84 results about "Kalman filtering algorithm" patented technology

The invention relates to the technical field of automatic control, and discloses a path planning method and device for mechanical arm and laser synchronous machining, and the method comprises the steps: obtaining a workpiece three-dimensional model, a material and technological parameters, and extracting an irregular curved surface to generate a discrete point set; the tail end pose of the mechanical arm is calculated through inverse kinematics, and a preliminary joint angle and a track sequence are obtained. The optimal laser parameters are matched in combination with the parameters, a change curve is generated and synchronously mapped to a joint angle sequence, and a synchronous joint angle sequence is obtained through spline interpolation fine adjustment. And fitting a smooth track by using a B spline curve, and generating a motion track control instruction in combination with the synchronous joint angle. And the mechanical arm and the laser move according to instructions, the track compensation coefficient is dynamically adjusted when the environment changes, online compensation is conducted through the self-adaptive Kalman filtering algorithm, and a correction control instruction is obtained. According to the method, irregular curved surface machining details are extracted, the mechanical arm and the laser working sequence are aligned, the machining environment is sensed in real time, the adaptability of path planning is improved, and therefore efficient and accurate machining is achieved.

The invention relates to the technical field of automatic control, and discloses a PID (Proportion Integration Differentiation) intelligent pressure reducer controller, which comprises the following steps of: judging whether an environment disturbance or a load sudden change event exists according to a system load change rate and sensor feedback stability, constructing a master control-controlled coupling characteristic matrix by adopting a dual-channel data fusion method, performing attribution splitting on each disturbance channel, and obtaining a PID (Proportion Integration Differentiation) intelligent pressure reducer. PID parameters are dynamically adjusted according to the principal component weight of a disturbance channel, a variable parameter PID adjustment matrix is constructed, filtering smoothing is carried out on sampled signals through an extended Kalman filtering algorithm, a feedback closed-loop correction vector based on sampling residual errors is constructed, a historical offset tracking and switching value boundary judgment mechanism is introduced, fine tuning suppression is carried out on control signals of a motor driving mechanism, and the control signals of the motor driving mechanism are output. Judging whether a mechanical limiting protection device needs to be activated or not; and if yes, automatically triggering voltage limiting protection logic and locking the current control quantity. The method has the advantage of improving efficient and stable operation.

A motion state observation method for autonomous navigation of a ship. The method uses a determination method for system matrixes Ad, Bd, Ed and Hd, and a determination method for a related measurement covariance matrix R and noise covariance matrix Q, and uses a discrete Kalman filtering algorithm to perform motion state observation and prediction for autonomous navigation of a smart ship. The method fully uses ship position and heading measurement information of a smart ship during autonomous navigation, uses a ship motion model and the discrete Kalman filtering algorithm, and selects an effective state observation variable, thus calculating reasonable and effective observer parameters. The state observation method can effectively calculate information such as the actual position, heading, speed and angular velocity of a smart ship at each time instant during autonomous navigation, and predict motion information at a next time instant, thus helping to reduce the difficulty of smart ship motion controller design, ensuring effective implementation of navigation control strategies, reducing ineffective thrust output and rotation of propulsion systems, and improving the autonomous navigation capability of smart ships.

The invention relates to an intelligent induction type cable quick connector plugging method, which comprises the steps of removing noise and errors in dust particle data through a signal processing module, constructing a first data set by adopting a Kalman filtering algorithm in combination with the processed dust particle data and ambient air pressure data, and storing the first data set according to a predefined format; by acquiring an output result of the particle swarm optimization model, analyzing to obtain a motion trend and concentration change of dust, monitoring an insulation state of the cable joint in real time through an insulation state sensor of the dust removal instrument, and constructing a second data set in combination with the motion trend and concentration change of dust particles and the insulation state of the cable joint; obtaining the working intensity of the dust removal instrument through a fuzzy logic controller in the dust removal instrument; according to the insulation state of the cable joint, judging whether the insulation performance of the cable joint reaches a safety standard; and collecting real-time plugging data in the plugging process, and evaluating the accuracy and safety of the plugging operation by analyzing the real-time plugging data.

The invention discloses a pumped storage unit fault diagnosis method based on multi-modal data fusion, and the method comprises the steps: collecting a voiceprint signal, an infrared thermal imaging image and historical operation data of a pumped storage unit, and carrying out the noise reduction of a voiceprint set through employing an improved unscented Kalman filtering algorithm, a COMRes + model is used to train the voiceprint signal after noise reduction and historical operation data, a future voiceprint signal is predicted, an improved Deeplabv3 + model is used to train an image set, and features of an infrared thermal imaging image are extracted; the voiceprint feature, the historical operation data text feature and the image feature are input into an improved CentraNet model for multi-modal data fusion, and a fault category is output through a classification recognition module, so that diagnosis of the pumped storage unit fault is completed; according to the method, through fusion of the multi-modal data and optimization of the deep learning model, the accuracy and efficiency of fault diagnosis can be effectively improved, and a powerful guarantee is provided for safe operation of the pumped storage unit.

The invention relates to the technical field of power system safety monitoring, in particular to a power transmission line online safety monitoring method, system and device based on quantum communication and a storage medium. The multi-modal fusion sensor synchronously collects multi-parameter electric signal data such as current, voltage and temperature of the power transmission line, and performs quantum encryption instantaneous recording on the data; based on quantum encryption instantaneous record extraction data, a deep learning adaptive Kalman filtering algorithm is combined with a neural network to predict the operation state of the power transmission line; generating a new Manchester code according to the multi-parameter electric signal data and the operation state predicted value, and encrypting by adopting quantum key distribution to generate a ciphertext; and transmitting the ciphertext to a client based on a terahertz communication channel, decrypting and recovering the data through the quantum key, and carrying out safety early warning. Theoretically uncracked data security assurance is realized by using the unclonable characteristic of the quantum state, high-rate data transmission is realized by terahertz communication, and the transformation from passive monitoring to active early warning is realized by a prediction algorithm.

The invention relates to the technical field of data signal processing, in particular to a conductivity meter data processing method based on sage-husa self-adaptive Kalman filtering, which comprises the following steps of: sampling conductivity and temperature original data of a measured object through a conductivity sensor probe; an improved sage-husa adaptive Kalman filtering model of the conductivity meter is established; the method comprises the following steps of: transmitting original data measured by a probe of a conductivity sensor into a sage-husa self-adaptive filter, and performing filtering and noise reduction processing on an original signal by using an improved sage-husa self-adaptive Kalman filtering algorithm; the sage-husa adaptive filter can dynamically adjust filter parameters according to the change of the system state and the difference of noise characteristics, a system dynamic model can be well reflected without setting an accurate initial value so as to obtain more accurate estimation data, the robustness of the system is enhanced through outlier judgment introduced during innovation calculation, and the robustness of the system is improved. The measurement precision is effectively improved, the risk that parameters are affected by outliers is reduced, and the real-time performance of the system is enhanced.

The invention discloses a hydropower station signal source side data anti-interference acquisition method and system. The method comprises the following steps: constructing a multi-source interference feature recognition network according to an environment state of a hydropower station, obtaining an interference source distribution feature spectrum, and obtaining an original signal through space-frequency domain joint filtering preprocessing based on the spectrum; performing recursion processing on the original signal by adopting an adaptive Kalman filtering algorithm to obtain a high-fidelity target signal; processing the high-fidelity target signal based on a multi-scale feature extraction and verification mechanism to obtain a credibility quantitative purification signal; and processing the credibility quantitative purification signal based on a compensation strategy to obtain a high-reliability output signal. According to the method, the influence of interference signals is gradually removed from interference source identification, filtering processing, signal reconstruction and quality evaluation to signal compensation, and finally, a result subjected to multiple anti-interference processing is obtained, so that various interferences in a complex environment of a hydropower station can be effectively resisted, and high-reliability data acquisition is realized.

The invention discloses a remaining oil distribution prediction method and system, a control device and a storage medium, and the method comprises the steps: obtaining a multi-scale fusion feature vector according to wide-area apparent resistivity data, seismic attribute data, optical fiber time sequence signal data, nano tracer tracking data and pore network point cloud data, inputting a physical constraint neural network to obtain a residual oil saturation initial value and a pressure gradient initial value of each region of the target layer, wherein a loss function comprises data fitting item loss, Darcy law physical constraint loss and mass conservation constraint loss; the real-time water content and the bottomhole flowing pressure of the production well serve as observation data, and a static parameter correction value, a remaining oil saturation correction value and a pressure gradient correction value are obtained through correction based on an ensemble Kalman filtering algorithm; the static parameter correction value and the pressure gradient correction value are fed back to a physical constraint neural network for loss function updating; and determining a remaining oil distribution result of the target layer. According to the invention, the accuracy and timeliness of residual oil distribution prediction can be improved.

The invention discloses a dynamic calibration control method for the detection precision of a non-standard automatic testing fixture, and particularly relates to the field of industrial automation and control, which comprises the following steps of: constructing a real-time dynamic error field model of the testing fixture; a distributed optical fiber sensing network and a miniature laser interferometer cooperatively collect real-time detection data and spatial-temporal variation errors under a multi-physics coupling effect, a precision degradation track is predicted based on spatial-temporal variation error data, a pre-compensation signal is generated, and a self-adaptive Kalman filtering algorithm is adopted to generate a precision compensation parameter. Environment temperature, vibration and electromagnetic interference data are monitored in real time, environment disturbance data are generated, a hybrid controller combining nonlinear model predictive control and reinforcement learning is adopted, multi-source data are fused, a calibration instruction is generated through multi-target optimization, and an execution mechanism is driven to achieve online dynamic calibration. And digital threads and transfer learning are integrated to realize full-life-cycle data closed-loop management and continuous optimization.

The invention discloses a multi-channel seed counting monitoring system based on data fusion, which is applied to the technical field of automatic seed counting of seed metering devices, and comprises a seed flow dispersion structure, a photoelectric sensor and a signal processing module, the signal processing module is used for processing output signals acquired by the photoelectric sensor from each seed monitoring channel, the method specifically comprises the following steps: determining a seed passing event based on digital signal level change of each seed monitoring channel, and updating a counter in combination with an event triggering interval threshold value to obtain a total seed count; in each time step length, performing state prediction by using an unscented Kalman filtering algorithm, generating Sigma points, propagating the Sigma points by using a system model, calculating a mean value and a covariance of a prediction state, and updating state estimation by combining the propagated Sigma points converted by an observation model; and adjusting the count of the seed monitoring channel and the total seed count based on the updated state estimation, and updating the counter. The seed counting accuracy is improved.

The invention discloses an intelligent human factor lighting method based on multi-sensing information, and the method comprises the following steps: collecting environment data and human physiological data, and carrying out the processing of the environment data and the human physiological data, so as to obtain the preprocessed environment data and human physiological data; an improved Kalman filtering algorithm is adopted to carry out fusion processing on the preprocessed environment data and human body physiological data, the fused data are input into an optimized BP neural network to be trained, an illumination control model is obtained, a corresponding illumination strategy is generated through the illumination control model according to the environment data and the human body physiological data, and illumination control is carried out on the environment data and the human body physiological data. And generating a control signal from the lighting strategy to drive an LED lamp module. According to the invention, the Kalman filtering algorithm is improved to process the data collected by the multi-sensor module, so that the module can adaptively cope with a noise environment with complex changes, the filtering precision is remarkably improved, and accurate data input is provided for next decision-making level data fusion.

The embodiment of the invention provides a temperature drift compensation device, method and equipment for contact impedance of a high-frequency probe card. The system is applied to the technical field of semiconductor testing and comprises a temperature monitoring module, an impedance modeling module, a compensation execution module and a cooperative control module which are connected in sequence. In this way, the technical problems that in the prior art, temperature drift compensation is slow in response and insufficient in precision are solved through the sub-millisecond-level temperature monitoring network, the nonlinear compensation model and multi-probe cooperative compensation in combination with the self-adaptive Kalman filtering algorithm.

The present application relates to a kind of unmanned aerial vehicle impact interception methods based on monocular vision, the azimuth measurement information and angle measurement information of target unmanned aerial vehicle are calculated by target detection algorithm based on monocular vision;Azimuth measurement information and angle measurement information are used, through orthogonal projection matrix and algebraic transformation, non-linear measurement is converted into pseudo-linear form, then pseudo-linear adaptive Kalman filtering algorithm is carried out to estimate the position of target unmanned aerial vehicle, finally, unmanned aerial vehicle maneuvering strategy is used to control impact unmanned aerial vehicle to realize the impact interception of target unmanned aerial vehicle.The method of the present application is obviously improved in system noise robustness, target state estimation accuracy and pursuit efficiency.The present application also relates to impact unmanned aerial vehicle guided by the above impact interception method.

The invention relates to the technical field of medical instrument volume detection, and discloses a medical instrument internal volume detection system. A continuous wave and a pulse modulation signal are simultaneously transmitted through an annular phased array microwave antenna array, three-dimensional dielectric constant distribution is reconstructed by adopting a compressed sensing algorithm, and Doppler time-frequency domain characteristics are extracted through short-time Fourier transform; and inputting the phase gradient field and Doppler velocity field features into a convolutional neural network to generate a four-dimensional space-time volume evolution model, separating a macroscopic volume trend and a microcosmic pulsation component by adopting an adaptive Kalman filtering algorithm, and predicting an abnormal hemodynamic event through a long-short-term memory network. The technical problem of real-time precise detection of the multi-dimensional volume parameters of the blood circulation medical instrument under the complex pulsating flow condition is solved, and accurate evaluation of the instantaneous volume value, the pulsation amount and the pulsating frequency spectrum and early warning of abnormal events are achieved.

The invention discloses a drainage basin water quality frequency conversion sampling method and device oriented to a runoff process, and belongs to the field of hydrological water environment monitoring, and the method comprises the steps: obtaining historical runoff observation data and rainfall data of a to-be-monitored drainage basin and hydrological observation data at the current moment; predicting runoff forecast data through a preset runoff prediction model in combination with the hydrological observation data and the rainfall data; the runoff prediction model is obtained by coupling a long short-term memory network model and a natural resource protection service curve number model; assimilating the hydrological observation data to runoff forecast data through an ensemble Kalman filtering algorithm; determining sampling time sites of a plurality of different runoff stages in a rainfall period corresponding to the runoff forecast data by combining the assimilated runoff forecast data through a segmented sampling method; and when the water level of the to-be-monitored drainage basin reaches a preset trigger water level, sampling the to-be-monitored drainage basin according to the sampling time site. Therefore, the sampling efficiency and the sampling result accuracy in the runoff process can be improved.

The invention discloses an open vocabulary multi-target tracking method based on confidence coefficient adjustment and wavelet convolution, belongs to the technical field of computer vision and target tracking, and can enhance the adaptive modeling capability of a target motion mode by designing a confidence coefficient weighted Kalman updating mechanism and developing a confidence coefficient noise adaptive Kalman filtering algorithm. And meanwhile, the Haar wavelet transform is introduced, so that the receptive field of convolution operation is remarkably expanded, and the capturing capability of high-frequency motion details and the maintaining capability of low-frequency contour information are effectively improved. And in combination with the joint cost matrix of the motion features and the appearance features, a more adaptive target tracking model is formed, and the tracking accuracy and stability are effectively improved. The method can effectively deal with rapid movement and temporary shielding of the target, accurately tracks various types of targets with undefined types, and is suitable for open vocabulary multi-target tracking tasks in various complex environments.

The present application provides a vehicle tracking method and apparatus, a communication unit, and a storage medium. The method includes the following: A vehicle position in a curvilinear coordinate system at a moment corresponding to a target data frame is predicted according to state information of a historical data frame; a beamforming vector is determined according to the vehicle position; a beam is sent based on the beamforming vector, and an echo is received; and state information of the target data frame is determined based on an extended Kalman filtering algorithm and a measurement parameter of the echo, where the state information of the target data frame includes the corrected vehicle position.

The invention discloses a buried gas pipeline multi-leakage parameter identification and concentration distribution rapid prediction method, and belongs to the technical field of buried gas pipeline safety detection. The method comprises the following steps: firstly, based on a porous medium flow control equation and a component transport equation, establishing a leakage diffusion physical model by utilizing computational fluid mechanics and generating a multi-working-condition sample data set; training a neural network model by using the sample data, and constructing a rapid concentration prediction agent model; in combination with measured data of a monitoring area, an ensemble Kalman filtering algorithm is adopted, the agent model serves as a forward prediction operator, and collaborative inversion of parameters such as the soil type, the leakage direction and the flow is achieved through iterative updating; and finally, outputting complete underground gas concentration distribution in the monitoring area based on the leakage parameters obtained through identification. According to the method, rapid identification of multiple leakage parameters and accurate prediction of concentration distribution are realized under the condition of limited measuring points, and the method has remarkable disaster early warning and engineering application values.

The invention relates to the technical field of energy spectrum data processing, in particular to a germanium spectrometer energy spectrum data adaptive optimization method and device, equipment and a storage medium. Performing feature extraction on the historical energy spectrum data set based on a wavelet transform method and a least square method; constructing a baseline initial fitting curve according to the corrected low-frequency coefficient features and the reconstructed high-frequency coefficient features; optimizing the baseline initial fitting curve based on a difference operation method and a preset Kalman filtering algorithm; carrying out median filtering processing on the real-time energy spectrum data set based on a median filtering method and a preset sliding window length; performing iterative optimization on the median filtering data set based on a preset particle swarm optimization algorithm and the baseline optimization fitting curve; according to the scheme, multiple algorithms are fused, baseline features are deeply mined, high and low frequency signals are separated, drift is corrected, spectral peak features are reserved, accuracy is improved, and the method is suitable for complex scenes and assists high-precision application of the germanium spectrometer.

The invention discloses a three-dimensional in-vivo radiotherapy dose monitoring method and system, and the method comprises the steps: collecting an image sequence {In} containing time dimensions at a theta t time interval through CBCT scanning equipment in a radiotherapy process; calculating target area displacement of adjacent frames of images in the image sequence {In} by using a bidirectional optical flow estimation method; state estimation and prediction are carried out on the displacement data based on an adaptive Kalman filtering algorithm, a dynamic displacement rule of the target area in the respiratory cycle is generated, and a respiratory movement prediction result is output; acquiring radiotherapy plan parameters from a radiotherapy plan system, and calculating planned dose distribution without considering respiratory movement; dynamically adjusting a ray transmission path and energy deposition according to a respiratory movement prediction result, and calculating real-time dose distribution of each respiratory time phase; mapping the dose distribution of each time phase to a reference time phase coordinate system to generate accumulated dose distribution; and comparing and verifying the accumulated dose distribution and the planned dose distribution.

The invention discloses a multi-sensor data fusion method for an intelligent transfer device of power equipment, and the method comprises the steps: carrying out the dynamic weight fusion of multi-source data through building a measurement model of a plurality of visual sensors, and generating a high-precision measurement sequence; in combination with an improved adaptive Kalman filtering algorithm, process noise and a measurement noise covariance matrix are estimated in real time, and the stability and the anti-interference capability of state estimation are optimized. And based on the fused high-precision state information, a real-time path planning instruction is generated, and the transfer device is controlled to complete obstacle avoidance and millimeter-level docking of the cabinet body. The problems that traditional manual transfer efficiency is low, the positioning error is large, and a single sensor is prone to electromagnetic interference are solved, the navigation precision and operation safety of heavy equipment in the complex electromagnetic environment are remarkably improved, and the method is suitable for intelligent transfer scenes of equipment such as transformer substation circuit breaker handcarts and the like.

The invention relates to the technical field of computer data processing and artificial intelligence, in particular to a data processing method and device for aircraft atmospheric turbulence early warning and a medium, and the method comprises the steps: carrying out the offline training of a first-scale meteorological risk assessment model and a second-scale turbulence disturbance model; in the flight process of the aircraft, time sequence flight state data and corresponding first-scale meteorological data are obtained in real time, and first-scale meteorological features and second-scale turbulence features are extracted; based on an ensemble Kalman filtering algorithm, taking the output of the first-scale meteorological risk assessment model as a background trend term, generating a nonlinear correction term from real-time data based on a key indication factor determined by the second-scale turbulence disturbance model, and fusing the background trend term and the nonlinear correction term to generate a forecast result of atmospheric turbulence intensity; and generating a turbulence early warning signal according to the forecast result. According to the invention, through multi-scale information fusion, accurate and early warning of atmospheric turbulence is realized, and the flight safety is effectively improved.

The invention relates to the technical field of agricultural non-point source pollution load evaluation, and provides an agricultural non-point source pollution real-time simulation method, device and equipment for an event rainfall. The method comprises the following steps: acquiring preset basic data of a target area, corresponding forecast rainfall of each hydrological response unit under a time scale, and initial conditions of each hydrological response unit before a rainfall event occurs; according to the forecast rainfall capacity, preset basic data and initial conditions of the model, prediction is carried out in combination with an agricultural non-point source pollution model of the rainfall event, and the flow and nitrogen and phosphorus load simulation values at the hydrological monitoring node at the current time point are obtained; and real-time correction simulation of the session event process is carried out based on an ensemble Kalman filtering algorithm and real-time monitored and transmitted flow and nitrogen and phosphorus load observation values. According to the method, the precision of the initial condition of the agricultural non-point source pollution model of the session rainfall event can be effectively improved, and real-time simulation of the process of the agricultural non-point source pollution model of the session rainfall event is realized.

A set Kalman filtering turbulence model assimilation method based on full field similarity criterion is disclosed, which comprises the following steps: sensitivity analysis is performed on the coefficients of the candidate turbulence model to obtain sensitive coefficients and their sampling range for Latin hypercube sampling, i.e., sampling interval; model constant sample combinations are extracted from the sampling interval by using a directional Latin hypercube sampling method, and target flow field results are calculated; a data assimilation algorithm based on the set Kalman filtering algorithm is used for the target flow field results and the model constant sample combinations, and after obtaining the optimized model constant combination, the model constant combination is updated and verified in cycles to obtain the optimal model parameter combination; finally, the improved turbulence model is used to simulate the internal flow field of the fuel assembly in the core of the pressurized water reactor with high similarity. The present application can provide an objective evaluation standard for the full physical field similarity of the data assimilation method, significantly improve the convergence speed of the algorithm, strengthen the robustness of the algorithm, and reduce the calculation cost of the algorithm.

The invention relates to a gamma radiation field reconstruction method based on Lagrange particle model parameter assimilation, and the method comprises the steps: determining the spatial positions of a plurality of radioactive aerosol particles based on a Lagrange particle model; constructing state parameters based on the plurality of spatial positions and the source intensity of each aerosol particle; forming a background state vector by using the state parameters, and dynamically updating the background state vector by adopting an ensemble Kalman filtering algorithm to obtain a corrected assimilation state; determining an optimized transportation track based on the spatial position of the radioactive aerosol particles in the corrected assimilation state; constructing a local concentration field corresponding to the radioactive aerosol particles by adopting a Gaussian kernel function; and determining a gamma radiation field corresponding to the radioactive aerosol particles. The three-dimensional radiation field reconstruction is carried out based on the background state vector and the real-time monitoring data by taking the parameters of the radioactive aerosol particles as the core, and the purposes of correcting the background state variable, obtaining the corrected assimilation variable and improving the accuracy of the three-dimensional radiation field reconstruction are achieved.

The invention provides an aircraft height multi-source observation fusion method and device. The method comprises the following steps: setting a simulation area; generating a meteorological initial field set and a meteorological boundary field set according to meteorological data in the simulation area; deriving a corresponding meteorological background field set based on the meteorological initial field set and the meteorological boundary field set; obtaining meteorological data of multi-source observation, wherein the meteorological data of multi-source observation comprises meteorological data of open source observation, meteorological data of field air pressure observation and meteorological data of aircraft GPS height observation; converting the meteorological background field set to an observation space through an observation operator model; and fusing meteorological data of multi-source observation through an ensemble Kalman filtering algorithm so as to correct the meteorological background field set in the observation space.

The invention discloses a data and model fusion driving-based space stack operation parameter correction and inversion method, which comprises the following steps of: 1, analyzing the influence degree of empirical constants in a space stack simulation model on an operation parameter prediction result, and determining empirical constants needing to be corrected and operation parameters; 2, predicting N groups of operation conditions containing empirical constant error disturbance by using the simulation model, obtaining N groups of simulation model estimation vectors, and forming an initial prediction matrix; 3, summarizing on-orbit actual operation data or ground test actual measurement data acquired by a space stack sensor, and providing an actual measurement data matrix for correcting space stack operation parameters; 4, fusing the characteristics of the prediction data and the actual measurement data of the simulation model by using an ensemble Kalman filtering algorithm to obtain a corrected empirical constant; 5, updating the empirical constants in the space stack simulation model into the corrected empirical constants to realize the correction of the simulation model; and 6, using the corrected simulation model to carry out operation condition analysis, and realizing operation parameter inversion calculation. The method is high in prediction efficiency and accurate in result.

The invention discloses a signal processing method based on a laser methane sensor. The method comprises the following steps: step 1, providing stable current for a laser through a constant current source circuit; step 2, acquiring the environment temperature in real time when the laser generates laser by using a temperature acquisition circuit; 3, processing the data of the temperature acquisition circuit by using a temperature control algorithm, and controlling the environment temperature to be the temperature most suitable for laser generation of the laser according to the preset optimal temperature parameter; step 4, receiving laser which is generated by the laser and passes through the methane gas by using a photodiode; and step 5, the amplified signal is collected through the single-chip microcomputer. In order to ensure that a laser of the sensor can stably emit laser, a PID algorithm is adopted to control and adjust the temperature of a light-emitting environment; and processing the acquired output signal of the amplifying circuit by using an improved Kalman filtering algorithm, removing noise and obtaining effective signal information.

The application discloses a kind of based on multimodal data fusion pumped storage unit fault diagnosis method, the voiceprint signal of pumped storage unit, infrared thermal image and historical operation data are collected, improved unscented Kalman filtering algorithm is used to denoising processing voiceprint set, the voiceprint signal after denoising and historical operation data are trained using COMRes+Model, future voiceprint signal is predicted, improved Deeplabv3+Model is used to train image set, and the features of infrared thermal image are extracted;Voiceprint feature, historical operation data text feature and image feature are input into improved CentralNet model for multimodal data fusion, and the fault category is output through classification identification module, so as to complete the diagnosis of pumped storage unit fault;The method can effectively improve the accuracy and efficiency of fault diagnosis by the fusion of multimodal data and the optimization of deep learning model, and provide strong guarantee for the safe operation of pumped storage unit.