1373results about How to "Prediction is accurate" patented technology

A predictive modeling method may include determining a time interval of time-series data; identifying one or more variables of the data as targets; determining a forecast range and a skip range associated with a prediction problem represented by the data; generating training data and testing data from the time-series data; fitting a predictive model to the training data; and testing the fitted model on the testing data. The forecast range may indicate a duration of a period for which values of the targets are to be predicted. The skip range may indicate a temporal lag between the time period corresponding to the data used to make predictions and the time period corresponding to the predictions. The skip range may separate input data subsets representing model inputs from subsets representing model outputs, and separate test data subsets representing model inputs from subsets representing validation data.

The invention discloses an SOC (State of Charge) and SOH (State of Health) prediction method of an electric vehicle-mounted lithium iron phosphate battery, which comprises the following steps of: (a) improving a Thevenin cell equivalent model; (b) determining the state equation and the output equation of a system; (c) identifying battery model parameters; (d) using a Kalman filter algorithm to iterate the state variables of the system, so that the predictive value of SOC is closer to the actual value; and (e) using a dual-channel Kalman filter algorithm to carry out the online predication of an internal resistance and capacity of the lithium iron phosphate battery, and simultaneously predicating the SOH of the battery according to the changes in the internal resistance and the capacity value of the battery in the current state and the initial state. With the method, the predication precision of SOH of the battery is effectively improved, the decline in battery performance can be determined more accurately, and the internal resistance and capacity information of the battery is combined to provide a basis for making the battery management strategy and maintaining and replacing the battery.

The invention discloses a mobile robot obstacle avoidance method based on obstacle movement prediction, comprising the following steps: in the robot moving process, a robot is taken as the center, and the space around the robot is be divided into three regions which comprise, from the outside to the inside, a path planning region, a conflict avoidance region and an emergency escape region; in accordance with the movement state of the obstacle, the robot judges at which region the obstacle is located and predicts the movement state of the obstacle in the next time; according to the region in which the obstacle is located, the robot carries out different obstacle avoidance strategies. The method adopts CV, CA and CS models to describe the movement of dynamic obstacles to reduce the restrictions on the movement state of dynamic obstacles. The method predicts the movement state of the obstacle through IMM algorithm and gets accurate prediction results, so that the mobile robot can carry out dynamic obstacle avoidance in a small scale, rather than large-scale deviation from the originally planned path.

The invention discloses a traffic state prediction method for the urban road network based on key road sections, which is characterized by comprising the steps of first, carrying out data preprocessing; second, establishing a spatial weight matrix of the road network; third, establishing a time correlation matrix; fourth, recognizing key road sections by using a time-space correlation matrix; andfifth, establishing a deep convolution neural network, predicting the state of the road network in the future, and carrying out evaluation on a prediction model. The traffic state prediction method predicts the urban traffic flow state from a level of the wide-range road network, thereby being conducive to guiding the traffic flow from a macroscopic perspective, and fully exploring time-space correlation characteristics of the traffic flow. The key road sections in the road network are recognized, so that the training time of the model can be greatly reduced compared with a method of taking historical states of all road sections as input data, and the prediction efficiency is improved; and the convolution neural network is adopted to serve as the prediction model, and the prediction resultis also more accurate.

The invention discloses an establishing method of a failure type prediction model. The establishing method comprises the following steps of: collecting a vibration signal of a fan, and meanwhile, collecting the audio signal of the fan; pretreating the vibration signal and carrying out feature extraction on the vibration signal after pretreating so as to obtain a vibration characteristic parameter, and pretreating the audio signal and carrying out feature extraction on the audio signal after pretreating so as to obtain an audio characteristic parameter; and substituting the vibration characteristic parameter to a failure type vibration template so as to obtain a failure type corresponding to the vibration characteristic parameter, and associating the audio characteristic parameter with the failure type so as to establish a failure type audio template. The invention also discloses an establishing system of the failure type prediction model and an establishing method and system of a malfunction probability prediction model as well as a fan monitoring pre-warning system and method. The fan monitoring early-warning system and method provided by the invention have the advantages that the sensitivity is high, and the failure monitoring is more exact.

The invention provides a sales volume forecasting method and a training method, a device and an electronic system of a model thereof. The training method of the sales volume forecasting model comprises the following steps: obtaining historical sales data of commodities; Generating sales characteristics related to the sales volume of the merchandise according to the historical sales data; Acquiringthe external characteristics of the corresponding time period of the historical sales data; External features include at least one of a time attribute, a weather characteristic, and a arrival crowd characteristic; determining The training samples according to the sales characteristics and external characteristics, and inputting the training sample into the preset machine learning model for training until the loss function value of the machine learning model converges, and finishing the training to obtain the sales forecast model. The present invention takes into account the sales characteristics related to the sales volume of commodities and the external characteristics that may affect the sales volume of various commodities, so that the prediction result of the sales volume of commodities is more objective and accurate, and contributes to the improvement of the sales volume and the flexibility of the capital turnover.

The invention discloses a sewage-disposal soft measurement method on the basis of an integrated neural network, and belongs to the field of sewage disposal. A sewage disposal process is high in nonlinearity, time-varying characteristics and complexity, and measurement for key water quality indexes is crucially significant in control of water pollution. In order to improve precision of simultaneous soft measurement for various key water quality parameters in a sewage-disposal soft measurement process by the sewage-disposal soft measurement method, an integrated neural network model is provided for measuring COD (chemical oxygen demand) of outlet water, BOD (biochemical oxygen demand) of the outlet water and TN (total nitrogen) of the outlet water, coupling relation between the three key water quality parameters is sufficiently utilized in the model, the integrated neural network model contains three feedforward neural sub-networks, and the various neural sub-networks are trained by particle swarm optimization, so that the optimal structure of each neural sub-network can be obtained. The COD of the outlet water, the BOD of the outlet water and the TN of the outlet water are predicted by the trained neural network finally, and prediction results are accurate.

The invention discloses a method for preventing boats from colliding, comprising the following steps: performing program initialization; reading information of any two boats, predicting the track of the boats; judging whether the predicted tracks of two boats intersect, if not, returning to step 2 to continue reading the information of any other two boats to predict the tracks of other boats, if yes, calculating the intersection point, and entering the next step; calculating the intersection point, and using a Mercator projection transformation formula to calculate the distance between the intersection boats and the intersection point and a time zone from the boats to the intersection point; judging whether the time zones of two boats have intersection, if not, returning to step 2 to continue reading the information of any other two boats to predict the tracks of other boats; if yes, entering the next step; and giving alarms. The invention uses the Mercator projection transformation, which ensures that the distance between any two points is more accurate and prediction results are more correct; in addition, the invention adopts the judging method of the time zone, which can comprehensively predict time for boats to travel to a designated site, so that the reliability of the prediction result can be improved.

Disclosed is a car chassis integrated controller hardware in-the-loop simulation test bed; a car multi-condition digital simulation mathematical model is established on a host machine, translated into C code format, received by a target machine through the Ethernet and then downloaded into the CPU of the target machine; an electronic control unit controls eight electromagnetic valves and receives the state of the electromagnetic valves through a PCL card to control the various car condition digital simulation model running and feedback information stored in the target machine. The car chassis integrated controller hardware in-the-loop simulation test bed has the advantages that the test bed realizes the hardware-in-the-loop of an ECU and an actuator so as to achieve more accurate prediction results of various control parameters; in the early stage of the research and development of the chassis electronic control system, the test bed optimizes the control parameters of extreme dangerous condition; the test bed also simplifies the test environment, and all the performances and optimized parameters obtained in the test are closer to that of real car testing.

The invention discloses a driving trajectory predicting system integrating a kinematic model and a behavioral cognition model. The driving trajectory predicting system is characterized in that an interactive mixing module subjects the prediction result of each prediction module at a last time moment to interactive mixing to output a mixed result used for prediction at a next time moment; the prediction module comprises a behavioral cognition trajectory prediction module based on behavioral cognition and a motion trajectory prediction module based on kinematics; the behavioral cognition trajectory prediction module and the motion trajectory prediction module performs prediction according to the mixed result output by the interactive mixing module and output respective prediction results including a vehicle position and a covariance matrix; a fusion updating module fuses a final prediction result according to the prediction results, updates a weight coefficient, and outputs a vehicle position and a covariance matrix at a certain time moment in the future. The driving trajectory predicting system can continuously estimate the position state and the driving behavior of the vehicle in a vehicle driving process, predicts a driving trajectory, and provides assistance for intelligent driving decision.

A depth learning-based road network traffic situation forecast method comprises the following steps of S1, acquiring multi-source traffic data and road network static configuration information, and building a traffic flow parameter model, wherein the multi-source traffic data comprises internet road segment speed data, detector flow data and signal machine control scheme data, and the road networkstatic configuration information comprises road network space geographical position information, intersection number, road segment class, road segment length, road segment number, lane number and lane function; S2, analyzing road network congestion relevancy, and building a basic forecast group; S3, building a dual-stage attention mechanism-based depth learning traffic situation forecast model; and S4, building a traffic situation forecast system. The forecast accuracy and the transportability are relatively good.

The invention relates to the field of short-term traffic flow control, in particular to a short-term traffic flow control method based on a graph convolution recurrent neural network. The method comprises the following steps of acquiring a data source; constructing a graph structure traffic flow sequence; constructing a recent component model of a space-time graph convolution recurrent network, adaily cycle component model of the space-time graph convolution recurrent network and a weekly cycle component model of the space-time graph convolution recurrent network according to the multilevel of the graph structure traffic flow sequence in the time dimension; fusing the results of the 3 models to obtain a short-term traffic flow prediction model; obtaining a prediction result according to the model; and counting the predicted data, sending a counted result to a traffic department, and controlling the traffic flow of each check point of a road network. The traffic flow recent, daily cycle and weekly cycle dependencies can be simultaneously modelled by utilizing the space-time graph convolution recurrent neural network, the short-term flow prediction model based on the space-time graph convolution recurrent neural network integrated with multiple component data is established, thereby reaching a precise prediction result.

The invention discloses a visual dialogue generation method based on a context perceptual map neural network. The visual dialogue generation method comprises the following steps of 1, preprocessing the text input in a visual dialogue and constructing a word list; 2, extracting the features of a dialogue image and the features of a dialogue text; 3, obtaining a context feature vector of the historical dialogue; 4, constructing a context perceptual map; 5, iteratively updating the context perceptual map; 6, carrying out attention processing on the nodes of the context perceptual map based on a current problem; 7, performing multi-modal semantic fusion and decoding to generate an answer feature sequence; 8, generating the parameter optimization of a network model based on the visual dialogueof the context perceptual map neural network; 9, generating a prediction answer. According to the method, the context perceptual map neural network is constructed on the visual dialogue, and the implicit relationship between different objects in the image can be reasoned by using the text semantic information with finer granularity, so that the reasonability and accuracy of the answers generated by an intelligent agent for question prediction are improved.

The invention relates to an energy storage device control method based on ensemble empirical mode decomposition and LSTM. The method comprises the following steps: firstly, performing normalization processing and ensemble empirical mode decomposition on historical short-term load data in first n + 2 time periods in continuous n + 2 time periods to obtain a subsequence or a residual component as aninput item; training an LSTM model by taking a subsequence or a residual component corresponding to historical short-term load data in a later time period as a theoretical output item; preprocessingthe data of the current time period and the historical short-term loads of n time periods closest to the current time period, and inputting the preprocessed data into the trained LSTM model; and afterthe trained LSTM model is utilized to output predicted values, all the predicted values are reconstructed and subjected to reverse normalization processing to obtain a prediction result, and finally,the energy storage device is controlled to charge and discharge according to the prediction result. The method has the advantages of high prediction precision and reasonable charge-discharge operation of the energy storage device.

The invention relates to an automobile double-clutch automatic speed changer hardware-in-the-loop simulation test bed; a host machine establishes a digital simulation mathematical model of multiple working conditions of the automobile and converts the digital simulation mathematical model into a C code format, the C code format is received and downloaded to a CPU by a target machine through an Ethernet, an electric control unit controls an executive motor of the two clutches and the executive motor of a synchronizer, and the displacement of the clutch is received through a PCL board card in order to control the operation of the digital simulation model of various working conditions of the automobile stored in the target machine and the information fed back by the target machine. The invention has the advantages of realizing the hardware-in-the-loop of a TCU and an actuator, the more accurate prediction results of various control parameters, optimizing the control parameters of extremely dangerous situations in the prophase of research and development of a DCT electronic control system, simplifying the test environment, and the tested various performances and gained optimization parameters closer to the real vehicle test.

The invention provides an electricity stealing risk prediction method and device based on deep learning. A solution is provided for the problems that in a traditional electricity larceny prediction method, an electricity larceny user cannot be positioned due to the fact that a topological structure in a power distribution network is not known, and the electricity larceny prediction accuracy is lowdue to the fact that a non-efficient identification feature is used in the traditional electricity larceny prediction method. The prediction method comprises the following steps: firstly, predictingthe daily power consumption mode quantity of a user by utilizing daily historical power consumption information of a low-voltage user; secondly, according to analysis of power utilization factors of the users in nearly three days of the day, an analytic hierarchy process is adopted to calculate the abnormal rate of power utilization, and finally, based on K <->, the abnormal rate of power utilization is calculated. And the Means method is used for grading the electricity stealing. Technical effects of improving prediction accuracy and realizing electricity larceny grade division are achieved.

The invention relates to a decision tree index-based neural network air quality prediction method. The method comprises the following steps of establishing a time sequence data set of related meteorological factors, air quality and atmospheric pollutant discharge; classifying the obtained training samples by using a decision tree DT algorithm to generate an optimal tree structure T alpha orientedby air quality characteristics and a corresponding classification result; according to the classification result, establishing a BP neural network model for each classification, and performing model training; inputting a prediction data set, performing classification indexing based on a decision tree, and selecting the trained DT-BP neural network model or the comprehensive BP neural network to predict the air quality; obtaining a continuous air quality prediction result based on an iterative algorithm; recording the frequency of occurrence of data sets which do not meet the decision tree classification matching rule, and automatically starting model updating when a set value is exceeded. The method is suitable for predicting and forecasting the air quality of conventional weather, abruptchange weather and heavy pollution weather.

The present invention discloses a multi-layer oil reservoir overall yield prediction method. The method comprises: dividing a multi-layer oil reservoir into a plurality of blocks, determining a reservoir stratum type, a formation factor Kh and an evaluated reserve volume of each small layer of each block; selecting, from the plurality of blocks, a block representing a multi-layer oil reservoir geological feature as a representative block, establishing a fine geological model of the representative block; establishing a corresponding fine numerical simulation model according to the fine geological model of the representative block, determining a typical mining curve of different types of reservoir strata in different development manners; determining a relationship curve between a single well injection amount and Kh in the multi-layer oil reservoir, and a relationship curve between a produced quantity and Kh; and predicting a multi-layer oil reservoir yield according to the typical mining curve, the relationship curve between the injection amount and Kh, the relationship curve between the produced quantity and Kh, and the reservoir stratum type, Kh and evaluation reserve volume of each small layer of each block. According to the multi-layer oil reservoir overall yield prediction method, the prediction accuracy of the multi-layer oil reservoir overall yield is improved, and further, working efficiency is improved.

The invention discloses a method for predicting a bearing fault based on Gaussian process regression. The method comprises the following five steps of: step 1, setting prediction system parameters, initializing a Gaussian process regression model; step 2, collecting a bearing vibration signal regularly, extracting characteristics of a vibration signal to obtain time domain characteristic parameters of the bearing vibration signal, and carrying out fault symptom judgment; step 3, judging whether a fault symptom exists; step 4, calculating and storing the characteristic parameters, and carrying out dynamic updating of the Gaussian process regression model; and step 5, predicting the fault of a bearing. According to an actual use condition of a product, small amount of data is collected, time that the product possibly has the fault is given quantificationally, a calculation speed and prediction accuracy are improved by using the Gaussian process regression, a whole life cycle of the bearing is divided into three time ranges, such as a health time range, a sub-health time range and a fault time range by use of an idea of health management, fault prediction is carried out in the sub-health state, usage management capacity of the bearing is improved.

The invention discloses a Bayesian network-based regional heat supply model predictive control system and method. The method includes the following steps that: step S1, physical layer heat network data sensing is performed: historical data are acquired from a source side, a network side and a building side in real time and are updated; step S2, a Bayesian network is constructed according to the historical data and on the basis of prior knowledge, and the load demand of a thermal station and the building side are predicted through the Bayesian network; step S3, the real-time control parametersof a secondary side, a primary side and the source side are obtained through the inference of the Bayesian network according to the load demand of the building side and on the basis of the historicaloperation data and real-time data; and step S4, time characteristic curves of source side adjustment, network side adjustment and building side adjustment are established according to the historical operation data and pipe network topology structures, and source side adjustment, network side valve and building side electric valve adjustment strategies are determined, control operations are performed according to the strategies, so that the hysteresis of heat network adjustment can be eliminated, real-time supply and demand balance can be met, and the accurate on-demand heat supply of a heat user side can be realized.

The invention discloses a time series data event prediction method and system based on a graph convolutional neural network and application of the time series data event prediction method and system,and the method comprises the steps: converting time series data after data cleaning into event sequence data at a preset time interval, and obtaining vector representation of an event and an event set; taking an event set contained in each piece of sequence sample data in the event sequence data at the last moment as a prediction target, and taking the prediction target as a corresponding sequencesample label to obtain labeled event sequence data; and when the graph convolutional neural network model is trained to meet a preset convergence condition, testing a model prediction effect by usingthe test set, and taking the model of the test effect as a final event prediction model. The method can make up for the defects that the traditional method has high requirements on data quantity andquality and cannot fully utilize the knowledge graph.

The invention belongs to the field of fault prediction of neural networks, and discloses a fault prediction method based on a migrated convolutional neural network. The method comprises the followingsteps: (a) numbering fault types, collecting a time-domain signal of a to-be-predicted object, acquiring initial fault type numbers, and converting time-domain signals into RGB images; (b) initializing an FC layer of a deep residual network model, and adding a classifier to obtain an improved network model; (c) inputting RGB images into the network model to train the FC layer and the classifier, continuously updating weight values of the FC layer, and when obtained fault type numbers are close to the initial fault type numbers, determining that the corresponding weight values are needed new weight values, and completing migration of the network model; and (d) inputting the RGB image of the to-be-predicted object into the migrated convolutional neural network model, and outputting a predicted fault type number. Through the method, a structure of the adopted migrated convolutional neural network model is simple, a prediction speed is high, and a prediction result is accurate.

The invention discloses a VMD-GRU-based short-term wind speed prediction method. The method is characterized is characterized in that it comprises, acquiring data of wind speeds at the current momentand n moments closest to the current moment to form a time sequence; and preprocessing the time sequence to obtain a plurality of sub-sequences and residual components, respectively inputting each sub-sequence and each residual component into respective corresponding trained GRU models, outputting predicted values by the trained GRU models, and performing post-processing on all the predicted values to obtain future wind speed prediction data at the next moment. According to the method, non-stationary wind speed data is decomposed into sub-sequences and residual components with different frequencies by adopting a variational mode decomposition method, the stability of the sub-sequences and the residual components is good, and better prediction is facilitated; the method has good predictionprecision for the wind speed with strong volatility, randomness and uncertainty, and the operation state of the wind power generation device can be adjusted more reasonably.

The invention relates to the technical field of rock tunnel construction and earthwork parameter calculation, in particular to a deep soft soil metro foundation pit soil body parameter inversion analyzing method. The method includes multiplying a least square function of a difference value between a finite element numerical calculation result and an actually monitored value by a specific weighting coefficient to obtain the function which serves as a target function, utilizing an improved Gaussian-Newton method to conduct optimal solution on the target function, and conducting repeated iterative computation to enable the finite element numerical calculation result to gradually approach the actually monitored value to determine an optimal solution of a parameter to be confirmed and build the relation between the monitored value and the foundation pit soil body parameter basic characteristic value. By means of the method, the actual observation value is combined to conduct constant soil body parameter inversion and correction to further predict the soil body deformation of the next stage, a prediction result is more and more accurate with the number increase of the monitored value, a very good monitoring and early warning method is provided for deep soft soil metro foundation pit project construction, and the method has high application value in the soil body parameter determination process.

The embodiment of the invention discloses a prediction method and device based on historical data, and the method and device effectively improve the precision of a prediction result. The method comprises the steps: obtaining the historical data of a prediction item, carrying out the training of the historical data through a neural network system, and obtaining a prediction model, wherein the historical data comprises the historical data of the prediction item and the related historical data of the prediction item, and the prediction model reflects the relation between the historical data of the prediction item and the related historical data of the prediction item; obtaining the related actual data of the prediction item, and achieving the prediction of the actual data based on the historical data of the prediction item according to the related actual data and the actual data, obtained by the prediction model, of the prediction item.

The invention discloses an ultra-short-term power load forecasting and early warning method based on a Kalman filter and wavelet echo state network. In order to solve the problem that noise and the like are contained in power load data, a Kalman filtering method is adopted to conduct real-time estimating on 'collected data', with the help of a forgetting factor, the weight of old-fashioned data is weakened, and prediction accuracy is improved. Before ultra-short-term load forecasting is conducted, firstly, a principal component is used for analyzing and determining main working procedures for influencing the change of a power load, the main working procedures are used as the input of a power load capacity prediction model, afterwards, wavelets are used for decomposing the loads of different spectral characteristics (high frequency, follow-up and stability) of the power load, echo state network singe power loads are respectively established for predicting and modeling, various forecasting components are integrated to obtain a total load variation trend, and ultimately an early warning test is conducted on the prediction model specified by a user.

The invention relates to the field of natural language processing, in particular to a bidirectional recursive neural network-based enterprise abbreviation extraction method. The method comprises the steps of serializing to-be-processed texts through word segmentation processing; selecting a certain number of the to-be-processed texts to perform manual annotation, segmentally annotating enterprise names in the to-be-processed texts as starting parts, keyword parts, industrial parts and organization form parts, and annotating data except the enterprise names as unrelated parts; inputting annotated training samples into a bidirectional recursive neural network to train the bidirectional recursive neural network; extracting word sequences belonging to the enterprise names through prediction of the bidirectional recursive neural network, and further extracting fields belonging to the keyword parts of the names as enterprise abbreviations; and establishing a corresponding enterprise abbreviation database. Therefore, powerful technical support is provided for related information analysis of informal texts.

The invention relates to a predicting method of the leakage rate of a bolted flange connection structure with an anti-loosing washer. Aiming at the defect that the leakage rate of the prior bolted flange connection structure with the anti-loosing washer can not be predicted, by establishing the deformation compatibility equation of the bolted flange connection with the anti-loosing washer and calculating the pretightening force and the working stress of a gasket, and according to the tightness principle, the predicting method of the leakage rate of the bolted flange connection structure with the anti-loosing washer is established, so that the prediction of the leakage rate of the connection structure is realized. The predicting method of the invention has simple calculation principle and method and accurate prediction result, and can be widely applied to the prediction and calculation of the leakage rate of the connection structure in the circumstance of high temperature and high pressure and larger fluctuation of temperature and pressure.

The invention discloses an urban road network generalized road right calculating method considering comprehensive traffic management measures. The method comprises road section time impedance calculation with traffic management measures being taken into consideration and intersection delay calculation under a traffic control strategy, and specifically comprises road section basic traffic capacity analysis, influence correction of the traffic management measures on road section traffic capacity, road section time impedance analysis and correction of the traffic management measures on road section time impedance; and intersection entrance road lane group division based on turning relation, lane group motor basic traffic capacity analysis, influence correction of the traffic management measures on lane group traffic capacity and intersection delay analysis for traffic assignment. The method can reflect independent road rights of different types of vehicles in an urban road network, can enable simulation and prediction results to be more accurate, quantifies difference of different traffic management strategy combination schemes and provides powerful support for planning of current urban traffic management and control and formulation of traffic policies.

The invention discloses a dynamic model-based unmanned car model prediction controller design method. According to the method, a novel vehicle dynamic model is established; an unmanned car model prediction controller on the basis of the novel vehicle dynamic model; (1) the novel vehicle dynamic model obtained through linearization at the current working point is utilized to predict an output stateof a system in a prediction time domain; (2) the obtained predicted output of the system and given reference output are utilized to construct an optimization problem; and (3) the optimization problemis solved, and a first component of U (k) is taken as an optimal control quantity of the current moment to act on the system. According to the method, vehicle speed is added to serve as the control quantity, and error compensation is introduced to make up accumulative errors caused by model precision, linearization and discretization in the prediction process, so that unmanned cars can have a better trajectory tracing effect.