43results about How to "Overcome the shortcoming of easy to fall into local optimum" patented technology

The invention discloses a multidimensional collaborative power grid planning method which includes the processes: acquiring geographic environmental information data and calculating comprehensive construction cost data by a rasterized map; building a power grid planning model; and calculating the solution of a power grid planning scheme by an optimal ant colony algorithm according to the acquired data and the power grid planning model. The method includes the steps: comprehensively considering substation locating and sizing, line selection and line corridor selection; and building a multidimensional collaborative power grid planning model considering complicated environmental influence. The multidimensional collaborative power grid planning method can accurately consider the influence of environmental factors on power grid planning. Line construction cost calculation precision is improved, and estimation difficulty is reduced. Catalytic elements are generated according to deviation of the local movement direction and the target movement direction of ants, so that the state transition rule of the ants is improved. The construction cost of lines on a mass raster and a substation is full-automatically and rapidly calculated, and calculation precision and efficiency are remarkably improved.

The invention relates to an antenna array spare construction method based on a quantum glowworm search mechanism. The antenna array spare construction method comprises the following steps: establishing an antenna spare array model, and determining a key parameter, which corresponds to the quantum glowworm search mechanism, of the antenna spare array model; substituting a quantum glowworm position into a fitness function to obtain a fitness value of the quantum glowworm position, and confirming a local optimal position and a global optimal position in a glowworm group; updating the fluorescein value and the learning neighborhood of each quantum glowworm; updating a quantum glowworm quantum position and the quantum glowworm position; updating the dynamic decision domain radius of the quantum glowworm; calculating the fitness value under a new quantum glowworm position, and confirming the local optimal position and the global optimal position in a quantum glowworm group again; and if a maximum iteration is achieved, outputting the global optimal position, and mapping to obtain a spare antenna array form.

The invention relates to a method for establishing a routing path through a Q learning on-board network based on fuzzy reasoning. The method comprises the following specific steps: (1) performing network initialization; (2) sending a greetings data packet in a broadcasting manner; (3) starting to send a request message through a source network node; (4) calculating the channel grade of an intermediate network node; (5) updating a Q value in a routing request data packet; (6) judging whether a current network node s is a destination network node or not, executing the step (7) if yes and otherwise, executing the step (4); (7) establishing positive routing information; (8) judging whether a routing reply data packet reaches the source network node or not, executing the step (9) if yes and otherwise, executing the step (7); (9) sending the data packet. According to the method, the combination of a fuzzy reasoning technology and a routing technology is realized, and the discount rate in a Q learning method is calculated according to the fuzzy reasoning and can be dynamically adjusted according to the network environment condition of the on-board network, so that the speed of establishing on-board network routing is accelerated.

The invention discloses a SOC estimation method for a lithium battery based on a state transition optimized RBF neural network. The method relates to the technical field of electric automobiles, wherein the method comprises the steps that: (1) offline training sample data are collected, normalization process is conducted on all training samples; (2) a SOC estimation model for a lithium battery based on a RBF neural network is established; (3) a STA optimization algorithm is adopted to optimize the established RBF neural network model; (4) the trained RBF neural network and each parameter are saved, the trained RBF network is used for conducting estimation on the SOC of a lithium iron phosphate battery; by means of the method, the SOC of the lithium battery is accurately estimated, the method has the advantages of high estimation precision, strong reliability, simple estimation model and the like, which can be widely applied in the technical field of power battery of the electric automobiles.

The invention discloses a routing method based on Q learning and a trust model in an Ad Hoc network, and mainly solves the problem of secure routing lookup in the Ad Hoc network. The routing method comprises the implementation steps that 1, a Q value table is generated; 2, the total number of adjacent nodes of each node is calculated; 3, each node evaluates trust values of all the nodes adjacent to the node; 4, trust awards are allocated; 5, instant awards are acquired; 6, aggregation awards are acquired; 7, Q values in the Q value table of the Ad Hoc network nodes are updated; 8, whether or not a current node in a routing request packet is a destination node is judged, if yes, the step 9 is executed, and otherwise, the step 6 is executed; 9, forward routing information is established; and10, data packets are sent. According to the routing method, combination of the trust model, a Q learning algorithm and a routing technology is achieved, a global optimal route can be dynamically looked up according to the environment of the Ad Hoc network, and the security and the stability of the network are effectively improved.

The invention discloses a Web API recommendation method based on topic model clustering. The method comprises the following steps: calculating semantic weight information of words according to context information so as to obtain a document-word semantic weight information matrix D; counting word co-occurrence information so that SPPMI matrix information is calculated; based on the obtained word frequency information matrix D of the words of the Mashup service document and the context SPPMI matrix M of the words, acquiring a word embedding information matrix by decomposing the M, combinding the two kinds of information , and calculating theme information of service; taking the obtained Mashup service theme features as spectral clustering input for clustering, segmenting a graph formed by all data points, wherein the sum of edge weights between different subgraphs after graph segmentation is made as low as possible, the sum of edge weights in the subgraphs is made as high as possible, and the clustering purpose is achieved; and combining GBDT and FM methods to predict and recommend the Web API service. Web API recommendation is effectively realized.

The invention relates to a combined running control method considering energy abandoning cost constraint for a sending-end power grid unit. The method comprises the following steps of 1) acquiring basic data; 2) building a unit combination optimization model, wherein the unit combination optimization model simultaneously considers energy abandoning constraint, peak adjustment constraint and frequency modulation constraint, and minimum sum of running cost, environmental cost and energy abandoning cost is used as a target function; and 3) solving the unit combination optimization model by an improved hybrid particle swarm optimization to obtain optimal output, and controlling a combined running sate of the unit according to the optimal output. Compared with the prior art, the method has theadvantages that the energy abandoning problem can be controlled from a running period, and the running efficiency is improved.

The invention provides a photovoltaic containing power distribution network voltage drop detection compensation method using UPQC implementation; the method comprises the following steps: 1, collecting voltage signals; 2, conditioning acquisition signals; 3, calculating a voltage drop compensation order; 4, using a multi-agent chaotic particle swarm optimization algorithm to set control parameters of a PI control unit; 5, using the set optimal PI control parameters to form and output PWM control signals; 6, outputting a compensation voltage so as to apply voltage compensation for the power distribution network. The method uses the multi-agent chaotic particle swarm optimization algorithm to set the PI control parameters of the UPQC, determines and uses the optimal PI control parameters to realize PI control, thus applying voltage compensation on the power distribution network, improving the UPQC voltage drop detection and compensation application precision and accuracy, and providing a novel method using the UPQC to manage the electric energy quality of the photovoltaic containing power distribution network.

A fast image segmentation algorithm based on artificial bee colony optimization fuzzy clustering is proposed and optimizes the sensitivity of a traditional FCM algorithm to the initialization of clustering centers by using the intelligent behavior of bee colony in nature. The algorithm starts with bees looking for food sources. An improved fitness function value which is described in the description is used to express the nectar content of the food source, according to the greedy algorithm, the old and new food sources are selected. After the bee finishes searching, the information is transmitted to the follower bee. The bee chooses a food source according to the probability P related to the nectar amount, and at the same time, the bee searches the neighborhood near the food source. When the nectar amount is not improved after limited searches in the vicinity of a food source, the nectar source is abandoned, and the bees associated with the food source are replaced by scouting bees toindependently and randomly search for the nectar source, and the location of each food source represents a possible solution of the optimal clustering center of the image to be segmented.

The invention discloses a new photovoltaic power prediction method based on AFSA-Elman. Firstly, the method decomposes the collected original power into low frequency trend component and high frequency detail component by wavelet decomposition, then forecasts the low frequency trend component and high frequency detail component with the collected weather data as the input of the model respectively, and finally reconstructs the corresponding forecasted value to obtain the final power forecasted value. An artificial fish swarm algorithm is used to optimize the weights and thresholds of Elman neural network, and the optimized Elman is applied to the short-term prediction of photovoltaic output power. The method can predict the output power of the next time according to the historical weatherand the output power of the photovoltaic station and the weather condition in a short period of time, and the prediction model can utilize the autocorrelation existing in the historical data without using a plurality of complex and tedious physical formulas for modeling.

The invention provides a novel codebook design method based on an ant colony clustering and genetic algorithm. The method includes the following steps: step one, training data are distributed in two-dimensional spaces of different dimensions randomly by using ants of different numbers, and an LF algorithm is adopted to perform generation of initial clustering; step two, clustering correction performed by the initial clustering according to the dimension of a codebook guarantees that the clustering number is the same as initial setup codebook dimensions; step three, under the premise that initial population is obtained successfully, individual selection, intersection and mutation operation are performed according to the basic procedure of a genetic algorithm until iteration is stopped and an optimal individual meeting requirements is obtained. The novel codebook design method based on the ant colony clustering and genetic algorithm overcomes the defect that correlation of initial selection and final design results is strong in an LBG algorithm, meanwhile, prevents similar LBG algorithm from getting into the inferior position of local optimum, and is suitable for the fields of large quantity processing, voice communication, mode recognition, internet protocol (IP) telephony and the like.

The invention discloses a simulated annealing optimization neural network-based visibility sensor and a detection method thereof, wherein the visibility sensor comprises a microprocessor module and apower supply circuit for supplying power for each electricity utilization module in the visibility sensor; the input end of the microprocessor module is connected with a visibility detection circuit;the visibility detection circuit comprises a particle concentration sensor and a temperature and humidity sensor; the microprocessor module, the particle concentration sensor and the temperature and humidity sensor are all connected with the output end of the power supply circuit. The detection method comprises the following steps of 1, performing data acquiring and transmitting; 2, carrying out data preprocessing; and 3, carrying out data processing to obtain a visibility detection value. The device is novel and reasonable in design, convenient to implement, low in cost and high in visibilitydetection precision, can be well applied to visibility detection, is high in practicability, good in use effect and convenient to popularize and use.

The invention discloses a photovoltaic inverter low voltage ride through detection method and system. The method comprises a voltage signal acquisition step, a Clarke step of a photovoltaic inverter,a microcontroller detection step of the photovoltaic inverter, specifically, setting PI control parameters of a phase-locked loop of an inverter by adopting a multi-agent chaotic particle swarm optimization algorithm, controlling the phase-locked loop by adopting the PI control parameters and repetitive control when a global optimal point of the system is reached, and detecting low-voltage ride through of the inverter, a step of calculating a symmetrical three-phase voltage drop value, subtracting a reference voltage udref from a d-axis component u'd obtained by Park transformation to obtain acorresponding value, and performing dq/abc inverse transformation of the corresponding value to obtain symmetrical three-phase voltage drop values uac, ubc and ucc, and an output step. The method isadvantaged in that tracking control of the phase-locked loop is carried out by using PI control and repetitive control in the photovoltaic inverter, PI control parameters are set by using an intelligent agent chaotic particle swarm optimization method, so detection accuracy and robustness are improved.

The invention discloses an electric power system economic dispatching method based on a quantum beetle antennae algorithm. The method comprises the following steps: (1) initializing the position and initial value of beetle antennae; (2) introducing a search mechanism of quantum behaviors into the beetle antennae algorithm, and establishing Delta quantum potential wells around the beetle antennae position to form a quantum beetle antennae algorithm; (3) establishing an economic dispatching model of the power system by taking the minimum total power generation power value of the generator set as a target function; and (4) optimizing the system economic dispatching model on the basis of a quantum beetle antennae algorithm under constraint conditions, and calculating the combination of the minimum power generation cost and the minimum pollutant emission of the generator set by using an optimal solution. According to the method, a search mechanism of quantum behaviors is introduced into the beetle antennae algorithm, the quantum Delta potential well is established around the beetle antennae position, the probability and uncertainty of the quantum theory are utilized, the convergence speed is effectively increased, the defect that a traditional search method is prone to falling into local optimum is overcome, and the accuracy of a search result is improved.

The invention discloses a two-stage power distribution network planning method based on a genetic algorithm, and belongs to the technical field of power distribution network frame planning. The implementation method comprises the following steps of optimizing a solution obtained by the first-stage genetic algorithm by adopting a second-stage dynamic adjustment iteration mode; optimizing an optimization scheme obtained by the first-stage genetic algorithm; obtaining effective information by fully utilizing a genetic algorithm; grading a to-be-planned line according to the occurrence frequency of the to-be-planned line in the initial scheme; providing correct criteria for edge adding and deleting operation in the second stage, regarding connected to-be-planned line lines as sub-trees, carrying out unified operation in the dynamic adjustment process such that excellent sub-trees can be found and stored in the algorithm in the second-stage dynamic adjustment process, the searching efficiency is improved, and a power distribution network planning scheme with lower cost is obtained. According to the method, the defect that an existing power distribution network planning method is prone to falling into local optimum can be overcome, and optimal power distribution network planning meeting constraint conditions is achieved on the basis that the search efficiency is improved.