294results about How to "Guaranteed prediction accuracy" patented technology

The invention discloses an interior noise analysis and prediction method of a high speed train. The method comprises the following steps: establishing a train reconditioning train body model, a statistical energy analysis model of a body in white structure and an interior and exterior vocal cavity statistical energy analysis model, and carrying out simplification and subsystem division; obtaining the statistical energy analysis parameters of the train body structure and an interior vocal cavity model, and loading the statistical energy analysis parameters onto a train body structure model plate subsystem and a vocal cavity model subsystem; and obtaining exterior vocal excitation source energy borne on the train body, applying the exterior vocal excitation source energy onto the exterior vocal cavity statistical energy analysis model, causing the exterior vocal excitation source energy to reach an interior vocal cavity after the exterior vocal excitation source energy is attenuated by the sound insulation property of a structural plate in a body in white structure model so as to obtain structural noise energy radiated into the train by the reconditioning train body under the function of the two-line suspension force of a compartment, and then, carrying out interior noise analysis and prediction. The problem that the interior noise of the train is difficult in prediction and the problems of the upper limit boundedness of a frequency domain, complex calculation flow, incomplete motivation consideration in the traditional method are overcome, calculation efficiency and prediction accuracy are improved, and development and test cost is lowered.

The invention discloses a convolutional neural network-based photovoltaic glass defect classification method and a convolutional neural network-based photovoltaic glass defect classification device. The method comprises the following steps of carrying out multi-angle and variable-illumination image acquisition on a plurality of defect samples to obtain a plurality of images; preprocessing the images to fuse the multi-channel information and generate a multi-channel-fused defect sample image; adopting a convolution neural network model which meets a preset condition, extracting a network according to defect category design features and constructing a feature extraction convolutional neural network; obtaining the number of layers of all-connected neural networks and the number of neurons ofeach layer, and constructing a feature classification network; under the condition that the cross entropy loss function is minimized, completing the training of the convolutional neural network; according to an input sample image, outputting a prediction result of a defect category through the trained convolutional neural network. Based on the method, the situation that training sets are insufficient can be effectively solved while the generalization ability and the prediction precision of the model are guaranteed. Meanwhile, the high classification precision can be achieved for a small amountof glass defect samples.

The invention provides a posting predication system based on a neural network technique. The system adopts phase-space reconstruction for realizing non-linear time sequence analysis and adopts a Browser / Server structure based on a J2EE platform. The system structurally includes a data pre-processing module, a predication analysis management module and predication modeling and simulation interface software and has characteristics of high convergence rate and low training error. The predication precision of established models is good. However, selection of a range of neural network training samples should be paid attention. Sample quantity can be reduced appropriately for predication of posting concerning to emergencies and samples of too early time may not be applied. In application, damage of models due to abnormal values should be paid attention and the abnormal values should be checked and adjusted, so that the model predication precision can be guaranteed. The established models adopt quantitative analysis and acquire a certain precision. Besides, the perception performance is good.

The invention discloses a short-term photovoltaic power prediction method based on VMD-IPSO-GRU, and belongs to the technical field of photovoltaic power generation and grid connection. Firstly, a historical photovoltaic power time sequence is decomposed into sub-sequences with different frequencies through variational mode decomposition, geographic information and component parameters contained in photovoltaic sequence data are fully mined, and signals and noise of original data are separated; secondly, main meteorological factors influencing photovoltaic output are determined through Spearman and Pearson correlation coefficients; and finally, gating cycle unit network models are established for the sub-sequences decomposed by the VMD respectively, and the GRU nerve is optimized through an improved particle swarm algorithm and an adaptive moment estimation algorithm, thereby improving the network convergence rate and the data fitting effect, accurately and efficiently finishing short-term photovoltaic power prediction, and avoiding errors caused by manual parameter adjustment.

The invention discloses a driving energy consumption prediction system and method, a storage medium and equipment. The prediction method comprises the steps of obtaining historical working condition data of a planned driving route; constructing a training sample data set based on the historical working condition data; performing data training on the training sample data set, and establishing a vehicle speed feature BP neural network model and a driving energy consumption BP neural network model; acquiring real-time working condition information on the planned driving route; and inputting the real-time working condition information into a vehicle speed characteristic BP neural network model for prediction to obtain vehicle speed characteristic data of future driving, and then inputting thevehicle speed characteristic data into a driving energy consumption BP neural network model for prediction to obtain future driving energy consumption data so as to realize online prediction of driving energy consumption. According to the invention, online effective prediction of the driving energy consumption under the driving working conditions of different road environments and traffic states can be realized, and the efficiency of intelligent energy management of the vehicle is improved.

The invention belonging to process and application field of medical images relates to an X-ray perspective image demarcating method of an operation navigation system. The method comprises obtaining X-ray perspective images containing marking point information in the operation, performing filtration and subtraction to obtain the images only containing marking point information, obtaining the center coordinate and the arrangement direction of the marking points by utilizing methods of template matching, cluster analysis and statistics of information entropy, correcting the marking point coordinate by using a B belt transect and calculating the demarcating parameters. A projection coordinate of the marking point is obtained according to the obtained imaging model parameters, all pixels of the image is corrected according to the distance between the projection coordinate and the real coordinate on the image to obtained a corrected X-ray perspective image for the operation navigation system. The method has characteristics of simple embodiment, reliable algorithm, convenience in clinical application and improvement of precision of the operation navigation system.

The invention relates to an artificial neural network based method for controlling online prediction of casting billet quality. The method comprises the steps of firstly, choosing prediction model variables and establishing a neural network model; selecting a training sample to study and train an established three-layer back propagation (BP) neural network after the structure of the network model and a target error are determined, and saving a weight and a threshold value after a network output layer error meets requirements; obtaining the trained and verified network model; and using the trained and verified BP neural network model to control the online prediction of intermediate cracks of continuous casting sheet billets. The method uses the classification ability of the neural network, evaluates prediction results of the network model through comparison of prediction values and measured values and guarantees the prediction precision of the model. By the aid of the method, the online automatic prediction of the casting billet quality during continuous casting production can be achieved, and the method is simple to operate, capable of predicting classifications of casting billet quality defects and defect grades and applicable to guiding of onsite production.

The invention discloses a lake and reservoir algal bloom predicating method based on multielement nonstationary time series analysis and neural network and support vector machine compensation, and belongs to the technical field of water quality monitoring. The method comprises the steps of characteristic factor nonstationary time series modeling, error influence factor kernel principal component analysis, neural network error modeling according to the situation of large sample data, support vector machine error modeling according to the situation of small sample data, final error compensation and predicating result obtaining. The problems that existing algal bloom predication precision is not high, and predication is hard to carry out according to the small sample data are solved, the description of the algal bloom forming process corresponds to reality better, and the result of algal bloom modeling predication is more accurate. The advantage compensation of a time series analysis method suitable for linear system modeling and a statistical learning method suitable for nonlinear system modeling is achieved, and the algal bloom predication accuracy is improved.

The invention provides a method for predicting discharge capacity of a lithium ion battery through partial discharge process by a BP neural network. The terminal voltage of the lithium ion battery in the at least previous 10 minute constant-current discharge process is taken as input, and a BP neural network model outputs the discharge capacity of the battery. The method solves the technical problems of long test period and high energy consumption in the conventional industrial method, also overcomes the defect that a laboratory method has complicated steps and is not suitable for massive industrial production, and guarantees that the average forecasting error is about 2.0 percent which is less than the error range of about 5percent allowable in the industrial production.

The invention discloses a real-time limit learning machine short-time traffic flow prediction method based on fusion. The method, for prediction technology in short-time unstable traffic flow scene, predicts short-time traffic flow on the basis of the realtimeness, the accuracy, and the reliability of short-time traffic flow and a fused real-time limit learning machine. The short-time traffic flow prediction method is based on a simplified single-implicit-strata feedforward neural network structure, may fast train historical data at a traffic flow peak value and updates reached data in an increment way, and saves learning time while guaranteeing certain prediction precision. Further, the method guarantees the stability and the robustness of short-time traffic flow prediction by using a fusing mechanism, performs reconstruction during a data missing and violent fluctuation period, and is short in training time. The root mean square error and the standard error percentage of a prediction result are both in a confidence region.

The invention relates to a system and method for production process self-adaption monitoring using an OCSVM. The system and method for production process self-adaption monitoring using the OCSVM is characterized in that the monitoring system comprises a data collecting module, a process monitoring module, an OCSVM model online update module and an alarming module; historical process variables and real-time process variables in an industrial production process are digitized through the data collecting module to historical process data and real-time process data and the historical process data and the real-time process data are transmitted to the process monitoring module; an original OCSVM monitoring model is established by the process monitoring module through the historical process data, the real-time process data are processed, so that an effective update sample is obtained, and the effective update sample and the OCSVM monitoring model are transmitted to the OCSVM model online update module together; after the OCSVM monitoring model is updated through the OCSVM model online update module, the OCSVM monitoring model is transmitted to the process monitoring module; when an abnormal sample is obtained by the process monitoring module, an alarm signal is generated and transmitted to the alarming module, so that an alarm is raised. The system and method for production process self-adaption monitoring using the OCSVM can be widely used for monitoring in the actual industrial production process.

The invention discloses a metro stray current leakage level prediction method based on the convolutional neural network and the BP neural network. The metro stray current leakage level prediction method includes the following steps that 1, the earth resistivity and polarization potential measuring position is determined in a section, and prediction model input data is collected; marticulated processing is conducted on the data of the three affecting factors including the earth resistivity, bury pipeline depth and bury pipeline polarization potential which affect stray current leakage; a convolutional neural network prediction model is used for predicting the single point kinetic stray current leakage level; the BP neural network is used for comprehensively predicting the stray current leakage level in the section; the prediction model based on the convolutional neural network model and the BP neural network is used for predicting test data. The stray current leakage level in the section can be effectively predict through easy-to-detect data, and the prediction precision of the system during long-time monitoring is ensured, which is of important actual significance in visually monitoring the stray current leakage condition.

The invention relates to a method for predicating composite material Pi-shaped non-planar glue joint strength based on triangular envelopes. The method includes the following steps that (1), according to parameters of a composite material Pi-shaped non-planar glue joint structure, a geometrical model is established; (2), according to actual working conditions of the composite material structure, loads and boundary conditions of the Pi joint geometrical model are determined; (3), grid partition is performed on the joint geometrical model, and a Pi joint three-dimensional finite element model is obtained; (4), on the basis of the Pi joint three-dimensional finite element model, a finite element stress analysis is performed; (5), according to the linear finite element stress analysis result, a curved edge triangular envelope route of a Pi negative moment steel padding region is set on the basis of the Pi joint three-dimensional finite element model, all stress component values on the curved edge triangular envelope route are extracted, and the average value of the stress component values is calculated and substituted to the failure criterion for predicating the strength. The method for predicating the composite material Pi-shaped non-planar glue joint strength based on the triangular envelopes is suitable for engineering application and can obviously shorten a development cycle and reduce experimentation cost.

The invention relates to the denitration field of coal-fired boilers in a heat-engine plant, in particular to an ammonia spraying control system of a boiler denitration system. The control system is mainly used for improving control precision of NOX concentration at a denitration exit. The control system comprises a multimode prediction module, a control part and an execution mechanism. The multimode prediction module is used for predicting the NOX concentration at the denitration exit. The control part is used for calculating to-be-compensated ammonia spraying quantity according to the predicted NOX concentration at the denitration exit. The execution mechanism is used for allowing a denitration system to acquire the to-be-compensated ammonia spraying quantity according to the instruction of the control part. According to the invention, through feedforward, override and feedback of the ammonia spraying quantity, the discharged pollution processing level of the boiler denitration system is reduced; by establishing the multimode prediction module of the NOX concentration at the denitration exit, prediction precision of the NOX concentration at the denitration exit can be ensured when a combustion system of the boiler is in different load states; and by adopting an ammonia spraying control mechanism combining the feedforward, the override and the feedback, precise control of the ammonia spraying quantity is achieved.

The embodiments of the invention disclose a power sale quantity prediction method and device based on X13 seasonal adjustment and factor regression. The method comprises the steps of preprocessing historic power sale quantity data, and decomposing the preprocessed power sale quantity sequence into a trend item, a season item and a random item by using an X13 seasonal adjustment algorithm; respectively performing prediction by adopting multiple prediction algorithms according to the influence factors and curve characteristics of sub sequences to ensure the prediction precision and robustness of the trend item; summing the prediction results of the sub sequences to reconstruct power sale quantity prediction results, and finally selecting a prediction result having the optimal performance from the multiple prediction results. Meanwhile, the embodiments of the invention further sufficiently consider the influence of some influencing factors on each decomposition item, so the precision of the prediction result obtained by the solution of the embodiment is higher.

The invention, which is applicable to the transportation field, provides a short-term traffic flow change prediction method and device, a computer equipment and a storage medium. The method comprisesthe following steps: acquiring traffic flow data of a predicted road segment in real time; on the basis of the traffic flow data, constructing a traffic flow prediction model with a forgetting factor;and according to a particle filtering algorithm, eliminating random noises of the traffic flow prediction model and acquiring and outputting an optimal short-term traffic flow change prediction value. According to the short-term traffic flow change prediction method provided by the embodiment of the invention, the traffic flow prediction model is constructed by introducing the forgetting factor;and with the model, the traffic flow data can be updated and corrected in real time and thus the traffic flow change is fitted well, so that a phenomenon that the prediction precision is affected by the early collected traffic flow data because of the time-varying of the traffic flow is avoided and thus the good prediction effect is realized. Meanwhile, the random noise of the model is eliminatedby particle filtering, so that the accuracy and feasibility of the prediction are improved.

The invention relates to a method for transferring a near infrared model of an organic fertilizer product. The method is characterized by comprising the following steps of: (1) selecting and setting main instruments and slave instruments and matching spectroscopic data between the two kinds of instruments; (2) matching spectrum arrays of the front main instrument and the rear main instrument and a sample chemical truth-value concentration array by using the spectroscopic data, establishing a leave-one-out full-interactive verification near infrared model based on a partial least squares method and comparing the predication result difference; (3) selecting a plurality of respective samples in correction set samples to form a standard sample set; (4) respectively calculating predication value concentration arrays of the standard sample sets of the main instruments and the slave instruments by using the near infrared correction module established by using the main instruments and carrying out least square fitting; (5) calculating a slave instrument verification set sample concentration array by using main instrument near infrared correction models and carrying out the correction by using a least square fitting relationship formula to obtain a slave instrument verification set sample concentration array after the model is transferred; and (6) estimating and analyzing a predication effect after the main instrument near infrared correction models are transferred.

The invention discloses a clustering and trend index-based power distribution network line load prediction method and device. The method comprises the steps of obtaining and cleaning load historical time sequence data of each transformer area in a power distribution network; dividing all courts into a plurality of clustering clusters through clustering according to the load historical data set, summing and reconstructing the load historical data set of each clustering cluster, and obtaining a plurality of load samples day by day; acquiring holiday information corresponding to each load sample,and calculating a load change trend index in the same period of the last year; training a corresponding long-term and short-term memory neural network load prediction model by using the load sample,holiday information and load change trend index of each clustering cluster; and predicting corresponding load data by using each type of trained load prediction model, and finally superposing prediction results of each type of load to obtain a prediction result of the total load of the power distribution network line. The method and device can improve the short-term load prediction precision of the power distribution network, so as to achieve the purpose of guiding the dispatching operation of the power distribution network.

A method for correcting the deformation of cerebral tissue in the navigation system of neurosurgical operation includes such steps as obtaining target tissue (cerebral tissue) by 3D automatic division algorithm based on MRI, generating the lattice of cerebral tissue, assigning the relative biomechanical attribute to each lattice unit, creating physical mode, tracking the movement of exposed cerebral cortex layer, finit element calculating to obtain the deformation for cerebral tissue, and updating the 3D data field before operation by an algorithm to direct the operation.

The invention relates to a deep learning cellular automaton model-based soil moisture content prediction method. According to the method, a machine learning and geographical phenomenon simulation are used in combination; different time-space prediction aspects of soil moisture content are improved; a soil moisture content prediction function local optimal solution can be obtained by means of deep learning; and a quantitative test is performed on the generalization ability of the model through using a model inspection mechanism, and a self-improvement mechanism of a cellular automaton is put forward, and therefore, the robustness of the model can be ensured better. The hybrid technology provided by the invention is expected to provide technical support for soil moisture content real-time monitoring in complex regions. With the prediction method adopted, prediction cost of the soil moisture content can be reduced, and prediction accuracy of the soil moisture content can be significantly improved. The prediction method has a wide industrial application prospect.

The invention relates to a traffic accident rate prediction system based on on-line variational Bayesian support vector regression, which comprises a data preprocessing module, an online variational Bayesian support vector regression model building module, an online variational bayesian support vector regression model training module, and an online variational bayesian support vector regression model prediction module. This method effectively solves the problem that the traditional support vector regression model predicts the speed of traffic accident rate is slow, the prediction result is inaccurate, and it is difficult to solve the problem on line and show its practical value.

The invention relates to a maximum power tracking control method for a small permanent-magnet direct-drive wind power generation system. According to the method, on the basis of collecting a lot of actual samples of wind speed vector-rotating speed-power, a firefly support vector machine regression prediction model of the wind speed is established, and wind speed estimation is performed by utilizing the model; next, the optimal rotating speed, corresponding to the maximum power point, of a draught fan is predicted through the optimum tip speed ratio method; afterwards, the rotating speed of the draught fan is adjusted to the predicted optimum rotating speed of the draught fan, and the maximum power of the draught fan is tracked based on the perturbation and observation method at a set perturbation step length with the rotating speed as an initial value. According to the maximum power tracking control method, wind speed estimation without sensors is achieved, the control cost of a power generation system is greatly reduced, the speed and accuracy for looking for the maximum power point are enhanced, and power loss during perturbation is lowered.

The invention provides a daily average power load prediction method based on a BP neural network. The method comprises the following steps of obtaining numerical values corresponding to a plurality ofmain meteorological factors of a prediction day and a prediction day date; obtaining historical day data used for load prediction model training according to the main meteorological factors and datesof the prediction days; inputting the historical daily data for training into a BP neural network, and optimizing the BP neural network to obtain an optimized BP neural network model; and inputting the main meteorological factors of the prediction day and the prediction day date into the optimized BP neural network model, and calculating to obtain the power load of the prediction day. According to the method, the similar daily algorithm is used for obtaining training data, training of the network is accelerated on the premise that the prediction precision is guaranteed, meanwhile, the weightof the BP neural network is optimized through the genetic algorithm, the problems that the BP neural network is trapped in a local minimum value in random initialization, convergence is difficult andthe like are solved, and the prediction precision of the model is improved.

The invention discloses an anti-collision pre-warning system and method for hoisting equipment. The anti-collision pre-warning system is based on a distributed type modular framework, and comprises aninformation collection system, an information transmission system and a field control system which are arranged in a distributed modularity mode, and all equipment and fixed obstacles are placed in the same coordinate system. The anti-collision pre-warning method includes the steps that the real-time position, the moving direction and speed information of moving parts of all the equipment are collected and preliminarily calculated by all the hoisting equipment in the field, then relevant information exchange is carried out, the probability of collision is calculated and pre-warning is carriedout, and the steps are cycled and repeated every 0.1 second. According to the anti-collision pre-warning system and method for the hoisting equipment, the distributed type anti-collision system framework is built, anti-collision calculation rules are made, the problem of collision among the hoisting equipment is decomposed into the problem of collision between the moving parts, a binary tree traversal recursive algorithm is adopted to ensure the prediction accuracy of the collision phenomenon, then the collision phenomenon is subjected to commonness processing, a collision pre-warning algorithm is compiled, and collision pre-warning is realized according to the anti-collision rules and algorithm compilation.

The invention discloses an online vehicle-mounted battery SOC (state of charge) prediction method based on big data and an extreme learning machine. The external characteristic parameters of the battery such as voltage, current, temperature and internal resistance are selected, the large amount of online acquired external characteristic parameters of the battery are integrated through a big data method, a big data system for SOC prediction is formed, the data thus can be effectively excavated later, and the prediction precision is ensured; and through the extreme learning machine method, effective data most closely related to SOC prediction are found out, and the SOC is further accurately predicted according to the excavated effective data. The method has the advantages of high predictionprecision and strong practicability and the like.

The invention provides a coupling large-scale data flow width learning rapid prediction intelligent algorithm based on network community detection and GCN. The algorithm comprises the following steps:step 1, community detection; step 2, space-time feature extraction; step 3, width learning rapid prediction; and step 4, large-scale real-time prediction of the space-time coupling width learning neural network. Compared with the prior art, the algorithm has the beneficial effects that intelligent community detection and GCN feature extraction are adopted, width learning is combined, the problemof large-scale node prediction is solved, and the algorithm has the advantages of being high in calculation speed, high in prediction precision, high in adaptive capacity and the like.

The invention discloses a maximum power tracking device for a mini permanent magnetic direct drive wind power generation system and a control method in the technical field of wind power generation, wherein the maximum power tracking device comprises a fan, an MPPT (maximum power point tracking) controller, a rectifier, n air velocity transducers, a revolution speed transducer, a voltage transducer, a current transducer, a DC-DC (direct current to direct current) converter, a driver module, a first capacitor, a second capacitor and a load; the control method comprises the following steps of acquiring air velocity vectors by utilizing the air velocity transducers arranged at different positions, collecting a large number of air velocity vector-optimal revolution speed practical samples, establishing an air velocity-optimal revolution speed prediction model by utilizing a support vector machine, and carrying out the maximum power tracking by combining the prediction model with a small-step perturbation and observation method. The maximum power tracking device for the mini permanent magnetic direct drive wind power generation system and the control method, disclosed by the invention, have the advantages that the tracking speed is increased, and the power loss of a perturbation process is effectively reduced; after the characteristics of a fan changes, the prediction precision is ensured by repeatedly collecting samples and training a new prediction model.

The invention discloses a predication model establishing method for a driving cycle of a plug-in hybrid electric vehicle and a plug-in hybrid electric vehicle energy management method based on the prediction model establishing method, wherein the methods realize online application based on a model prediction control energy management strategy. Based on a basic principle of model prediction, through changing the time scale of the future driving condition of the predicated vehicle, controlling for the future vehicle speed predication precision is realized. Furthermore a predicated time domain transforming principle and a dynamic planning algorithm are introduced into a model predicating control frame, thereby forming a variable-time-domain model prediction energy management method for aimingat the plug-in hybrid electric vehicle. Particularly the prediction model establishing method comprises the steps of predicating a settling solution of condition loss problems which may occur in theactual driving process of the vehicle, predicating a process precision defining formula in real time, and forming a prediction energy management method based on a variable time domain model through apredicated time domain transforming principle which is composed of main component analysis, clustering analysis, relativity analysis and the like and introducing the principle and a dynamic planning algorithm into the model prediction control frame.

The invention relates to a soil organic carbon storage amount estimation method based on soil genetic horizon thickness prediction. Better technological ideas are provided for continuous prediction of soil mass distributed in the horizontal dimension in consideration of soil attributes through unstructured information encapsulation of the genetic horizon; with the adoption of the technologies of genetic horizon merging, prediction and recalculation, the limitation of a conventional prediction method in continuous description of soil mass is corrected while character information of the soil genetic horizon is not missed, a universal soil organic carbon storage amount estimation technology for standard description and accurate prediction is realized, and the method has wide industrial application prospect in engineering survey of agricultural application, environmental protection, territorial resources and other relevant departments.

The invention discloses a deep convolutional neural network model adaptive quantization method, and designs a deep convolutional deep network low-bit quantization algorithm suitable for FPGA calculation, which mainly comprises preprocessing of network model parameters and a grouping adaptive quantization method of a parameter set. the method includes: acquiring dynamic thresholds to perform coarse-grained cutting on the original parameters of the model; constructing an initial clustering center point set suitable for FPGA shift calculation; grouping and clustering the preprocessed model parameters based on a mode length minimization method; finally, overlaying the clustering center point set with the non-null parameter class, achieving self-adaptive low-bit quantization of different networks through optimization; the quantization algorithm is moderate in complexity and quite conforms to the calculation characteristics of the FPGA, hardware resource consumption on the FPGA is reduced, and the model reasoning speed is increased while the model reasoning precision is guaranteed.