26 results about "Penalty coefficient" patented technology

The invention relates to a dynamic density guided generative adversarial network unbalanced data processing method (DAG-WGAN). The DAG-WGAN realizes unbalanced data processing through data preprocessing, dynamic density estimation and weight distribution, potential structure learning based on a variational auto-encoder (VAE), and density guide generation and dynamic feedback optimization based on WGAN-GP. The DAG-WGAN adaptively evaluates the sample density by using kernel density estimation (KDE) and a Gaussian kernel function, and allocates a weight for a generation process, thereby emphatically enhancing the low density and discriminating the sample generation of a difficult region. The VAE learns a potential manifold structure of a minority class of samples, realizes density-guided generation of a potential space under a WGAN-GP framework, and ensures diversity and manifold consistency of generated samples. In addition, a dynamic feedback mechanism is introduced, the weight and the gradient penalty coefficient are adaptively adjusted and generated, and the training stability and the sample generation robustness are improved.

The invention discloses a bus shift extraction optimization method based on multi-dimensional score driving, and relates to the field of urban public traffic scheduling management, and the method comprises the following steps: initializing shift extraction parameters, determining a shift extraction target, defining time period features, and setting a score weight and a penalty coefficient; constructing a shift extraction optimization mathematical model, including defining decision variables and establishing an objective function and constraint conditions thereof; respectively constructing a multi-dimensional scoring system and a constraint verification strategy by utilizing the objective function and the constraint conditions thereof; generating candidate shift drawing schemes, calculating a comprehensive score and selecting an optimal shift drawing scheme; performing iteration processing on the optimal shift extraction scheme in combination with loop control and a scheme rollback strategy to obtain a final shift extraction scheme; and verifying and counting the final shift drawing scheme, and generating a corresponding shift drawing report. On the premise that operation balance and service integrity are ensured, global optimal selection of the shift drawing scheme is achieved, and the intelligent level and operation efficiency of bus dispatching are remarkably improved.

The invention provides an asset allocation method and device based on a global optimization algorithm and electronic equipment. The method comprises the following steps: constructing an asset allocation model which aims at maximizing the asset yield and takes the allocation proportion of various assets as a variable under each constraint condition, wherein each constraint condition comprises the limitation on the value interval of the variable; determining a penalty coefficient of each constraint condition according to a business rule, and constructing an optimization objective function of the asset allocation model according to each penalty coefficient; and performing iterative solution on the optimization objective function by adopting a global optimization algorithm until an optimal resource allocation scheme is obtained, summing current values of all variables in each iterative solution process to obtain a judgment metric, and judging whether optimization process calibration is executed or not according to the judgment metric and a threshold value corresponding to the judgment metric. The optimization process calibration is used to avoid the global optimization algorithm falling into a locally optimal solution.

The application discloses a power and electricity balance method based on fuzzy opportunity constraint conversion and a computer device, and the method comprises the following steps: constructing a tolerance degree based on power shortage risk, taking a comprehensive objective function of conventional power generation cost, standby capacity cost, new energy abandoned electricity cost and uncertainty penalty term, taking system constraints, unit constraints, unit electricity constraints, cross-section safety constraints and energy storage constraints as constraint conditions, and improving the accuracy of a medium and long term power and electricity balance model under the consideration of single modeling of new energy uncertainty, thereby improving the economy and safety of the power system. Through the establishment of the uncertainty modeling mode of long-period power and electricity balance, the fuzzy opportunity constraint and the gradient penalty coefficient, the economy and reliability can be flexibly balanced, the redundant cost of standby capacity is reduced by converting the fuzzy boundary into a deterministic boundary, and through risk quantification, the decision transparency can be supported and the interpretability can be ensured.

The invention provides a scheduling method and system of an integrated energy system and a storage medium, and the method comprises the steps: constructing a mixed integer nonlinear programming model with the minimum total operation cost as a target, carrying out the iterative solving of the model until the upper and lower bounds converge, solving a main problem to obtain a tentative solution of a unit start-stop scheme and the lower bound of the current iteration, and carrying out the scheduling of the unit start-stop scheme. The main problem comprises a regularization item and is constrained by a cut plane, and a penalty coefficient of the regularization item is adjusted according to a distance between a tentative solution and a historical optimal solution and a lower bound gain rate; fixing a tentative solution of the discrete variable, setting solution precision according to a current upper and lower bound gap and a convergence rate, and solving an energy flow sub-problem to obtain a continuous variable solution and an upper bound; if the sub-problem is feasible, weighted averaging is carried out based on dual solutions in historical iteration to generate an optimal cut plane, and the optimal cut plane is added to the main problem; and if the sub-problem is not feasible, identifying a key constraint cluster causing infeasibility, generating a feasible cut plane and adding the feasible cut plane to the main problem.

A Multiple-Input Multiple Output (MIMO) device (200) determines a minimum output vector produced by a quadratic unconstrained minimization function. The quadratic unconstrained minimization function comprises a peak power factor, an average power factor, an Error Vector Magnitude (EVM) constraint factor for meeting an EVM constraint, and a plurality of penalty coefficients that penalizes outcomes that violate the EVM constraint. The MIMO device (200) minimizes the peak power factor and maximizes the average power factor. The minimizing and maximizing are each performed within the EVM constraint. The MIMO device (200) derives a Quantum Unconstrained Binary Optimization (QUBO) from the minimum output vector. The MIMO device (200) then configures the quantum precoder (210) to generate a configuration of qubits that are representative of a precoding vector that meets the EVM constraint and minimizes Peak to Average Power Ratio (PAPR) of a transmission from the MIMO device (200). The configuring comprises embedding the QUBO on the quantum precoder (210).

The application discloses an automatic theorem proving method based on neural network guidance and quantum optimization, comprising the following steps: extracting the logical structure features of a propositional logic task, and predicting the applicability weight of natural deduction rules by using a preconfigured neural network; performing a forward reasoning iteration process, dynamically identifying potential intermediate conclusions to determine a variable space, and constructing a quadratic unconstrained binary optimization (QUBO) model according to the variable space; in the construction process, the applicability weight is used to adjust the penalty coefficient of the rule constraint term, and the energy topography of the solution space is reshaped; according to the problem size, the computing resources are adaptively scheduled, the model is mapped to a coherent Ising machine or a classical simulation backend for solving, and the proof path is reconstructed through double verification. The application effectively solves the problems that the traditional method is blind in rule selection and the static optimization model is difficult to handle dynamic multi-step reasoning, and the efficiency and scalability of the proof are improved.

The invention provides a rock high slope mechanical parameter inversion method based on an IPSO-LSSVM-MCMC coupling model, and the method comprises the following steps: optimizing a penalty coefficient gamma and a kernel function parameter sigma of an LSSVM algorithm through employing an IPSO algorithm, and enabling an error between a monitoring point displacement value predicted by the LSSVM algorithm and a numerical simulation value to be minimum; the IPSO-LSSVM coupling model is substituted into an MCMC algorithm, slope monitoring displacement data are input, and corresponding rock mass mechanical parameters are obtained through inversion of the IPSO-LSSVM-MCMC coupling model. The method has the advantages that the precision and efficiency of rock mass mechanical parameter inversion are remarkably improved, and innovative technical support is provided for safety evaluation of major geotechnical engineering.

A microgrid multi-resource collaborative autonomous optimization method and system based on coupling constraint relaxation abstracts each physical unit with power regulation capability within the microgrid as an identical regulation agent. The minimum sum of the local objective functions of all regulation agents is used as the global objective function. Relaxation variables are introduced to correct the total power balance constraint satisfied by the global objective function. A global penalty term common to all regulation agents is constructed based on the relaxation variables and penalty coefficients, and this global penalty term is superimposed on the global objective function to obtain the collaborative autonomous optimization objective. Based on the collaborative autonomous optimization objective, a distributed gradient projection method is used to iteratively predict the output of each regulation agent. When the iterative convergence criterion is met, the iteration stops and the predicted output of each regulation agent is output as the result of the microgrid multi-resource collaborative autonomous optimization. This forms a microgrid internal collaborative optimization system that is decoupled from resource types, has no external dependencies, and possesses adaptive capabilities, realizing autonomous collaboration among multiple units under the constraint of coupling.

The present invention provides a calculation program, calculation method, and information processing device that can appropriately adjust the penalty coefficient. [Solution] The computer is instructed to search for a solution using a cost function and penalty term obtained by incorporating continuous relaxation into a discrete optimization problem, and in the process of doing so, it performs a process of changing the penalty coefficient of the penalty term using the gradients of the cost function and the penalty term.

The invention relates to the technical field of data statistics, in particular to a multi-dimensional index data statistics method and platform based on machine learning, and the method comprises the steps: extracting key information from an original delivery record, combining the key information to form a unified strategy unit set, and converting the strategy unit set into a corresponding vector set; constructing a path index set based on the delivery performance indexes, aggregating sample sets belonging to the same path structure in the strategy unit set, constructing a path index distribution difference score function to screen strategy paths, finally outputting output path sets meeting screening conditions, and combining the output path sets into a structure diagram; respectively outputting an index quantile structure and a fluctuation penalty coefficient corresponding to each path through a kernel weighted empirical quantile function and a distribution fluctuation metric function; and filtering the candidate recommendation path set through a comprehensive scoring function and a confidence constraint function, and outputting a final putting path suggestion.

The invention discloses a method for realizing accurate classification of power system network security data by using an SVM algorithm, and belongs to the technical field of power system network security. In order to solve the problem that the classification precision is insufficient when high-dimensional heterogeneous and class imbalance data is processed by a traditional method, a multi-level technical architecture is constructed; dimensionality reduction is carried out through an adaptive weighted feature selection algorithm fused by information gain and correlation; designing an improved hybrid kernel SVM model with adaptive density adjustment, and dynamically fusing a polynomial kernel and an RBF kernel; and introducing a dual-objective optimization model and a differential penalty coefficient strategy. According to actual measurement of a provincial power grid dispatching center, the attack sample detection rate is increased to 95.1% from 68.5%, the overall classification accuracy rate reaches 98.7%, the false alarm rate is reduced to 0.9%, and the training time lt of one hundred thousand samples is shortened; the method effectively solves the problem of power system network security data classification, and improves the detection precision and real-time performance.

This application relates to a method, apparatus, equipment, medium, and product for the coordinated operation of a power distribution network and buildings. The method includes: constructing an augmented Lagrangian function based on a pre-built two-layer model and consistency constraints on coupled variables; iteratively solving the two-layer model with the objective of minimizing the augmented Lagrangian function to obtain the optimal solutions for the upper-layer and lower-layer decision variables; determining a coordinated control strategy based on the optimal solutions for the upper-layer and lower-layer decision variables; and controlling the coordinated operation between the power distribution network and each building according to the coordinated control strategy. The upper-layer model is coupled to each lower-layer model through coupling variables, which include the target power of the upper-layer model and the actual response power of the lower-layer model. The augmented Lagrangian function includes an economic coordination term based on Lagrange multipliers and a quadratic penalty term based on penalty coefficients. This method balances reliability and computational efficiency.

The invention provides a power distribution network contact diagram generation method and a power distribution network contact diagram generation device. The method comprises the following steps: firstly, based on a topological structure of a power distribution network, generating an initial population meeting radial topological constraints, and setting constraint conditions; performing load flow calculation on each individual in the initial population to obtain network loss and constraint violation degree; adjusting a penalty coefficient for the proportion of individuals violating the constraint condition in the initial population, and constructing a fitness function by combining the penalty coefficient, network loss and constraint violation degree; and finally, carrying out genetic operation on the initial population based on a fitness function until a power distribution network connection diagram which meets constraint conditions and has minimum network loss is generated. According to the invention, the generation efficiency of the power distribution network contact diagram is improved.

To appropriately and efficiently set a magnitude of a penalty coefficient in representing a constraint condition about a logic operation as a penalty function.SOLUTION: According to one desirable aspect of the present invention, an information processing device includes a processor and a storage device, and has a penalty coefficient setting part for using a solution function for calculating a solution of a combination optimization problem by using a cost function and a constraint condition, and being realized by the processor and the storage device. In the information processing device, the penalty coefficient setting part sets a penalty function and a penalty coefficient such that a solution of the combination optimization problem satisfies the constraint condition on the basis of the constraint condition about a logic operation applied to two variables of the cost function and a value of a model coefficient of the cost function, and executes solution search of the combination optimization problem on the basis of the penalty function and the penalty coefficient.SELECTED DRAWING: Figure 12

There is provided a non-transitory computer-readable medium storing a calculation program for causing a computer to execute a process. The process includes, in searching for a solution using a cost function and a penalty term obtained by introducing continuous relaxation into a discrete optimization problem, changing a penalty coefficient of the penalty term by using a gradient of the cost function and a gradient of the penalty term.

The application relates to a fast time-domain frequency-domain anti-clutter waveform generation method, device, equipment and medium, a frequency-domain anti-clutter optimization problem model is established by taking the maximum frequency-domain signal-to-clutter ratio as a criterion, considering a waveform constant module constraint and a weighted integrated sidelobe level constraint, a penalty coefficient is introduced, the problem is converted into an optimization problem of minimizing clutter energy and a weighted integrated sidelobe level, a weight vector is introduced to convert the constraint optimization problem into an approximately equivalent problem model, the problem model is introduced into a complex circle flow space to convert the constraint optimization problem in the Euclidean space into an unconstrained optimization problem in the flow space, then, a Euclidean gradient and a Hessian matrix of the unconstrained optimization problem are calculated, based on a Riemannian gradient and a Riemannian Hessian matrix obtained through projection, the unconstrained optimization problem is subjected to a confidence interval descent in the Riemannian complex circle flow space, an iterative point is updated until convergence, and finally, a radar transmitting waveform is output. The method generates a radar transmitting waveform based on the difference between a target and clutter in the frequency domain, and can effectively suppress the clutter.

The invention discloses a dam abnormal data generation method based on a dynamic penalty weight CGAN, and belongs to the technical field of hydraulic structure safety monitoring and intelligent analysis. Aiming at the problems of unstable training, fixed penalty coefficient and insufficient authenticity of generated samples in dam monitoring data modeling of a traditional generative adversarial network, a dynamic penalty weight mechanism is introduced into a Wasserstein adversarial training framework, gradient penalty intensity is adaptively adjusted according to fluctuation of environmental variables such as water level and air temperature, and the dynamic penalty weight mechanism is introduced into the Wasserstein adversarial training framework. Keeping balance training of the model under different working conditions; and meanwhile, a multi-head self-attention mechanism and a bidirectional long-short-term memory network are combined in the generator to realize joint modeling of global association and local time sequence characteristics, so that a high-quality abnormal sample conforming to a dam physical rule is generated. Experiments show that the method can significantly improve the diversity and authenticity of generated data, enhances the accuracy, recall rate and AUC index of a downstream anomaly detection model, and has high engineering application value and popularization potential.

This application discloses an artificial intelligence-based dynamic control optimization method for warehouse environments, mainly relating to the field of control optimization technology. It addresses the problems of existing solutions where feature representations fail to reflect the local fluctuation patterns and phase differences of different microenvironments within the warehouse. The method includes: adjusting the convergence coefficient, adaptive adjustment factor, and spiral shape parameters involved in the whale optimization algorithm by introducing a population diversity index; employing a fitness function that integrates the total energy consumption target and constraint violation penalties, and using a dynamic penalty coefficient adjustment strategy based on the total historical violations, combined with fuzzy logic to smoothly change the penalty during the iteration process; constructing the next generation of whale population after each iteration through elite retention, crossover operations, and diversity maintenance; and outputting the current globally optimal whale individual position when a preset termination condition is met.

The invention discloses a cargo collection and distribution management system for multiple suppliers in a region, and belongs to the technical field of e-commerce and logistics. In order to solve the defect that multiple logistics costs and dynamic changes cannot be systematically considered in an artificial planning mode in the prior art, the system firstly obtains supplier and cargo information and logistics cost data; then, algorithm hyper-parameters such as a configurable penalty coefficient and a time discount coefficient are set, and rigid limiting conditions including cargo-container matching, capacity and distribution uniqueness are constructed; further, on the basis of the parameters and constraints, a comprehensive optimization target capable of quantifying hidden costs such as collection and distribution, sea transportation, allocation and various delays and space waste is constructed; and finally, inputting all the elements into a cargo collection and distribution heuristic engine, automatically solving an optimal solution considering the cost, the time efficiency and the resource utilization rate by the engine under the constraint by taking a minimization optimization target as a guide, and outputting an accurate collection and distribution, boxing and allocation scheme.

The application relates to the technical field of computer resource scheduling, in particular to an AI computing power optimization method based on multi-agent cooperation, which comprises the following steps: multi-modal task features are extracted, and a cooperative representation vector is generated in combination with double prediction heads; based on the vector, node resource occupation and a risk penalty coefficient generated by a load prediction distribution entropy value, a bilateral preference score is calculated; a bilateral stable matching algorithm with a capacity constraint is run to output a distribution scheme; for unallocated tasks, a marginal acceptance threshold is used to guide the tasks to be re-matched after one-way relaxation of constraints in an elastic interval, and a bottom distribution is set; and the prediction model parameters are updated online based on actual load deviation. Through multi-modal semantic perception, dynamic risk matching and an elastic negotiation mechanism, the application reduces node overload and display memory fragmentation, alleviates the scheduling congestion problem in a resource shortage state, and further improves resource utilization and system operation robustness in a high-concurrency environment.

The invention discloses an automatic theorem proving method based on neural network guidance and quantum optimization, and the method comprises the steps: extracting the logic structure features of a propositional logic task, and predicting the adaptability weight of a natural deduction rule through a pre-configured neural network; executing a forward reasoning iteration process, dynamically identifying a potential intermediate conclusion to determine a variable space, and constructing a quadratic unconstrained binary optimization QUBO model according to the variable space; in the construction process, the penalty coefficient of the constraint term of the rule is adjusted by using the adaptability weight, and the energy landform of the solution space is remodeled; computing resources are adaptively scheduled according to the problem scale, the model is mapped to a coherent Isin machine or a classical simulation back-end of the coherent Isin machine to be solved, and a proof path is reconstructed through dual verification. According to the method, the problems that rule selection is blind and a static optimization model is difficult to process dynamic multi-step reasoning in a traditional method are effectively solved, and the proving efficiency and the expandability are improved.

The invention discloses a reliability structure-control coupling design optimization method, and particularly relates to the technical field of electromechanical equipment optimization, and the method comprises the steps: building a mathematical model for electromechanical equipment deterministic structure-control coupling design optimization, and solving to obtain optimal structure physical design parameters; fixing the optimal control variable, and performing reliability analysis on the optimal structure physical design parameters to obtain In-MPP points and constraint function values in an index circle; the initial penalty coefficient is used as a starting point in subsequent iteration, and the penalty coefficient is updated in combination with the change of the target function value and the satisfaction change of the constraint function at the In-MPP point; and in the iteration process, calculating a translation vector of an inequality constraint function, constructing and solving a decoupled deterministic structure-control coupling design problem in combination with a penalty coefficient, and repeating the steps to finally obtain a convergent design scheme meeting the reliability requirement. According to the method, the reliability of an intermediate result in the optimization process is improved, the convergence efficiency is improved, and the application range of the reliability structure-control coupling design optimization method is expanded.

This invention relates to the intersection of power system optimization scheduling and quantum computing technology, proposing a method, device, and system for transmission line cross-section search based on optical quantum computing. The invention first constructs a power network model; then, based on the optimization objective, it constructs a QUBO model as the cross-section search model, with the objective being to minimize the objective function of the transmission lines. The cross-section search model is solved using an optical quantum computer to obtain the power grid cross-sections. The cross-section search model is the sum of the difference between the objective function and the cross-section weights, and the sum of the total constraint difference and the relaxation variables, multiplied by the corresponding penalty coefficients. The cross-section weights are the target value for the number of lines on the transmission line cross-section. This invention utilizes the topology reconstruction characteristics of optical quantum processors to encode power grid partition constraints into qubit combinations, generating a set of candidate cross-sections that satisfy the ground-state power flow in parallel. This solves the problem of missed selection caused by combinatorial explosion in traditional algorithms, improving the comprehensiveness and efficiency of the search.

The application discloses a bearing fault early warning method and system based on IWAOA improved SVMD noise reduction and LSTM-CNN classification, and the optimal decomposition layer number and penalty coefficient of successive variation modal decomposition SVMD are adaptively obtained by using IWAOA, the bearing vibration signal is decomposed by SVMD, the signal is reconstructed and denoised by screening effective modal components through a correlation coefficient, the acceleration peak value and kurtosis index of the reconstructed signal are extracted to form a feature vector, after dimension reduction by PCA, the feature vector is input into an LSTM_CNN classification model combined with a long short-term memory network and a convolutional neural network for training and classification, the time sequence dependence is captured, and the accurate identification of the normal and fault states of the bearing is realized. The application has better denoising effect, faster calculation speed, and the bearing fault early warning accuracy reaches 100%, and is suitable for fault early warning of different types of bearings and under different working conditions.