81results about How to "Achieving joint optimization" patented technology

The invention discloses an image retrieval method based on variable-length depth hash learning and mainly relates to the field of image retrieval and depth learning. According to the method, learning of hash codes is modeled into the process of similarity learning. Specifically, the method utilizes a training image for generating a batch of ternary image sets. Each ternary image set comprises two images with the same label and one image with the different label. The purpose of model training is to space image pairs matched to the maximum and the unmatched image pairs in hamming space. The depth convolution nerve network is introduced into the learning part of the method, the image characteristics and hash functions are optimized in a combined mode, and the end-to-end training process is achieved; hash codes output by the convolution network have different weights. For the different retrieval task, a user can regular the length of the hash codes by disconnecting unimportant bits; meanwhile, according to the method, the discrimination of the hash codes can be kept effectively under the circumstance that the hash codes are short.

The invention belongs to the field of micro-grid optimization planning and design and relates to a double-layer optimization planning and designing method for a combined cooling, heating and power micro-grid system. The smallest whole life circle economic present value and the minimum CO2 emission are taken as the optimization target of the method, a multi-target genetic algorithm based on NSGA-II is adopted for an outer layer optimization module, and the optimal equipment combination and the equipment capacity of the system are sought according to input equipment physical parameters of the system, input economic parameters and the optimization scheduling result output by an inner layer optimization module; the hybrid integer linear programming algorithm is adopted for the inner layer optimization module, with the lowest system operation cost as the optimization target, the optimal operation strategy on a typical day of the system is obtained according to the optimization configuration scheme given by the outer layer optimization module, the optimization scheduling result of the system is output by the inner layer optimization module, and the system operation cost is transmitted to the outer layer optimization module. The method can achieve combined optimization of the configuration and the operation strategy of the combined cooling, heating and power micro-grid system.

The invention discloses a complex step reservoir group water storage period optimized scheduling method which comprises the steps of S1, collecting basic information of a river basin; S2, dividing a complex step reservoir group into a plurality of sub-river-basin areas, and respectively calculating an after flood season of each sub-river-basin area; S3, estimating a probability that the flow of a branch flood prevention node exceeds a safe flow in each sub-river-basin area, and demonstrating feasibility of seasonal flood end water storage scheduling optimization by means of the probability; S4, constructing a multi-target water storage scheduling optimization model; S5, solving the multi-target water storage scheduling optimization model by means of a Pareto dynamic dimension searching method, and obtaining a certainty optimization scheduling result of a seasonal flood end water storage period; S6, extracting an optimization scheduling rule of the seasonal flood end water storage period by means of a large system polymerization / decomposition model; and S7, performing flood prevention safety examination on the optimization scheduling rule through using river basin designed flood as an input. The complex step reservoir group water storage period optimized scheduling method is a new method for complex step reservoir seasonal flood end water storage period optimized scheduling under a precondition that a flood prevention standard of the basin is not reduced.

The invention discloses a software defined wireless network system architecture and method for LTE small base station ultra dense networking. The method comprises the following steps: a mobile terminal is accessed to an LTE small base station, wireless resources of the small base station is forwarded to an SDN controller of the data center through an SDN switch, the interface for mutual communication between the LTE small base station and the SDN controller controlled by the wireless network is a southbound interface API, and a northbound interface API can flexibly combine the functions in the controller according to different new service deployment requests. The SDN controller realizes the separation between the control surface and the data surface through a software defined mode, achieves the complete control of the wireless access network part, and achieves the division of the services of the core network; the system capacity of the wireless access network can be increased and the coverage of the wireless access network can be enhanced through interference coordination, load balancing and other technologies; and the internal flow table of the switch can be flexibly configured by monitoring the port of the SDN switch in the dense network, and thus the backhaul network can reach a load balancing state.

The invention discloses a word2vec improvement method of related factor training combining parts of speech and word orders. The method provides a model of Structured word2vec on POS. The model includes a CWindow-POS (CWP) model and a Structured Skip gram-POS (SSGP) model. According to each of the two models, part-of-speech label information and the word orders are used as influencing factors for combined optimization, and part-of-speech correlation information is utilized for modeling inherent syntactic relationships between words in context windows; and weighted calculation is carried out on context word sequences through part-of-speech correlation weights, then vector inner-product calculation is carried out according to the word position orders, and a stochastic gradient descent (SGD) algorithm is used for learning the related weights and word embedding in a combined manner. The method directionally embeds words according to the position orders thereof, realizes combined optimization for word vectors and weighted matrices related to the parts of speech, and has high efficiency in all of word analogy tasks, word similarity tasks and qualitative analysis.

The invention discloses a calculation method for the security constrained unit commitment of an alternating current and direct current hybrid power grid. The method comprises the following steps that:establishing the active power equivalent models of an alternating current electric transmission line and a direct current electric transmission line, which participate in optimal scheduling, in a multi-area alternating current and direct current hybrid power grid; on the basis of the active power equivalent models, establishing a security constrained unit commitment optimal model which considerspartitioning load balance; solving the security constrained unit commitment optimal model; according to an optimization result, considering that all network monitoring elements are subjected to security correction; and if no new monitoring element power flow is out of limit, outputting an optimization result. Meanwhile, the invention discloses a corresponding system. By use of the method, power generation resources can be optimally regulated and controlled from a large range, a multi-area unit commitment scheme and the power plan of a cross-area alternating current and direct current electrictransmission line can be subjected to combined optimization, and large power grid ruling ability and electric power resource optimal configuration ability are improved.

The invention provides a method and a device for selecting terminal equipment nodes in a federated learning system. The method comprises the following steps: determining a test accuracy optimization target model selected corresponding to a terminal device node; and obtaining current environment state data corresponding to each federated learning iteration process, and inputting the current environment state data into a terminal equipment node selection model to obtain a terminal equipment node selection strategy corresponding to test accuracy and time delay constraints output by the terminal equipment node selection model. By adopting the method for selecting the terminal equipment nodes in the federated learning system, the federated learning performance can be ensured, the joint optimization of the resource utilization rate and the energy consumption of the terminal equipment is realized, the terminal equipment set is reasonably selected to participate in the aggregation of the federated learning model, and the efficiency and the accuracy of the federated learning are effectively improved.

The invention discloses a compressive spectrum sensing method. The method is applied to the spectrum hole detection in a cognitive radio system, and belongs to a novel method for carrying out spectrum sensing by a compressive sensing technology. The method carries out condition setting by aiming at observing data, the detecting efficiency in the compressive spectrum sensing is improved, when acceptable conditions are met, the estimation value is reconstructed, when the acceptable conditions are not met, the observing times are increased, and the adaptive process of the matrix observation is realized. The method realizes the integral optimization through reducing the sparseness among the observing matrix array vectors, reducing the correlation among the line vectors and combining the matrix adaptivity observation in a united way. Compared with the ordinary compressive spectrum sensing, the compressive spectrum sensing method has lower mean square errors generated during the spectrum reconstruction, and the spectrum detecting efficiency is higher under the condition of the same observing times; the observing times required by the method is fewer when the same receiving operation performance.

The invention provides a park energy internet energy equipment capacity optimal allocation method. The method includes the following specific steps that: photovoltaic output and electricity / heat / natural gas load scene models are constructed; an upper layer model for energy equipment capacity optimization is established, the optimal allocation capacity of energy equipment is solved; a lower layer model for park energy internet operation optimization is established, the optimal operation mode of a park energy internet under a determined scene and a determine energy equipment capacity is solved;and a genetic algorithm is adopted to solve the upper layer model, and a cplex solver is adopted to solve the lower layer model, and the optimal allocation capacity of the energy equipment is obtained. With the park energy internet energy equipment capacity optimal allocation method of the invention adopted, the joint optimization of the allocation and operation of the park energy internet can berealized; and the photovoltaic output and load scenes are considered, so that planning results better accord with actual conditions.

The present invention provides a large-scale intermittent energy access mixed energy multi-target robust optimization method, belonging to the electric power system automation technical field. Aiming at the nondeterminacy problem of the intermittent energy generation, the concept of the uncertain cost is introduced, the economy, the environmental protection and the nondeterminacy of the electric power system are taken as a target, and the nondeterminacy robust optimization theory is combined to provide a large-scale multi-energy and multi-target robust optimization model. The large-scale system decomposition coordination optimization theory is employed to decompose a link optimization model into a plurality of subsystem multi-target optimization models and perform grade division of the risk degree possibly caused by each uncertainty scheme set, and the scheme with the lowest risk degree is selected as an optimal scheme so as to obtain the optimal solution of each subsystem or a Pareto scheme set; the optimal scheme of each subsystem or the scheme set are fused in the optimal Pareto solution set of the whole system to provide an improved decision support for decision makers.

The invention provides a system capacity optimization method and device for a satellite communication constellation, and relates to the technical field of satellite communication. The satellite communication constellation comprises a plurality of non-geosynchronous orbit satellites, and the method comprises the steps of obtaining the constraint conditions of the system capacity performance in a satellite communication constellation, wherein the constraint conditions comprise the constraint conditions of the mapping relation between wave beams and earth stations, the constraint conditions of transmitting power and the constraint conditions of interference values, and the interference values are represented through the mapping relation and the transmitting power; calculating to obtain an optimal solution of the transmitting power and an optimal solution of the mapping relation according to the constraint condition; and communicating according to the optimal solution of the transmitting power and the optimal solution of the mapping relation, so that the maximum system capacity is achieved under the condition that the same-frequency geosynchronous orbit satellite communication system is not interfered.

The invention belongs to the field of hydropower energy optimization operation, discloses a multi-objective hydropower purchase optimization method and system of a provincial power network under time-of-use electricity price. The method establishes and solves a multi-objective time-of-use electricity price model based on the response of user demand to the time-of-use electricity price. The conditional value-at-risk (VaR) based on portfolio theory is introduced as a risk index, and a multi-objective optimization model is proposed to maximize the expected profit and minimize the risk of power purchase. The power purchase ratio scheme and the power purchase risk scheme are optimized to obtain the inter-provincial power purchase risk decision and the optimal power purchase ratio. The inventioncombines the sensitivity of the user to the change of the electricity price, considers the minimization of the peak and valley load of the system, maximizes a plurality of objectives of the user satisfaction, establishes a multi-objective time-of-use electricity price model based on the user response, improves the feasibility of the time-of-use electricity price scheme, and lays a foundation forestablishing a maximum expected benefit model of power purchase.

The invention relates to a virtual network function deployment optimization algorithm based on deep reinforcement learning, and belongs to the technical field of mobile communication. According to themethod, VNF sharing is considered under the constraint of a physical layer CPU, bandwidth resources and SFC end-to-end time delay, and the total cost of a service provider and the SFC end-to-end timedelay are jointly optimized by deploying VNF and allocating CPU resources; secondly, since the state space and the action space of the scheme are continuous value sets, a VNF intelligent deployment algorithm based on deep reinforcement learning is provided, and thus an approximately optimal VNF deployment and resource allocation strategy is obtained. And on each discrete time slot, the VNF is deployed to a proper destination server according to the arrival rate of the SFC, the remaining CPU resources of the universal server and the remaining bandwidth resources of the physical link, and the VNF is allocated to the CPU resources of the destination server. According to the VNF deployment optimization algorithm provided by the invention, the compromise between the total cost of the service provider and the end-to-end delay of the SFC can be realized, and the resource utilization of the physical network is improved.

The invention discloses a partial reconstruction-based quality optimization method for scalable video coding. The method comprises the following steps of: setting a truncation degree and a layer weight; performing partial reconstruction or entropy coding on a residual error conversion coefficient; calculating macro block distortion of a base layer and an enhancement layer; selecting an optimal mode for determining base layer macro blocks; traversing all macro blocks in a coding frame on the base layer so as to obtain the code stream of the base layer; selecting an optimal mode for determining enhancement layer macro blocks; traversing all macro blocks in a coding frame on the enhancement layer so as to obtain the code stream of the enhancement layer; and calculating and storing error diffusion distortion and error diffusion sensitivity. In the method, the residual error conversion coefficient is selectively truncated in a coding process, so that combined optimization of video quality at different target code rates is realized, and the problem of video quality reduction caused by the conventional code stream extraction method is solved; and simultaneously, an improved code stream extraction algorithm is used in a combined way, so that video quality is further enhanced.

The invention discloses a real-time power generation plan optimization method considering northern Shaanxi cascade section limit linkage adjustment. The method comprises the following steps: based onvarious basic data of a real-time power generation plan, constructing a network-province two-stage real-time power generation plan coordination optimization model in a generalized tie line mode, and providing a unit initial plan value and a section initial tide value for joint optimization of the real-time power generation plan and Shaanbei cascade section limit. According to the method, based onvarious basic data of a real-time power generation plan, a unit initial plan and a section initial tidal current value, a real-time power generation plan optimization model considering northern Shaanxi cascade section limit linkage adjustment is constructed; an optimization target is introduced into the Shaanbei virtual subarea, so that the new energy wind curtailment amount is minimum; constraints such as a Shaanbei subarea cascade section and a the central Shaanxi plain subarea admitting ability are introduced into constraint conditions, combined optimization of unit output and stable section limits is achieved, the Shaanbei subarea new energy consumption potential of Shaanxi province is fully excavated, the utilization rate of the Shaanbei cascade section is increased, and the lean level of northwest regional power grid new energy scheduling is improved.

The invention belongs to the related technical field of structure optimization, and discloses a gradient dot matrix sandwich structure optimization method driven by an agent model, which comprises thefollowing steps of: (a) constructing a prototype dot matrix and a sample dot matrix based on a level set function, calculating an elastic tensor of the sample dot matrix and using the elastic tensorto build the agent model to predict mechanical properties of the dot matrix unit; (b) optimizing the density of each slice layer by adopting a variable thickness method based on the agent model, and determining the thicknesses of an upper panel, a middle sandwich and a lower panel of the sandwich structure according to the optimized slice layer density distribution; (c) optimizing density distribution of the dot matrix units in the sandwich layer by adopting a free material optimization model based on the agent model; and (d) performing shape interpolation on the dot matrix units in the sandwich layer based on the relative density of the dot matrix units in the sandwich layer to obtain the specific configuration of the gradient dot matrix in the sandwich layer, thereby realizing a topological optimization process and improving the mechanical properties of the sandwich structure.

The embodiment of the invention provides a communication, caching and computing resource joint optimization method and device for a vehicle-mounted network and electronic equipment. The method comprises the following steps that: request information sent by a requester is received, wherein the request information carries an expected target content file; based on the target content file, a pluralityof contracts of a target collaboration party are generated by utilizing a preset optimization problem model; and the plurality of contracts are sent to the requester, so that the requester screens out a target contract with the maximum utility value from the plurality of contracts, and receives the target content file sent by the target collaboration party based on the target contract. Accordingto the communication, caching and computing resource joint optimization method and device for the vehicle-mounted network and the electronic equipment of the embodiments of the invention, a corresponding incentive mechanism is established by using a contract theory, the problem of asymmetric information possibly existing in a scene is solved, reasonable resource allocation is realized, joint optimization of communication, calculation and cache resources is realized, and radio resources are efficiently managed.

The invention discloses a hierarchical reinforcement learning anti-interference algorithm for joint optimization of channel selection and transmission time. The algorithm comprises a wireless communication network consisting of a transmitter, a receiver and an interference machine, wherein the interference machine generates an interference signal according to any one of a frequency sweeping mode,a comb mode and a random mode; On a data channel, a transmitter and a receiver carry out data communication in a dynamic spectrum access mode so as to withstand Communication interference to the transmitter / receiver caused by interference signals released by the jammer; And on the control channel, the transmitter and the receiver realize coordination of the dynamic spectrum of the transmitting andreceiving end through information interaction. In a dynamic interference scene, channel selection optimization based on fast reinforcement learning is carried out with a small time granularity, transmission time length optimization based on a random automatic learning machine is carried out with a large time granularity, and loop execution is carried out until the data transmission time length isconverged or reaches the maximum iteration frequency. The throughput of the wireless communication network system is improved.

The present invention discloses a wavelength assignment method for inhibition of WDM-QKD (Wavelength Division Multiplexing-Quantum Key Distribution) system four-wave mixing and Raman scattering noise.Aiming at interference problems in a WDM-QKD system, the spectrum assignment features of the four-wave mixing and Raman scattering noise in the system is fully considered, and a channel interleavingscheme for inhibition of four-wave mixing noise and an optimal channel interval scheme for inhibition of Raman scattering noise are provided so as to achieve combined inhibition of the four-wave mixing noise and the Raman scattering noise, effectively improve the system capacity and save the fiber resource. The method provided by the invention is low in interface and easy to enforce and provides theoretical basis for carrying of quantum signals in the classic light network.

The invention provides an end-to-end long-time speech recognition method. The method comprises the following steps: selecting a corpus as a training data set, and carrying out data preprocessing and feature extraction on voice data in the training data set to generate voice features; constructing an improved RNN-T model fusing an external language model and a long-term speech recognition algorithm, and inputting the speech features into the RNN-T model for training to obtain a trained improved RNN-T model; and taking the trained improved RNN-T model as a teacher model in a mutual learning knowledge distillation algorithm, training a student model in the mutual learning knowledge distillation algorithm by using the mutual learning knowledge distillation algorithm, identifying long-term voice data to be identified by using the trained and verified student model, and outputting a voice identification result. According to the method, the external language model, the long-term speech recognition algorithm module and the RNN-T model are fused, so that the robustness and generalization ability of long-term speech recognition of the model are improved, and the time and space complexity of the algorithm is optimized.

The invention discloses a control information and service separate layered isomerism millimeter wave wireless local area network framework which is characterized in that a 2.4 Ghz / 5 Ghz omni-direction control subnet is used as a foundation, a plurality of 60 Ghz orientation service subnets are borne on the foundation, the control subnet comprises an access point and a plurality of users, the access point and the users are provided with intelligent antennas which are antenna arrays with beam forming capacity, and the antenna arrays can form certain antenna beams which can achieve orientation sending and receiving. The framework has the advantages that control information and service are separated, 60 Ghz millimeter wave high-speed service data transmission is achieved, network service capacity is greatly improved, meanwhile, basic operation of the whole network is guaranteed through omni-direction control information of 2.4 Ghz / 5 Ghz frequency bands is guaranteed, unnecessary energy loss caused by high-frequency-band transmission is avoided, and joint optimization of energy efficiency and spectrum resources is achieved finally.

The invention discloses a distributed method for dynamic access point selection in a wireless network. through a distributed power calculation method based on a power updating factor, greedy access point selection is realized, and thus, only a small part of power updating factors needs to be exchanged among receivers, the remaining power updating factors can be obtained through automatic derivation based on local information, global optimization on pairing between a receiver and a transmitter, access point selection and power control becomes simpler, and thus, the computational cost and the signaling overhead are significantly reduced.

The invention discloses an alternating current and direct current interconnected power system decentralized robust optimization scheduling method, which comprises the following steps: constructing operation constraints according to the operation characteristics of a direct current tie line; constructing an upper-layer coordination optimization model and a lower-layer robust optimization model, wherein the lower-layer robust optimization model establishes a double-layer iterative solution framework based on a target cascade analysis method, and then a column sum constraint generation algorithm is introduced to carry out conversion processing and solution on the two-stage robust optimization model of each interconnection subsystem; feeding the active plan of the virtual unit back to the upper-layer coordination optimization model; enabling the upper-layer coordination optimization model to optimize and update a direct-current tie line power transmission scheme according to the virtual unit active plan obtained by optimization of the lower-layer robust optimization model, and transmit an optimization result as a reference value of next optimization back to the lower-layer robust optimization model; and carrying out iterative interactive solution on the upper and lower double-layer models until a convergence standard is met, and stopping iteration. According to the invention, joint optimization of the scheduling plan and the standby scheme of the large-scale alternating current and direct current interconnection system can be realized.

The invention relates to a multi-view clustering machine learning method with missing kernel. The method fuses filling and clustering, fills missing kernel under the guidance of clustering, and clusters with the filled kernel. The method comprises the following steps: 1) obtaining a target data sample, mapping the target data sample to a multi-kernel space; 2) establishing a missing multi-core k-means clustering optimization objective function; 3) adopting a three-step alternating method to solve that missing multi-core k-means clustering optimization objective function to realize clustering.Compared with the prior art, the invention considers the joint optimization of filling and clustering, and has the advantages of good clustering effect and the like.

The embodiment of the invention discloses a system exception prediction model training method, a system state prediction method and device, equipment and a storage medium. The method comprises the steps of acquiring sample features; initializing the system anomaly prediction model according to set weight parameters; processing the sample features through the system anomaly prediction model to obtain prediction energy; constructing a target function based on the predicted energy; and in back propagation, updating weight parameters of the system anomaly prediction model through the objective function. Therefore, the abnormal condition of the system operation state can be predicted.

The invention discloses a regional integrated energy system optimization scheduling method and device based on a Stackelberg game, and the method comprises the steps of issuing an electricity price signal with economic benefit maximization as an optimization objective function according to a system manager, and feeding back a thermoelectric load with operation cost minimization as an optimization objective function by a user; in response to the obtained user feedback thermoelectric load, enabling the system manager to judge whether to update an electricity price signal based on a thermoelectric load demand-electricity price sensitive relationship; if the system manager judges that the electricity price signal is not updated, enabling the system manager to send a production control instruction to the capacity base; in response to the acquired production control instruction, enabling the capacity base to judge whether the wind power integration power reaches a target value or not; and if the capacity base judges that the wind power integration power does not reach the target value, controlling the system manager to update the electricity price signal. The wind power integration power can be obviously improved by adopting a Stackelberg game optimization strategy, so that the wind power integration demand of a productivity base is ensured.

The invention discloses an SVC video transmission method based on an intelligent edge and the intelligent edge, and the method comprises the steps: determining the video priority corresponding to a video block when the intelligent edge receives a video block downloading request; enabling the intelligent edge to acquire network state information of a network where the video block downloading request is located; enabling the intelligent edge to determine a video layer number corresponding to the video block according to the network state information and the video priority, and acquiring a videolayer corresponding to the video block from a server according to the video layer number; and enabling the intelligent edge to issue the acquired video layer to each user side in a user cluster corresponding to the video block. In this way, the intelligent edge is assembled on the network edge of the network, and the intelligent edge issues different numbers of SVC video layers to each user side by using the network state information and the video priority to provide video streams with different qualities for different users, so the QoE joint optimization of SVC video users is realized, and the income of a content provider is improved.

The invention provides a frequency domain multichannel reflection coefficient joint inversion method and system based on an MWV model. The method comprises the steps of acquiring preprocessed post-stack seismic data; acquiring frequency domain multichannel reflection coefficient joint inversion parameter prior information according to the post-stack seismic data; extracting seismic wavelets from the post-stack seismic data; extracting frequency spectrum information of the seismic wavelet based on Fourier transform; constructing a forward operator matrix of frequency domain multichannel reflection coefficient joint inversion according to the frequency domain multichannel reflection coefficient joint inversion parameter prior information; extracting frequency spectrum information of the post-stack seismic data channel by channel through Fourier transform; and constructing an objective function of frequency domain multichannel reflection coefficient joint inversion based on the frequencyspectrum information, and solving the objective function through an MMV model and a block sparse Bayesian learning theory to obtain reflection coefficient model parameters. According to the invention,the precision and stability of the inversion result of the frequency domain reflection coefficient are effectively improved.

The invention discloses a weather detection method and system based on a multi-band weather radar, and a storage medium. The method comprises the steps of: obtaining radar echo signals of at least two frequency bands, the radar echo signals comprising a differential propagation phase shift rate and an echo intensity value; calculating according to the differential propagation phase shift rates of all the frequency bands to obtain an intensity correction coefficient of each frequency band; and calculating according to the intensity correction coefficient of each frequency band and the echo intensity value to obtain a detection result. According to the method, the radar echo signals of the multiple frequency bands are obtained, the intensity correction coefficient is calculated and determined by adopting the differential propagation phase shift rates in the radar echo signals, the intensity correction coefficient is adopted to correct the intensity value, and the detection results of the multiple frequency bands are obtained. Therefore, joint optimization of detection data of multiple frequency bands is realized, and the data quality of radar detection is improved. Meanwhile, accurate inversion of radar detection echo data is realized, and the accuracy of radar detection is ensured.

The invention discloses an approximate minimum bit error rate (BER) power distribution method for a hybrid carrier system, relating to an approximate minimum BER power distribution technology for the hybrid carrier system, and aiming to improve the BER performance of the hybrid carrier system and increasing the achievable rate of the system. The method comprises the following steps: at a transmitting terminal, performing alpha-order weighted fractional Fourier transform for data on each band to obtain the data in a frequency domain; in the frequency domain, performing power distribution operations for the data obtained in step 1, performing IFFT for the distributed data, transforming the data to a time domain, and sending the data to a channel; at a receiving terminal, performing FFT for the received data, and transforming the data back to the frequency domain; performing channel equalization in the frequency domain; after equalization, performing inverse operations of power distribution in the frequency domain; and performing -alpha-order fractional Fourier transform demodulation, and outputting the data. The approximate minimum BER power distribution method disclosed by the invention is suitable for powder distribution occasions in wireless communication.