193 results about "Back propagation neural network" patented technology

The invention relates to a method and a system for detecting and defending multichannel network intrusion, wherein a half-polling manner is used in a network card driver at a network side to capture a data packet and from a rule for judging abnormal traffic. At an operating system side, secure access control is realized by additionally setting authority control in a Capability module of Linux, meanwhile, monitoring and control treatment is performed on a kernel layer of an operating system, and the monitoring of Trojan and other abnormal operations or virus destruction is realized by carrying out credential privilege arbitration on processes, operating an i-node and carrying out secure control on an application layer. Formed feature data are gathered and sent to a Bayes model for classification, an improved backward propagation neural network (BPNN) is guided in to carry out data training so as to make the produced rule capable of defending corresponding attacks. The system disclosed by the invention comprises three modules and five submodules, wherein the three modules include a network data packet processing module, an operating system layer detecting module ad a detection center module.

The invention discloses a combination forecast modeling method of wind farm power by using gray correlation analysis, belonging to the technical field of wind power generation modeling. In particular, the invention is related to a weighted combination forecast method of wind power based on a least square support vector machine and an error back propagation neural network. The forecast method comprises that forecasted values of wind speed and wind direction are acquired in advance from meteorological departments while real-time output power is acquired from a wind farm data acquiring system; that the forecasted values of wind speed and wind direction and the real-time output power are inputted into a data processing module for data analyzing extraction and data normalization, and then normalized data is loaded to a database server; processed data in the database server is extracted by a combination forecast algorithm server to carry out model training and power forecast, and the wind farm sends running data to the data processing module in real time to realize rolling forecasting. The method of the invention achieves the goal of combination forecast of wind farm output in a short time. The method not only maximally utilizes advantages of two algorithms but also increases forecast efficiency by saving computing resources and shortening computing time.

The invention discloses an on-line transmission line lightning shielding failure trip early-warning method. The method comprises the following steps of: performing statistics on historical lightning shielding failure trip information to obtain transmission line lightning shielding failure trip probability distribution by a two-dimensional information diffusion theory and a conditional probability method; selecting radar forecast data, such as echo intensity, echo tops and vertical accumulated liquid water content; establishing a lightning current magnitude prediction model based on a back propagation neural network; and sending real-time early warning and an early warning grade of transmission line lightning shielding failure trip probability according to the predicted lightning current magnitude and the side distance to lightning stroke and by virtue of the transmission line lightning shielding failure trip probability distribution model. According to the real-time forecast data of a meteorological radar, the method provided by the invention can predict the trip probability of a transmission line and send an early warning signal, thereby providing reference for decision-making analysis of grid dispatching operators, making a transmission line dispatching strategy in time, improving power supply reliability, lowering economic loss of a grid, and improving the reliable running ability of the grid.

The invention relates to a single-channel music singing separation method based on a deep relief network. The method comprises the steps that firstly, the deep relief network (DBN) is utilized for extracting high-rise abstract features facilitating accompany and singing separation, then the Back-Propagation neural network is utilized for separating accompany and singing features, and finally an overlap-add method is utilized for obtaining time-domain accompany and singing signals. According to the single-channel music singing separation method, fusion music to be separated is divided into short segments, and therefore the defect of the DBN for processing music signals of different time scales is made up; meanwhile, the capacity of the DBN for rapidly extracting the high-rise abstract features is utilized for extracting the high-rise abstract features facilitating accompany and singing separation, and finally due to the facts that the single-channel music signals belong to high-dimensional data, and the neural network has the specific processing capacity according to the problem of processing high-dimensional input and high-dimensional output, the BP neural network is selected to be used as the final singing and accompany separator. The method is easy and flexible to implement and has high practicability.

The invention discloses an artery coordination signal control method based on dynamic O-D matrix estimation. According to the method, road segment flow of the detecting of import approaches and export approaches of crossings of an artery is adopted, dynamic O-D matrixes of crossings are estimated by the adoption of a Kalman filtering and a back propagation neural network algorithm, a Bayes combination method is designed to improve accuracy and stability of the estimated result, a single-crossing multi-target signal control model is built on the basis of the estimated result, and the calculated maximum of crossing signal cycles is served as a public cycle of the artery. The artery coordination signal control method that the maximum of the obstructing-free rate of artery vehicles serves as the target function is further designed, the green split of all crossing artery directions and the phase difference of the adjacent crossings are acquired through solving, and an artery coordination signal control scheme is formed. The artery coordination signal control method based on dynamic O-D matrix estimation has the advantages that on the premise of guaranteeing the optimizing passing of the artery vehicles, the passing efficiency of all the single crossings is balanced, the problem that the control scheme can not be adjusted timely according to the traffic flow in the prior art is solved, advantages of being high in accuracy, online in application and the like are achieved.

The invention relates to the technical field of network security evaluation, in particular to a network security situation prediction method based on a combination of the chaos theory and a neural network. The method comprises the following steps: carrying out processing of normalized network security situation value sequence sets through the mutual information method and the cao method to obtain the optimum embedded dimensions of network security situation sample values, carrying out phase-space reconstruction, and analyzing the maximum Lyapunov exponent of reconstructed samples to determine whether the evaluated samples have chaos predictability or not; determining the numbers of nodes of an output layer and a hidden layer of a BPNN according to characteristics of a nonlinear time sequence and experience; carrying out parameter optimization through an improved firefly algorithm, so as to determine network weights and offset values and establish a network security situation prediction model; and inputting test set samples into the BP neutral network for prediction, and carrying out denormalization of obtained prediction values. The method provided by the invention has the advantages that a network security situation can be more precisely predicted, and the network security situation prediction convergence rate can be increased.

The invention discloses a server resource prediction method and device, computer equipment and a storage medium. The method comprises the following steps: taking a user selected in a user list as a target user, and acquiring historical performance parameters correspondingly consumed by the target user by using a cloud server in a preset historical time period according to a preset acquisition period, so as to obtain a historical performance parameter set corresponding to the target user; performing model training on the to-be-trained back propagation neural network according to the historicalperformance parameter set to obtain a back propagation neural network for predicting performance parameter values; correspondingly obtaining a current input sequence according to the historical performance parameter set and the received to-be-predicted time point; and inputting the current input sequence into the back propagation neural network to obtain a predicted value corresponding to the to-be-predicted time point. According to the method, a server resource prediction model is established by using historical data, and the future usage amount of cloud server resources is predicted.

The invention discloses a method for dynamically fusing, counting and forecasting air quality based on dynamic and thermal factors. The method comprises the following steps of: collecting data; introducing dynamic and thermal influence factors; performing empirical orthogonal decomposition of a vector matrix composed of influence factors having the significance level alpha, which is equal to 0.01, and selecting a principal component, the cumulative variance contribution of which is beyond 98%; establishing a regression equation by using the principal component; establishing a neural network model by utilizing a back-propagation neural network algorithm; performing evaluation check of a fitting result of the regression equation and the neural network model and the historical forecasting accuracy; calculating the final fusing and forecasting result by using a weighted average algorithm; performing evaluation check of the accuracy of the final fusing and forecasting result; and, adding new data into a historical data set in real time, and dynamically updating a forecasting model according to a check evaluation result. Compared with the existing method, the method disclosed by the invention has the advantages that: the relative error of various pollutant concentration forecasts is reduced by 3-11%; and the level forecasting accuracy rate is increased by 4-8%.

The invention discloses a combined estimation method for a road junction dynamic steering proportion based on Bayes weighting. According to the method, three sub algorithms of an improved Kalman filtering algorithm, an improved back-propagation neural network algorithm and a genetic algorithm are designed to solve the road junction dynamic steering proportion by utilizing road segment traffic detected by all inlet roads and outlet roads of road junctions, historical data are combined based on the road junction dynamic steering proportion, correction on historical and current estimation deviation is considered comprehensively, calibration is carried out by utilizing a Bayes formula and weight is updated dynamically, and obtained results through the three sub algorithms are weighted to obtain the dynamic steering proportion estimated by the combined method. Aiming at different traffic flow situations, the dynamic steering proportions estimated by existing methods all have advantages and disadvantages in the aspects of precision and efficiency, the combined estimation method can embody the advantages of all the methods on the whole, local oversize deviation is avoided, the combined estimation method has the advantages of being strong in adaptability, high in precision, good in stability and optimal in entirety, and can provide basic data supporting for signal control and other real-time traffic management and information service systems.

The invention relates to a wind turbine generator short-term reliability prediction method considering the operating state, which comprises the steps of acquiring state parameters of a wind turbine generator through a state monitoring and data acquisition system, classifying the state monitoring parameters into two categories; establishing a back-propagation neural network based state parameter prediction model in allusion to equipment temperature parameters, calculating the protective action probability based on prediction residual distribution characteristics, and calculating the protective action probability according to the out-of-limit time in allusion to other parameters; carrying out statistics on the outage times and the speed at an outage moment of the wind turbine generator according to wind power station operation and maintenance data and SCADA (Supervisory Control And Data Acquisition) data, and building a wind speed dependent wind turbine generator statistical outage model; and calculating the wind turbine generator short-term outage probability by combining outage statistical information and state parameter out-of-limit information. The method provided by the invention can accurately evaluate the outage risk of a wind turbine generator in a short term, the accuracy of a wind turbine generator short-term outage model is greatly improved, and technical reference is provided for short-term reliability evaluation and safe and economical operation of the whole wind power station.

The invention, which belongs to the battery thermal management system technology field, relates to an artificial neural network-based highest surface temperature prediction method of a secondary battery. A secondary battery is placed in a high and low temperature test box and a charge and discharge testing machine is connected; the battery is discharged and then charging is carried out; a changing situation of highest surface temperatures of the battery during the charging process is monitored; input, output, the neuron number, the number of layers, a transfer function, and a training algorithm of a Back-Propagation neural network model are set so as to complete construction of the model; those data are used for model training, so that the model can be applied to prediction; and highest surface temperatures of the battery during charging processes under other environmental temperatures can be predicted by the model. According to the invention, the above-mentioned model can be applied simply; parameters are easy to control; and results have practical values; because highest surface temperatures of the battery during working processes under different environmental temperatures can be predicted, guarantees are provided for effective work of a battery thermal management system and safety of the battery.

The invention discloses a resistance spot welding nugget nucleation dynamic quality nondestructive testing method which comprises the following steps of: acquiring the structure load acoustic emission signal in the resistance spot welding process, and drawing a dynamic curve chart; extracting the acoustic emission signals of the electrode loading stage, nugget nucleation stage and crack generation stage from the dynamic curve; counting the ringing number and total energy of the acoustic emission signals respectively; splitting the resistance spot welding sample nugget, and measuring the diameter, height and crack length of the actual nugget; establishing a sample pair to form a training set; establishing a nugget size artificial neural network and a crack length artificial neural network respectively, and training the obtained sample by use of the Back Propagation neural network algorithm; and applying the trained model to the real-time detection of the nugget size and the crack generation. According to the invention, online testing of the quality of multiple common resistance spot welding nuggets of metal structure material can be realized relatively accurately.

The invention provides a low-cost calibration method for PM2.5 monitoring nodes. The method includes the following steps that the nodes are deployed near an air quality inspection station, and training samples consistent in time and space are obtained; models are built to show the relationship between a number read at each node and a PM 2.5 true value; training data is preprocessed, wherein parts of characteristics are standardized and a training sample set and a testing sample set are determined by means of a set-aside method; for a linear invariant model, a three-layered back-propagation neural network is adopted to train a multiple linear regression module on the training sample set, and verification of accuracy of the models is completed on the testing sample set; for a linear variable model, in the time interval, the training samples are fitted simply by means of the least square method to obtain a linear parameter, in different time periods, linear parameters, average values of node readings and average values of the sensitive characteristics data serve as new training samples, post-pruning-strategy-based CART regression tree training is adopted on the new training samples, and verification of reliability of the models is completed on the testing sample set; an off-line model which is verified to be accurate is written to a node program.

The invention discloses an oilfield pumping unit oil pumping energy saving and production increasing optimization method based on the back propagation neural network (BPNN) and the strength Pareto evolutionary algorithm 2 (SPEA2). The method is characterized by including the following steps: step 1, calculating decision variables X; step 2, collecting samples of power consumption and samples of oil production Y to acquire a sample matrix; step 3, building a process model of oil pumping of a pumping unit; step 4, optimizing each decision variable in the range of an upper limit and a lower limit of each decision variable by using the SPEA2 based on a BPNN model; step 5, guiding actual production if the power consumption is reduced and the oil production is improved, and if not, returning the process to the step 1, changing S1 decision variables X on purpose and screening the decision variables X again; and step 6, assigning S1+1 to the S1, and returning the process to the step 1 if the combination of the set S1 decision variables X can not enable the power consumption to be reduced and the oil production to be improved. The oilfield pumping unit oil pumping energy saving and production increasing optimization method based on the BPNN and the SPEA2 has the advantages that an optimal value of technological parameters can be determined, and actual production guiding can be carried out according to the optimized technological parameter optimal value.

The invention provides a multiple graphics processing unit (GPU)s-based back-propagation neural network (BPNN) training method and apparatus. The method comprises the following steps: S1, controlling all GPUs to carry out BPNN forward calculation and synchronizing forward calculation outputs among all GPUs; S2, controlling all GPUs to carry out BPNN backward error calculation and synchronizing backward error calculation outputs among all GPUs; S3, controlling all GPUs to update the weight of the BPNN according to the forward calculation outputs obtained by synchronization and backward error calculation outputs obtained by synchronization. According to the invention, data synchronization costs of multiple GPUs during the BPNN training can be lowered; and the BPNN training efficiency of the multiple GPUs can be improved.

The invention discloses a big data-based electrical load prediction system. According to the system, a prediction is carried out on the electric load capacity of one place in the future by an error back-propagation neural network method; the error back-propagation neural network method comprises an input layer, a hidden layer and an output layer; a mathematic model of the error back-propagation neural network method can be established through a function relationship among the input layer, the hidden layer and the output layer; input layer information corresponds to historical data of the electric load capacity; output layer information corresponds to predicted data of the electric load capacity; and the hidden layer corresponds to the function relationship between the predicted data and the historical data. According to the big data-based electrical load prediction system, the prediction is carried out the electric load capacity in the future by the error back-propagation neural network method; and the mathematic model of the error back-propagation neural network method is established, so that the prediction result is relatively accurate; the error can be accurately calculated; and a great convenience is brought to life.

The invention belongs to the technical field of logistics express delivery business volume prediction, discloses a daily short-term express delivery business volume prediction method for logistics enterprises, uses a particle swarm optimization algorithm with an improved inertia weight to optimize the back propagation neural network, and adopts a new horizontal data selection method to select andinput data for the BP neural network. The optimized BP neural network is used to predict the daily short-term express delivery business volume of the logistics company, a proper amount of cloud computing resources can be dynamically applied in different time periods to process the express parcel data and monitor the parcel transportation process. The method can predict the daily short-term expressdelivery business volume, apply for a proper amount of cloud resources, and can process the business data of all express parcels on time without causing waste of cloudy resources. The method can be applied to the daily short-term express delivery business volume forecast of the logistics company, and is of great significance for reducing the cost of the logistics company and improving the qualityof the user service.

The invention belongs to the technical field of wireless communication and indoor positioning and discloses a 3D indoor positioning method based on a spectral clustering and weighted back-propagationneural network. The 3D indoor positioning method is divided into an offline phase and an online phase, wherein the offline phase comprises the steps of dividing a reference point into NC clusters by using spectral clustering, and training a back-propagation neural network model by using reception signal intensity and corresponding position information in each cluster; and the online phase comprises the steps of estimating the position of a to-be-detected point by adopting a weighted back-propagation neural network (BPNN) algorithm, determining a weight of fingerprint of the to-be-detected point in each cluster, obtaining NC coordinates by utilizing the NC trained BPNN models, and performing weighted estimation on the position of the o-be-detected point by using the NC coordinates. According to the 3D indoor positioning method disclosed by the invention, the equipment complexity and layout cost are reduced, high positioning accuracy and positioning stability are provided, and the training time is shortened.

The invention discloses an on-line condition process monitoring method for a plastic injection moulding process, and belongs to the industrial monitoring and fault diagnosis field. The on-line condition process monitoring method comprises the steps: S1 utilizing a sensor to collect the data under various conditions, and forming a training sample set X for modeling; S2 performing data pre processing and normalization to enable the mean value of the training sample set X to be 0 and enable the variance to be 1, and then obtaining a matrix X'; S3 according to the matrix X', applying Gaussian kernel function to calculate and obtain a distance matrix W; S4 standardizing the distance matrix W, obtaining a Markov matrix P (1), obtaining P(t) by making the P(1) to migrate t times, and performing spectral decomposition of the obtained characteristic matrix X' based on the P(t); S5 inputting the characteristic matrix X' and the condition Tq corresponding to various samples into an error back propagation neural network in pairs to receive training, and preserving the neural network model with the highest prediction accuracy as the model for monitoring; and S6 performing practical monitoring. The on-line condition process monitoring method successively realizes on-line monitoring of high dimension data.

The invention discloses a program scoring system based on EEG (Electroencephalogram) emotion recognition, comprising an EEG signal acquisition module, an EEG signal pre-processing module, an EEG signal analysis module, an emotion recognition analysis module, a program effect analysis module and a program scoring module which are connected in order, wherein EEG signals of audiences are acquired bythe EEG signal acquisition module when the audiences are watching different slices and after completing watching for a period of time, then filtering operation is performed through the EEG signal pre-processing module, and feature extraction of the EEG signals is achieved through wavelet packet decomposition; after that, emotion recognition is performed on the extracted feature by the emotion recognition analysis module through a back propagation neural network; the program effect analysis module is used for monitoring duration of emotion influences of the audiences; and finally, scoring operation is performed on the corresponding program by the program scoring module. Through adoption of the program scoring system of the invention, actual influence of the program on the audiences can be analyzed and calculated from the EEG signals of the audiences, so as to perform scoring, therefore, outside interference can be eliminated, and the most actual effect of the program can be reflected.

The invention discloses a metallurgical gas integrated dynamic balance scheduling system and method, and belongs to the technical field of multiple gases balance scheduling control. Based on real-time state data and production plan data, and on the basis of a combined model of a K-Nearest Neighbor(KNN) algorithm model, a back-propagation neural network algorithm model, a RBF neural network learning algorithm model, prediction is conducted on gas production/consumption value, gas tank place value, and pipe network fluctuating value in a future time period, and corresponding prediction values are obtained, and in accordance with a gas balance scheduling expert database, dynamic balance scheduling on the gas is completed in combination with the aforementioned prediction values. The system and the method of the invention are advantaged in that on the basis of the completion of the gas balance scheduling, functions of online adding algorithm model are expanded, and functions of online correcting balance scheduling expert database is expanded, and in cases that stable and safe operation of a gas system, real-time correction of system defects are achieved, so that gas can be more reasonably utilized and usage efficiency of gas is provided.

The embodiment of the invention provides a software quality prediction method, a device, a terminal and a computer-readable storage medium. The method comprises the following steps: acquiring characteristic data of a plurality of categories of software to be predicted in a software development flow; preprocessing the characteristic data of each kind and inputting the same into the quality prediction model. Quality prediction model is based on back propagation neural network and multi-class historical data. The quality prediction model processes all kinds of characteristic data and outputs theprediction results of the software to be predicted. The embodiment of the invention synthesizes the characteristic data of each category in the software development flow according to the related indexes affecting the software quality, and carries out the software quality prediction quickly and comprehensively, so as to ensure the prediction accuracy.

The invention discloses a face complex expression recognition method based on a neural network. On the basis of a general back-propagation neural network, a face recognition technology combining a method based on face feature sub-graph extraction and a method based on biometric parameter extraction is employed. In the method based on face feature sub-graph extraction, four sub-images of the eyes, nose and mouth are extracted and fed to one general back-propagation neural network. In the method based on biometric parameter extraction, seven measured distances between the facial feature points are fed to the other general back-propagation neural network. The network used in the method based on face feature sub-graph extraction is selected as a main neural network, while the network used in the method based on biometric parameter extraction serves as an auxiliary neural network. If a main classifier fails to recognize a complex expression, an auxiliary classifier is used instead.