41 results about "Narx neural network" patented technology

The invention provides a lithium ion battery remaining useful life prediction method and system based on a PCA-NARX neural network, so as to solve the problems of large difficulty and high redundancyin lithium ion battery remaining useful life prediction health index measurement. The method comprises steps: 1) a constant current discharge voltage change law of the lithium ion battery at differentdischarge periods is analyzed, and parameters that reflect the degradation of battery performance are extracted; 2) the correlation between the extracted parameters and the correlation between the extracted parameters and the lithium ion battery capacity are verified, a PCA algorithm is used to remove redundancy of the parameters, and a main component obtained after redundancy removal is used asa health indicator for the lithium ion battery; and 3) the obtained health indicator for the lithium ion battery is inputted to an NARX neural network, and lithium ion battery capacity estimation andremaining useful life prediction are carried out. An experiment result proves that the method is high in prediction precision and can be used for accurate prediction on the lithium ion battery remaining useful life.

The invention discloses an intelligent monitoring system of ammonia gas in a cow barn environment based on a wireless sensor network. The intelligent monitoring system is characterized in that the intelligent monitoring system consists of a cow barn environment parameter intelligent detection platform based on the wireless sensor network, a cow barn environment ammonia gas intelligent detection module and an NARX neural network ammonia gas combined prediction module. According to the invention, a problem is solved that ammonia gas concentration in the cow barn is monitored and acquired only by a device and there is no effective measurement module for giving out an early warning for hazard gas in the cow barn in the prior art.

The invention relates to a passenger flow prediction method based on a time-lag NARX neural network and mainly aims to solve the technical problem of low prediction precision in the prior art. The method comprises the steps that n pieces of historical data is collected from an automatic fare collection system to serve as original samples, and preprocessing is performed to obtain preprocessed samples; an NARX short-time passenger flow prediction model p(t) about a time sequence is established according to a nonlinear autoregression network with external input, wherein the external input is an external impression factor u(t), and the nonlinear autoregression network with the external input comprises an input layer, input lag time, a hidden layer, an output layer and output lag time; and real-time passenger flow prediction is performed according to the NARX short-time passenger flow prediction model, an excitation function and a training algorithm, wherein real-time passenger flow prediction comprises short-time passenger flow prediction, peak prediction and representative passenger flow distribution site prediction. Through the technical scheme, the technical problem is solved, and the method can be applied to rail passenger flow prediction.

The invention discloses a wind speed prediction method and an apparatus based on an NARX neural network. The method includes: acquiring historical data of related parameters required by wind speed prediction, wherein the related parameters including wind speed, pitch angle, rotating speed, and power; performing normalization processing of the acquired data; inputting the processed data to the NARX neural network as a training sample for training; and inputting a test sample to the trained NARX neural network, performing reverse normalization of an output value, and obtaining a practical prediction value. According to the method and the apparatus, the model is built for the prediction of the wind speed by employing the NARX neural network, input parameters of the network are selected by employing formulas and principles of fan aerodynamics, the prediction is convenient, the accuracy is high, the wind energy capturing capability and the generating capacity are increased, and the method and the apparatus are applicable to be promoted and applied.

The invention discloses a method for predicting rate of wheel load reduction in the technical field of railroad safety. The method comprises the following steps: first, collecting left track longitudinal irregularity, left track direction irregularity, right track longitudinal irregularity and right track direction irregularity data by adopting a track inspection car; then, simulating the data by using professional software ADAMS (automatic dynamic analysis of mechanical systems) / RAIL to obtain rail-wheel load data, namely vertical rail-wheel load and horizontal rail-wheel load, thereby obtaining and normalizing the rate of wheel load reduction; selecting a training sample to train a NARX (nonlinear auto-regression with exogenous input) neural network model; testing the trained NARX neural network prediction model and outputting the rate of wheel load reduction data after the test; and analyzing the rate of wheel load reduction data in the training sample and the rate of wheel load reduction data obtained from the neural network after the test and valuing performance of the NARX neural network prediction model. By using the method, derailment coefficient is precisely predicted and accuracy of railway operation safety evaluation is improved. Therefore, the method has important realistic meanings to railway traffic safety control.

The invention discloses a model prediction controlling methodachieving data center energy conservation temperature control combined with machine learning. The model prediction controlling method achieving data center energy conservation temperature control combined with a machine learning combines artificial neural network and a model prediction control algorithm to adjust a heating ventilation air conditioning system in a data center; and applies the artificial neural network to analyze data including the outside temperature, time, and energy consumption and the like to estimate the inside optimum temperature; and then inputs the estimating temperature to the model prediction control algorithm to operate, control and adjust. The selected artificial neural network model is an NARX neural network arithmetic. The model prediction controlling method achieving data center energy conservation temperature control combined with machine learning is used in the data center. Model algorithm of self-learning model prediction control based on the energy conservation and the temperature can solve prior problems that the temperature requirement is not met and the consumption of the heating ventilation air conditioning system is not minimized.

The invention discloses a short-term ionosphere forecasting method and device based on NARX. The method comprises the steps that: TEC grid data in a continuous period of time are obtained from a historical TEC data file; then, according to the longitude and latitude of an observation station, a single-point TEC time sequence in the period of time is obtained in a bilinear interpolation mode; an NARX neural network model is established, the NARX model is trained by using the TEC time sequence, wherein input parameters comprise the TEC time sequence and a time sequence of external time parameters, and output is a TEC prediction value of the next moment; and finally, real-time prediction is carried out according to the trained NARX model to obtain a TEC prediction value at a future moment. With the method provided by the invention, the prediction precision of the TEC of the ionosphere can be improved, and the positioning precision can be improved when the method is applied to GNSS positioning.

The invention discloses a method for dynamically predicting drainage discharges at drainage outlets of urban rainwater system. The method comprises the following steps of: (1) simulating a rainfall-runoff by utilizing a rainstorm flood management model pair, and taking drainage flow duration curves at outlets of a plurality of drainage pipe networks as training samples; (2) establishing an RBF neural network to carry out training, and in the training process, optimizing network hidden nodes and center widths Spread; (3) establishing an NARX neural network to carry out training; and (4) coupling the trained NARX neural network and the RBF neural network to obtain a coupled network, carrying out prediction, calculating mean square errors between the coupled network and the samples, returninga flow value with the minimum mean square error as an optimized coupling locus, randomly selecting rainfall data to input into the coupled network so as to obtain a predicted drainage flow duration curve. According to the method, advantages and characteristics of different neural networks are organically combined, the prediction results well accord with SWMM simulation, and the mean square errorsof the curves are 0.000458, so that favorable prediction precision is provided.

The invention discloses an intelligent vehicle lane changing decision-making method based on an incomplete information game, and the method comprises the steps: carrying out the modeling of the lane changing intention and lane changing feasibility of an intelligent vehicle in time and space; establishing a revenue matrix according to the state information of the lane changing vehicle and the surrounding vehicles; solving the revenue matrix by adopting Pareto optimum and Nash equilibrium in the game theory; in order to solve the problem that behavior information of an incomplete information non-cooperative dynamic game is unknown, a lane changing vehicle carries out longitudinal behavior prediction of speed and acceleration on vehicles with game conflicts, and an NARX neural network model is adopted; and the vehicles around the lane changing vehicle predict the transverse movement of the lane changing vehicle through a continuous hidden Markov model. And the game participant corrects the decision of the income matrix by predicting the future behavior trend of other vehicles, and the decision is executed when the decision reaches an execution threshold.

The invention discloses a method for predicting derailment coefficients in the technical field of railway safety. The method comprises the following steps of: firstly, acquiring left-rail height irregularity data, left-rail rail direction irregularity data, right-rail height irregularity data and right-rail rail direction irregularity data of rails by using a rail inspection vehicle; secondly, by using professional automatic dynamic analysis of mechanical system (ADAMS) / Rail software, simulating the acquired data to obtain data of wheel-rail forces including a vertical wheel-rail force and a horizontal wheel-rail force so as to obtain the derailment coefficients, and normalizing the derailment coefficients; thirdly, by using a selected training sample, training a non-linear auto-regressive with exogenous input (NARX) neural network prediction model; fourthly, testing the trained NARX neural network prediction model, and outputting derailment coefficient data which are tested; and finally, analyzing the derailment coefficient data in a test sample and the derailment coefficient data which are obtained through a tested neural network, and evaluating the performance of the NARX neural network prediction model. By adoption of the method, the derailment coefficients can be accurately predicted, the accuracy in evaluation of railway running safety is improved, and great practical significance is provided for rail traffic safety control.

The invention discloses a CA-NARX water quality prediction method based on meteorological factors, and belongs to the technical field of intelligent water quality prediction data application. The method comprises the following steps: 1, ; data standardization, 2, creating a sample matrix, 3, determining an initial clustering center according to the quantile; (4) initial clustering is carried out according to the Euclidean distance; (5) the mean value of each class is used as a new clustering center; (6) clustering is carried out in batches according to the Mahalanobis distance between each sample and the clustering center; (7) clustering number screening is carried out; (8) the best clustering number is selected; according to the method, the problems of high cost and low prediction accuracy of water quality prediction of small and medium-sized reservoirs are mainly solved. Meanwhile, the problem that a traditional clustering algorithm is inapplicable to data heterogeneity and differentvariances is solved, and the training accuracy of the NARX neural network is improved to a certain extent.

The invention discloses a neural network short-term and temporary rainfall forecasting method integrating foundation GNSS water vapor and meteorological elements. The method comprises the following steps: (1) acquiring the foundation GNSS water vapor; (2) calculating an atmospheric stability index; (3) preprocessing data, including gross error data elimination, data interpolation and data normalization processing; (4) identifying rainfall forecasting factors; (5) carrying out NARX neural network design: taking the rainfall forecasting factor and the actual rainfall data determined in the step(4) as an input layer, taking the predicted rainfall data as an output layer, and adopting default values or initial parameters for the number of hidden layers, the number of hidden layer neurons, input and output delay orders and a neural network algorithm; (6) carrying out neural network training; (7) optimizing input parameters, and constructing a multi-factor short-term and temporary rainfallforecasting model; and (8) evaluating the precision of the newly constructed multi-factor short temporary rainfall forecasting model by utilizing the reserved verification data set. According to the method, a reasonable and accurate multi-factor short temporary rainfall forecasting model is established, so that the short temporary rainfall can be accurately forecasted.

The invention discloses a short-term load forecasting method based on improved HS-NARX neural network, the method comprises S1 collecting data and preprocessing; S2, establishing NARX neural network;the neural network is trained with the preprocessed data. 3, determining a fitness function of the HS algorithm; S4, setting parameters of the harmony search algorithm; S5 initialization parameters; S6 generates HMCR and PAR according to HMCRmean and PARmean, and the pitch adjustment bandwidth is (BWmax, BWmin); S7 generating (0, 1) random numbers, generating new harmonic vectors, and uses improved pitch adjustment rules and adaptive parameter tuning method to generate new harmonics; S8, comparing the generated new solution with the worst solution in the harmonic memory bank, if the new solution is better than the worst solution, replacing the worst solution, otherwise, not operating, recording HMCR and PAR again; S9 returning to S7 if the number of iterations is not reached, otherwise, the optimal solution is outputted; S10 mapping the optimal solution to the neural network, obtaining the weights W and the threshold theta of each layer of the network, and training the network and theload forecasting.

The invention discloses a photovoltaic power generation power short-term prediction method and device. The photovoltaic power generation power short-term prediction device comprises a data input, output and processing module, a model parameter optimization module and a prediction model training and application module, wherein the data input / output and processing module comprises an original data normalization and prediction result reverse normalization unit, a GA coding unit and a GA fitness function calculation unit; the model parameter optimization module comprises an RBM initial parameter optimization unit and an RBM model training and NARX parameter initialization unit; the prediction model training and application module is an NARX neural network training and application unit.

The invention discloses a risk prediction system and method for optimizing NARX neural network through an ant lion algorithm, wherein the system comprises a user side and a server side, the user side comprises an information acquisition module and a risk prediction initiation module, and the server side comprises an information processing module, a database and a risk prediction module; the information acquisition module is used for a user to acquire customer data and integrate the customer data into customer data; the risk prediction initiating module is used for a user to initiate a risk prediction application request; the information processing module is used for acquiring customer data and storing the customer data in a database, and is also used for acquiring and auditing the risk prediction application request and generating auditing information transmitted to the user side and the risk prediction module; the database is used for storing customer data; and the risk prediction module is used for acquiring audit information and acquiring customer data in the database according to the audit information, and is also used for performing risk prediction on the customer data to obtain customer overdue risk prediction data. The method comprises steps A1-A6.

The invention discloses a gas turbine anomaly detection method based on an NARX network-box line graph and constant mode extraction, and the method comprises the steps: training an NARX neural networkthrough the data of a training set, and obtaining an exhaust temperature prediction value of training data and a trained NARX neural network model; calculating a residual error between the exhaust temperature prediction value and the corresponding exhaust temperature true value, and inputting the residual error into an improved box line graph algorithm to obtain a residual error detection threshold value; calculating a residual error between a turbine exhaust temperature value predicted by a model obtained by inputting to-be-detected data into a trained NARX neural network model and an actualturbine exhaust temperature value and judging whether the residual error is within a residual error detection threshold value or not. According to the method, the problem that in the prior art, abnormity detection of the gas turbine cannot be achieved under the condition that only a large amount of normal historical data exists is solved, online detection can be achieved, and the method has important significance in safe and reliable operation of the gas turbine.

The present invention discloses a nonlinear delay dynamic system model intelligent identification method. The method includes the following steps that: an NARX neural network model difference equation is assumed; nonlinear dynamic system delay in a set NARX neural network model is determined; the input and output order of a nonlinear dynamic system is determined; the number of hidden layer neurons of a three-layer single-output NARX neural network is determined; a three-layer NARX neural network model is determined; and finally, the validity of the three-layer NARX neural network model is verified, if the validity of the three-layer NARX neural network model is successfully verified, the method terminates, otherwise, the input and output order of the three-layer NARX neural network model is adjusted. With the method of the invention adopted, the problems of high complexity and instability which is caused by severe fluctuation of the switching process of a plurality of local linear identification methods of an existing nonlinear delay dynamic system identification method which adopts the plurality of local linear identification methods to perform identification can be solved.

The invention discloses an NARX neural network assisted integrated navigation method, and relates to the technical field of navigation. The method comprises the steps that during the normal working period of a GPS, a neural network receives IMU output information, SINS position speed increment is used as neural network input, and GPS position speed increment is used as network target output to carry out neural network training; during the short-time failure period of the GPS, the NARX neural network gives a predicted value of the position and speed increment of the GPS according to the input IMU information and the position and speed increment of the SINS, and the predicted value and the position and speed increment of the SINS are subtracted to obtain the quantity measurement required by Kalman filtering. According to the method, the measurement updating is ensured to be smoothly carried out, and the influence of short-time failure of the GPS on the integrated navigation precision is counteracted.

The invention provides an electric power information network security detection system and method based on an NARX neural network. The system comprises a data acquisition module used for acquiring power grid measurement information, state information and load prediction data; the optimal power flow module is used for calculating the node active power, the node reactive power, the branch active power, the branch reactive power and the node voltage amplitude of the power grid on the basis of the acquired data under the optimal power flow operation condition; the NARX neural network design module is used for modeling and training an NARX neural network based on the acquired data; the state vector prediction module is used for predicting a state vector and calculating a residual vector based on the data output by the optimal power flow module and the constructed NARX neural network; and the attack judgment module is used for carrying out 2-norm detection and maximum standardized residual detection on the residual vector, and judging whether the measurement information contains bad data or not based on comparison between a detection value and a threshold value. The method is crucial to safe and stable operation of a power system.

The invention relates to the field of automatic production, and discloses an air injection control system. An NARX neural network model 1 and an NARX neural network model 2 are used for controlling air flow and predicting an air flow value respectively, and an NARX neural network establishes a dynamic recursive network of the models by introducing a time delay module and outputting feedback, input and output vector delay feedback is introduced into network training to form a new input vector, good nonlinear mapping capability is achieved, the input of the NARX neural network not only comprises the error of original gas flow, the control quantity and the input data of actual gas flow, but also comprises the trained corresponding output data; the generalization ability of the network is improved, so that the system has better prediction precision and adaptive ability in prediction of corresponding parameters of the gas flow compared with a traditional static neural network, and the single-chip microcomputer controller improves the precision, robustness and reliability of the control system.

The invention discloses a neural network inverse control method for an SCR denitration system of a coal-fired unit. The method comprises the steps of establishment of an SCR denitration system mechanism model, training of an inverse model and application of a feedforward controller based on the inverse model; establishing a SCR denitration system mechanism model, including establishing a SCR denitration system model based on a Langmuir-Hinshelwod mechanism and an Eley-Rideal mechanism,; by analyzing field real data, a particle swarm or a genetic algorithm is used for optimizing and obtaining optimal parameters of a mechanism model, and the model can basically represent a real system; wherein the training of the inverse model comprises the steps of giving a group of random numbers to an input end based on a constructed accurate mechanism model of the SCR denitration system, setting simulation time as long as possible, and generating random output of the model; using the generated randomdata, and using an NARX neural network to train an inverse model; the application of the feedforward controller based on the inverse model comprises: the generated inverse model is combined with an original cascade PID control system of the coal-fired unit to form an SCR denitration inverse control system.

The invention discloses a logging curve reconstruction method based on a nonlinear autoregressive neural network model. The logging curve reconstruction method comprises the following steps: dividing an existing logging curve data acquisition depth into training curve data and test curve data; establishing an NARX neural network model according to the training curve data; performing initial optimization on the NARX neural network model through a particle swarm algorithm; substituting training curve data into the NARX neural network model by using a Levenberg-Marquardt algorithm so that training can be completed; substituting the test curve data into the NARX neural network model for testing; and using the tested NARX neural network model to obtain reconstructed logging curve data. According to the method, the problem of falling into local optimum is effectively avoided, a nonlinear logging curve reconstruction system can be approximated with high precision, the nonlinear and sequential characteristics of logging data are fully utilized, the corresponding relation between curves can be accurately reflected, and the method has a good logging curve reconstruction capability.

The invention provides an AC / DC power grid autonomous capability evaluation method based on an NARX neural network, and the method comprises the steps: extracting sub-item evaluation indexes of training AC / DC power grid data from the perspective of source grid load storage; determining an original evaluation matrix and a reference sequence according to the subitem evaluation indexes; determining an objective weight of each subitem evaluation index by using an entropy method; determining a comprehensive grey correlation degree of the training power grid data according to the original evaluationmatrix, the reference sequence and the objective weight; training an evaluation model of the NARX neural network according to the sub-item evaluation indexes and the comprehensive grey correlation degree; and based on the trained evaluation model of the NARX neural network, evaluating the autonomous capability level of the AC / DC power grid. Compared with an index feature selected by a traditionalmethod, the method is more global and higher in evaluation precision.

The invention discloses a conduction oil circulation fault diagnosis system, and belongs to the technical field of boiler fault diagnosis. According to the conduction oil circulation fault diagnosis system, a multi-modular monitoring method is adopted to refine a monitoring object, a monitoring module including multiple fault modes such as conduction oil leakage, conduction oil degradation, conduction oil overtemperature and overpressure, pipe wall scaling and the like is arranged, system operation states are recognized and divided, fault prediction is achieved according to different states, daily maintenance and fault processing of the system are achieved, the monitoring coverage is wide, the monitoring efficiency is high, and the system can be suitable for various fault conditions of the steam-conduction oil double-medium boiler system; and meanwhile, an NARX neural network training method is adopted for prediction, large sample data of the boiler system are effectively learned and processed, and prediction accuracy and efficiency are improved.

The invention discloses a method for dynamically predicting the drainage flow of an urban rainwater system drainage outlet. In step (1), the rainstorm and flood management model is used to simulate the rainfall-runoff, and the drainage flow process lines of multiple sets of drainage pipe network outlets are used as training samples. Step (2), set up RBF neural network and train, carry out the optimization of network hidden layer node and center width Spread in the training process; Step (3), set up NARX neural network and train; Step (4), will finish training The NARX neural network and the RBF neural network are coupled to obtain the coupling network, and then predict, calculate the mean square error of the coupling network and the sample, return the flow value with the smallest mean square error as the optimized coupling point, and randomly select the rainfall data to input the coupling network. Obtain the predicted drainage flow hydrograph. The invention organically combines the advantages and characteristics of different neural networks, the prediction result is in good agreement with SWMM simulation, the mean square error of the curve is 0.000458, and has good prediction accuracy.

The invention belongs to the technical field of internal combustion engine noise prediction, and particularly relates to an internal combustion engine noise prediction method based on VMD and NARX. In the internal combustion engine noise signal processing process, signal separation is achieved through the variational mode decomposition technology, signals with different frequency characteristics are obtained, the characteristics of the internal combustion engine noise signals are systematically analyzed, and all mode component signals are predicted through the NARX neural network to obtain signal characteristics at the non-occurrence moment. According to the method, the internal combustion engine noise signal stabilization processing flow is simplified, and the precision and timeliness are effectively improved; and the method combines an optimization algorithm, improves the prediction efficiency, has better applicability, and can more accurately predict the noise value of the internal combustion engine at the next moment.

The invention discloses a short-term ionospheric forecast method and device based on NARX. The method first obtains TEC grid data for a continuous period of time from historical TEC data files; Obtain the single-point TEC time series in this period of time by means of the method; then establish the NARX neural network model, use the TEC time series to train the NARX model, the input parameters include the TEC time series and the time series of external time parameters, and the output is the next moment The predicted value of TEC; finally, real-time prediction is performed according to the trained NARX model to obtain the predicted value of TEC in the future. Using the method proposed by the invention can improve the prediction accuracy of ionospheric TEC, and can improve the positioning accuracy when applied to GNSS positioning.

The invention relates to the field of automatic production, and discloses a shading control system, which is characterized in that an NARX neural network model 1, an NARX neural network model 2 and an NARX neural network model 3 are utilized to predict the displacement error, the control quantity and the actual displacement value of a lifting frame respectively, and a dynamic recursive network of the models is established by introducing a delay module and outputting feedback through an NARX neural network; input and output vector delay feedback is introduced into network training, a new input vector is formed, good non-linear mapping ability is achieved, input data including original lifting frame displacement errors, control quantity and actual displacement values and corresponding output data after training are input, the generalization ability of the network is improved. Compared with a traditional static neural network, the single-chip microcomputer controller has better prediction precision and adaptive capacity in the prediction of the corresponding parameters of the lifting frame, and the precision, robustness and reliability of the control system are improved through the single-chip microcomputer controller.