56results about How to "Not easy to fall into local optimum" patented technology

The invention discloses an improved ICP object point cloud splicing method for fusing fast point characteristic histogram. The method comprises the steps of projecting a standard sinusoidal digital grating onto the surface of the object to be measured, photographing stripe images of the surface of the object projected with the standard sinusoidal digital grating from different angles of view by aCCD camera, and obtaining photographing point clouds from multiple angles of view; for two image point clouds that need to be stitched together, building a k-D tree and interpolate to obtain that interpolated point cloud; for the two interpolated point clouds to be spliced, computing the fast point feature histogram, and obtaining the point cloud by random sampling consistent transformation; usingthe improved iterative nearest point method to obtain the first interpolated point cloud which is precisely registered; overlaying point cloud and mesh to realize the mosaic of two different angles of view of the shooting point cloud. The invention has low requirement for the initial position of the splice point cloud, the robustness is remarkably improved, the local optimization is not easy to fall into, the splice accuracy is improved, and the precise splicing of the point cloud under multi-view angles is realized, so that the practical industrial application requirements can be met.

The invention discloses a fault diagnosis method of a high-voltage circuit breaker based on a depth belief network, which comprises the following steps: step 1, selecting a data sample required by anexperiment, and dividing the unified standardized sample data into a test sample and a training sample according to a specific proportion; Step 2: building and initializing the DBN deep belief networkfault diagnosis model; Step 3, inputting a large number of unlabeled samples or unlabeled samples in the pre-training set from the bottom of the model, and pre-training the RBM in the model by usinglayer-by-layer unsupervised greedy learning; Step 4: the whole model being fine-tuned by genetic algorithm; Step 5, the fault diagnosis model of the high-voltage circuit breaker obtained by training being classified to the fault samples of the test set in step 1, so as to obtain the fault classification result, and the diagnosis accuracy rate of the model being counted. The invention discloses a fault diagnosis method of a high-voltage circuit breaker based on a depth belief network, which can train a large amount of data samples to realize the fault diagnosis function of the high-voltage circuit breaker.

The invention discloses a self-adaptive large neighborhood search method for AEOS (agile earth observation satellite) scheduling. The method comprises following steps: Step 1, a constraint satisfaction model of AEOS scheduling problems is established; Step 2, an initial liquid is solved and taken as a current liquid; Step 3, adjustment is performed in a neighborhood of the current liquid: a deletion operator is selected, part of masks is deleted, and a destruction liquid is obtained; an insertion operator is selected, masks are inserted into the destruction liquid, and a new liquid is obtained; Step 4, the benefit of the new liquid is calculated and compared with benefits of the current liquid and an optimal liquid, and the current liquid and the optimal liquid are updated; Step 5, operator scores are updated according to operator performance; operator weights are updated after iteration is performed preset times each time; Step 6, the optimal liquid is output if an end condition is met, otherwise, the Step 3 is executed. Aiming at the time-dependent characteristic of the AEOS scheduling problems, the constraint satisfaction model of the AEOS scheduling problems is established by adopting the maximum observation benefit as a target, a liquid framework based on self-adaptive large neighborhood search is provided, and a conflict-free scheduling scheme is generated.

The invention provides a ship route navigational speed multi-task comprehensive optimization method. The ship route navigational speed multi-task comprehensive optimization method includes the steps:obtaining basic information including meteorological and sea conditions, geographic conditions, a recommended initial route and ship basic information during navigation; preprocessing the data; dividing each route segment of the recommended route at equal longitude; obtaining a plurality of equal division points as to-be-optimized steering points; establishing a multi-objective optimization modelwith a wave height penalty function by taking oil consumption and navigation time as objectives; and inputting related data into the model, solving the model by using a multi-objective evolutionary algorithm, obtaining a Pareto solution set of the optimal air route by adjusting the latitude position of each steering point to be optimized and optimizing the water navigational speed of each segment,and finally obtaining the optimal air route according to the requirements of a client. Compared with a traditional route optimization method, the route optimized through the ship route navigational speed multi-task comprehensive optimization method can effectively avoid meteorological severe areas, navigation risks are reduced while navigation oil consumption and navigation time are optimized, and navigation cost is reduced.

The invention is an unmanned aerial vehicle cluster formation control method based on non-inferior solution pigeon colony optimization, which comprises the steps of 1, unmanned aerial vehicle clusterformation modeling, 2, unmanned aerial vehicle cluster formation state prediction, 3, non-inferior solution pigeon colony optimization method parameter initialization, 4, design of a non-inferior solution pigeon colony optimization-based method, 5, design of a non-inferior solution pigeon colony optimization-based unmanned aerial vehicle cluster formation RHC controller, and 6, unmanned aerial vehicle cluster formation control method result output. Thus, a real-time and online unmanned aerial vehicle cluster formation controller optimization method is provided, and thus, the unmanned aerial vehicle cluster formation control level in a complex battlefield environment can be effectively improved.

The invention relates to an improved particle swarm algorithm based inverse kinematics calculation method for a permanent magnetic spherical motor. The method is carried out by the following steps of: step 1: determining a coordinate position of an output shaft of a rotated rotor according to initial position coordinates of the output shaft of the rotor and an obtained Euler angle, and taking a distance between the given coordinate position of the output shaft of the rotated rotor and an actually obtained coordinate position as a fitness function; and step 2: calculating the Euler angle corresponding to inverse kinematics of the permanent magnetic spherical motor by applying an improved particle swarm algorithm based on a simulated annealing algorithm. The method can effectively calculate a local optimal solution and has relatively high calculation precision.

The invention provides a multi-target coal distribution method based on uniform design. The method is carried out according to the following steps that the non-linear mapping relation between single-coal-type information and multiple-coal-type information is built through a BP neural network, and coal quality information of mixed coal is predicted; target factors are selected, corresponding weights are designed through uniform design, and a multi-target function is established; the predicted coal quality information of the mixed coal is substituted into the multi-target function to obtain target function values, the maximum target function value is found out through the genetic algorithm, and an optimal coal mixing plan is acquired; the fuel coal consumption corresponding to the optimal plan is calculated. The multi-target coal distribution method based on uniform design solves the problems that a single-target model can not meet the requirements of being economical and environmentally friendly, saving resources, achieving efficient operation of boilers and the like.

The invention discloses a non-invasive household appliance load identification method based on a bee colony algorithm. According to the method, a non-intrusive load identification device is used for carrying out real-time load input removal event detection at a home-entry place; when a load input event is detected, electrical parameters on a bus are recorded; wherein the data include current effective values, active power, reactive power, current harmonics and the like, the device sends the data to the cloud after acquiring the data, and the cloud matches the data with data in a database through an artificial bee colony algorithm and sends an identification result back to the device, so that the purpose of household appliance load identification is achieved. The method is high in flexibility and high in reliability, the misjudgment rate and the missed judgment rate of the load can be effectively reduced, and powerful technical support is provided for load management of a power grid side and a user side.

The invention discloses a shafting dynamic balance multi-target optimization method based on a differential search algorithm. The method is characterized by utilizing the accuracy, stability and rapidness features of the DS (differential search) algorithm to the multi-target optimization of rotation mechanical shafting balance. The method utilizes three targets of the residual oscillation quadratic sum, the residual oscillation maximum value and the difference of the residual oscillation maximum value and residual oscillation minimum value to weigh the machine residual oscillation state; the multi-target optimization is carried out through the DS algorithm; and the quality of the balance effect is comprehensively evaluated, and the defect that with the influence factor method in the past, only the residual oscillation quadratic sum minimum is served as a single balance target in the dynamic balance is solved.

The invention discloses a model-free adaptive control method based on a basic loop of an ore grinding process. The control method is applied to a cyclone ore feeding concentration control loop in theore grinding process; a grey wolf optimization algorithm is combined with model-free adaptive control; and meanwhile, the model-free adaptive control algorithm and the grey wolf optimization algorithmare improved. The related parameters of the IMFAC algorithm are optimized by adopting the IGWO algorithm, so that the control precision of the IMFAC algorithm and the optimal value of parameter selection are guaranteed; the cyclone ore feeding concentration is ensured to be stabilized near an expected value, so that the cyclone ore feeding concentration in the ore grinding process can be better tracked; and the method is better in control effect, wider in applicability and relatively strong in robustness. In the actual ore grinding process, the link of manual parameter adjustment is removed,so that the control process is more efficient, and the applicability is wider.

The invention provides a task scheduling method for a cloud computing platform. The method comprises the steps that all tasks nodes in a DAG (directed acyclic graph) are traversed, and static priorities of all node tasks are sequentially obtained through calculation; the tasks are arranged according to the descending order of the static priorities, and the tasks are sequentially placed into a taskpriority queue; and for each task in the task priority queue, as long as execution starting time of the current task can be advanced and scheduled tasks are not delayed, a parent task of the currenttask is copied instead of only copying a key parent task of the current task. Compared with a traditional algorithm, the starting stage of task duplication is advanced, virtual machines are selected after the tasks are duplicated, therefore, each task can be executed on the virtual machine enabling the task to be completed fastest, and selection of the virtual machines is more reasonable; and loadbalance is considered at the selection stage of the virtual machines, so that good load balance of a cloud computing system is realized, and the resource utilization rate of the cloud computing system is increased.

The present invention discloses an improved crisscross optimization algorithm-based multi-objective reactive power optimization method and system. The method comprises the steps of calculating target values of each particle in an initial population, wherein the target values at least comprise target values of an active power network loss, a voltage offset and a voltage stability margin; performing horizontal cross and vertical cross on the initial population so as to generate sub-generation W and sub-generation R; screening the sub-generation R to obtain an excellent particle population; and combining the initial population, the sub-generation W and the excellent particle population so as to generate a population pool, selecting a new generation of population by using non-dominated sorting and crowding distance, and outputting a final result when an iteration number of times is greater than a preset threshold. In the method, the active power network loss, voltage offset and voltage stability margin are all considered in reactive power optimization of the system, and the system is optimized by using the improved crisscross optimization algorithm, so that multi-objective reaction power optimization is realized, and the algorithm is less likely to optimize locally.

The invention relates to a cloud workflow task execution time prediction method based on limit gradient improvement, and belongs to the technical field of cloud computing. According to the method, influence factors of task execution time are classified from three levels of workflow task composition, resources on which task operation depends and a physical execution environment of the resources, and comprehensive modeling of the influence factors of the task execution time is achieved. Secondly, aiming at the condition that the sample data set has a data missing value, the data set with the missing value is complemented by adopting a machine learning method; and finally, by means of the multi-type data processing capability of the extreme gradient lifting algorithm, the parameter design isrelatively simple, the calculated amount is small, the advantages of a serial learner and a parallel learner are combined, and a cloud workflow task execution time prediction model is trained by adopting the extreme gradient lifting algorithm. Compared with an existing prediction model, limitation on the sample data type is reduced, prediction errors are reduced, and the generalization ability ofthe model is further improved.

The invention discloses a power system load modeling parameter identification method based on a grey wolf algorithm. Aiming at the disadvantage that the identification accuracy of an existing load model parameter is insufficient, the grey wolf algorithm is adopted for optimizing to-be-identified load parameters; based on the original polynomial model, the consideration of frequency variation is added to ensure that the load characteristics can be described accurately even when the load voltage deviates from the reference voltage. Compared with a traditional static model which is too optimistic, through the input voltage and the output response of the actual system, a comprehensive load model can better reflect the load characteristics and has the characteristics of fast convergence speed and more accurate identification results. In addition, optimization performed through the grey wolf algorithm can improve the accuracy and speed of parameter identification. The grey wolf algorithm converges faster than a particle swarm optimization algorithm, and contains wobble factor, so it is not easy for the grey wolf algorithm to fall into local optimization. At the same time, the parametersneeding to be adjusted by the grey wolf algorithm are fewer. Compared with other algorithms, the grey wolf algorithm is simpler and more flexible, and can effectively improve the accuracy of parameteridentification.

The invention discloses an SAR (Synthetic Aperture Radar) image change detection method based on Memetic kernel clustering; and the SAR image change detection method mainly solves the problems that the conventional algorithm is high in time complexity and easy to fall into a local optimum value, and has bad performance in region consistency and edge retaining. The implementation process of the SAR image change detection method comprises the steps of: (1) inputting two SAR images at different times, and performing median filtering on the two images; (2) computing the logarithmic ratio difference striograph of the two time-phase images subjected to change detection; (3) setting initial conditions; (4) carrying out kernel clustering and computing a fitness function fk; (5) selecting optimal individuals after performing clone and dual mutation operation on current individuals; (6) selecting optimal individuals after performing clone and crossover operation on the optimal individuals obtained in the step (5); (7) selecting a final individual by an elitist strategy; and (8) judging stop conditions, and outputting clustering results if the conditions are satisfied, otherwise returning to the step (4). The SAR image change detection method based on Memetic kernel clustering has the advantages of rapid convergence rate, high detection precision and accuracy in edge retaining, and can be applied to target identification and change detection of the images in the image processing field.

The invention discloses a multi-agent multicast routing method based on adjacent immune clonal selection, and mainly aims to overcome the shortcomings of low convergence rate and low searching capability of the conventional method when multicast routing problems are solved. The method is implemented by the following steps of: 1, generating a network model; 2, initializing antibody populations, memory unit populations and optimized running parameters; 3, calculating the affinities of all antibodies, finding an optimal antibody and extracting a vaccine; 4, judging whether termination conditionsare met or not, outputting an optimal individual if the termination conditions are met, otherwise turning to the step 5; 5, performing an immune colonization operation on all individuals in a currentpopulation; 6, performing an agent adjacent competition operation on the population obtained by the step 5, and updating the current population; and 7, extracting a better antibody updating memory unit from the antibody population obtained by the step 6, finding the optimal individual and returning to the step 4. The method has the advantages of high convergence rate and high searching capability, and can be used for solving the multicast routing problems of delay limitations.

The invention discloses a selective clustering integration method based on data stability. The method comprises the following steps of 1) inputting a data set and carrying out preprocessing; 2) carrying out clustering result set generation on the data set; 3) carrying out clustering result screening and acquiring a clustering subset; 4) carrying out sample division, and dividing the data set intoa stable subset and an unstable subset; 5) making a target function based on the stable subset and the unstable subset, and further screening the clustering subset; and 6) fusing the final clusteringsubset and acquiring a clustering result. Compared with a traditional method, the method has the following innovation points that multi-view clustering is realized so as to enhance diversities; an appropriate clustering algorithm is automatically screened and a problem that a data assumption does not match is avoided; the target function based on data stability is designed and high adaptability isachieved; and through an index increase degree, a multi-target genetic algorithm convergence direction is controlled and a convergence speed and accuracy are increased.

The invention discloses an NSGAII-based multi-objective optimization method for capacitor module component configuration in the field of capacitor module components. The method comprises the followingsteps: determining a constraint condition according to a capacity value and a quality value of a capacitor module required by a functional index by taking the lowest output voltage drop, the lowest cost and the smallest size of the capacitor module formed by each capacitor component as objectives; and establishing a multi-objective optimization model, and then solving the multi-objective optimization model based on a multi-objective genetic algorithm NSGA-II. According to the method, the NSGA-II multi-objective genetic algorithm is used for solving a problem of capacitor module component configuration, a non-inferior solution set meeting requirements is obtained, guidance is provided for application implementation, and good performance is obtained on the whole. Meanwhile, a constraint violation value method is introduced into constraint processing, feasible solutions are included in the scheme search process, some excellent approximate solutions are reserved, local optimum is not likely to happen in scheme search, more comprehensive scheme decision support can be provided, and the actual engineering design requirement is met.

The invention discloses an energy station equipment configuration and pipeline planning method considering multi-region interconnection cooperation, and the method comprises the following steps: firstly, building a to-be-planned distributed energy station comprehensive model; secondly, for a multi-region energy station interconnection coordination system, establishing an interconnection power linemodel and a heat distribution pipeline model considering energy loss, and expanding a distributed energy station comprehensive model considering interconnection coordination; thirdly, establishing adistributed energy station site selection planning method based on an improved P-median model, and determining the construction positions of the multi-region energy stations, the actual lengths of interconnection pipelines to be constructed among the energy stations and the supply loads of the energy stations.Furthermore, a distributed energy station equipment configuration and pipeline type selection planning method considering multi-region interconnection cooperation is established by taking the lowest total cost of the multi-region system as a target, and integration of various equipment configurations of the distributed energy station, type selection planning of inter-station interconnection pipelines and overall operation optimization of the multi-region system are realized.

The invention provides a distributed flow shop scheduling method and system with a batch delivery constraint. The method comprises the steps of taking the minimization of manufacturing time and totalenergy consumption as a target, enabling the manufacturing time to comprise the time of a processing stage and the time of a batch delivery stage, and enabling the total energy consumption to be the energy consumption of a truck transportation stage, solving the target to obtain a generated factory allocation vector, a generated job priority vector and a generated batch allocation vector, and performing distributed flow shop scheduling according to the vectors, and realizing minimization of the manufacturing time and the total energy consumption.

The invention discloses a bus station moving flow prediction method and system based on GW and SVR and a storage medium. The SVR is used for predicting the movement flow of the long-distance bus station, and the optimal parameters of the SVR are optimized through the grey wolf optimization algorithm, so that the tedious manual parameter selection process of the SVR is omitted, and the movement flow of the long-distance bus station is accurately predicted. The invention has the advantages that (1) the SVR algorithm is used for predicting the mobile network flow of the long-distance bus station,so that the mobile flow of the bus station is accurately predicted, and the network security and the experience of the bus station with large pedestrian flow in holidays and festivals are guaranteed;and (2) the advanced meta-heuristic optimization algorithm is used for optimizing the optimal parameters of the SVR, the GW optimization algorithm selected by the invention not only inherits the advantages of the meta-heuristic optimization algorithm, but also has the advantages of strong search capability and difficulty in falling into local optimum, and the complex manual parameter selection process of the SVR algorithm is omitted.

The invention aims at the problem of software module clustering in software system reconstruction and discloses a software module clustering method for probability selection. The software module clustering method comprises the steps that a module dependency relation drawing of a software system is extracted and obtained from a software system source program, then nodes are subjected to local merging operation based on probability selection by taking sparse points in the drawing as starting points so that initial module clusters of the software system can be obtained, then the cluster to whicheach node belongs is dynamically adjusted according to the correlation number of the nodes and all the modules based on the probability, and the clustering result of software modules is obtained. Thesoftware module clustering method provides a simple engineering method with high convergence rate and good clustering effect for the clustering problem of the software modules, and is used for reconstructing a software system structure and improving the understandability of the software system.

The invention discloses a reservoir group scheduling method and system based on a multi-population cooperative particle swarm algorithm, and belongs to the field of reservoir scheduling. According tothe method, the global optimization capability and the local optimization capability of the particle swarm algorithm are fully explored by introducing a plurality of particle swarms of which the inertia weights are gradually reduced step by step, and meanwhile, the population is differentiated through attraction factors defined by the population optimal values obtained by multi-population searching, so that the algorithm is not easy to fall into the local optimal values, and the particle swarm optimization capability is maximized; and maximizing the optimization capability in the solution of reservoir group scheduling, wherein the solution is a global optimal value. According to the method, the optimal positions are transmitted among the populations step by step in a multi-population cooperation mode, and convergence to the global optimal value is quickly realized in the multi-population cooperation mode, so that the optimization and convergence speeds of the particle swarm algorithm are increased, and the reservoir group scheduling problem is solved in a relatively short time.

The invention discloses an algorithm for simultaneously reconstructing flame three-dimensional temperature and smoke black volume fraction distribution, which comprises the following steps of: inputting emergent spectrum radiation intensity information of flame radiation light rays in different directions, dispersing a flame infinitesimal body, and establishing a flame radiation transmission equation; and setting an objective function, performing iterative computation by using a constructed simultaneous reconstruction algorithm, and performing reconstruction to obtain flame three-dimensional temperature and smoke black volume fraction distribution. According to the algorithm, in the iteration process, the non-negative least square algorithm is nested in the simulated annealing algorithm, so that the searching efficiency of the SA algorithm in the target function searching process is remarkably improved. And meanwhile, the SA algorithm has the advantages of high quality, strong initialvalue robustness, universality, easiness in implementation and the like, so that the NNLSSA algorithm has good global search characteristics, is not easy to fall into a local optimal value, and has relatively high precision and relatively good convergence in a process of simultaneously solving flame three-dimensional temperature and smoke black volume fraction distribution.

The invention provides an improved particle swarm optimization-based economic load dispatching method for a power system, relates to the technical field of power systems, and aims at solving the technical problem of optimization of economic load dispatching of the power system. According to the method, a swarm management rule is introduced into a particle swarm optimization algorithm to form an effective swarm utilization strategy-based particle swarm algorithm; the quantity of particles is effectively changed through a global optimal value change of the swarm; the improved algorithm is high in convergence rate; and the convergence precision is high. The method provided by the invention is applied to economic load dispatching of the power system comprising different units, achieves performance improvement of economic load dispatching of the power system and obtains a good optimization effect.

The invention discloses a double-chain quantum genetic algorithm for structural optimization design. The design result obtained through the method better meets the actual requirement, and the engineering applicability is higher. Firstly, selection of an initial scheme of structure optimization is improved on the basis of a superposition state thought in quantum computing, and dependence of a traditional method on an initial value is avoided. A quantum genetic algorithm mechanism is introduced in the structure optimization process (size / topology), the search range of a solution space is expanded, and therefore the optimal configuration more suitable for actual engineering requirements is obtained. The method has the advantages that a new thought of a double-chain quantum genetic algorithm is introduced into structural optimization design; on one hand, the dependence of a traditional gradient optimization method on an initial value is overcome, on the other hand, the search range of theunderstanding space is expanded, compared with a common evolutionary algorithm, the method is not prone to falling into local optimum, higher global optimization capacity is achieved, and a new tool and means are provided for the structure optimization design problem in the field of aircraft design.

The invention discloses a spectrum decision-making multi-objective optimization method based on an adaptive population search algorithm. The method comprises the following steps that a spectrum parameter decision-making model is established; initialization is carried out; the optimum individual in a population performs a discovery policy, and the other individuals select performing polices; the individuals in the population perform sequence pairing in pairs and perform single-point cross operation; linear arrangement is carried out on the individuals in the population; the individuals in the population perform direction mutation operation; current target function values of all individuals in the population are updated; and whether the current iteration times reaches the preset maximum times or not is judged, if the current iteration times reaches the preset maximum times, the optimum solution is output, and if the current iteration times does not reach the preset maximum times, the step of carrying out the linear arrangement on the individuals in the population is carried out. According to the optimization method, the minimum bit error rate, the minimum transmitting power and the maximum data rate of a cognitive radio system are optimized at the same time.

The invention discloses a welding workshop comprehensive scheduling method based on an improved firework algorithm. The method specifically comprises the steps: firstly building a mathematic model ofwelding workshop comprehensive scheduling, and obtaining a welding workshop comprehensive scheduling result based on the optimization target that the maximum completion time of a workshop processing complex welding product is minimum and the machine load is reasonable; meanwhile, considering constraint conditions such as before and after a product process and occupation of machine resources, and constructing a welding workshop comprehensive scheduling model; finally, adopting an improved firework algorithm for solving, in the solving process, designing firework explosion and mutation operatorscapable of meeting the process machining sequence; it is guaranteed that no illegal solution is generated in the whole solving process. According to the method, the influence of complex product process constraints and different processing machine types on actual welding workshop scheduling is fully considered, so that a scheduling solution scheme is more reasonable; according to the method, different search purposes are guaranteed, population diversity is guaranteed, the solving process is not prone to falling into local optimum, and the method has more superiority compared with a genetic algorithm and other heuristic methods.

The present invention provides a task scheduling method oriented to a cloud computing platform, comprising: traversing all task nodes in a DAG graph, and sequentially calculating the static priority of each node task; arranging tasks in descending order according to the static priority, and sequentially Put it into the task priority queue; for each task in the task priority queue, as long as the start execution time of the current task can be advanced, and the task that has been scheduled is not delayed, the parent task of the current task will be copied instead of only Copy only the critical parent task of the current task. Compared with the traditional algorithm, the start stage of task replication is advanced, and the task is copied first, and then the virtual machine is selected, so that the task can be executed on the virtual machine that can complete it the fastest, and the selection of virtual machine is more reasonable. Considering the load balance in the selection stage of the virtual machine, the good load balance of the cloud computing system is realized, and the resource utilization rate of the cloud computing system is improved.