46 results about "Subtractive clustering" patented technology

The invention discloses a super-short-term prediction method of photovoltaic power station irradiance. The method includes the steps that irradiance data are extracted from a history database, data of a night time quantum are removed, corresponding extraterrestrial theoretical irradiance is calculated, data abnormal detection is carried out based on the preceding operations, and the data are normalized in the difference value ratio method of an extraterrestrial irradiance theoretical value and practical irradiance; a training sample set is extracted according to input and output dimensionality of a model; a model of an irradiance time sequence is built through an ANFIS, a the rule quantity and an initial parameter of the ANFIS model are determined in a subtractive clustering method, and a fuzzy model parameter is optimized in a counter propagation algorithm and a least square method; a prediction sample is input, and a prediction value is obtained through calculation; the prediction value is added to form a new sample set, and multiple steps of prediction are achieved in a cycling mode; counter normalization processing is carried out on the prediction value. Super-short-term prediction of the irradiance can be achieved only by means of a history irradiance time sequence, prediction accuracy is good and the method is easy to carry out.

A data analysis system and / or method, e.g., a data mining / data exploration method, using subtractive clustering is used to explore the similarities and differences between two or more multi-dimensional data sets, e.g., generated using a flow cytometer, an image analysis system, gene expression, or protein microarrays.

The invention provides a method for forecasting a hybrid neural network and recognizing scenic spot meteorological elements. The method includes the steps of firstly, collecting and conducting normalization processing on data banks of meteorological stations; secondly, determining the number of RBF network hidden nodes established by the main meteorological elements of the meteorological stations through a subtractive clustering algorithm according to the data banks of the n meteorological stations; thirdly, obtaining RBF network model parameters of the m meteorological elements established by the n meteorological stations respectively through chaotic particle swarm optimization algorithm; fourthly, forecasting future meteorological element values of an assigned number of days of the n meteorological stations through optimum RBF network prediction models of the elements obtained by the n meteorological stations; fifthly, conducting autoregression adjustment on soft factor information of a certain scenic spot according to the n meteorological elements and forecasting the meteorological element values of the scenic spot; sixthly, establishing an ART2 network to recognize and record weather phenomena of the scenic spot. The method has the advantages that the hybrid neural network prediction models have good generalization performance, are high in accuracy for forecasting the weather in the scenic spot and have application value.

The invention provides a generalized predicative control method of a high-speed train based on an adaptive network-based fuzzy inference system (ANFIS) model. The method utilizes a data-driven modeling method to build the ANFIS model in a running process of the high-speed train according to acquired high-speed train running data; adopts subtractive clustering to determine rule number and initial parameters of a fuzzy model, and adopts a back-propagation algorithm and a least square method to optimize parameters of the fuzzy model. The predictive tracking control method of electric multiple unit running speed on the basis of the ANFIS model obtains accurate controlled quantity through multistep predication and circular rolling so as to change blindness of adjustment by experience, enables the high-speed train running speed to track a target curve accurately, solves the problem of large lag, achieves on-schedule, safe and effective running of the train, and guarantees safety of passengers. The method is simple, practical, capable of achieving automatic drive control of the high-speed train and suitable for on-line monitoring and automatic control of a running process of the high-speed train.

The invention relates to a micro-grid load prediction system and method based on electricity purchased on-line and a dynamic correlation factor. The system includes an electric quantity purchased on line module, a load characteristic analysis module, a short period load prediction module and a prediction result output module. The method comprises the steps: pushing an initially-drafted order of electric quantity and a reference electricity price to an electric energy user by a micro-grid; correcting the initially-drafted order of electric quantity, and feeding back the corrected order of electric quantity to the micro-grid by the user; counting the statistical values of electric quantity purchased on line and the historical load data for various load users, determining the load type of the micro-grid and the correlation factor of the load type; establishing an RBF neural network mathematic model; utilizing a subtractive clustering K-means optimization algorithm based on the input data and the output data to acquire initial network parameters of the RBF neural network mathematic model; utilizing a quantum particle swarm optimization algorithm to optimize the initial network parameters; calculating the final predicted values of various loads of the micro-grid and the final predicted value of the total load; and outputting the final predicted values of various loads of the micro-grid and the final predicted value of the total load of the micro-grid.

The invention discloses an intelligent graphic retrieval method, which is characterized in that: extracting features of graphics to generate feature set by the method of Fourier change, training RBF neural network classification model by taking one part of the feature set as training set, indexing the graphics using the classification result given by the classification model; client of a retrieval system extracts the features of retrieval graphics, gives a category by the trained classification model and computes the similarity distance between the retrieval graphics and each graphics in the feature set of the same category; sorting the similarity distance, returning to the graphics according to the number made by the system and further revising RBF neural network classification model by relevant feedback methods. The invention improves the intelligence of search process, effectively determines the RBF neural network classification model by improved algorithm of subtractive clustering, greatly improves retrieval precision, speeds up retrieval speed and upgrades retrieval performance.

The invention discloses a method for building adaptive soft sensor. The method comprises the following steps. The input and schedule vectors are constructed, and a novel learning algorithm that uses online subtractive clustering is used to recursively update the structure and parameters of a local model network. Three rules are proposed for updating centers and local model coefficients of existing clusters, for generating new clusters and new models as well as for merging existing clusters and their corresponding models. Once verified, the online inferential model can be created to generate the predicted value of process. Thus, it does not need much memory space to process the method and can be easily applied to any other machine.

The present invention discloses a global illumination real-time rendering method based on radial basis function neural network fitting, and belongs to the field of realistic graphics real-time rendering. The method comprises training-data acquisition, neural network construction, neural network training, indirect illumination value fitting, direct illumination value computation and global illumination rendering. An offline rendering mode is used to perform pre-computation to obtain training data; a center of a radial basis function is determined by using a subtractive clustering method; a supervised learning method is used to perform training; and a nonlinear relationship between a viewpoint position, a light source position, an object surface normal vector in a scene and the like in indirect illumination and an indirect illumination value is fitted to replace a traditional global illumination model to complete indirect illumination computation, thereby avoiding a plurality of intersection of lights. According to the global illumination real-time rendering method based on radial basis function neural network fitting provided by the present invention, redundant data can be effectively reduced; convergence occurs at a higher speed; the offline rendering effect is well fit; and global illumination real-time rendering is completed.

The invention discloses a real-time prediction system and method for a boiler entering heat value of a circulating fluidized bed household garbage incineration boiler. By means of hidden knowledge in operation historical data and an operational mechanism of the circulating fluidized bed household garbage incineration boiler and through an integrated modeling method of a particle swarm optimization (PSO) algorithm, a subtractive clustering algorithm and an adaptive neuro-fuzzy inference system (ANFIS) algorithm, the system and the method which are fast and economical are established for real-time prediction of the heat value of boiler entering garbage, the complicated work of off-line measurement for garbage components is avoided, a new path is provided for boiler operation operators and relevant management staff of a power plant to judge the heat value of the boiler, and meanwhile, a heat value judgment signal can be provided for an automatic control system of the power plant.

The invention discloses a method and system for predicting the bed temperature of a circulating fluidized bed municipal solid waste incineration boiler. Based on tacit knowledge in operation mechanisms and operation history data of the circulating fluidized bed municipal solid waste incineration boiler, a modeling method is integrated by a Gamma Test algorithm, a PSO (particle swarm optimization) algorithm, a subtractive clustering algorithm and an ANFIS (adaptive neuro-fuzzy inference system) algorithm for predicting a bed temperature of the boiler in real time to avoid from complex mechanism modeling work. The whole modeling process is clear in logic, few in set parameters, high in modeling automation degree and easy to master and popularize. Meanwhile, an excellently trained ANFIS bed temperature diction model can serve control algorithms based on the model to provide help for ACC (automatic combustion control) system actuation of the circulating fluidized bed municipal solid waste incineration boiler.

The invention discloses a short-term power load prediction method based on similar days and an RBF neural network, and relates to the field of power load prediction, a proper similar day is selected as a training sample in short-term power load prediction, so that the training process can be simplified and the prediction precision can be improved; and in order to reduce the influence of summer accumulated temperature effect on similar day selection, the similarity of the temperature and other load influence factors is calculated respectively, so that a similar day is selected according to thecalculated comprehensive similarity; besides, the prediction effect of the RBF neural network is improved; the training samples are clustered by using subtractive clustering, the initial value of fuzzy c-means clustering is set according to the clustering result, the hidden layer parameters of the RBF neural network are optimized by applying fuzzy c-means clustering, and finally short-term power load prediction is carried out by combining the similar day and the improved RBF neural network, so that the accuracy of short-term power load prediction can be obviously improved.

An intelligent auxiliary medical treatment decision supporting method of a two-stage mixed model mainly comprises the following steps that based on a subtractive clustering method, real medical treatment sample data information is subjected to weighting preprocessing, and a nonlinear indivisible feature space is converted into a divisible linear data feature space; based on an efficient extremity learning machine model, weighting feature data obtained by preprocessing are used, and a medical treatment decision supporting model is established by data self learning; and based on the established learning model, real medical treatment data are used, a study object is subjected to classification forecasting, and accordingly the fact that reliable efficient auxiliary forecasting support is provided for a target user is achieved. The intelligent auxiliary medical treatment decision supporting method has the main advantages that the method is simple and easy to realize, and forecasting results with high accuracy can be generated; the number of parameters involved in the method is small, parameter influence is low, namely needed human intervention is low, and operation is convenient; and the method is high in computing speed and efficiency.

The invention designs a magneto-rheological damper hybrid modeling method. The method mainly comprises the four steps of: performing mechanical property experiment on a magneto-rheological damper, analyzing the power indication and speed characteristics, and obtaining the original experiment data of the piston displacement, the speed, the current and the damping force; performing numerical filtering and normalization processing on the original experimental data, and creating an input and output sample set; modeling the magneto-rheological damper by adopting a self-adaptive neural fuzzy reasoning system, and introducing a subtractive clustering technology to construct a system rule base; and optimizing the clustering parameters by using a genetic algorithm, establishing a magneto-rheological damper model, and analyzing the model precision, the hysteresis characteristic and the like. According to the method, a subtractive clustering technology is introduced to construct a rule base of a magneto-rheological damper model, a genetic algorithm is utilized to optimize clustering parameters, and the optimal fuzzy rule quantity and structure for depicting the system behavior are obtained, so that the modeling accuracy is effectively improved, the hysteresis characteristic and the low-speed region behavior of the system are better described, and the development and application of a semi-active control technology are promoted.

The invention provides a software-based measuring method for the content of elements obtained by rare-earth cascaded extraction separation, including the steps of data collecting and preprocessing, building software-based measuring model and correcting model, etc., and characterized in: as building the software measuring model, classifying the filtered sample data by subtraction clustering algorithm; finding model parameters by genetic algorithm to obtain a forecasting model of the content of elements. The invention also provides a software-based system for implementing the soft measuring method, comprising main program module, algorithm module, database and four interfaces and using the model station computer of an intelligent control system for rare-earth cascaded extraction separation as a hardware platform. The software-based measuring method makes data collection, pretreatment and modeling on extraction production of the elements La, Ce and Pr and the forecasting of the content of elements in the products shows that the results are all within the preset error range; and the forecasting of the content of elements is successfully made on the Yi production line by the software-based measuring model.

The invention discloses a communication signal recognition method based on a subtractive clustering algorithm and a fuzzy clustering algorithm. The method comprises the following steps: initializing parameters; setting different initial neighborhood radius values for different subcarriers, and clustering constellation points of a received communication signal by the subtractive clustering algorithm; when the number of subtractive clustering centers is less than a first preset threshold, reducing a neighborhood radius and continually performing subtractive clustering; clustering the constellation points of the communication signal again by the fuzzy clustering algorithm in a way of taking a first preset threshold number of subtractive clustering centers with a higher density in the subtractive clustering centers as an initial center; specifying the initial clustering number of fuzzy clustering, evaluating the rationality of clustering in conjunction with a Xie-Beni index and a relativeradius of a clustered constellation diagram, and iterating the initial clustering number if the clustering is irrational; and comparing a relative radius with the radius of a standard constellation diagram to get the modulation way of the signal, so that a type to which a standard modulation signal belongs is the type of the communication signal.

The invention discloses a signal identification method and system based on constellation graph reconstruction by clustering and a particle swarm. A data point of collected data is obtained by using a subtraction clustering algorithm; a clustering center is optimized by using a particle swarm algorithm; and then a reconstructed constellation graph is obtained. Distances between all constellation data points and original data points are calculated; all constellation data points are divided to circles with different radiuses; and a ratio of a largest radius to a smallest radius is compared with a characteristic range of a standard constellation graph circle radius to identify an MQAM signal. Compared with the traditional subtraction clustering algorithm, the anti-interference capability is high; the signal identification rate is high on the condition of the low signal to noise ratio; and the method and system can be applied to the practice well.

The invention discloses a circuit breaker opening / closing coil current signal identification method comprising the steps as follows: collecting an opening / closing coil current signal under normal operation of a circuit breaker operating mechanism, and extracting a feature data set; establishing a benchmark mode library of the circuit breaker opening / closing coil current signal; calculating the degree of fitting between a sample to be identified and the clustering centers in the benchmark mode library, and finding out the clustering center with the maximum degree of fitting; and identifying the operation rate of the circuit breaker operating mechanism according to the residual ratio, and realizing early warning for failure. According to the invention, a benchmark mode library is established for the circuit breaker opening / closing coil current signal based on a subtractive clustering algorithm, and when the circuit breaker is in an operation or maintenance state, the circuit breaker opening / closing coil current signal is compared with the benchmark mode library online to identify the operation state of the circuit breaker. Therefore, effective early warning for failure is realized, the need for fast, safe and efficient circuit breaker detection is satisfied, the shortage of the existing circuit breaker mechanical characteristic test method is solved effectively, and safe and stable operation of circuit breakers is guaranteed.

The invention discloses a green pepper greenhouse environment intelligent monitoring system based on a subtractive clustering classifier. The intelligent monitoring system is characterized by comprising a green pepper greenhouse environment parameter detection platform based on a wireless sensor network and a green pepper greenhouse yield intelligent early warning system. In the prior art, only adevice is adopted for monitoring the green pepper greenhouse environment parameter, and green pepper greenhouse yield cannot be warned according to green pepper greenhouse environment temperature andsunlight. The green pepper greenhouse environment intelligent monitoring system solve the above problem.

The invention discloses a neural-network learning algorithm based on particle swarm optimization. Firstly, data are acquired to be used as training sample data of an RBF (radial-basis function) neural-network; subtractive clustering is carried out on the training sample data, and the number of basis-function centers is determined; a particle swarm is initialized; fitness of each particle in the particle swarm is processed, the current fitness of the particle is compared with historical optimal fitness, and pid is updated if the current fitness of the particle is better; the fitness of each particle is compared with fitness of a best position experienced by the population, and pgd is updated if the fitness of the particle is better; velocity and positions of particles are repeatedly adjusted until a requirement is met; values of a best position experienced by the entire population are used as parameters of the RBF neural-network after decoding, and training, accuracy verification and prediction of the neural network are carried out; and operations are stopped. The method has the advantage of enabling performance of the RBF neural-network to be better through introducing a particle swarm optimization algorithm of neural-network improvement.

The invention discloses a method for building adaptive soft sensor. The method comprises the following steps. The input and schedule vectors are constructed, and a novel learning algorithm that uses online subtractive clustering is used to recursively update the structure and parameters of a local model network. Three rules are proposed for updating centers and local model coefficients of existing clusters, for generating new clusters and new models as well as for merging existing clusters and their corresponding models. Once verified, the online inferential model can be created to generate the predicted value of process. Thus, it does not need much memory space to process the method and can be easily applied to any other machine.

The invention discloses a subtractive clustering based rapid image segmentation method. The method comprises the following steps: firstly, normalizing all pixel points to a hypercube, and carrying out equal-interval uniform sampling and then carrying out restructuring on all the pixels to be clustered; and in the restructured pixels, calculating a density weight matrix and an inverse matrix thereof between every two sampling pixel points and density weight matrixes among sampling pixels and the rest of non-sampling pixels; then, calculating a density weight matrix between every two approximate non-sampling pixel points, and calculating the density values of all the pixels; and finally, calculating the maximum density value of all the pixels and obtaining a clustering center, and for finding out a new clustering center, necessarily carrying out attenuation on the density value of each pixel point, carrying out incremental iteration on the process, and stopping iterating according to termination conditions. Compared with the classical subtractive clustering method, the method disclosed by the invention greatly improves the real-time property of a subtractive clustering method for large-scale data sets under the condition of not affecting clustering results.

The invention relates to the technical field of industrial control anomaly detection, in particular to an RBF improvement method, device and equipment based on industrial control anomaly detection, and the method comprises the steps: collecting network data of an industrial control system, carrying out the preprocessing of the network data to obtain sample network data, determining a clustering center of the hidden node in the sample network data based on a declustering algorithm, determining an expansion constant of the hidden node according to the clustering center, further determining an output weight of the hidden node based on a grey wolf algorithm, and determining an improved RBF optimization model according to the clustering center, the expansion constant and the output weight. According to the technical scheme provided by the invention, the clustering center, the expansion constant, the output weight and other network parameters of the RBF model are optimized through the subtractive clustering algorithm and the grey wolf optimization algorithm, the minimum value is avoided, the operation efficiency and the classification accuracy are improved, and the method, device and equipment are suitable for a high-dimensional redundant industrial control data set. Whether the network behavior of the industrial control system is abnormal or not can be quickly judged, and losses caused by network attacks are avoided.

The invention discloses a variable static pressure adaptive fuzzy control method of a variable air volume air conditioner system. Before variable static pressure fuzzy control is achieved, through Matlab / Simulink software, the variable air volume air conditioner air supply system is subjected to analogue simulation, and data are collected; clustering subtraction is used for estimating the number of clusters and the cluster center position in one set of input data; screened input and output data are trained, and a Takigi-Sugeno fuzzy reasoning system capable of simulating data behaviors can begenerated; and a fuzzy rule obtained through training of a subtractive clustering method and an adaptive fuzzy reasoning system is applied to a variable static pressure fuzzy control system model. Through the method, the control process is more stable, anti-interference performance and energy saving performance are achieved, and the difficulty of rule obtaining through manual debugging can be solved to a certain degree.

The invention provides an intelligent detection method and system for sow parturition based on a nesting behavior. The method comprises the steps that from the time before sow parturition to the timewhen parturition is completed, a distance measuring module is adopted for acquiring distance data; the distance data is classified through a subtractive clustering algorithm; the cluster where the nesting behavior is located is obtained by collecting video information, nesting behavior data and non-nesting behavior data are saved in a database; the sow behavior is monitored in real time, and the cluster where the behavior is located is determined, when the sow behavior is located in the cluster of the nesting behavior and the lasting time exceeds a set threshold, an alarm is given. After the method is applied, the nesting behavior before sow parturition can be monitored in real time for 24 hours, no manual guarding is needed, abnormal situations can be found in time, and corresponding measures can be taken, so that the loss caused by the abnormal situations of a breeding factory is reduced, and the work efficiency of breeding personnel is effectively improved.

The invention relates to a big data clustering algorithm for reducing the risk of customer losing. The method comprises the following steps that: (1), related attributes are selected by using an axiomatic fuzzy set theory and a fuzzy concept is expressed by using a membership function and logical operation; (2), according to the calculated membership degree, a neighborhood radius and a weight coefficient of a subtractive clustering algorithm are automatically determined; (3), with the subtractive clustering algorithm, a mountain function is selected and updated to calculate a clustering numberand centroid, wherein the subtractive clustering algorithm and the axiomatic fuzzy set are integrated into a semantic-driven subtractive clustering method; and (4), on the basis of a K-means algorithm, the clustering of the clustering centroid obtained by the semantic-driven subtractive clustering method is calculated. Therefore, with the semantic-driven subtractive clustering method (SDSCM) based on the subtractive clustering algorithm and the axiomatic fuzzy set, the clustering precision of the subtractive clustering algorithm and K-means method is improved; and because of the novel algorithm, the risk of inaccuracy of operational management based on the axiomatic fuzzy sets (AFS) is reduced.

The invention relates to a nonlinear time sequence prediction method based on a small-world scale-free network. The method comprises: replacing a random network based on a small-world scale-free network to obtain a small-world scale-free echo state network SSESN, clustering random distribution neurons by adopting a subtractive clustering optimization algorithm, adaptively obtaining an optimal clustering scheme, and improving the clustering performance of an SSESN storage pool network; constructing a small-world scale-free network connection model through intra-cluster connection and inter-cluster connection according to the clustering condition of neurons; extracting connection information of each node, calculating an internal connection matrix of the reserve pool, constructing a nonlinear time sequence prediction model based on the small-world scale-free network according to the generated SSESN reserve pool network and the input / output training sample set, and the prediction precision of the nonlinear time sequence prediction model is improved.

The invention discloses a super-short-term prediction method of photovoltaic power station irradiance. The method includes the steps that irradiance data are extracted from a history database, data of a night time quantum are removed, corresponding extraterrestrial theoretical irradiance is calculated, data abnormal detection is carried out based on the preceding operations, and the data are normalized in the difference value ratio method of an extraterrestrial irradiance theoretical value and practical irradiance; a training sample set is extracted according to input and output dimensionality of a model; a model of an irradiance time sequence is built through an ANFIS, a the rule quantity and an initial parameter of the ANFIS model are determined in a subtractive clustering method, and a fuzzy model parameter is optimized in a counter propagation algorithm and a least square method; a prediction sample is input, and a prediction value is obtained through calculation; the prediction value is added to form a new sample set, and multiple steps of prediction are achieved in a cycling mode; counter normalization processing is carried out on the prediction value. Super-short-term prediction of the irradiance can be achieved only by means of a history irradiance time sequence, prediction accuracy is good and the method is easy to carry out.

The invention discloses a cam profile fitting method based on the RBF neural network. The method comprises the steps of establishing an RBF neural network model, conducting subtractive clustering learning on the RBF model, training the RBF model, and conducting model error index evaluation. According to the cam profile fitting method based on the RBF neural network, the number of hidden layer neuron nodes can be determined conveniently and quickly by means of the subtractive clustering algorithm, and the number of times of iterations of the algorithm is effectively reduced. Compared with the BP neural network, the RBF network has the advantages that calculation amount is small, calculation speed is high, generalization ability is high, robustness is high, and the RBF network has high application value in the field of cam profile fitting and other mechanical part surface shape fitting fields.

The invention discloses a critical heat flux density prediction method based on a deep learning support vector machine. The method is specifically implemented according to the following steps: step 1,dividing collected critical heat flux density original data into two parts, wherein 70% of the original data serves as a training set and is represented by X = {(x1, y1), (x2, y2),... (xi, yi)... (xn, yn)}, normalization processing is conducted on xi in the obtained training data set through linear transformation, and a data point x'i obtained after normalization processing is obtained; the other30% of the original data is taken as a test set for testing the prediction accuracy of the prediction model obtained by training; step 2, selecting experimental data from the normalized data points x'i obtained in the step 1 by using subtraction clustering added with information potential, wherein i is equal to 1, 2,..., n; and step 3, optimizing parameters of a support vector machine by performing cross validation on the experimental data obtained in the step 2 through a one-step method, and performing training by adopting a restricted Boltzmann machine in deep learning to obtain an optimalprediction model and optimal parameters. The prediction method can predict the critical heat flux density more accurately.

The invention provides a detergent dosage control method and device, a storage medium and a washing machine, and the method comprises the steps: obtaining various sample data of to-be-washed clothes, and taking the various sample data as an input layer; calculating by adopting a subtraction clustering algorithm to obtain the number of neurons corresponding to each kind of sample data, and taking the number of neurons corresponding to various kinds of sample data as a hidden layer; according to the plurality of sample data and the number of neurons corresponding to the plurality of sample data, calculating to obtain a plurality of detergent dosage output data, and taking the plurality of detergent dosage output data as an output layer; constructing a detergent dosage neural network model according to the input layer, the hidden layer and the output layer; and obtaining at least one type of target data, and inputting the at least one type of target data into the detergent dosage neural network model to determine target detergent dosage data. The detergent dosage neural network model is constructed by combining the subtraction clustering algorithm and the neural network, and the detergent dosage putting accuracy can be improved.