48 results about "Sum squared error" patented technology

The invention provides a wind direction clustering method, device and electronic equipment, wherein the method includes: acquiring periodic wind direction data of each wind turbine in a wind farm; determining a similarity measure index suitable for the periodic wind direction data according to the periodic wind direction data ; Based on the similarity measurement index, select k cluster centers, and divide the periodic wind direction data into k clusters; determine the error sum of squares criterion function based on the periodic wind direction data; The clustering effect is evaluated to determine the final clustering effect. Among them, the similarity measurement index and the error sum of squares criterion function suitable for the periodic wind direction data are determined by the periodic wind direction data, which solves the disadvantage that the traditional K-means algorithm is not suitable for the periodic wind direction data clustering, and ensures the clustering results. reliability and authenticity, and reliable and reasonable fan division can be obtained according to the clustering results, avoiding the subjectivity of artificial division of sectors.

The invention discloses a method for processing the autocorrelation function at the zero-time-delay point of radar based on damping fitting, which mainly solves the problem that the existing incoherent scattering radar cannot accurately obtain the information at the zero-time-delay point when detecting the ionosphere. The implementation plan is: 1) use the algorithm to calculate the combined measured autocorrelation value L; 2) establish the damping function model, and set the value range and search step size of each parameter in the model; 3) obtain the combined fitting according to the damping function model Autocorrelation value S expression, select parameters multiple times within the parameter value range, calculate the corresponding residual error sum of squares by L and S, and compare to obtain the minimum residual error sum of squares; 4) The parameters corresponding to the minimum residual error sum of squares The value determines the final damping function model, so that the zero-time-delay fitted autocorrelation value is calculated. The processing process of the invention is simple and easy to realize, and the data information at the zero time delay is reserved, and the deficiency of the existing non-coherent scattering radar in the processing of the zero time delay autocorrelation function is made up.

The invention discloses a fuel cell optimization modeling method with an adaptive genetic policy RNA-GA. The method comprises the steps of 1) obtaining sampling data of an input current and an output voltage of a proton exchange membrane fuel cell through field operation or experiment; 2) taking an error square sum of sampling data of an estimated output and an actual output of a fuel cell model as an objective function during RNA-GA optimization search; 3) setting algorithm running parameters; and 4) running the RNA-GA to estimate unknown parameters in the fuel cell model, obtaining estimated values of the unknown parameters in the model by minimizing the objective function, and substituting the estimated values of the unknown parameters into the fuel cell model, thereby forming a mathematic model. According to the method, a decision is made to execute crossover or mutation operation by applying the adaptive genetic policy, so that the population diversity is effectively kept and the convergence speed of an algorithm to a global optimal solution is increased; the obtained fuel cell model parameters are reliable; and the method is also suitable for optimization modeling of other complex chemical reaction processes.

The invention discloses a method for predicting upstream and downstream water levels of a cascade power station, which includes the following steps: Step 1, selecting input variables and output variables; Step 2, standardizing the data to eliminate the influence of dimensions; Step 3, determining the input Vector dimension, LSTM layer number, output vector dimension, and time step; step 4, LSTM forward propagation process and error back propagation process; forward propagation process is input into LSTM network in sequence according to time step, and corresponding output value is obtained ; Using the sum of the squares of the error between the output value and the true value as the loss function, the error is backpropagated along time to update the parameters; step 5, use the trained model for multi-moment continuous prediction. Applying LSTM to the water level prediction of cascade power stations can capture the lagging influence information of upstream power stations on downstream power stations, improve prediction accuracy, and provide more reliable theoretical support for scientific scheduling decisions.

Provided by the present invention is a MapReduce-based distributed cluster processing method for large-scale data, which comprises: sampling large-scale data according to an equal-scale non-repetition principle; inputting the sampled data into a MapReduce distributed parallel framework, and calculating the local density and average density of the sampled data; finding all sampled data having a local density greater than the average density to serve as a candidate point set of initial cluster center points for each cluster, and feeding the candidate point set back to a master node, wherein every two adjacent candidate points at a distance from each other which is greater than twice that of a set range are selected to serve as the initial cluster center points; using the MapReduce distributed parallel framework to perform a parallel clustering task, wherein an average value of the distance between the data is calculated for each cluster in order to update the cluster center points; child nodes applying an error sum of squares criterion function so as to determine whether to continue iteration; the child nodes performing clustering on the large-scale data according to the cluster center points. By means of the present invention, parallel clustering is implemented, thereby reducing the number of clustering iterations, while increasing clustering accuracy and the efficiency of parallel clustering.

An HEVC inter-frame coding processing method is used to speed up the judgment speed of the skip mode and improve HEVC coding efficiency. In some feasible implementations of the present invention, the method includes: obtaining the predicted value and original value of the current prediction unit PU; calculating the error sum of squares between the predicted value and the original value of the PU; comparing the error sum with a threshold Comparing, judging whether the prediction mode of the PU is a skip mode according to the comparison result. The embodiment of the present invention also provides a corresponding device.

The invention discloses an image target detection method based on DC-SPP-YOLO. Firstly, a data enhancement method is used to preprocess the training image samples and a training sample set is constructed, and a k-means clustering algorithm is used to select a target bounding box prediction The prior candidate frame of the YOLOv2 model; then the convolutional layer connection method of the YOLOv2 model is improved from layer-by-layer connection to dense connection, and at the same time, spatial pyramid pooling is introduced between the convolution module and the target detection layer to establish DC-SPP-YOLO target detection Model; finally, the loss function is constructed by the sum of squared errors between the predicted value and the real value, and the weight parameters of the model are updated iteratively to make the loss function converge, and the DC‑SPP‑YOLO model is obtained and used for target detection. The present invention considers the "gradient disappearance" caused by deepening the convolutional network and the YOLOv2 model does not fully use the multi-scale local area features, and constructs the improved DC-SPP-YOLO target detection model based on the dense connection of the convolutional layer and the spatial pyramid pooling. Improved object detection accuracy.

The invention discloses a modeling method for a support vector machine for a bridge crane through bulged ring crossover operation RNA-GA, and belongs to the field of intelligent modeling. The modelingmethod comprises the following steps: 1) acquiring the sampling data input and output by a bridge crane system through an experiment or field acquisition, and taking the error square sum of the expected output and the actual output as the target function of the RNA-GA; 2) abstracting out the modeling method for the support vector machine for the bridge crane through the bulged ring crossover operation RNA-GA through the enlightening of the molecular bulged ring structure of RNA; 3) setting the running parameters of an algorithm; and 4) running the bulged ring crossover RNA-GA to acquire the optimal values of the parameter estimation of a support vector machine model with the least square estimation model for the position of the bridge crane and a support vector machine model with the least square estimation model for the oscillating angle of the bridge crane separately, and substituting the optimal values into the support vector machine models with the least square estimation models to form a bridge crane model. The modeling method for the support vector machine has the advantages of being high in convergence rate, high in accuracy, and the like, and is also applicable to modelingfor other complex systems.

The invention discloses a Lamb wave spatial sampling signal fitting method based on a Morlet mother wavelet, and belongs to the technical field of engineering structure health monitoring. This method first sets the center frequency, sampling frequency, and sampling time of the Morlet mother wavelet; then constructs the Lamb wave excitation simulation signal according to the number of Lamb wave excitation signal peaks; secondly, according to the Morlet mother wavelet fitting waveform and Lamb wave Excite the correlation coefficient of the simulated signal to obtain the frequency bandwidth parameter of the Morlet mother wavelet; finally, according to the error sum of the squares of the Morlet wavelet function fitting waveform and the Lamb wave spatial sampling signal under different scale factors and displacement factors, obtain the Morlet wavelet function approximation At this time, the corresponding Morlet wavelet function fitting waveform is the Morlet wavelet function fitting waveform of the Lamb wave spatial sampling signal. The invention improves the spatial resolution and length of the Lamb wave spatial sampling signal, thereby helping to promote the application of the space-wavenumber domain signal processing method in the field of engineering structure health monitoring.

An ARAIM subset selection method based on a BeiDou constellation includes: (a) deeming a current airspace as a two-dimensional plane and satellite distribution at some point as point distribution on the two-dimensional plane; (b) selecting coordinates of K satellites as original centroids, assigning each satellite to the closest original centroid by means of a K-Means algorithm to form K classification clusters, and recalculating a centroid of each classification cluster to obtain a centroid coordinate; (c) re-selecting the coordinates of the K satellites as the original centroids, repeating the calculation step (b) to calculate the sum of squared errors of the centroid obtained every time, and selecting the centroid with a minimum sum of squared errors as an output result; and (d) performing outlier detection on the output result of the step (c), and if an outlier appears, not performing fault assumption on a satellite at the outlier.