119 results about "Learning factor" patented technology

The invention relates to the field of solar photovoltaic power generation and discloses a particle swarm optimization based control method and device of photovoltaic maximum power point tracking. A duty ratio of a converter is used as particles, the particles are initialized and uniformly dispersed at possible extreme value points, inertia weight and learning factor are adjusted linearly, an anti-tangent function is introduced to modify conventional particle swarm speed updating, and speed of each particle is limited solely. Convergence rate of a particle swarm algorithm is increased, oscillation during tracking is reduced, and new maximum power points can be quickly tracked during shadow changing.

The invention discloses an improved kernel-related filtering tracking method based on ultra-pixel optical flow and self-adaptive learning factor. The appearance reconstruction of target can be realized through ultra-pixel analysis, and the target is divided into ultra-pixel blocks which are clustered into an ultra-pixel center. The displacement change of the optical flow analysis pixel point of each ultra-pixel center is calculated, and the movement offset and scale change of the target can be detected. Based on the predicted parameter, cycled sampling is conducted on each new-frame image, andan improved and gauss kernel-based filtering target tracking method which introduces the self-adaptive learning factor is adopted by each sample, and the accurate position and scale of the target canbe detected. The detection result is detected and corrected through an on-line SVM detection model, and the position with low confidence is corrected, and finally the target position can be accurately positioned and the target accurate scale can be obtained. The invention is advantageous in that the tracking problems like scale change, shielding, deforming, and motion blur, which exit in the target tracking process can be overcome, and real-time and highly-precise target tracking can be realized.

The invention designs a blind channel balancing method based on an improved PSO (Particle Swarm Optimization) BP (Back Propagation) neural network. In the process of solving the blind balancing problem on the basis of a BP neural network, determination of an initial weight and a threshold of the BP neural network is lack of the theoretical basis and has the defects of low convergence speed, easiness for falling into a local minimal value and the like so as to cause a poor channel blind balancing effect. In order to overcome the defects of the BP neural network and improving the channel blind balancing effect, the invention discloses a blink balancing method based on the improved PSO-BP neural network. According to the method, firstly, defects of a basic particle swarm algorithm are overcome, parameters of the basic particle swarm are improved, and an inertia weight and a learning factor are adaptively regulated; secondly, the initial weight and the threshold of the neural network are optimized by utilizing the advantage of high global searching capacity of the improved particle swarm, and then more accurate searching is carried out in such local space by utilizing a BP algorithm soas to obtain an optimal connection weight and threshold of the neural network; and finally, blind balancing based on the the improved PSO-BP neural network is implemented.

The invention relates to the technical field of reactive powder optimization of a power distribution network of a power system, and particularly relates to a tabu particle swarm algorithm based reactive power optimization method of a power distribution network. According to the situation that a basic particle swarm algorithm in the optimization process can be easily trapped in local optimization, the invention discloses the improved method by the combination of a tabu search algorithm, and the defect that the particle swarm algorithm can be easily trapped in local optimum is overcome by utilizing the memory function and the characteristic of high climbing ability of the search algorithm; meanwhile, learning factors c1 and c2 which change as the increase of iterations and an inertia weight coefficient Omega are introduced in a particle position and a speed upgrading equation of the particle swarm algorithm, and the problem that the particle swarm algorithm can be easily trapped into the local optimum is further solved. By the combination of the two intelligent optimization algorithms, the optimization capability is improved greatly; the tabu particle swarm algorithm based reactive power optimization method is much suitable for departments relevant to a power system and the like to implement reactive power optimization of the power distribution network.

The invention provides a transmission and transformation project construction cost assessment method and device. The transmission and transformation project construction cost assessment method comprises the following steps: input historical sample data of a transmission and transformation project are received; iterations, inertia weight, learning factors, particle velocity of a chaos particle swarm and the population size of the particle swarm are initialized to build a chaos particle swarm model; according to the chaos particle swarm optimization, parameters of the chaos particle swarm model are optimized; according to the historical sample data and the optimized chaos particle swarm model, optimal values of the iterations, inertia weight and learning factors of the chaos particle swarm model are determined; according to the determined optimal values of the iterations, inertia weight and learning factors, penalty coefficients, insensitive coefficients and kernel function parameters of a least square support vector machine model are determined respectively to build the least square support vector machine model; input actual sample data of the transmission and transformation project are received; according to the actual sample data of the transmission and transformation project and the built least square support vector machine model, a construction cost assessment result of the transmission and transformation project is generated.

The invention relates to an AC / DC (Alternating Current / Direct Current) mixed micro-grid optimized operating method based on a main-slave game model. The method comprises the following steps: acquiring weather data of an area in which a micro-grid is located such as illumination intensity, temperature and cloud cover, counting historical data of a photovoltaic power generation amount in the micro-grid, and predicting power outputted by a photovoltaic cell in an AC / DC mixed micro-grid in a next day by adopting a prediction method based on a typical trend; acquiring transmission power data at two sides of a bidirectional converter and a tendency controller in the AC / DC mixed micro-grid, and establishing a mathematic model of various devices in the AC / DC mixed micro-grid; dividing a day into various dispatching time periods, establishing a main-slave game model of cooperation of a photovoltaic operator and a grid company at each dispatching time period, wherein the main-slave game model comprises a photovoltaic utilization rate model maximizing the benefit of the photovoltaic operator and an AC / DC mixed micro-grid loss model minimizing the loss of the representative grid company; and improving the global searching performance of an algorithm by adopting a two-order oscillation particle swarm algorithm of an asynchronous variation learning factor. The AC / DC mixed micro-grid optimized operating method is high in applicability and high in prediction precision.

The invention discloses a cloud computing resource scheduling system and method. The cloud computing resource scheduling method includes that building a resource distribution model according to task number and resource number, and defining a resource distribution matrix, a resource scheduling time matrix and a scheduling target; generating codes of an initial population according to a code rule; computing the fitness value of the population according to fitness functions; selecting a dominant population according to the fitness value; updating the population through learning factors, and generating the next generation of new population according to the updating result; judging whether conforming to stop conditions, if so, stopping, and obtaining the optimal solution; otherwise, returning to compute the fitness value; decoding the optimal solution according to a decoding rule to obtain the resource distribution result. The cloud computing resource scheduling system and method are capable of well solving the cloud computing resource distribution problem.

The invention discloses a method for predicting a line loss rate of a power distribution network. The method comprises the steps of determining multiple electrical characteristic parameters, influencing the line loss rate, of the power distribution network, performing standardization processing on parameter values of the electrical characteristic parameters, and performing normalization processingon the line loss rate; by taking the parameter values of the electrical characteristic parameters as inputs of an input layer, and taking the value of the line loss rate as an output of an output layer, building an initial neural network model, wherein the initial neural network model comprises at least one hidden layer; determining the number of hidden layer nodes; by dynamically adjusting an inertial factor and a learning factor of a particle swarm algorithm, improving the particle swarm algorithm; by utilizing the improved particle swarm algorithm, optimizing a weight value and a thresholdvalue of the initial neural network model to determine an optimized neural network model; and inputting the electrical characteristic parameters of a power distribution network line to the optimizedneural network model, and predicting the line loss rate corresponding to the electrical characteristic parameters by utilizing the optimized neural network model.

The invention discloses a cognitive radio hostile interference resisting method based on Q learning and relates to cognitive radio. The cognitive radio hostile interference resisting method includes that 1), a cognitive source node initializes a learning factor gamma and a discount factor and sets values of a Q value table and a V value table to be 0; 2), the cognitive source node senses state s of k moment and selects optimal behavior a according to the Q value table; 3), the cognitive source node observes state s' of k+1 moment and calculates immediate reward us of information transmission this time; 4), the Q value table and the V value table of a Q learning algorithm are updated; 5), the cognitive source node repeats the steps 2)-4) until an optimal strategy is found. Power control is realized on the basis of Q learning, the cognitive source node and an intelligent jammer adjust own emitting power in realtime by constantly learning environment, and benefit optimization is realized along with continuation of iterative learning.

The invention discloses a multi-robot cooperation positioning method based on an MAPSOPF algorithm. The method includes the steps that an actual operating environment is abstracted into a planar graph, and according to the position of a road sign and the paths of robots, an environmental map is built; relevant information and blackboard information of the robots are initialized; the MAPSOPF algorithm is adopted, the subsequent positions of the robots are predicted according to a fitness value, the predicted positions and the blackboard information are updated, and accurate positioning is completed. By adjusting weighing coefficients and learning factors, a competition mechanism is introduced so that the optimization effect of the algorithm can be improved, and finally, multi-robot coordination positioning accuracy is improved.

The invention proposes a pedestrian detection and tracking method based on a head-shoulder contour and a BP neural network. The method comprises the steps: firstly extracting a moving human body target in a video sequence through employing an adaptive mixed Gaussian background updating algorithm, and improving the background estimation precision through changing a learning factor of a mixed Gaussian model; secondly extracting an initial contour of an original target through employing a Canny operator, and carrying out contour clustering through combining a Mean shift algorithm, so as to obtain a completer body contour; thirdly building a head-shoulder contour model through combining a head-shoulder width-height ratio of a human body, extracting a head-shoulder contour characteristic vector, inputting the characteristic vector of the head-shoulder contour model into the BP neural network, clustering a plurality of human body head-shoulder models, and carrying out human body recognition; and finally tracking a detected pedestrian target through employing a particle filter. The method avoids misjudgment and wrong judgment because of the incompletion of a recognition target, improves the recognition accuracy of the pedestrian target, and reduces the calculation amount.

The invention provides an infrared image enhancement method based on particle swarm optimization, global gamma correction is carried out by adopting a particle swarm optimization algorithm in combination with an image gamma correction method, and the method comprises the following steps: firstly, carrying out graying on an infrared image and initializing a population; Next, for each particle, carrying out Gamma correction on the grayed image to obtain an intermediate enhanced image.; calculating the entropy value of the intermediate imagecalculated; calculating fitness values of the particlesthrough weighted fusion of edge contents and gray scale standard variances, iteratively updating individual and group optimal fitness values, iteratively updating particle speeds and positions by timely adjusting learning factors, and continuously iteratively optimizing to find out final group optimal positions as optimal gamma values to enhance images. When the low-contrast infrared image is enhanced, the entropy value of the image is maximized, the edge is clear, the structure information of the original image is reserved, and the enhancement result is more natural; When the highlight area is enhanced, noise and artifacts are not easily generated, and the infrared image can be obviously improved.

The invention belongs to the field of software test and particularly relates to a test case generating method based on a modified particle swarm algorithm, comprising: introducing a learning factor with a weight function; allowing the learning factor to change in gradient increase or decrease manner correspondingly along with linear or nonlinear changes of inertial weight; balancing global searching and local mining capacities of the algorithm through the interaction of the learning factor and the weight function. Secondary search and reverse learning are introduced herein, so that solving precision can be improved and improvements are brought to population diversity and the like. A design method of a fitness function is analyzed at a test case generating module, excellence degrees of different branch nodes are considered, and more reasonable fitness function evaluation is designed.

The invention mainly belongs to the technical field of multi-target optimization, and particularly relates to a novel multi-target particle swarm optimization method. The method is based on shared learning and Cauthy mutation. Shared learning factors are adopted in the method to change the speed and position updating formula of particles, and the optimal position of the particle individuals and external files are updated by adopting Cauthy mutation operators. According to the method, the global search capability and the local optimization search precision of the particles can be enhanced, and the algorithm is enabled to fast approach Pareto leading edge and premature convergence of the algorithm can be avoided. According to the method, solution convergence, diversity and distributivity of the multi-target particle swarm algorithm for processing the problem of multi-target optimization can be enhanced.

The invention discloses a resource scheduling method in a cloud computing system. According to the method, the position of an updating frog is computed by the aid of a leapfrog updating formula when each sub-population is locally searched, the fitness of the updating frog is computed to judge whether the fitness of the updating frog is superior to that of the worst frog or not, the position of the updating frog replaces that of the worst frog if the fitness of the updating frog is superior to that of the worst frog, the position of the optimal frog in the whole population replaces that of the worst frog if not, whether the fitness of the updated frog is superior to that of the worst frog or not is judged, the position of the updated frog replaces that of the worst frog if the fitness of the updated frog is superior to that of the worst frog, a new step length is generated by a double learning factor formula if not, the new step length is mutated according to mutation probability to obtain the position of the updated frog and replace the position of the worst frog, and the fitness of the updated frog is computed. The method can achieve good performances in optimal time span and load balance for task scheduling. The invention further discloses a resource scheduling system in the cloud computing system.

The invention discloses a wireless sensor network congestion control method based on a PID controller by adopting a computer program. The method comprises the steps of: defining and initializing related parameters; embedding a PID queue management congestion algorithm obtained by combination of a PID control technology and an active queue management method into a wireless sensor network environment; setting fixed parameters (K, K and K<d>) of a PID controller by using a mono-neuron control technology; performing online optimization of the initial parameters (K<P0>, K<I0> and K<d0>) of a neuron PID controller and the neuron learning rates (eta<1>, eta<2> and eta<3>) by using an improved particle swarm optimization; and, after three parameters of the PID controller adaptive to the dynamic wireless sensor network environment are obtained, calculating the abandon probability (P), and abandoning a data packet, wherein the learning factor for optimization by adopting the particle swarm optimization is as follows: C<1>=0.95+0.1*rand, C<2>=C<1>; and the value of a weighting coefficient (w(k)) is adjusted by using a guide Hebb learning algorithm. Thereby, the parameters of the PID queue congestion algorithm are adjusted online; therefore, the parameters are suitable for characteristics of the wireless sensor network; and thus, the purpose of relieving the wireless sensor network congestion is achieved.

The invention discloses a global self-adaptive gray level image enhancement method based on double gamma correction, which adopts a particle swarm optimization algorithm combined with an image doublegamma function to carry out global double gamma correction, and mainly comprises the following steps of: inputting a gray level image and initializing parameters of the particle swarm optimization algorithm; performing double-gamma correction on the input image by adopting each particle position to obtain a preliminary enhanced image, calculating a corresponding fitness value, and updating a historical optimal fitness value and an optimal position of the particle individual and the group; judging whether an iterative optimization termination condition is met or not, and if not, updating the inertia weight, the learning factor, the speed and the position of each particle and continuing iteration; otherwise, performing double-gamma correction on the input image by using the final group optimal position to obtain a final enhanced image. When the low-illumination grayscale image is enhanced, the contrast of the image can be effectively improved, excessive enhancement of a local bright areais avoided, the texture and detail information of the enhanced grayscale image are clear and complete, and the overall visual effect is improved.

The invention discloses a short-period share price prediction algorithm based on an IPSO-BP neural network. The algorithm includes the following steps: sampling historical data and constructing a sample data set; conducting normalization processing on the sample data set and mapping the data to the same interval; conducting fitting and cyclic iteration on the normalized data, and predicting future sample data; determining the number of layers of a network, the number of neurons of each layer, and a BP network parameter and an activation function, dynamically changing inertia weights and learning factors in order from the highest to the lowest, and constructing an IPSO-BP neural network; and according to the future sample data, utilizing the IPSO-BP neural network to predict stock data. The short-period share price prediction algorithm improves the precision, and can effectively increase the convergence speed of the overall network.

The invention discloses a multilayer-perceptron training method based on a bee colony algorithm with a learning factor. In the method, an idea of asynchronous adjustment of the learning factor is introduced to update a nectar source position of following bees so that the self-adaptive optimization capability of a bee colony is strengthened and single extreme values and a global extreme value of the colony are used to improve a search mode in the manual bee colony algorithm so that search efficiency and performance are optimized so as to achieve an optimal optimization effect of an initial weight and threshold of a multilayer perceptron and improve the predication precision of the multilayer perceptron.

An air conditioner includes: a blower configured to discharge air, the blower being connected to an outdoor unit, a parameter generator configured to generate at least one parameter during a time period for which the air conditioner is operated with a first cooling capacity based on a set temperature, a learning unit configured to receive the generated at least one parameter as a learning factor and generate operation mode information, an operation mode controller configured to control at least one of the blower or the outdoor unit based on the generated operation mode information, and a central controller configured to control the parameter generator, the learning unit, and the operation mode controller. The air conditioner is operated with a second cooling capacity after the air conditioner is operated with the first cooling capacity for the time period, the second cooling capacity being different from the first cooling capacity.

The invention discloses a robot path planning method based on spider colony algorithm, comprising of: modeling the map environment by the grid method; initializing the basic parameters of the spider colony algorithm; randomly assigning the initial spider colony; the spider colony completing the collaborative process according to the learning strategy of the spider colony algorithm; outputting fitness value and determining whether the fitness value is less than the set minimum fitness value; and outputting a shortest path and a length. The invention optimizes the neighborhood search and the global search by introducing the inertia weight and the learning factor, and increases the interaction between the inertia weight and the learning factor to balance the global search and local development capabilities of the method. The improved method has more stable convergence results, improves convergence speed and search accuracy, and palns the best shortest path.

The invention provides an economic optimization method of a microgrid containing wind power and photovoltaic power generation. The method comprises the following steps: constructing a microgrid operation data set; establishing an operation optimization target function; establishing a constrain condition of the operation optimization target function; constructing an improved particle swarm optimization algorithm model, and outputting an optimal location; and obtaining the least cost of the optimal solution according to the optimal location and the operation optimization target function, and accomplishing the microgrid economic optimization. Through the microgrid economic optimization method provided by the invention, the diversity of the particle is increased, the capacity of searching theglobal optimum is increased, and the method is hard to trap into the local optimum. And meanwhile, the searching capacity on the optimal solution by each particle is further improved by adopting self-adaptive inertia weight and learning factors, better optimization effect can be acquired when performing economic optimization on this cooling-heating-power cogeneration microgrid, the problem that the economic optimization problem is easy to trap into the local optimum is effectively solved, and the better economic optimization effect is acquired.

The invention discloses a sensor node coverage method based on an improved swarm algorithm. Swarm individuals of different initial points are constructed according to different layout of nodes in a sensor area, the area coverage of the swarm individuals is used as an optimized target function, the asynchronous change idea of learning factors is integrated to the traditional swarm algorithm in the search process, and a search population is guided to jump out of local optimum by sufficiently using historical information and global information of a swarm in the search process, so that the optimization search efficiency and performance are greatly improved, premature convergence of the algorithm is avoided, and intelligent layout optimization of sensor network nodes is realized.

The invention provides an optimized scheduling method and device for a micro-energy grid of an intelligent agricultural greenhouse. The method comprises steps as follows: a scheduling optimizing model of the micro-energy grid is established on the basis of energy flowing models of micro-energy in the micro-energy grid of the intelligent photovoltaic facility agricultural greenhouse with the purpose that the comprehensive operation cost of the micro-energy grid in a day is lowest; the scheduling optimizing model is solved with a dynamic learning factor type second-order oscillation cultural particle swarm algorithm, and the operation scheduling strategy of the micro-energy grid is obtained. According to the optimized scheduling method and device for the micro-energy grid of the intelligent agricultural greenhouse, the scheduling optimizing model is solved with the dynamic learning factor type second-order oscillation cultural particle swarm algorithm, so that the scheduling strategy is obtained, the method and the device have higher operating speed, higher global searching ability and better convergence, and economical operation of the micro-energy grid of the intelligent photovoltaic facility agricultural greenhouse can be realized.

The invention discloses a virtualized wireless network function editing method based on economic benefits. According to the method, a virtualized wireless network function arrangement mathematical optimization model is established, and a virtualized wireless network function arrangement process (VNFPSO) based on economic benefits is provided. Aiming at the openness of a virtualized wireless accessnetwork architecture, the discreteness of a network function and the exponential appearance of a network load, the classical particle swarm optimization algorithm is corrected in the aspects of inertia weight, particle variation, learning factors and the like, and the solving speed of a global approximate optimal solution is increased. According to the invention, the service rejection rate of thevirtualized wireless access network is reduced, and the utilization rate of network system resources is improved.

The invention provides a DV-Hop positioning method based on anchor node selection and a random sampling particle swarm. The method comprises the following steps: enabling an anchor node to broadcast adata packet, selecting an anchor node closest to a target anchor node in an anchor node set as an intermediate anchor node, and estimating a first distance between an unknown node and the target anchor node through the intermediate anchor node; solving a second distance between the unknown node and the anchor node according to a DV-Hop positioning algorithm; performing weighted summation on the calculated first distance and second distance to obtain a final estimated distance; determining the inertia weight of the particle swarm algorithm and the value interval of the two learning factors according to the convergence of the particle swarm algorithm, randomly sampling values in the value space to update the speed position of the particle swarm, and obtaining an optimized particle swarm algorithm; and establishing an objective function by using the final estimated distance, and solving coordinates of unknown nodes by using an optimized particle swarm algorithm. According to the method,the positioning precision is obviously superior to that of an existing algorithm, and the positioning stability is obviously improved.

Provided is a multi-target integrated scheduling method for multi-place transportation in a cloud manufacturing environment. Most of the current cloud manufacturing resource scheduling methods often consider only one optimization objective and ignore the impact of transportation factors on scheduling plans. The method of the invention takes service time, service cost, quality of service and satisfaction evaluation as an objective function, and fully considers cross-region manufacturing resource transfer factors. The method mainly includes the steps as follows: carrying out initialization through a process and resource-based double-layer encoding mode; adding a dynamic learning factor to guide the population search range and accuracy; and updating a global optimal solution set through non-dominated sorting, and introducing a dynamic selection factor to improve the distribution uniformity of a non-dominated solution set. Through the method, a detailed processing plan for a processing task can be obtained. The method is used to solve the multi-target scheduling plan problem in a cloud manufacturing environment. The utilization efficiency of resources is improved, and the processing cost of users is reduced.

The present invention relates to the technical field of reactive power compensation and particularly relates to a low voltage reactive power compensation calculation method based on improved particleswarm algorithm. Based on the optimization of parameters of an inertia weight omega and a learning factor c in a multi-objective particle swarm optimization, a power factor COS phi and a voltage distortion ratio THD are used as objective functions, an optimized compensation capacity Qc is calculated, and then the calculated compensation capacity Qc is inputted into a reactive power compensation device. The amplification of harmonic waves by the reactive power compensation can be reduced, the influence on grid parameters is reduced, and thus the effect of reactive power compensation is improved.

A remains 3D reconstruction method based on a frontal image is characterized by a single picture 3D face reconstruction technology based on a statistical model. The method comprises creating a 3D facestatistical model based on a 3D face database in advance, and training a 2D face feature point parameter model by using a SDM algorithm based on a 2D face database. During the face reconstruction based on a single frontal image, the face feature points are extracted first by using a two-dimensional image feature point parameter model; then, an energy function is subjected to iterative optimization by an learning factor adaptive gradient descent method based on the three-dimensional face statistics model to obtain the parameterized vector of the statistical model, that is, the three-dimensional face model corresponding to the two-dimensional face image is obtained. Results show that the face reconstructed by the method has high similarity.

The invention discloses a screw type material discharging device based on variable-rate learning and a controller of the screw type material discharging device. The discharging device comprises a rack, a discharging bin, a screw conveyor, a metering hopper, a weighing module, a discharging valve, a material mixing hopper, a controller, a material storage in and a feeding pump. The upper portion ofthe metering hopper is provided with a funnel-shaped material distributor, and the discharging bin is internally provided with a distance sensor and a stirrer. The controller is used for controllingthe screw conveyor through iterative learning based on the discharging error of the single time and the accumulated discharging error, and dynamic adjustment is conducted on the learning factor of thedischarging error of the single time and the learning factor of the accumulated discharging error based on the changes of the discharging error in the iterative process. By means of the screw type material discharging device based on variable-rate learning and the controller of the screw type material discharging device, the distance sensor and the stirrer are used for detecting and the adjustingmaterial accumulation in the discharging bin, stability of compactness of materials is ensured, and the learning factor can be automatically adjusted and optimized. Compared with the prior art, repeated parameter cut and try are not needed, an iterative approach good in convergence performance can be rapidly obtained, and falling materials can be effectively utilized in the iterative learning process.