectrum amplitude-frequency correction method applicable to high-speed rail and multi-industry engineering signals. The steps are as follows: Step 1: Spectrum peak search and extraction; Step 2: Adaptive correction judgment; Step 3: Hybrid correction calculation. This method constructs an adaptive discrimination mechanism and simplifies the correction calculation process. While controlling computational complexity, it compensates for spectral amplitude and frequency deviations, thus balancing correction accuracy and computational efficiency. The method is suitable for low-computing-power embedded processing platforms such as microcontrollers, enabling the correction of spectral amplitude and frequency parameters under limited computing resources.

The application relates to an AI large model-based supercomputing center peak load shifting and valley filling electric management platform, and relates to the technical field of automatic control of a power system, which comprises a user interaction module, an AI large model module, a pricing module and a scheduling module. The user interaction module receives a calculation task uploaded by a user and feeds back a calculation price of the calculation task to the user; the AI large model module predicts the computing power consumption of a new calculation task by analyzing historical task data; the pricing module generates an estimated price in combination with a time-of-use electricity price; and the scheduling module dynamically schedules the calculation task to be executed in a valley value period or a peak value period by setting a periodization capacity threshold value. The application provides an estimated price before the calculation task is executed, guides the user to select a valley value period with a lower cost, effectively reduces the calculation cost of the user, balances the load of the supercomputing center, and achieves the effect of "peak load shifting and valley filling".

The present invention provides a satellite communication system and method for managing radio resources of a non-terrestrial network. The method comprises: transmitting a first resource scheduling allocation by a first satellite when leaving a service area of ​​the non-terrestrial network; receiving a second resource scheduling allocation corresponding to the first resource scheduling allocation by a second satellite when entering the service area; and accessing radio resources by the second satellite according to the second resource scheduling allocation.

The application discloses a fuzzy PI PWM control method applied to a BOOST circuit and belongs to the technical field of power electronics. According to the method, the initial values of a group of proportional coefficients and integral coefficients with the highest boost efficiency are debugged in advance, fuzzy PI control parameters are generated, proportional coefficient correction amounts and integral coefficient correction amounts corresponding to different output voltage deviation amounts and output voltage change rates are generated and stored in a memory in a table form, then the output voltage is sampled at a fixed time, the output voltage deviation amount and the output voltage change rate are calculated, the proportional coefficient correction amount and the integral coefficient correction amount are quickly obtained by table lookup, and then the next round of proportional coefficient and integral coefficient are calculated and used for PI control of a voltage loop. The method improves the tracking precision of the output voltage, avoids the need for fuzzy PI recalculation each time, significantly reduces the number of floating point operations in adaptive fuzzy PI control, and reduces the operation burden of a single-chip microcomputer.

The present application relates to the field of computer technology, and in particular to a log statement collection method, anomaly detection method and system, and related devices. The log statement collection method includes: obtaining a log method in the source code of a target system, screening an associated method in the source code of the target system that is associated with the obtained log method; based on the associated method and the log method, collecting log statements of the target system; serializing the log statements to obtain log sequences, and assigning category labels to the log sequences; wherein the log sequence categories are associated with the target system analysis requirements. The use of this method can collect the computational burden of log statements in the source code and is conducive to improving the comprehensiveness of the collection, thereby realizing anomaly detection of the storage system based on the operation log of the storage system.

This invention discloses a tracked flipping machine and a flipping method. The tracked flipping machine includes a plate thickness detection unit, a conveying unit, a flipping mechanism, and a control system. The plate thickness detection unit is located at the front end of the conveying unit and is used to detect the plate thickness of the circuit board product. The conveying unit is used to convey the product. The flipping mechanism is located at the rear end of the conveying unit and includes a base, a drive motor, at least two sprockets, a chain, and multiple sets of adapter slots. The chain is composed of chain links, and one end of the adapter slot is detachably connected to a chain link, while the other end has a card slot groove. The width of the card slot grooves in each set increases sequentially. The control system is electrically connected to the plate thickness detection unit and the drive motor. Based on the detected plate thickness, the control system controls the motor to drive the sprockets to rotate by a predetermined angle, so that the predetermined adapter slot of the corresponding width moves to the rear of the conveying unit to adapt to and accommodate the product. The tracked flipping machine of this invention has a simple structure, low modification cost, reliable control, and can accurately adapt to the product.

This application discloses a photovoltaic panel vibration monitoring method based on heterogeneous fusion network communication. The method includes: after each sensor terminal completes vibration data acquisition and processing within the same acquisition cycle, it sends the local relative timestamp generated by the free-running timer to the gateway base station along with the uplink data packet; the gateway base station calculates the relative transmission delay difference between any pair of nodes based on the packet reception time and the local relative timestamp, and constructs a frequency domain phase correction factor at each discrete frequency point accordingly to correct the uncorrected cross-spectral density, obtaining the corrected cross-spectral density; further, it calculates the coherence function value by combining the self-spectral density of each node, and locates the vibration anomaly area based on the spatial distribution of the coherence function value on the photovoltaic array topology. This invention can improve the accuracy of multi-node vibration joint analysis and the reliability of anomaly location, and is suitable for online monitoring of photovoltaic power plants.

The invention discloses a frequency modulation control method and device for a network tracking / constructing hybrid photovoltaic power station. The method comprises the following steps: respectively establishing a network tracking type photovoltaic equivalent unit and a network constructing type photovoltaic equivalent unit; calculating a frequency modulation power amplitude limiting value of the grid-following type photovoltaic equivalent unit by taking the short-circuit ratio as a reference; when the system is subjected to power disturbance, distributing a corresponding first network following type frequency modulation instruction value and a corresponding first network constructing type frequency modulation instruction value to the overall frequency modulation power requirement of the photovoltaic power station according to the steady-state operation power proportion of the network following type photovoltaic equivalent unit and the network constructing type photovoltaic equivalent unit respectively; if the first tracking network type frequency modulation instruction value is greater than the frequency modulation power amplitude limiting value, the tracking network type photovoltaic equivalent unit participates in system frequency modulation according to the frequency modulation power amplitude limiting value; and adding the difference value between the first grid-following type frequency modulation instruction value and the frequency modulation power amplitude limiting value and the first grid-constructing type frequency modulation instruction value to serve as a final frequency modulation instruction value of the grid-constructing type photovoltaic equivalent unit, so that the mixed type photovoltaic power station stably supports the system frequency under the low short circuit ratio.

This invention belongs to the field of power electronic conversion and microgrid control technology, specifically involving a synchronous control method for grid-connected / off-grid switching of an islanded microgrid based on a virtual generator. The method includes transforming the three-phase voltage signals from the microgrid side and the grid side from a three-phase stationary coordinate system to a two-phase stationary coordinate system, obtaining orthogonal components of the microgrid voltage and grid voltage. These orthogonal components are then substituted into a calculation formula to obtain a real-time phase difference signal. This real-time phase difference signal is then fed into a regulator for proportional-integral and filtering processing to generate a phase compensation coefficient. This phase compensation coefficient is then superimposed onto the active frequency control loop of the virtual synchronous generator to correct the phase angle of the virtual synchronous generator's output voltage. This achieves the goal of maintaining the simplicity of the non-phase-locked loop structure while reducing the inrush current during grid-connected / off-grid switching.

The application discloses a double-Hall limit position sensing method for preventing over-travel failure, belongs to the technical field of Hall limit position sensing, and comprises an ascending edge detection module for input preprocessing, filtering of signal burrs, response to a real trigger edge, output of sequence judgment timing, a sequence latching module for judging the movement direction of a reading head through Hall trigger sequence judgment, output of latching or resetting basis, a conditional trigger module for generating a control signal in an effective scene where both Halls are triggered, to avoid false triggering, a limit latching module for continuously latching a high level after over-limiting, and an operation output module for outputting a high level when a first Hall switch is triggered when no over-travel occurs, still outputting a high level when the Hall becomes low, outputting a low level when the Hall becomes low after limit resetting, and enabling the limit protection of both sides in linear motion to be realized by selecting single-pole Hall switches with different magnetic pole actions.