78 results about "Neuron network" patented technology

The invention discloses a novel power system primary frequency modulation optimization method based on deep learning, and relates to the technical field of power optimization. The method comprises the following steps: collecting state data of a hydroelectric generating set running in a power grid, constructing a three-dimensional feature tensor, and extracting spatio-temporal features through a TCN-GRU hybrid network; constructing a deep reinforcement learning decision layer, and defining an action space and a reward function; constructing a PID neural network with a time-varying forgetting factor, and outputting a dynamic weight; establishing a dual-time scale updating mechanism; a safety verification module is arranged, and when the system frequency deviation value is larger than a threshold value, the traditional PID mode is switched. Characteristics are extracted from state data of operation of a hydroelectric generating set in a power grid, the time sequence dependency relation of the characteristics is processed through deep learning, the frequency change trend of the power grid is captured, the optimal control strategy is explored by applying the DRL technology, PID controller parameters are adjusted in a dynamic environment, and the optimal control strategy is obtained. And the primary frequency modulation response speed and the control precision of the hydroelectric generating set during power grid frequency fluctuation are improved through self-adaptive control.

A system and method for real-time time series forecasting using a compound large codeword model with integrated supervisory neurons. The system processes diverse inputs through adaptive codebook generation and codeword allocation. A projection network fuses different data types for a latent transformer-based machine learning core. A hierarchical supervisory network, comprising low-level, mid-level, and high-level nodes, monitors local neural network regions, performing real-time statistical analysis and implementing structural modifications. The system efficiently handles multi-modal data, capturing complex relationships between input types. An adaptive codebook generation method, coupled with the supervisory architecture, ensures responsiveness to evolving data patterns and task requirements. This approach provides accurate and timely forecasts by leveraging diverse data types in a sophisticated, integrated manner, while continuously adapting its structure during operation to maintain optimal performance.

The invention discloses a method and device for automatic roasting and process intelligent control of a belt type roasting machine, and relates to the technical field of pellet roasting of the belt type roasting machine in the metallurgical iron and steel industry. The method comprises the four steps of historical data slice acquisition and modeling, target process parameter determination, feedback verification and online correction and rolling prediction control, a model historical library is established through service slices and time slices, modeling is performed based on a three-layer feed-forward neural network, flexible weighted deviation calculation combined with dynamic weight and rolling optimization of a softening coefficient are combined, and the rolling prediction control is performed. And accurate regulation and control of equipment parameters are realized. The automatic control system is mainly used for automatically controlling the pellet production process of the belt type roasting machine, the product quality stability is improved, the energy consumption is reduced, the labor intensity of workers is relieved, and support is provided for unmanned and intelligent production.

Disclosed is a method for determining a vehicle clutch temperature of a vehicle clutch by a neuron network. The method includes determining at least one input value representing a power supplied to the vehicle clutch, such that the at least one input value is determined on the basis of processing consecutive values of power supplied to the vehicle clutch. At least one input value and at least one value of an operating parameter are input as input data into the neuron network. A clutch temperature is determined by the neuron network on the basis of the input data and of a relationship learned by the neuron network between a time variation of the input data and the clutch temperature. A method is also disclosed for generating a training data set for a neuron network, and a control unit for determining a vehicle clutch temperature using a neuron network.

The invention discloses a wireless communication network intelligent optimization method and system based on neuron collaboration, and relates to the technical field of wireless communication. The method comprises the following steps: mapping a communication node into a bottom layer sensing neuron and constructing a neural state variable set; constructing a node neural situation function based on the variables; when the local threshold value is exceeded, excitation pulses are generated and uploaded to middle-layer convergence neurons; the middle layer carries out pulse space aggregation and extreme value search, and outputs a selection strategy and an adjustment strategy; and reporting to a top layer to execute whole network neural synchronization index analysis, and optimizing the wireless communication network. The technical problems of low spectrum resource utilization rate and unstable network performance caused by the fact that a traditional wireless communication network cannot realize high-efficiency spectrum allocation and dynamic topology reconstruction under user mobility change are solved, and the purposes of realizing local quick response and global collaborative optimization by constructing a layered neural network and improving the network performance are achieved. And the spectrum resource utilization rate is improved, and the network dynamic adaptive capability is enhanced, so that the user service quality is guaranteed.

The embodiment of the invention discloses a business scene digital transformation method and device, equipment and a storage medium, and relates to the technical field of business scene digital transformation information. According to the scheme, digital plans of business scenes with different granularities are aggregated, features and resources of the business scenes are fused, multi-modal data are associated, key features are extracted, and the semantic comprehension ability of the digital plans and a large language model is intelligently matched; the neural network of the service scene is calculated according to the service scene attribution, the associated data and the positioning information, and the disposal scheme corresponding to the service scene is generated, so that subjective deviation of a traditional manual compiling solution is avoided, sudden working conditions and various service scenes are flexibly adapted, and the adaptability of complex scenes is enhanced.

The invention discloses an intelligent detection navigation system and method for an inland ship. The method comprises the following steps: step 1, acquiring multi-modal sensing data in the navigation process of the inland ship; 2, constructing an initial cognitive map; step 3, carrying out dynamic modeling on each semantic state node by adopting an improved liquid neural network, embedding symbolic logic constraints in state transition, and generating a navigation control bias steering quantity; step 4, calculating a tension value of a causal directed edge and executing a structure evolution operation to generate a self-evolution cognitive map; 5, constructing a candidate path set; step 6, calculating an atlas winding degree index of each candidate path; 7, constructing a path decision vector, executing a path backstepping subduction operation, and screening to obtain a target optimal path; and 8, converting the target optimal path into a navigation control instruction, and outputting the navigation control instruction to a ship control system. According to the invention, cognitive map modeling and neural network reasoning are fused, and intelligent navigation control of inland ships is realized.

The present invention discloses a microfluidic bio-interconnected neural network chip and a preparation method thereof. The chip consists of a microelectrode array, a microfluidic structure, and a cell culture ring structure. The microelectrode array can detect and stimulate the activity of neurons; the microfluidic structure limits the growth direction of axons between cell culture partitions by setting cell culture partitions, microfluidic channel dimensions, microfluidic channel shapes, and fillet distribution, thereby achieving controllable low-power transmission of neuronal electrical information; the cell culture ring structure provides nutrients for cells in the cell culture chamber. The present invention provides a method for manufacturing a microfluidic bio-interconnected neural network chip, using fully exposed positive photoresist as a sacrificial layer for microfluidic structure peeling, thereby solving the problem of difficult demolding of thin-film microfluidic structures. The chip can be used to explore the response characteristics of neuronal networks to external stimuli, study the internal information transmission mode of neuronal networks, and other fields.

The invention discloses a substation monitoring and early warning method based on a neural network and a related device. The method comprises the following steps: predicting predicted values of key parameters of a substation in a future period of time based on the neural network; comparing the predicted value with an actual measured value, and when the deviation between the predicted value and the actual measured value is greater than or equal to a preset deviation, triggering and sending out a preliminary abnormal prompt; when the preliminary abnormal prompt is received, the preset value is compared with a preset safe operation domain of the preset value, when the preset safe operation domain is exceeded, an early warning signal is generated, and the method and the related device can improve the reliability, the economical efficiency and the intelligent level of operation of the transformer substation.

The invention discloses a BIM (Building Information Modeling) and Internet of Things integrated multi-parameter collaborative intelligent dosing system and device, and relates to the technical field of sewage treatment intellectualization, and the system comprises a BIM modeling unit, a data acquisition unit, a multi-parameter collaborative analysis unit, a dosing execution unit and a water quality analysis unit; according to the method, a 3D building information model is exported through a Revit interface to obtain a water treatment BIM model, a unique identifier is allocated to each component in the water treatment BIM model, a water quality prediction model is constructed based on a neural network algorithm, and the water quality prediction model is constructed by combining pipeline layout and equipment position information in the water treatment BIM model, dynamically calculating the dosage demand and adopting a fuzzy PID control strategy. The frequency of a dosing pump is automatically adjusted according to the deviation between real-time water quality parameters and target values, the dosing pump, a disinfectant preparation device and a proportioning system are driven through a PLC, a dosing adjustment instruction is executed, component state labels are updated, historical component state labels are obtained, water quality state differences are evaluated according to component state difference values, and a water quality evaluation report is generated.

The invention relates to an unmanned aerial vehicle intelligent path planning method and system based on visual signals, and belongs to the field of path planning. The method comprises the following steps: optimizing a neural network through progressive pruning; fuzzy coordinates and scene types are provided for the unmanned aerial vehicle, the unmanned aerial vehicle flies to a fuzzy coordinate area, and when the unmanned aerial vehicle is close to the fuzzy coordinate area, the unmanned aerial vehicle is switched to a pure vision mode to perform environment perception through a neural network, and a three-dimensional semantic map is constructed; performing real-time path planning based on the three-dimensional semantic map to obtain a flight path; the method comprises the following steps: performing autonomous flight along a flight path, when approaching an operation target, performing autonomous operation to obtain real-time monitoring data, storing the real-time monitoring data in airborne storage equipment, and returning the real-time monitoring data after returning. And an unmanned aerial vehicle intelligent path planning system with high robustness, high autonomy and high task adaptability is constructed.

The application discloses a kind of isomorphic neuron system for generating hidden coexisting attractor, it is related to neuromorphic computing and nonlinear circuit field, the system includes: memristor, first neuron network and second neuron network;The structure of the first neuron network and the second neuron network is identical;Synaptic connection is carried out between the first neuron network and the second neuron network by the memristor, to simulate the chaotic dynamic behavior of biological neuron, to generate hidden coexisting attractor.The application can accurately simulate the synaptic transmission characteristics of biological neuron, improve the authenticity of bionics, successfully reproduce the chaotic dynamic behavior of biological neuron, efficiently generate hidden coexisting attractor, and enrich the dynamic performance of neuron system.

The invention relates to a gamma photon positioning method and device for an orthogonal strip-shaped tellurium-zinc-cadmium detector. The method comprises the following steps: acquiring current signals respectively generated by interaction of a plurality of gamma rays and the orthogonal strip-shaped tellurium-zinc-cadmium detector; vectorizing the plurality of current signals to obtain an input vector set; an initialized neural network is constructed based on the positions of a plurality of electrode strips in the orthogonal strip-shaped cadmium zinc telluride detector, and the number of initialized neurons is the same as that of the plurality of electrode strips; performing weight vectorization processing on the plurality of initialized neurons to construct an initialized weight vector set; and based on the input vector set and the initialized weight vector set, iteratively reducing the range of the initialized weight vector set through a self-organizing mapping algorithm to obtain a target weight vector, and determining the position of the gamma photon. The purpose of determining the position of the gamma photon based on the neural network and the self-organizing mapping algorithm is achieved, and therefore the technical effects of improving the accuracy of the position of the gamma photon and improving the position obtaining efficiency are achieved.

A method is provided for correcting a reconstruction artefact of a three-dimensional tomographic image. The method includes the steps of providing an acquired three-dimensional tomographic image from a cell group, and applying the image to a neural network trained in advance to determine a corrected tomographic image.

The present application relates to a kind of bottle neck edge shrink model, neuron network and its construction method, bottle neck edge shrink model is arranged in bottle neck edge shrink module, the bottle neck edge shrink model at least includes two 1*1 convolution layer and a layer m*m convolution layer, wherein m*m convolution layer is arranged between two 1*1 convolution layer, the edge shrink method of the bottle neck edge shrink module includes: G (x) = c_2 (sigma (conv (sigma (c_1 (x))))) ) + r (x) ;Wherein, conv (X) indicates that m*m convolution transformation is carried out to image matrix x without padding, and m>1;C_1 (x) and c_2 (x) indicate that 1*1 convolution transformation is carried out to image matrix x, sigma (x) indicates that nonlinear transformation is carried out to image matrix x;R (x) indicates that edge is shrunk to image matrix x.The bottle neck edge shrink model of the present application reduces the channel number of image matrix by 1×1 convolution layer c_1, so that the parameter number and the amount of calculation of 3×3 convolution kernel are greatly reduced, so that the parameter number of bottle neck edge shrink module is less than the corresponding basic edge shrink module, while the corresponding network depth is increased by 50%, by reducing the parameter number and improving the stability of entire neuron network.

The invention discloses a distributed building heating parameter intelligent optimization method and system, and the method comprises the steps: carrying out the multi-scale feature extraction of a heating region through an improved SSD algorithm, building a population distribution dynamic data sequence through a heating time sequence and a population optimization model, collecting heating system parameters, and carrying out the normalization processing, and the data are input into an optimization decision model, an optimization combination scheme is calculated and output through a multi-layer neural network, heating parameters are adjusted in real time according to the optimization combination scheme, and operation data are fed back to an iterative optimization model in a set period. The system comprises a data acquisition unit, a feature extraction unit, a time sequence and population analysis unit, an optimization decision unit, a parameter adjustment unit and a feedback updating unit which work cooperatively. According to the method and system, environment and population dynamic changes can be fully considered, parameter association is deeply excavated, intelligent optimization of heating parameters is achieved, and the heating effect and the energy utilization efficiency are improved.

The invention discloses an isomorphic neuron system for generating hidden coexisting attractors, and relates to the field of neuromorphic calculation and nonlinear circuits, and the system comprises a memristor, a first neuron network and a second neuron network. The structures of the first neural network and the second neural network are the same; the first neuron network and the second neuron network are in synaptic connection through the memristor so as to simulate chaotic dynamic behaviors of biological neurons and generate hidden coexisting attractors. According to the invention, synaptic transfer characteristics of the biological neurons can be accurately simulated, bionic authenticity is improved, chaotic dynamic behaviors of the biological neurons are successfully reproduced, hidden coexisting attractors are efficiently generated, and dynamic performance of a neuron system is enriched.

The invention provides an output voltage estimation system and method for a dielectric barrier discharge pulse resonant converter. The system comprises a pulse resonant conversion module, a data processing module, a dielectric barrier discharge load output voltage prediction neural network model and an upper computer. Wherein the pulse resonant conversion module is used for converting a direct-current bus voltage into a high-frequency high-voltage pulse voltage and providing the high-frequency high-voltage pulse voltage to the dielectric barrier discharge load; the data processing module is used for collecting the input voltage of the pulse resonant transformation module and transmitting the input voltage to the dielectric barrier discharge load output voltage prediction neural network model; the output voltage prediction neural network model of the dielectric barrier discharge load estimates the peak output voltage of the dielectric barrier discharge load by taking the acquired input voltage and the current control switching frequency as input characteristics; and the upper computer is used for setting parameters and receiving and displaying the estimated value of the peak output voltage.

The invention provides a mouse neuron network construction method and system based on two-photon imaging, and the method comprises the steps: collecting neuron imaging data through a two-photon calcium imaging system, and extracting a fluorescence change time sequence of each neuron from the neuron imaging data; constructing a function connection matrix according to the fluorescence change time sequence, and then calculating a small world coefficient and a network core degree of the neural network based on the function connection matrix; and based on the small-world coefficient and the network core degree, using a liquid state machine to establish a brain-like neuron network. The problems that in the prior art, the network connection mode between single neurons cannot be researched at the cellular level, and effective simulation of a spatial-temporal information processing mechanism of a biological nervous system is lacked are solved.

This invention discloses a method and system for intelligent optimization of heating parameters in distributed buildings. The method uses an improved SSD algorithm to extract multi-scale features from the heating area, establishes a dynamic population distribution data sequence using a heating time-series and population optimization model, collects and normalizes heating system parameters, inputs the data into an optimization decision model, and outputs an optimized combination scheme through a multi-layer neural network. Heating parameters are adjusted in real time based on this, and the operating data is fed back to the optimization model at a set period. The system includes data acquisition, feature extraction, time-series and population analysis, optimization decision-making, parameter adjustment, and feedback update units, all of which work collaboratively. This method and system can fully consider dynamic changes in the environment and population, deeply explore parameter correlations, achieve intelligent optimization of heating parameters, and improve heating performance and energy efficiency.

The invention discloses a high-precision temperature and humidity control method for a clean room. The high-precision temperature and humidity control method comprises the steps that the flow of chilled water introduced into a primary cooling section, the flow of pure water introduced into a humidifying section, the flow of medium-temperature water introduced into a heating section and the flow of chilled water introduced into a secondary cooling section are controlled through regulating valves V1-V4 correspondingly; constructing a PID (Proportion Integration Differentiation) neuron network controller, connecting the regulating valve with the PID neuron network controller, and setting temperature and humidity control target values T0 and H0 of the clean room and an initial network weight; acquiring measured actual temperature and humidity values T and H of the clean room through a temperature and humidity sensor; the PID neural network controller calculates and obtains valve position control values of the regulating valves V1-V4; comparing difference values between T0 and H0 and the adjusted T and H, and if the difference values meet requirements, continuously monitoring the temperature and humidity of the clean room; otherwise, correcting the network weight. The PID neural network control technology is adopted, the control quantity can be automatically adjusted, the temperature and the humidity serve as independent output quantities respectively for decoupling control, and therefore high-precision control over the temperature and the humidity of the clean room is achieved, and meanwhile system energy consumption is reduced.

The invention provides a capacitor abnormal discharge detection method and device, a storage medium and computer equipment, and relates to the technical field of capacitor state monitoring, and the method comprises the steps: collecting multiple paths of voltages and multiple paths of currents of a capacitor to be detected, and carrying out the pulse coding of each path of voltage and each path of current, and obtaining each path of pulse sequence; a spiking neural network is determined, the number of input channels of the spiking neural network is the same as the number of the pulse sequences, and the spiking neural network establishes a relation with the characteristic response of abnormal discharge by adopting a specific neuron group; and inputting each pulse sequence into the pulse neuron network to obtain the abnormal discharge type of the capacitor to be detected. Therefore, the adaptive learning capability of the complex pulse sequence can be realized, so that the health state of the capacitor can be accurately and comprehensively evaluated.

The application discloses a small-scale Hopfield neural network circuit based on a memristor, which comprises an activation function module circuit, a memristor model circuit and a Hopfield neural network circuit; the Hopfield neural network circuit comprises three neuron networks X1, X2 and X3; and a connection weight in the Hopfeild neural network is replaced by the memristor module circuit to obtain a new neural network. By replacing the synaptic weight of the Hopfield neural network system, it is found that the originally stable system exhibits rich dynamic behaviors, including chaos, hyperchaos, quasi-periodicity, double-vortex attractor and the like. The circuit is simple in design, and can adjust the weight parameters Rr and the parameter w32 to realize various dynamic behaviors. Therefore, the circuit can be used in the field of secret communication and is helpful for the research on the nervous system.

The invention discloses a low-delay pulse neural network conversion method and system based on threshold value correction and BPTT fine tuning, and the method comprises the steps: introducing a nonlinear threshold value correction function in a training stage of a source ANN, forcibly mapping a trainable threshold value parameter into a strict positive value, and carrying out the calculation of the positive value; the risks of excessive activity and function collapse of the neural network caused by a negative threshold value are thoroughly eliminated from a physical mechanism; and then, constructing a fine tuning framework based on BPTT, and carrying out time sequence dynamics optimization on the converted SNN by using a teacher-student architecture and a dynamically weighted composite loss function. The method not only ensures the stability of the conversion process, but also guides the network to learn a high-efficiency coding strategy of'front-end sparse and rear-end focusing ', and realizes high-precision and high-energy efficiency reasoning under ultra-low delay.

The invention discloses a low-cost in-vitro brain-computer interface chip batch preparation method and a low-cost in-vitro brain-computer interface chip batch preparation device, and belongs to the technical field of biosensor and micro-nano manufacturing. The method comprises the following steps: cleaning, photoetching, sputtering a metal layer, stripping, depositing an insulating layer, overlay, etching and exposing an electrode site and a bonding pad on a glass or silicon insulating substrate in sequence, preparing a 3D electrode by combining secondary overlay with electrochemical deposition, and finally separating a plurality of independent microelectrode arrays through scribing or laser cutting. The microelectrode bonding pad and the adapter plate are connected by adopting an aluminum wire pressure welding technology, and an aluminum wire is embedded by high-temperature-resistant insulating glue and adheres to the cell culture ring. According to the method, the microelectrode arrays of various specifications are prepared on a single substrate in batches, the process compatibility is high, the cost is low, the efficiency is high, microscopic observation and 3D electrode integration are facilitated, and the method is suitable for in-vitro neuron network detection and organoid electrophysiology and electrochemical research.

The application belongs to the field of artificial intelligence, and provides a training method and device of a training architecture based on a blockchain and electronic equipment, the method comprising: generating a training task through a task management contract of a blockchain in the training architecture, the training task comprising training task parameters, the training task parameters comprising: a stimulus category and an input sequence; executing neural training based on the received training task through a plurality of biological neuron network nodes in the training architecture to generate node response results of the plurality of biological neuron network nodes; and dynamically adjusting the training task parameters and a stimulation scheme based on an evaluation result and an aggregation result through a training coordinator in the training architecture. The training method provided in the application embodiment can not only execute neural training based on the received training task to generate node response results of the plurality of biological neuron network nodes, but also dynamically adjust the training task parameters and the stimulation scheme based on the evaluation result and the aggregation result.

The invention provides an intelligent control method and system based on an in-vitro culture biological neuron network, and the method comprises the steps: carrying out the extraction and in-vitro culture of neurons of an organism, and constructing a biological chip through a microelectrode array chip; selecting an electrode area of the biochip based on the emission rate scanned from the channel of the biochip; determining a relative distance based on the position of the body in the real-time task picture to be judged and the position of the target body, and selecting an electrode area I or an electrode area II for applying stimulation to the biochip and stimulation intensity based on the relative distance; the method comprises the following steps: acquiring feedback of a biochip to stimulation, constructing an initial feedback vector, calculating a saliency parameter for each channel in the initial feedback vector, selecting each channel based on the saliency parameter of each channel, and constructing a channel vector based on the selected channel; and inputting the channel vector into a pre-trained classification model, and outputting an action through the classification model.

The application provides an output voltage estimation system and method of a dielectric barrier discharge pulse resonant converter, which comprises a pulse resonant conversion module, a data processing module, a dielectric barrier discharge load output voltage prediction neuron network model and a host computer; the pulse resonant conversion module is used for converting a DC bus voltage into a high-frequency high-voltage pulse voltage and providing the high-frequency high-voltage pulse voltage to a dielectric barrier discharge load; the data processing module is used for collecting an input voltage of the pulse resonant conversion module and transmitting the input voltage to the dielectric barrier discharge load output voltage prediction neuron network model; the dielectric barrier discharge load output voltage prediction neuron network model takes the collected input voltage and a current control switch frequency as input features and estimates a peak output voltage of the dielectric barrier discharge load; and the host computer is used for parameter setting and receiving and displaying the peak output voltage estimation value.

The application provides a protein molecular machine motion mode simulation method, comprising the following steps: S10, obtaining a protein molecular machine structure data file; S20, simulating the molecular dynamics of the protein molecular machine to obtain a complex and its receptor, ligand motion topological file and coordinate file, and calculating the binding free energy thereof; S30, predicting the motion trajectory of the protein molecular machine based on the built NRI model and the coordinate file, wherein the NRI model comprises an encoder for predicting the interaction of a given dynamic system trajectory, and a decoder for predicting the given dynamic system trajectory in the interaction graph; and S40, observing and recording the three-dimensional structure and motion trajectory of the protein molecular machine based on a VR device. The application realizes accurate simulation and prediction of the motion mode of the protein molecular machine by combining molecular dynamics simulation and deep neural network.

The application particularly relates to a multi-energy hybrid energy storage system and a coordinated energy management strategy for an underwater vehicle. The system comprises a lithium ion battery, a super capacitor and a fuel cell using a proton exchange membrane, which are connected to a DC bus respectively. The method of the coordinated energy management strategy comprises: obtaining current motion parameters and current energy parameters of the underwater vehicle; inputting the current motion parameters, the current energy parameters and a desired speed into a load prediction model based on a BP neuron network to estimate a desired total power at a next time; determining a corresponding action strategy according to a current operation mode of the underwater vehicle, and performing power distribution on the desired total power based on the action strategy. The scheme can enable the hybrid energy system to perform smooth operation mode switching and optimal power distribution under different working conditions.