39results about How to "High precision prediction" patented technology

An environment detection unit (40) detects the positions, motion states, and movement states of moving objects, and also detects multiple types of road-division regions and stationary-object regions. A mapping generation unit (42) generates an existence-possibility mapping in which existence possibilities are assigned to the detected road-division regions and stationary-object regions. On the basis of the detected positions, motion states, and movement states of the abovementioned moving objects, a moving-object generation unit (44) generates a moving-object position distribution and movement-state distribution and records said distributions in the existence-possibility mapping. On the basis of said moving-object movement-state distribution, a position update unit (46) moves the moving-object position distribution. A distribution change unit (48) changes the moved position distribution on the basis of the existence possibilities in the existence-possibility mapping, and a future position distribution for the moving objects in the existence-possibility mapping is predicted. This makes it possible to predict future positions of moving objects in a variety of situations with high precision.

The present invention relates to an image processing apparatus and method that can generate a high-accuracy prediction image with a small amount of control information.A motion compensation circuit 51 specifies a macroblock corresponding to a prediction image in a reference frame other than a current frame using a motion vector supplied from a prediction mode determination circuit 41. The motion compensation circuit 51 reads an image of the specified macroblock from a frame memory 19, and extracts the read image as a motion compensation image. An intra-prediction circuit 52 performs intra-prediction on the current frame using an arbitrary method to generate an intra-prediction image IP. The present invention can be applied to, for example, an encoding apparatus and a decoding apparatus.

The invention relates to a position control method for a sensorless steering wheel. According to the position control method, continuous, smooth and stepless change of a steering ratio at different car speeds is realized by a regulation motor for controlling an ECU (Electronic Control Unit) gear ratio; meanwhile, the change rule is calculated by utilizing a correlation formula, and rotation angle compensation is carried out according to the calculated change rule; reverse steering adjustment is actively carried out during car oversteer, and the regulation motor and an assisting motor are flexibly controlled and mutually compensated so as to perfectly combine the flexibility, the portability and the control stability of the steering, so that the control performance and the safety are improved.

To provide a river water level predicting apparatus which predicts the water level of a river with accuracy. The river water level predicting apparatus 1 comprises a water level measuring section 5 for measuring the water level of the river 4 and storing therein a measured value, rainfall measuring sections 7, 8 for measuring rainfall quantities of a river 2 and the river 4 and storing therein the quantities, and a water level predicting model 20 for predicting the water level of the river, based on the measured values. The water level predicting model 20 has an autoregressive portion and an FIR model portion, and the model 20 is identified by a model identifying section 21. The model identifying section separately calculates a parameter for the autoregressive section and a parameter for the FIR model section, based on the measured values.

The invention provides a method for predicting the tail gas pollution distribution in a city road network. The method comprises the following steps: acquiring multi-source heterogeneous data; carryingout stack-type self-encode features dimension reduction, and constructing a multi-layer sparse self-encoder network structure to extract the features of the multi-source heterogeneous data; generating sequential data based on spatio-temporal semi-supervised learning; pre-training a deep spatio-temporal network model replacing the corrected model data with the telemetry data of the real monitoringpoints, and re-training the corrected regional tail gas emission prediction model; determining the weighted parameters of the model to obtain a deep spatio-temporal network model, and inputting the multi-source heterogeneous data t to obtain a predicted regional tail gas pollution emission result. The invention is based on a stack-type self-encoder dimension reduction feature extraction method, which can learn essential feature mapping between road network information, meteorological information, traffic flow information, POIs information and regional tail gas emission, and can realize higherprecision regional tail gas prediction on real telemetry data.

The present invention provides a railroad train operation management system and program for automatically giving various types of operation guidance such as the guidance on the stop station, the variation from the scheduled time, warning of the train speed limit to the motorman at an adequate timing at each check point arbitrary set on a predetermined line to support the motorman and realizing operation management with high precision by predicting the position of the train with high precision at low cost even in a region where the GPS signal is hard to receive. The system comprises line data storage means (2) for storing line specific position data and check point position data, operation schedule data storage means (3) for storing operation schedule data, GPS signal receiving means (6) for receiving the GPS signal, operation management means (9) for performing operation management of trains according to the GPS data and operation schedule data on the trains computed from the GPS signal, and operation guidance output means (7) for giving the motorman the information acquired by the operation management means.

The invention belongs to the technical field of thermal error predicating methods, and provides a thermal error predicating method for a rolling ball screw feeding system of a numerical-control machine tool. Thermal errors of the rolling ball screw feeding system are predicated through a self-adaptive real-time model (ARTM). The thermal error predicating method comprises the following steps of: firstly, establishing experiments for measuring temperature changes, along with time, of the surface of a workbench; secondly, establishing the self-adaptive real-time model for predicating transient temperature distribution and thermal error distribution of rolling ball screws; thirdly, determining a heating rate of two bearings, a movable nut and a slide rail of the rolling ball screw feeding system through a finite element combined monte carlo method (MC); fourthly, establishing an exponential equation of feeding speed and time according to finite element calculating data for describing and measuring temperature difference changes, along with time, between surface points and kinematic pair centers; and finally, establishing a numerical value predicating algorithm for predicating thermal errors of the rolling ball screw feeding system. Corresponding thermal errors are quickly predicated with high precision by the numerical value predicating algorithm in a mode of monitoring surface temperatures of two bearing bases and a motion nut side surface.

An article residual value predicting device of the invention comprises an article residual value predicting computer, a first data memory device connected to the article residual value predicting computer to store, as basal record data, respective items such as article names, used article values for each article type, new article values for each article type, and year and month data to which the used article value is applied, a second data memory device connected to the article residual value predicting computer to store item category scores. The article residual value predicting computer comprises article residual rate proven-value calculating means for reading out the used article value and new article value for each article type stored in the first data memory device, calculating article residual rate proven-value from the ratio of the used article value to the new article value, and storing a calculated result thus obtained as an article residual rate proven-value in the first data memory device, category score calculating means for reading out the article name, article residual rate proven-value, year data to which the used article value is applied and month data to which the used article value is applied, which are stored in the first data memory device, and calculating an item category score by performing a regression analysis based on the qualification theory I using the readout article residual rate proven-value as an objective variable and the readout article name, the year to which the used article value is applied as an explanatory variable and the month to which the used article value is applied as an explanatory variable, and storing a calculated score thus obtained in the second data memory device, article residual rate predictive-value calculating means for reading out the score stored in the second data memory device with respect to a specified item category and adopting a year-classified score relative to the year at some future point to be predicted as the year-classified score to calculate an article residual rate predictive-value from an equation “(article residual rate predictive-value)=(item-classified score)+(year-classified score)+(month-classified score)+(constant value)”, and article residual rate calculating means for multiplying the article residual rate predictive-value by a new article value to calculate an article residual value. The first data memory device serves to store maker-classified new article sales quantity or article name-classified new article sales quantity before elapsed years. The article residual value predicting computer further comprises a first weight coefficient calculating means for reading out the maker-classified new article sales quantity or article name-classified new article sales quantity before elapsed years stored in the first data memory device, calculating a weight coefficient from an equation “(maker-classified new article sales quantity before elapsed years) / (maker-classified record number)” or “(article name-classified new article sales quantity before elapsed years) / (article name-classified record number)”, and storing the weight coefficient based on the calculated new article sales quantity in the first data memory device, and weighting means for reading out the weight coefficient based on the calculated new article sales quantity from the first data memory device and duplicating the number of relevant records stored in the first data memory device corresponding to the weight coefficient based on the readout new article sales quantity and storing the record numbers increased by duplicating. The category score calculating means serves to perform the aforementioned regression analysis using concurrently all the relevant records weighted by the weighting means collectively.

The invention provides a blood glucose level prediction device whereby a blood glucose level of a user in future can be predicted at a high accuracy. The blood glucose level prediction device 1 comprises: an acquisition section 11 for acquiring a measured blood glucose level of a user, a measured HbA1c value of the user and health examination result of the user; a layer determination section 12 for determining the type of the user from among normal type, boundary type and diabetes type on the basis of the measured blood glucose level, measured HbA1c value and health examination result as described above; and a prediction section 13 for predicting a fasting blood glucose level of the user in future using the determination result and a fasting blood glucose level of the user measured in thepast.

The invention discloses a wind turbine active power prediction and error correction method based on neural network. The method includes the following steps: firstly using a wavelet spectrum analyzing method to extract a significant period sequence hidden in an active power time sequence of a wind turbine and perform separation to obtain a residual sequence, then separately predicting the prominent period sequence and the residual sequence by using a neural network model, wherein the prominent period sequence uses a BP neural network which is optimized based on particle swarm optimization and laminated with error correction to perform prediction, the residual sequence uses a RBF neural network which is optimized based on particle swarm optimization and laminated with error correction to perform prediction, and acquiring a final result of wind turbine active power prediction based on the prediction results of the significant period sequence and the residual sequence. According to the invention, the method can predict the active power of each wind turbine in a wind power plant in a precise manner so as to effectively increase short-term output prediction level of the entire wind power plant.

The invention discloses a multi-classification deep learning short-term wind power prediction method based on pitch angle classification. The method comprises the following steps: equally dividing original fan data into four data sets according to pitch angle intervals; respectively carrying out Pearson correlation analysis on the four data sets to determine the correlation degree between each variable in each data set and the wind power; according to the relevancy of the variables in each data set, selecting several variables with high relevancy with the wind power as the input of the deep neural network model, and taking the wind power as the output; and dividing each data set into a training set and a test set in proportion, and training and testing the deep neural network model by using the training set and the test set to obtain a final deep neural network model. According to the prediction method, the problem of low wind power prediction accuracy in the prior art is solved.

Provided is a breakdown prediction device which is provided with a vibration measurement unit for measuring vibrations generated from a monitored target apparatus and a signal processing unit for performing a breakdown prediction in the case where the vibration measurement unit measures specific vibrations, the maximum amplitude value of which exceeds an upper limit amplitude threshold value, and an amplitude value of which at the time when a specified period of time elapses from when the upper limit amplitude threshold value is exceeded falls below a lower limit amplitude threshold value.

With conventional methods, there is a reduction in the prediction accuracy of arriving wind information, such as arriving wind speed or wind direction. This wind measurement LIDAR device 1B: is mounted on a windmill 2; transmits, through the atmosphere, transmission light, which is pulsed laser light, in a plurality of beam directions determined with respect to the front direction of the windmill 2; and measures the wind speed in the beam directions at a plurality of distances from the windmill, from the Doppler frequency deviation with respect to transmission light of reflected light, which is the transmission light reflected by particles that move together with the atmosphere. This invention comprises: spectrum integration units 12c, 12e that obtain an integrated spectrum, which is an integration, for each wind speed measurement section consisting of a combination of a beam direction and a time section, of a spectrum obtained from split reception signals of a plurality of pulses transmitted after the wind speed was calculated previously; a wind speed calculation unit 12h that calculates a wind speed for each wind speed measurement section, for an integrated spectrum with an SN ratio that is at least a first threshold value; and an arriving wind information prediction unit 16 that, on the basis of the wind speed for each wind speed measurement section, predicts arriving wind information, namely information of arriving wind, which is wind that will arrive at the windmill 2.

The invention discloses an algorithm for performing flood-caused diarrhea outbreak risk remote sensing diagnosis by utilizing remote sensing data and an expert knowledge base. The algorithm disclosed by the invention comprises the following steps: (1) dividing a flooded area by utilizing multi-temporal and different resolution remote sensing data; (2) obtaining flood continuous distribution characteristics through synthetic aperture radar remote sensing data based on long time series and a GIS Overlay analysis technology; (3) analyzing by utilizing an inverse distance weight difference algorithm and resolving to obtain dissolved oxygen, infected people, death toll and other spatial distribution characteristics; (4) establishing a flood-caused diarrhea outbreak risk fitting model by utilizing an expert system diagnosis algorithm, analyzing a relation among all variables, resolving to obtain model parameters and obtaining a final flood-caused diarrhea outbreak risk remote sensing diagnosis model. The invention forms a model method capable of quickly providing diarrhea outbreak risk forecast for residents in a flood prone area. Through the method, high-precision forecast can be performed on diarrhea outbreak risks, so that the morbidity and the mortality of patients in a disaster area are reduced.

The invention relates to the technical field of electric vehicle lithium batteries, and discloses an electric vehicle lithium battery residual life prediction method based on an XGBoost-LSTM optimization model, and the method comprises an electric vehicle lithium battery information online collection technology. Experiments are carried out through actual electric vehicle charging data, classification can be carried out based on the appropriate data size, different and appropriate training models are selected, the RUL prediction problem in the electric vehicle lithium battery coverage total attenuation process is solved, and the method has remarkable significance in improving the long-term prediction performance and prediction accuracy of the battery RUL and has good application prospects. The accumulated feature influence and the meta-reinforcement learning algorithm are introduced into battery RUL prediction, battery health state information hidden in the accumulated features and the change rule of the battery health state information are fully excavated, meanwhile, the high small sample learning capacity of the meta-reinforcement learning algorithm is exerted, and high-precision prediction can be achieved in the whole life process of the battery.

An Internet of Things system and a modeling method for predicting a soil state of an agricultural land. The system includes at least one calculation module, which is the main control device of the system and used to control the overall functions of the system. The calculation module further includes an analysis unit and a machine learning unit, wherein the analysis unit is used to analyze and build a prediction model according to its analysis information. The machine learning unit has at least one appropriate calculation function to establish a corresponding learning model. The computing module is also electrically connected to an Internet of Things module to serve as an intermediary role for information transmission. The Internet of Things module is electrically connected to at least onedetection unit, which is used to build the environment and soil conditions in the target environmental soil and area and then return the same to the calculation module for subsequent analysis.

The invention discloses an AP self-adaptive optimization selection method based on machine learning, and the method is applied to the process of establishing WIFI connection between a mobile device and an AP and switching an Internet of Vehicles adaptive network, and the method comprises the steps: collecting connection device data in a current environment, establishing a training data set and a feature set, and determining a threshold value; determining whether the tree is a single-knot tree according to the data set and an ID3 algorithm; if the tree is not a single-knot tree, segmenting thesubset to construct a sub-node spanning tree; performing recursive call until a complete decision tree is generated so as to classify the APs into a FAST set and an SLOW set, and selecting the fastestAP in the FAST set to establish a connection. According to the invention, the AP access point is selected according to the machine learning model to shorten the connection time and reduce the WIFI connection setting time cost.

The invention provides a meteorological AI platform based on big data, which comprises a hardware layer, a system layer, a software interface layer and a meteorological service layer. The meteorological service layer comprises an intelligent client and an intelligent weather model; the intelligent weather model is used for analyzing a field sequence based on weather in a latest set time length ofa forecast area in the meteorological and geographic information data, inputting to a depth neural network model based on training, and obtaining a weather forecast field sequence of the forecast areawithin the forecast duration from the current moment; the field sequence analysis based on weather within the latest set time length is as follows: setting a field sequence analysis based on weatherwithin a time length from the current moment to the front; and the intelligent client is used for realizing data interaction between the weather AI platform and external intelligent equipment. The prediction result is more accurate, and the prediction result can be used as a reference basis for the forecast of a forecaster together with the result of the mode forecast, so that the forecaster can provide more accurate weather forecast.

The invention discloses a region tail gas migration prediction method and system based on domain adaptation and a storage medium. The method comprises the steps: obtaining and processing historical tail gas data and external factor data of a source region and a target region, taking monitoring points as nodes for the source region data and the target region data, and enabling the source region data and the target region data to be connected pairwise; constructing graph structure data by taking the weight as the reciprocal of the monitoring point distance, and dividing a time sequence set according to the tail gas concentration change characteristics of the source region and the target region; constructing a tail gas spatio-temporal feature extraction module, and performing shallow feature extraction and fusion on the time sequence data of the source region and the target region; constructing an automatic encoder, and utilizing the encoder to non-linearly map shallow spatio-temporal features of the source domain and the target domain belonging to different feature spaces to the same feature space; performing depth extraction on the shallow layer features, and outputting a prediction result. According to the method, efficient utilization of the source domain data is realized by utilizing the domain adaptation method, so that higher-precision regional tail gas prediction of the target domain lacking data is realized.

The invention discloses a high-precision weld joint morphology prediction method suitable for myriawatt-level laser welding, and aims to solve the problem that the weld joint morphology prediction precision of myriawatt-level laser welding is low due to the fact that a weld pool and plume coupling behavior cannot be considered in an existing weld joint morphology prediction method. According to the method, a compressible two-phase flow numerical calculation method based on pressure is adopted to solve the coupling behavior of the molten pool and plume, and therefore high-precision prediction of the myriawatt-level laser welding seam morphology is achieved. Firstly, a welding seam morphology function and welding parameters at the initial moment are input; secondly, a compressible two-phase flow numerical calculation method based on pressure is adopted to obtain a welding seam morphology function at the next moment; and drawing a welding seam morphology function at the next moment, and extracting welding seam morphology and welding seam morphology characteristics. Compared with an existing weld joint morphology prediction method, the weld pool and plume coupling behavior in myriawatt-level laser welding can be accurately calculated, the algorithm is simple and easy to implement, the calculation efficiency is high, the physical conservation is good, and high-precision prediction of the myriawatt-level laser welding weld joint morphology can be achieved.

The invention discloses an ultrasonic synthetic aperture imaging method and device introducing machine learning, and the method comprises the steps: collecting a synthetic aperture echo signal in a sparse scanning mode through a one-dimensional linear ultrasonic transducer array; then, machine learning is adopted to predict related signals to solve the problem of sparse spatial arrangement of echo signals caused by a sparse scanning mode; and finally, realizing ultrasonic synthetic aperture focusing imaging based on a delay superposition algorithm (DAS). After optimization processing of machine learning, compared with a traditional synthetic aperture focusing technology (SAFT), the efficiency and the imaging quality of synthetic aperture focusing imaging are greatly improved, and meanwhile, the influence of side lobes and the like on an imaging result is reduced. The method has the characteristics of high scanning speed and high image reconstruction precision, and is beneficial to popularization and application of the ultrasonic synthetic aperture imaging technology in nondestructive testing and medical diagnosis.