604results about How to "Reduce forecast error" patented technology

The invention discloses a shortest path planning method for dynamic origins, and belongs to the technical field of dynamic path guidance in an intelligent traffic system. The shortest path planning method is used for optimizing vehicle traveling routes. The shortest path planning method always takes a current vehicle position as a starting point and predicts road condition information such as traffic flow of relevant road sections and road intersections, and the like for many times according to vehicle running time to calculate a shortest path according to a path planning algorithm in a no-request mode at regular time, thereby acquiring a shortest time path in accordance with the real-time conditions of a road network better and directly displaying the shortest time path on the screen of a vehicle navigation system; and meanwhile, by the method, the update operation amount is less so as to shorten calculation time.

The invention relates to an interframe prediction encoding method, an interframe prediction decoding method and equipment. The interframe prediction encoding method comprises the following steps of: taking a motion vector predicted value obtained according to the motion information of an encoded macro block as a center, and acquiring a reference area most matched with the content features of the current encoded macro block; dividing the current encoded macro block into sub-blocks according to the dividing mode of the reference area; dividing the reference area according to the content features; and performing motion estimation on the sub-blocks, and searching the optimized motion vector for interframe prediction encoding. By using stronger time relevance between a reference image and an encoded image, a reference area which is matched with the content features of the encoded macro block is found in the reference image, the dividing mode or code rate allocation mode of the encoded macro block is determined according to the content features of the reference area, and the accuracy of the macro block division is improved. Therefore, the prediction error is reduced, and the accuracy of the interframe estimation value of the macro block is improved.

A large-scale real-time traffic flow prediction method of the present invention is based on fuzzy logic and deep LSTM, which relates to a technical field of urban intelligent traffic management. The method includes steps of: selecting an urban road network scene to collect color images of real-time traffic flow congestion information; obtaining congestion levels of multiple intersections according to the color images, which are used in a data training set; and forming a data sensing end of FDFP through a fuzzy mechanism; establishing a deep LSTM neural network, performing deep learning on the training data set, and constructing a prediction end of the FDFP; construct a graph of road intersections and formulate a k-nearest neighbors-based discounted averaging for obtaining congestion on the edges; and inputting real-time traffic information received from a server into an FDFP model.

The invention discloses a power system short-term load probability forecasting method, a device and a system. The short-term load probability density forecasting model of Gaussian process quantile regression is established by selecting an optimal input variable set affecting the load. Firstly, the importance score of input variables is given by stochastic forest algorithm, and the influence degreeof each input variable is sorted. Secondly, particle swarm optimization algorithm is used to search the super-parameters of the model to form the optimal Gaussian process quantile regression prediction model, avoiding the adverse effect of artificial experience setting initial parameters on the prediction performance of the model. The invention can avoid the shortcomings of manual experience selection, the load forecasting model established in the optimal input variable set has low error, which further reduces the forecasting error, and overcomes the problems that the common conjugate gradient method is easy to fall into the local optimal solution, the iterative number is difficult to determine, and the optimization performance is greatly affected by the initial value selection, so that the self-searching and group cognitive ability can be brought into full play.

The invention discloses a real-time closed loop predictive tracking method of a maneuvering target, which is a closed loop real-time self-adaptive processing method of on-line predictive immediate tracking in a maneuvering small target imaging tracking system and is mainly used for fields of photoelectric imaging tracking, robot vision, intelligent traffic control and the like. Due to the adoption of the method, a captured target can be extracted to to establish a flight track, the target flight track is filtered, the position of a target at a next collection time is predicted, a platform is processed in real time on line with high performance of a DSP main processor and a FPGA coprocessor, a prediction algorithm which can cope with target maneuver with higher accuracy is adopted to predict the motion state of the target in real time and a prediction result is utilized to drive a piezoelectric ceramic motor two-dimensional motion station to carry out overcompensation, thereby the self-adaptive predictive tracking is realized. The invention has the advantages that the method can overcome the defect of a largened tracking error caused by system delay and can still carry out continuous and stable tracking when the target maneuvers or is temporarily sheltered.

The invention relates to a network security situation forecasting method and system. The network security situation forecasting method comprises the steps that an obtained network security situation value sequence set serves as training data; a back propagation neural network structure is initialized, and the initialization operation comprises the step of setting the number M of nerve cells of an input layer and the number N of nerve cells of an output layer; real number encoding is carried out on the training data and the training data with the maximum fitness are found; a security situation value corresponding to the number M of the nerve cells of the input layer in the training data with the maximum fitness serves as an input value, a security situation value corresponding to the number N of the nerve cells of the output layer serve as a desired output value, the back propagation neural network is trained, and a forecasting model of the network security situation is established; the security situation value corresponding to the number M of the nerve cells of the input layer serves as the input value, and the security situation value corresponding to the number N of the nerve cells of the output layer is forecast according to the forecasting model. The network security situation forecasting method can improve the convergence rate of network security situation forecasting, shorten training time and reduce forecasting errors.

The invention relates to a rapid locking method for a full digital phase-locked loop, which is used for locking the frequency of the full digital phase-locked loop comprising a phase detection discriminator, a time-to-digit converter, a digital filter, a digital controlled oscillator and a frequency divider in a short time. The method is characterized by comprising the following steps: finding a control word controlling the frequency of the digital controlled oscillator by a designed algorithm; dividing the frequency of a clock output by the digital controlled oscillator which is controlled bythe control word to obtain a divided-frequency clock with the frequency approximate to a reference clock frequency; and then, controlling the digital controlled oscillator to lock based on a phase difference between the reference clock and the divided-frequency clock which are distinguished by the phase detection discriminator. The full digital phase-locked loop is provided with a rapid frequencycapture loop and a phase-locked loop which alternately work, i.e. firstly, the rapid frequency capture loop finishes frequency capture, and then, the phase-locked loop finishes accurate locking.

The invention discloses a new short-term wind speed forecasting method of a wind farm, which comprises the following steps: collecting wind speed data of the wind farm, forming the time sequence of the historical wind speed and carrying out normalization treatment; applying the chaos analysis method for analyzing the time sequence of the historical wind speed after the normalization treatment for obtaining phase space reconstruction parameters of a wind power system in the area located by the wind farm, wherein the parameters are delay time and embedding dimension of the time sequence; utilizing the parameters for carrying out treatment on the time sequence of the historical wind speed after the normalization treatment, and obtaining a training sample set required by a support vector regression model for wind speed forecasting; adopting the training sample set for training the support vector regression model; utilizing the support vector regression model after training for carrying out short-term wind speed forecasting on the wind farm, and obtaining the normalized result of the short-term wind speed forecasting of the wind farm; and carrying out anti-normalization treatment on the obtained normalized result of the short-term wind speed forecasting of the wind farm, and obtaining the short-term wind speed forecasting result of the wind farm.

The invention discloses a continuous blood pressure measuring device based on pulse waves. The continuous blood pressure measuring device comprises a pulse wave signal collecting module, a pulse wave signal denoising module, a pulse wave waveform extraction module, a pulse wave feature parameter collecting module and a blood pressure prediction module. Errors during the estimation of diastolic pressure can be reduced by adopting a dicrotic wave feature parameter method; during the denoising processing, the baseline drift and the high-frequency noise are removed, so that the interference is effectively reduced; the pulse wave time domain features can be further obtained only through collecting one path of pulse wave signals and only through detecting the starting point of the pulse waves; the feature obtaining method is simple; the detection speed is high; the calculation quantity is small; and the waveform quality is high. A blood pressure regression model can be conveniently built; the systolic pressure and the diastolic pressure can be well predicted; the prediction errors are small; and the important basis can be provided for clinic diagnosis and family health care.

The invention relates to a landslide space-time predicting method of a river basin scale, and aims to provide a river basin landslide space-time predicting method under a rainfall effect. The method comprises the following steps of: calculating a safety factor (SF) of any three-dimensional position in a river basin by using an InHM hydrological model, an unlimited side slope stability model and a landslide calculation module, fitting the safety factor into a specified time river basin safety factor distribution map and a specified time river basin landslide depth map through a visual module, and displaying the maps on display equipment; and if the landslide exists, selectively displaying the river basin landslide time distribution map. By the method, the river basin landslide can be analyzed and predicted effectively; the defect that a conventional model needs depression pretreatment in complex terrain is overcome; natural river basin geomorphic and hydrological response characteristics are maintained; a landslide mass is closer to an actual landslide mass shape; and a predicting error is reduced. Simultaneously, the precision and the applicability of a model are improved. The method is suitable for the landslide prediction of a simple side slope, a complex side slope and the river basin scale.

The invention discloses a short-term wind speed forecasting method based on a deep neural network transfer model. The method comprises the following steps that (1) normalization preprocessing and division of sample sets are carried out on data of two or more wind power plants, (2) the deep neural network transfer model is established, (3) layered training is started from bottom to top in an unsupervised learning mode, (4) supervised learning is carried out from top to bottom on the basis of the third step, (5) weight parameters of connection of a top layer and hidden layers are finely adjusted so as to obtain an output layer, corresponding to the wind power plants, in a deep neural network, and (6) inverse normalization is carried out on the result output by a deep neural network so as to obtain the predicted value of wind speed. Transfer learning is introduced to the wind speed forecasting field, knowledge of other wind power plants rich in data is transferred to target wind power plants, and the problem that the newly built wind power plants have few data is solved effectively. By means of the effective transfer scheme based on the deep neural network, the wind speed prediction accuracy of the target wind power plants is greatly improved.

The invention provides an interpolation method of missing data. The method includes the following steps that a dataset in the preset time region is acquired, and the missing data in the preset time region and the time corresponding to the missing data are searched for; based on the time corresponding to the missing data, a DA multiple interpolation model is built; based on the multiple interpolation model, a plurality of intermediate interpolation values of the time corresponding to the missing data are calculated; the interpolation value with the minimal prediction error is acquired from theintermediate interpolation values and used as a final interpolation value corresponding to the missing data; in the position of the time corresponding to the missing data, the missing data is replacedwith the final interpolation value. Interpolation is conducted on the missing data by means of the multiple interpolation model, the intermediate interpolation values are generated in the multiple interpolation process, the uncertainty of the missing data is reflected by means of the variability among the interpolation values, and the interpolation value replacing the missing data is the interpolation value with the minimal prediction error, so that the error is small compared with original data.

The invention provides a method for constructing a photovoltaic power station generation capacity short-term prediction model based on multiple neural network combinational algorithms and belongs to the technical field of photovoltaic power generation, power grid connection technology and solar energy photovoltaic forecasting. The method overcomes the problem that a usually-used algorithm for constructing the photovoltaic power station generation capacity short-term prediction model is single and is likely to fall into local optimization, further resulting in big measurement error of the prediction model. The technical construction method of the invention is realized as follows: firstly using four different neural network algorithms to construct sub-models for neural network prediction; secondly screening and classifying weather information and analyzing the suitability of the various sub-models for neural network prediction; giving weighted parameter values of the sub-models in a combined model according to the suitability to further make the combined neural network model for prediction suitable for different weather conditions and then completing the construction of the photovoltaic power station generation capacity short-term prediction model. The method is mainly used for photovoltaic power station grid connection short-term prediction.

The invention provides a short-term traffic flow prediction method based on a nerve network combination model. The method is used to construct a counterpropagation nerve network combination prediction model and the short-term traffic flow prediction method is provided based on the model. Aiming at a characteristic of a traffic flow, a fuzzy C mean value clustering algorithm is used to cluster the traffic flow. For bunch generated through clustering, a counterpropagation nerve network prediction model is constructed. According to grade of membership, a weighted sum of prediction model prediction results is calculated and is taken as a final prediction result. In order to increase prediction precision, a taguchi method is used to carry out test designing so as to test influences of different structure parameters on prediction model prediction precision, and an optimum structure parameter is used as an initial structure of the prediction model. By using the method in the invention, the prediction precision of the short-term traffic flow can be effectively increased, an influence of a noise on the prediction precision in training data is reduced and operation time is reasonable.

The invention discloses a power consumer load interval prediction method based on deep learning. The method comprises the following steps of (1) establishing a large consumer historical load data preprocessing model; (2) establishing a load point prediction model based on an LSTM time recurrent neural network; and (3) adopting a load interval prediction algorithm of a point prediction value scaling coefficient. In this way, according to the method, a user load preprocessing model based on a state vector machine method is established to carry out preprocessing analysis on the single user historical data; and according to the processed historical data, an LSTM machine learning method is adopted to find a prediction model for reducing a user load prediction error to the maximum extent, and the load interval prediction of a single user is carried out by using a point prediction value scaling coefficient load interval prediction algorithm, so that the accurate load interval prediction can be carried out on the load of the single power user with strong random fluctuation, and the prediction accuracy of the user load is obviously superior to that of a traditional method.

The invention relates to multi-virtual reality helmet anti - vertigo systems and methods based on an acceleration sensor, belongs to the field of virtual reality interactive technologies. The system includes four line acceleration sensors, helmet position tracking device and computer. The working steps of the system are:through a acceleration sensor on the helmet ,gets the angular acceleration rotating around head and the line acceleration of movement before and after; 6-DOF position information of head movement presented by the position tracker , is given effective forecasting of location by the computer; by placing another acceleration sensor on movement platform of the human body set, it can realize position measurement of the corresponding freedom of movement platform, and gain an effective forecasting of the relative position. With the invention of high-performance, strong practicability, low-cost features, users can effectively alleviate vertigo symptoms of disease caused by long wearing virtual reality helmet, such as simulation.

The invention discloses a double-rate Kalman filtering method based on GNSS / INS deep integrated navigation. The method comprises the following steps of 1, building a state equation according to the initial position, the rate and the attitude information of a carrier, and initializing the parameters of the Kalman filtering; 2, performing state prediction updating on M step lengths and obtaining a predicted value of a prior state quantity which is described in the specification; 3 correcting the prior state quantity which is described in the specification to obtain a predicted value of a posterior state quantity which is described in the specification; 4, adaptively updating the errors of the state quantities and a systematic error covariance matrix, and compensating an inertial navigation result by using the predicted value of the posterior state quantity which is described in the specification to obtain the position, the rate and attitude information of the carrier; and 5, updating thepredicted value of a posterior state quantity which is described in the specification after the compensation. The method can reduce a truncation error caused by the low update frequency of GNSS satellite data or the losing lock of the satellite data during a data fusion algorithm of the GNSS / INS deep integrated navigation, and simultaneously solve a navigation positioning error caused by the non-synchronization of INS data and GNSS data.

The invention discloses a hybrid prediction method for network traffic. The method is characterized in that a network traffic prediction method with an extreme learning machine (ELM) being compensated by autoregressive integrated moving average (ARIMA), namely, a network traffic prediction method with the ELM being compensated by a Farctal autoregressive integrated moving average model is provided through a self-similarity analysis of a network traffic sequence. The method comprises the following steps: firstly, predicting the network traffic sequence with the ELM; secondly, correcting an error sequence of network traffic prediction through an ARIMA model; and lastly, overlaying an ELM predicted value and an ARIMA model correction value to obtain a final predicted value. According to the method, prediction error data is fitted by the ARIMA model, and the predicted value of the ELM is overlaid with a residual error of ARIMA prediction to obtain the final predicted value. Residual error compensation is performed through the ARIMA model, thereby effectively increasing the prediction accuracy.

The invention relates to an Internet of Things data uncertainty measurement, prediction and outlier-removing method based on the Gaussian process. The method is a dynamical system method of estimating and collecting the standard deviation of Internet of Things perception sensor measurement errors and combining the Gaussian process modeling theory with autoregression model representations; prediction values and uncertainty measurement of observation data effective time sequence data are given, whether the data are missing values or outlier data is judged according to the information, and data supplement is correspondingly carried out. The method is a non-parameterized probability prediction method. Due to the fact that training set learning has the feature of tracing system dynamic states, judgment, early-warning and data supplement can be carried out on data exception and data missing phenomena in time according to the prediction value uncertainty and the sensor calibration standard deviation, the prediction error is small, and the accuracy is high. The Internet of Things data uncertainty measurement, prediction and outlier-removing method is used for controlling the quality of Internet of Things automatic observation data, and can ensure accuracy of collected data.

The invention discloses a hydrofractured water production shale gas well shale gas yield predicting method. The hydrofractured water production shale gas well shale gas yield predicting method comprises the following steps of S1, establishing a shale matrix gas seepage model; S2, establishing a shale fracture seepage mathematic model, S3, according to the shale matrix gas seepage model and the shale fracture seepage mathematic model, acquiring a water production shale gas well seepage mathematic model; S4, solving the water production shale gas well seepage mathematic model through a numericalsimulation method to obtain the shale gas yield of a water production shale gas well. Therefore, by means of the shale matrix gas seepage model and the shale fracture seepage mathematic model, the water production shale gas well seepage mathematic model, which takes into consideration of formation pressure drop influence as well as water saturation increase influence, can be established; by solving the water production shale gas well seepage mathematic model, the shale gas yield of the water production shale gas well can be accurately predicted, and the predicting errors of the shale gas yield can be greatly reduced.

The invention relates to a CAN bus-based pig house environmental temperature intelligent monitoring system. The system is characterized in that the intelligent monitoring system is composed of a CAN bus-based pig house environment parameter acquisition and intelligent prediction platform, a pig house environment multi-point temperature fusion model and a pig house environment intelligent prediction model. With the CAN bus-based pig house environmental temperature intelligent monitoring system of the present invention adopted, an existing pig house monitoring system cannot intelligently monitor and predict the temperature of the environment of a pig house according to the non-linearity and large lag of the environmental temperature change of the pig house, the large area of the pig house, complicated temperature change and other characteristics, as a result, control and adjustment of the environmental temperature of the pig house are seriously affected, while, with the CAN bus-based pig house environmental temperature intelligent monitoring system adopted, the above problem can be solved.

The invention discloses a video coding and decoding method and a device using a synthetic reference frame, which uses coded image information as the reference frame and code and decodes the image frame of a successive code. The reference frame includes the synthetic reference frame which is reset and renewed by a reconstructive frame in each frame and then renewed by the reconstructive image information of a following inter-frame prediction frame according to a macro-block or a sub block of the macro-block. The renewed synthetic reference frame is used as a reference frame for the prediction during the coding of the successive image frame and is renewed continuously. The video coding and decoding method and the device of the invention can increase the accuracy of the video sequence motion estimation in video monitoring, reducing prediction errors and improving coding efficiency without the increasing of the complexity.

The invention discloses a pretreatment method for multi-viewpoint image, which carries out the extremum detection to the object image and the source image in the scale space via the invariant scale characteristics transform algorithm, picks up the invariant scale characteristics transform eigenvector of the object image and the source image key points, acquires the accurate matching key point pair set of the object image and the source image by the feature matching algorithm and corrects the color of the source image by computing the multiplicative error and the additive error of the key point pair set, carries out the geometric calibration to the rectifying image after color correction by establishing the affine transformation between the key points, the advantages of which lies in that under the prerequisite of ensuring the accuracy of the multi-viewpoint image color correction and the geometric calibration, the invention greatly improves the robustness and the mapping accuracy of the color correction, reduces the complicatedness of the geometric calibration and enhances the coding performance of the multi-viewpoint image.

The invention relates to a full digital phase-locked loop applying a rapid frequency capture method for locking the frequency of the full digital phase-locked loop comprising a phase detection discriminator, a time-to-digit converter, a digital filter, a digital controlled oscillator and a frequency divider in a short time. The full digital phase-locked loop is characterized by comprising the following steps: finding a control word controlling the frequency of the digital controlled oscillator by a designed algorithm; and dividing a frequency output by the digital controlled oscillator which is controlled by the control word to obtain a divided-frequency clock with the frequency approximate to a reference clock frequency. The full digital phase-locked loop is provided with a rapid frequency capture loop and a phase-locked loop which alternately work, i.e. firstly, the rapid frequency capture loop finishes frequency capture, and then, the phase-locked loop finishes accurate locking.

The invention relates to a steelmaking end point prediction method based on neural networks. The neural networks are used for replacing manual experience to predict the end point of steelmaking of a medium and small sized converter, a plurality of sets of production parameters are collected to serve as independent variables, flame temperature information and flame spectral information in the converter are collected through optical fibers to serve as independent variables, the multiple sets are used for training three layers of BP neural networks, MIV is adopted to screen the independent variables, the independent variable most affecting the end point is selected out to serve as an input parameter, an input parameter of the converter to be measured is selected to be inputted to the trained neural networks, and predicted converter end point temperature and the end point carbon content are obtained; and the shortcoming of manual experience prediction of the end point of steelmaking of the medium and small sized converter is overcome, optical fiber conduction is used for precisely measuring flame information inside the converter, and the accurate neutral networks are obtained.

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.

Relating to maintenance technologies of lead-acid batteries, the invention provides an intelligent maintenance device of a multifunctional lead-acid battery and a capacity prediction method. The intelligent maintenance device of a multifunctional lead-acid battery in the invention is used for on-line test of lead-acid battery capacity, automatic constant current charge and discharge, as well as capacity recovery of a degraded battery. The device comprises a charging unit, a discharging unit, a master control module and a measurement control circuit. The provided lead-acid battery capacity prediction algorithm through BP neural network optimization based on genetic annealing can be realized in the control module and an upper computer, and the capacity prediction error is smaller than a 15%prediction error of present domestic battery capacity rapid testers. The device of the invention can be used in on-line health status test of lead-acid battery capacity and can realize capacity recovery of a degraded battery, as well as extend the battery service life.

The embodiment of the invention provides a network flow prediction method and device, and relates to the technical field of networks. The method comprises the steps of determining the prediction moment of network flow; determining a first feature vector of the network flow at the prediction moment according to the network flow information at the first historical moment, wherein the first feature vector comprises a first flow feature and a change feature of the first flow feature; inputting the first feature vector into a network flow prediction model, and obtaining a flow prediction value at aprediction moment, wherein the network traffic prediction model is model obtained by training a preset regression model by adopting a sample feature vector and a flow actual value of the sample network flow at a sample moment; the sample eigenvector is an eigenvector determined according to the network flow information at a second historical moment, the sample moment is an actual generation moment of the sample network flow, and the sample eigenvector comprises a sample flow characteristic and a change characteristic of the sample flow characteristic. By applying the scheme provided by the embodiment of the invention, the accuracy of network flow prediction is improved.