68results about How to "Avoid local optimum problems" patented technology

The invention relates to a breaker fault diagnosis method based on separating / closing coil current signals, which includes the following steps: 1, collecting a primary fault dataset and performing normalization processing to the collected data; 2, adopting the algorithm P-KFCM combined by the algorithm PSO (particle swarm optimization) and the algorithm KFCM (kernel fuzzy C-mean clustering) to classify the data set, so as to obtain a primary fault sample membership matrix and classifying the dataset into C types according to the degree of membership; 3, verifying the cluster number C through the cluster effectiveness index MPC, and setting the cluster number C corresponding to the maximum value of MPC as the optimal cluster number; 4, building an SVM predicating trainer through data samples of every two types in the optimal classification, so as to obtain C*(C-1) / 2 SVM training models; 5, inputting the collected test samples in the SVM training models, and obtaining the primary testing results; 6, obtaining the final fault diagnosis by combining all the testing results. Through the adoption of the method, the accuracy of fault classification can be improved, and the on-line detection on fault types of breakers can be realized.

The invention discloses a fast color image segmentation method based on combination of clustering analysis and an image pyramid. The method comprises the following steps: 1) constructing a Gauss and Laplace image pyramid according to a source image, and obtaining a low-resolution image and an edge image; 2) transforming the low-resolution image to an HSV (hue, saturation and value) space, obtaining the clustering number and an initial clustering center through histogram analysis, and performing clustering segmentation on the low-resolution image in the HSV space by use of a clustering algorithm; 3) performing up-sampling on the segmentation result obtained in the step 2), projecting to the original resolution, performing spatial filtering and removing the excessively small area to obtain the original resolution area segmentation result; and 4) integrating the edge image obtained in the step 1) and the area segmentation result obtained in the step 3) to obtain the final segmentation result. The method disclosed by the invention guarantees the segmentation quality while increasing the color image segmentation speed, and is suitable for a digital image processing system with high requirements on real-time property.

The present invention relates to the technical field of electrical engineering, and more particularly, to a method based on CEEMD-VMD-GA. The short-term wind speed prediction method based on ORELM model firstly obtains the historical wind speed data and pretreats the data, then decomposes the historical wind speed data into a series of discrete modes with specific sparse attributes by using the complementary empirical mode decomposition and variational mode decomposition. Then genetic algorithm is used to optimize the outlier robust limit learning machine prediction model to predict all the subsequences in one step. Finally, all the predicted values of the sub-sequences are superposed and the actual predicted results are obtained. The invention utilizes two-layer decomposition of complementary empirical mode decomposition and variational mode decomposition to reduce non-stationarity and non-linearity of wind speed series, Genetic algorithm is used to optimize the outlier robust limit learning machine to form a hybrid model for single-step prediction, which reduces the influence of complex characteristics of wind speed series on prediction results, improves the accuracy of short-term wind speed prediction, and solves the problem of local optimization of neural network.

The invention discloses an electric vehicle power battery SOC intelligent detection device which is characterized in that the intelligent detection device includes a battery parameter acquisition platform and a battery SOC estimation system, the battery parameter acquisition platform collects real-time parameters of voltage, current and temperature of an electric vehicle power battery group, the battery SOC estimation system estimates an SOC value through the collected parameters, and a battery SOC system is a nonlinear, time-delay and multivariable coupling complex real-time system with a high requirement. According to the device, a problem that a conventional detection device can not obtain an ideal effect of the intelligent detection of the electric vehicle power battery SOC.

The invention discloses a resource allocation optimization method based on reinforcement learning. The method comprises the following steps: acquiring a bandwidth value of a downlink, acquiring the number of physical resource blocks which can be called within a single transmission time interval according to the bandwidth value, and acquiring the number of user services to be transmitted, the characteristics of the user services to be transmitted on an nth physical resource block at the current t moment, and Characteristics of the whole downlink at the t-1 moment; judging whether the bandwidthutilization rate of the downlink needs to be improved; and if the bandwidth utilization rate of the downlink needs to be improved, whether the fairness of the downlink needs to be improved or not, orthe trade-off of the bandwidth utilization rate and the fairness of the downlink needs to be realized, and if the bandwidth utilization rate of the downlink needs to be improved, inputting the characteristics into the trained bandwidth utilization rate reinforcement learning model to obtain the metric value of the ith user service on the nth resource block. The technical problem that the scheduling performance is affected due to the fact that an existing algorithm only considers the local optimal solution condition caused by optimal allocation of a single resource block can be solved.

The invention discloses a parking AGV path planning method based on an improved ant colony algorithm. The method comprises the following steps of: 1) modeling a parking environment, 2) initializing system parameters, 3) improving a pheromone updating strategy, 4) improving heuristic information strength, and 5) using an improved ant colony algorithm for path planning on the parking AGV. Accordingto the method, a time optimal path from a starting point to a target point is provided for searching of the parking AGV, and an AGV system is ensured to accurately and rapidly complete vehicle storing, taking and parking tasks in a short time.

The invention discloses a PSS4B parameter setting method based on a hybrid particle swarm optimization algorithm. First, a PSS4B double-input model is transformed into a single-input model with speed deviation being an input signal, and then a phase angle of a transfer function of the single-input model is solved, namely phase-frequency characteristics of PSS4B; different optimization variables are selectively set targeting a three-frequency-band two-order lead-lag phase compensation link time constant for a conventional self-excited excitation system and a three-machine excitation system with large uncompensated phase-frequency characteristic lag, and a PSS4B parameter optimization model is established; and the hybrid particle swarm optimization algorithm is utilized to perform PSS4B parameter setting, particles in a specified number are selected according to hybrid probability and placed into a hybrid pool in every iteration, every two particles are hybridized to generate filial particles in the same number, and the filial particles are used for replacing parental particles to improve optimizing capacity. Through the method, phase compensation requirements of different types of excitation systems can be met efficiently and conveniently.

The invention discloses a multi-unmanned aerial vehicle collaborative target tracking method combining an improved APF and segmented Bezier. The method comprises the steps that firstly, the position of a target and the position of an obstacle are detected by utilizing an airborne camera and a laser radar; secondly, a model of target gravity and obstacle repulsive force currently borne by unmannedaerial vehicles is established, and a self repulsive force potential field of each unmanned aerial vehicle is built; thirdly, the resultant force borne by the unmanned aerial vehicles is worked out according to the gravity and repulsive force currently borne by the unmanned aerial vehicles, and the unmanned aerial vehicles with paths trapped in local optimum are enabled to escape local optimal points through a virtual obstacle; fourthly, the flight angle of the unmanned aerial vehicles at the next moment is worked out, and next waypoint positions of the unmanned aerial vehicles are calculated;finally, by using a segmented Bezier curve, online smooth optimization is carried out on an air route, optimized next waypoint positions of the unmanned aerial vehicles are obtained, and the steps are repeated till all the unmanned aerial vehicles track the target. The method mainly solves the problem of collisions among the unmanned aerial vehicles in the multi-unmanned aerial vehicle collaborative target tracking process, and meanwhile, the phenomenon of air route oscillation in the tracking process is eliminated.

The invention discloses an electric automobile power battery SOC (State of Charge) detection system. The characteristics lie in that the detection system comprises a battery parameter acquisition platform and a battery SOC estimation system, the battery parameter acquisition platform is responsible for real-time parameter acquisition for the voltage, current and temperature of an automobile power battery pack, and the battery SOC estimation system can accurately estimate a battery SOC value through the acquired parameters; and the battery SOC is a non-linear, delayed, multivariable coupling and complex highly demanding real-time system. The detection system effectively solves a problem that the traditional automobile battery SOC estimation method is difficult to achieve an ideal effect.

The invention discloses a fault line selection method for a power distribution network containing distributed generation based on a deep belief network. The method comprises the steps of acquiring zero-sequence current and voltage signals of the head end of each feeder line under different single-phase grounding fault conditions; extracting steady-state and transient-state characteristic quantities contained in the zero-sequence current and voltage signals; performing data and polarity pretreatment on the original characteristic quantity; and performing fault line selection and positioning byadopting a deep belief network algorithm. According to the invention, the deep learning theory is applied to small-current grounding fault positioning containing the distributed generation, better fault tolerance is achieved, the accuracy of small-current grounding fault line selection can be improved, and the reliability and the safety of the operation of the power distribution network are improved at the same time.

The invention discloses a self-organizing TS-type fuzzy network modeling method applied to infrared flame recognition, comprising the following steps: (1) collecting time-domain signal data of different flames and interference sources, and preprocessing the time-domain signal data to obtain frequency-domain signal data; (2) extracting the characteristic information from the time-domain and frequency-domain signal data of the waveform, obtaining the characteristic vector of the flame, and composing a sample set; (3) dividing the sample set into training set, verification set and test set; (4) building TS-RBF fuzzy neural network; (5) setting TS-RBF fuzzy neural network parameters initial value, using the training set of samples to TS-RBF fuzzy neural network training, structure, parameter learning; (6) using verification sets to test the training TS- Verification and Model Selection of RBF Fuzzy Neural Network; (7) inputting the test set into the trained TS-RBF fuzzy neural network, theresults as the final evaluation of the model.

The invention discloses an intelligent SOC (State of Charge) prediction device for an electric vehicle power battery, which is characterized by comprising a battery parameter acquisition platform and a battery SOC prediction system, wherein the battery parameter acquisition platform is used to acquire real-time parameters of voltage, current, temperature, and ambient temperature of the vehicle power battery pack; and the battery SOC prediction system is used to predict the battery SOC value through the acquired real-time parameters. The battery SOC is a nonlinear, delayed, multivariable-coupling, and complex real-time system with extremely high real-time performance requirements. The problem that the conventional prediction device can not achieve ideal battery SOC prediction precision effects can be effectively solved.

The invention provides optimization processing for a search target task of multiple unmanned aerial vehicles, wherein the optimization processing comprises the steps: enabling the prior probability distribution of a search target of single unmanned aerial vehicles to serve as a prior search graph of a search region, carrying out iterative updating of the prior search graph, and obtaining updated search graphs of the single unmanned aerial vehicles; calculating the communication probability between the unmanned aerial vehicles, determining the communicability between any two unmanned aerial vehicles based on the communication probability, and determining an unmanned aerial vehicle communication network; fusing the search graphs among the unmanned aerial vehicles based on the communication probability among the unmanned aerial vehicles to obtain a fused search graph; updating the fused search graph in combination with the target motion to obtain a fused updated search graph; and enablingthe single unmanned aerial vehicles to update the search graph based on fusion, optimizing the positions of the unmanned aerial vehicles and guiding the unmanned aerial vehicles to fly. The method comprehensively considers and optimizes the communication performance and the search performance in the search task of the multiple unmanned aerial vehicles, and is more suitable for the search of a moving target, especially the target search task in the actual complex marine environment.

The invention discloses a fault diagnosis method for an oil immersed reactor based on an IFOA optimized SVM model. The method comprises the following steps: preprocessing fault diagnosis data of the reactor, constructing training samples and test samples; establishing a fault diagnosis model of the reactor based on a support vector machine; improving a fruit fly algorithm; establishing a fault diagnosis model of the oil immersed reactor based on IFOA-SVM; and substituting sample data into the fault diagnosis model to achieve fault diagnosis of the reactor. The fault diagnosis method for the oil immersed reactor based on the IFOA optimized SVM model proposed by the invention is suitable for training of small data set of samples. The improved fruit fly algorithm realizes dynamic balance between the global search capability and the local optimization ability, and can effectively avoid the problem of local optimum. The improved fruit fly algorithm is applied to the selection of a penalty parameter C and a kernel function g of the support vector machine, which can realize the optimal selection and adaptive selection of parameters.

The invention discloses a remote sensing mass image automatic screening method based on an optimized genetic algorithm, and the method comprises the steps: automatically screening an image solution meeting the demands through employing the genetic algorithm according to the requirements of a user for the coverage area, the resolution coverage rate, the cloud amount coverage rate and the timelinesscoverage rate; firstly, obtaining an initial image set, and then conducting coarse screening; then, modeling in a binary coding mode, and generating an image solution set through initialization; andthen calculating a comprehensive score, and calculating a probability according to the score for selection; solving a new image solution set for the selected image solution set through crossover variation; circulating the process, and selecting a final solution when a circulation ending condition is reached. According to the method, the problem of local optimization caused by a traditional greedyalgorithm is solved, the problem that the solving overlapping rate is too high or the coverage rate is low is avoided, weight configuration can be carried out on preferences in multiple aspects such as the coverage area, the cloud coverage rate and the timeliness coverage rate, and different requirements of different users for final screening results are met.

The invention discloses a WRSN multi-mobile charger optimal scheduling method based on reinforcement learning, and belongs to the field of Internet of Things wireless rechargeable sensor networks. According to the invention, the neural network and the multi-agent reinforcement learning method are introduced into the charging path planning under the scene of multiple mobile chargers of the wireless rechargeable sensor network, and the multi-agent reinforcement learning is mainly utilized to solve the problems of efficient cooperation and optimal scheduling of the multiple mobile chargers in the wireless rechargeable sensor network. Under the condition that the energy of the mobile chargers and the sensor is limited, the charging paths of the chargers are optimized through mutual cooperation of the multiple mobile chargers, and the sensor node with the low electric quantity is charged in time. On the premise of ensuring that the sensor nodes do not die due to power shortage, the total moving path of each mobile charger is the shortest, and the charging efficiency is optimized on the whole.

The invention relates to the technical field of wind power prediction, in particular to a wind power output prediction method based on a historical predicted value. The method comprises the following specific steps: S1, building a BP neural network model, and carrying out the data quality analysis of historical measured data; s2, verifying BP neural network example indexes according to the data RMSE and nMAE; s3, after the index verification is passed, fragmenting the collected long-term historical data according to time, optimizing a neuron threshold by using a particle swarm algorithm, and performing network iteration neuron parameter correction by using a historical data fragmentation iteration process; and S4, after the parameter model is corrected, determining a predicted output time period according to an actual demand, carrying out time sequence series connection on historical data and a wind power prediction value of the wind power plant, taking a series connection result as input of the neural network model, and carrying out actual prediction value output and verification. The predicted value is closer to the actual value, and the requirements of current development are met.

The invention relates to a long-distance high-timeliness economical cruise vehicle speed planning method, which comprises the following steps of: firstly, acquiring road coordinates and altitude dataof a future distance through a global positioning system and a geographic information system according to travel planning; then, adopting a composite second-order cubic Bezier curve for fitting to generate a position-based road gradient and an initial vehicle speed curve; secondly, encoding and decoding the vehicle speed curve by adopting a genetic algorithm, and taking a cruise control curve as an initial population; randomly generating a vehicle speed curve by adopting selection, crossover and mutation algorithms, and evaluating the fuel consumption fitness; and obtaining an optimal vehiclespeed curve through iterative computation, and storing the optimal vehicle speed curve in a vehicle control unit memory. Finally, a PID controller is adopted for optimal vehicle speed tracking control, a vehicle speed curve is randomly generated through a genetic algorithm, fuel consumption evaluation is conducted, and dynamic planning or quadratic planning is avoided.

The invention provides a method and system for realizing global optimization of AODV routing, and a medium. The method comprises an updating process of an initiating node Q table and a process of calculating approximate global optimization routing by a destination node, wherein the updating process of the initiating node Q table is as follows: a distribution value function is introduced into an initiating node to optimize an updating strategy of a Q value, and a weight parameter is calculated by establishing a trajectory prediction model by combining Kalman filtering with a Gaussian filteringalgorithm, so the updating process of the initiating node Q table is completed; the process of calculating the approximate global optimization route by the destination node is as follows: the destination node returns the route with the maximum average Q value in the cache within a certain time to the source node. According to the invention, the robustness and the information transmission success rate of the network are improved, and the end-to-end delay of the network is reduced.

The invention discloses a method for accurately obtaining the relationship between the output current and the output power of a fuel cell system. The method comprises the steps: selecting a proper curve function equation to be fitted according to an output current output power characteristic curve obtained by the calibration experiment; combining an experiment and simulink fuel cell modeling simulation to obtain an output current and an output power value; and performing parameter solution of a to-be-fitted curve on the output current and the output power by using a particle swarm algorithm to obtain a relation curve of the output current and the output power of the fuel cell system. The method has the advantages that the reliability of data and models is improved, output current and output power values under different working conditions can be obtained by changing experimental conditions in simulation, the method is used for multi-working-condition research, the adaptability is improved, and the research cost is reduced. A particle swarm algorithm is adopted to solve parameters of a to-be-fitted curve, so the calculation efficiency and the curve precision are improved, and the accuracy of output power control by utilizing the curve is further improved.

The invention discloses an end-to-end indoor mobile robot autonomous navigation method in a non-structural environment. The end-to-end indoor mobile robot autonomous navigation method comprises the following steps of 1) acquiring a plane layoutdiagram in a strange building, 2) carrying out topologization on the layout diagram, and creating an undirected graph, 3) generating a shortest node path by using the undirected graph and combining with the target room, 4) inputting the first visual angle image of the robot into an anti-collision navigation module, and generating an anti-collision road point, (5) generating an anti-collision path through cubic splines, and tracking through model predictive control, and (6) judging whether the target point is reached or not, if so, ending navigation, otherwise, returning to the step (3). By adopting the autonomous navigation method of the robot, autonomous navigation can be realized only by using a single monocular RGB camera as visual input in a complex medical environment. The method has the advantages of simplicity, flexibility, economy, practicability, high accuracy and strong robustness.

The invention relates to a wind power prediction method. The method comprises the following steps: S1, decomposing an original wind power time sequence according to local mean decomposition; S2, constructing a respective training data set and a respective test data set for each PF component and margin; S3, establishing a wind power prediction model of an improved differential algorithm optimization extreme learning machine for each PF component and margin; S4, inputting the training samples of the plurality of PF components and the allowance in the training data set into a wind power prediction model of an improved differential algorithm optimization extreme learning machine one by one for training to obtain respective corresponding wind power prediction sub-models; S5, inputting the testdata into the corresponding wind power prediction sub-models for prediction; and S6, performing combination superposition processing on the prediction output value of each wind power prediction sub-model. According to the wind power prediction method, the influence of strong nonlinearity of the original wind power data on the prediction result is effectively reduced, and the wind power predictionresult with higher precision is obtained.

The invention relates to a G protein-coupled receptor topology calculation prediction method based on a feedback type conditional random field. The prediction method comprises the steps of 1, data set preparation, 2, data preprocessing, and 3, the feedback type conditional random field which comprises 31 a basic conditional random field model, 32, a feedback type conditional random field frame and 33, a feedback mechanism and algorithm. The GPCR topology structure prediction method is based on the FCRF; the joint probability model based on the conditional random field does not need to carry out independence assumption on an observation sequence, so that the local optimum problem of a hidden Markov model is solved; in addition, the basic conditional random field is improved, the feedback mechanism is introduced, and the modeling ability of the conditional random field is improved in the uninterrupted feedback, so that finally, GPCR topology prediction precision is improved.

The invention discloses a fuzzy reasoning-based in-vehicle network routing establishment method which is mainly used for solving the problem that in the prior art, the data transmission performance is low due to the fact that a reliable communication link is not selected. According to the scheme, the method comprises the steps that 1, a network is initialized; 2, neighbor node movement information is acquired; 3, a source node sends a routing request packet; 4, node-to-node link reliability values are computed by means of fuzzy reasoning according to the step 2, and the link reliability values are updated; 5, whether a current node is a target node or not is judged, if yes, the step 6 is executed, and otherwise, the step 4 is executed; 6, the target node selects the communication link with the highest reliability value and sends a routing reply packet; and 7, whether a routing reply data packet reaches the source node or not is judged, if yes, a routing path is established to complete data transmission, and otherwise, the step 6 is executed. According to the method, the data packet delivery success rate of the network is increased, the network transmission time delay is shortened, the stability and high efficiency of the communication links are guaranteed, and the method can be applied to data communication.

The invention relates to an ant colony graph matching method based on a G-W distance. The method comprises the following steps: selecting feature points of a to-be-matched graph; generating an H matrix under a G-W distance through the sampling information data of the feature points; establishing a secondary convex optimization model for the minimum difference degree of the G-W distance between thefeature points, and proposing the hypothesis of the model in combination with the ant colony; stipulating physical constraint conditions of the ant colony in a G-W distance minimum difference degreeoptimization process, so that the ant colony meets a one-to-one correspondence matching principle; accumulating distances among city nodes traversed by ants; optimizing the quadratic convex optimization problem under the G-W distance, and solving a shortest path and city nodes on the path; and defining the shortest path as the overall difference degree between the two manifold curved surfaces, andcalculating the matching relationship between the feature points by utilizing the relationship between the city nodes on the shortest path and the serial numbers of the feature points to complete graph matching. The method is the optimal solution closest to the real theory, and the dependence of traditional numerical iteration optimization on the initial solution is overcome.

The invention discloses a 5G communication adaptive transmission method based on long-short-term multi-threshold channel state prediction. The method comprises the following steps: 1, collecting historical channel statistical property samples and channel state samples corresponding to the historical channel statistical property samples through 5G radio equipment; 2, constructing a neural network based on LSTM; 3, updating neural network parameters of the LSTM through an adaptive gradient descent algorithm; 4, proposing a multi-threshold algorithm to optimize the network; 5, dynamically selecting a redundant block according to the predicted channel state to improve the reliability of data transmission. According to the invention, the channel prediction takes time correlation into consideration while fitting the transmitting signal model to the receiving signal model, so that the prediction precision can be improved, the proportion of redundant blocks can be dynamically adjusted according to the channel state, and the reliability of data transmission in 5G communication is improved.

The invention relates to a pulmonary nodule benign and malignant identification method, and particularly relates to a pulmonary nodule benign and malignant identification method based on support vector machine sample reduction. The method comprises the steps that an original sample set S0 of malignant and benign pulmonary nodules is acquired; sample reduction is carried out for the original sample set S0 of malignant and benign pulmonary nodules, so as to acquire a final train set S2 of malignant and benign pulmonary nodules of a support vector machine; support vector machine train is carried out on the final train set S2 after reduction, so as to acquire a final classification decision function; and support vector machine prediction is carried out on an unknown pulmonary nodule sample xi', so as to acquire a pulmonary nodule benign and malignant identification result. According to the invention, the method of support vector machine sample reduction is provided to improve the train speed of the support vector machine; a space storage requirement is reduced; the pulmonary nodule benign and malignant identification time is reduced; and the diagnosis efficiency and the objective consistency of doctors are improved.

The invention provides an air target potential threat degree prediction method and system based on SSA-BP, and the method comprises the steps: carrying out the information collection of a target, and obtaining the real-time feature data of the air target; predicting to obtain future state data of the air target; identifying emotional information of our pilot at the current moment; constructing a threat degree data set of the air target in two-level fusion of the emotion of the pilot and the features of the air target, and carrying out quantification and normalization processing to obtain a standard target feature-emotion data set; and optimizing a BP neural network structure through a sparrow search algorithm, carrying out fusion analysis on future state data of the target and emotion information of a commander, and obtaining the threat degree of the target at the future moment from the situation levels of the two parties. The efficient threat degree prediction method is constructed from the aspects of two parties, and the method is simple, high in tracking speed, high in interactivity, easy to implement and capable of effectively coping with interference of aerial targets.

The invention discloses an improved ShuffleNet-based radar signal RD image interference identification method and system, and the method comprises the steps: constructing a radar signal RD image data set under an interference condition, and dividing the radar signal RD image data set into a training sample and a test sample; constructing an attention-enhanced ShuffleNet recognition model, and setting a loss function of a classification task of the attention-enhanced ShuffleNet recognition model; training a ShuffleNet recognition model of which the attention is enhanced by utilizing the training sample; and inputting a test sample into the trained attention-enhanced ShuffleNet recognition model, predicting and outputting a test sample category and a confidence score, and completing radar signal RD image interference recognition. The attention enhancement module is introduced, interaction between feature channels is enhanced from the channel information interaction level, the feature discrimination is improved, and therefore the RD image recognition precision of the image radar signal under interference is improved.

The invention relates to a gear reducer optimization method based on an FOA (FruitFly Optimization Algorithm). The method comprises the steps of taking a two-stage gear reducer as a design research object, and establishing a mathematic model with a minimum volume of the two-stage gear reducer as an optimization target; introducing a correction factor described in the specification into the FOA in combination with the FOA to obtain a corrective FOA; and then optimizing the mathematic model established in step S1. The gear reducer optimization method based on the FOA has the advantages of rapid convergence speed and effective avoidance of a local optimum problem on the premise of guaranteeing safety and reliability.