10 results about "Momentum factor" patented technology

The invention discloses an index method for fitting a complex water level and flow relationship of a river channel section in a middle and downstream region of a drainage basin, and the method comprises the following steps: collecting long-time sequence data of a target river channel section, and carrying out the preprocessing of the long-time sequence data; calculating the change rate of water level to time and the change rate of flow to time; based on a one-dimensional river hydrodynamic model, solving a flow velocity v and a water depth h through a Saint-Venant equation set, and calculating a Froude number Fr; the water level Z, the change rate dZ / dt of the water level to the time, the change rate dQ / dt of the flow to the time and the Froude number Fr serve as four input neurons of the neural network, and the flow Q serves as an output neuron; setting a hidden layer activation function, the number of neurons, a learning rate and a momentum factor parameter, and training a neural network through a back propagation algorithm; the trained neural network is used for simulating the flow of the target section, and a complex water level flow relation curve is output; the simulation precision is greatly improved.

The invention discloses a document collaborative analysis method and system based on deep learning, and relates to the technical field of deep learning. The method comprises the following steps: extracting quantitative characteristics from literatures, screening and constructing a literature internal analysis event set, and determining an external interference event set at the same time; the method comprises the following steps: determining initial states of two types of events by adopting a single index method, constructing an event influence matrix, and determining a cross influence coefficient according to a mapping rule by calculating a co-occurrence correlation factor, a domain specificity factor and a trend momentum factor; and on the basis of the event influence matrix, meeting normalization constraint, calculating a time development index to quantify the event comprehensive influence, and updating the event state to predict the development change trend of the analysis event in the literature in the time domain. And fitting the initial state of the event with the predicted state value, and drawing a development trend chart to assist literature research. According to the method, internal and external factors are comprehensively considered, the complex relation of the literature is mined, the literature development trend is accurately predicted, and the defects of a traditional method are overcome.

The application provides a power system inertia analysis method, device, equipment and storage medium, wherein the method comprises: estimating the equivalent inertia of the power system using a physical model; correcting the equivalent inertia of the power system using a pre-trained data model to obtain updated equivalent inertia, wherein the data model adopts a denoising autoencoder, training labels are generated by including fine models of various types of inertia sources in the training process, and the update speed is adjusted by a momentum factor when updating the gradient; input the updated equivalent inertia into a simulation platform, add perturbations to each inertia node in the simulation platform to obtain the frequency of each inertia node, and calculate the distribution of each inertia node based on the updated equivalent inertia, the frequency of each inertia node and the center frequency of the inertia of the power system. The application proposes a hybrid intelligent driving scheme guided by a physical model and corrected by a data model, which can balance the calculation efficiency and prediction accuracy as a whole.

The application provides a new energy control parameter self-adaptive identification method based on double-layer particle swarm optimization, and belongs to the technical field of power system simulation analysis. Typical control strategies of heterogeneous new energy units are comprehensively used to construct a strategy database and perform verification; test data are acquired for processing and feature extraction, so that high and low threshold values, current initial states and test feature data are obtained; a fitness function is constructed based on the strategy database, a discrete particle swarm algorithm is used in the global optimization layer to select an optimal strategy combination from the strategy database, and in the local optimization layer, parameters of the selected strategy are continuously optimized in combination with a momentum factor and a learning rate. The application uses actual unit test data characteristic values as inputs of the double-layer particle swarm algorithm, determines an optimal strategy-parameter pair through a total error index weighting, verifies the accuracy of the strategy-parameter pair, and effectively improves the efficiency and precision of modeling of heterogeneous new energy equipment.

The invention relates to the technical field of artificial intelligence driven complex dynamic system model characterization and state prediction, in particular to a flue gas desulfurization system state prediction method based on a self-adaptive radial basis function neural network, which comprises the following steps: acquiring real-time working parameters of a target flue gas desulfurization system; the real-time working parameters are input into a preset state prediction model, the real-time flue gas desulfurization state of the target flue gas desulfurization system is predicted, the state prediction model is obtained based on training of a training set, and the training set comprises historical data of different desulfurization processes of the flue gas desulfurization system. The state prediction model is constructed by introducing a weight memory mechanism and adaptive hyper-parameters into a radial basis function neural network. According to the method, multi-step information is efficiently extracted by utilizing a model weight information memory mechanism, a learning rate and a momentum factor self-adaptive updating mechanism, the sensing and prediction precision of the prediction model on the concentration state of the SO2 in the outlet gas is improved, and the calculation complexity of the model is reduced.

The application relates to the technical field of high-dimensional data, in particular to a high-dimensional data fitting algorithm and system based on an improved gradient descent, the algorithm is deeply coupled with a long short-term memory (LSTM) network, a gated differential adaptive momentum variable gradient descent (AM-VGD) is designed for high-dimensional time series data, high-precision fitting of high-dimensional data is realized, the gated differential adaptive momentum variable gradient descent (AM-VGD) is proposed, different momentum factors and gradient decoupling items are designed according to the gradient characteristics of the LSTM forget gate, input gate and output gate, the gradient propagation efficiency is improved by 40% under high dimension, the model convergence iteration number is reduced to <=500 times, the convergence speed is improved by more than 2 times, a stacked auto-encoder (SAE) is fused to decouple and extract high-dimensional features, a high-dimensional decoupling regular term of the gradient descent is combined, the collinearity interference between features is eliminated, the fitting determination coefficient R2 of high-dimensional time series data is greater than or equal to 0.98, the mean square error (MSE) is reduced by more than 60%, and there is no overfitting / underfitting phenomenon.

The invention relates to the technical field of mobile terminal safety monitoring, and discloses a risk signal screening monitoring method and system for a mobile terminal. The method comprises the following steps: collecting intensity change sequences of various risk signals of the mobile terminal, calculating cosine similarity and propagation momentum factors of deviation rate vectors, counting time sequence mutual information of the risk signals to determine a causal propagation relationship, constructing a directed propagation graph to predict a future risk evolution value, and calculating the risk evolution value of the mobile terminal. And calculating a weighted distance, dynamically adjusting a judgment threshold and judging a multi-step progressive attack. The technical problems that in the prior art, staged progressive attacks cannot be recognized, risk evolution cannot be pre-judged in advance, the detection sensitivity cannot be adaptively adjusted, and the response strategy is extensive and single are solved. According to the method and the device, the recognition accuracy and the response timeliness of the multi-step progressive Trojan horse attack are improved.

The invention discloses a subjective and objective combined equipment health evaluation method and system, and the method comprises the steps: synchronously obtaining objective sensor data and subjective expert evaluation data of equipment operation through a multi-source data collection module, carrying out the preprocessing, and carrying out the parallel calculation of an objective weight and a subjective weight through an improved entropy weight method and a group decision AHP algorithm, dynamically fusing the two types of weights by using a time-varying combination coefficient strategy, and introducing a momentum factor to smooth weight mutation; real-time data processing and complex calculation separation are realized through a cloud edge collaborative architecture, weight full-life-cycle traceability is realized in combination with block chain evidence storage, and evaluation accuracy is ensured by a built-in multi-dimensional verification mechanism. The method achieves the complementation of subjective and objective advantages, solves the problems of high subjectivity, poor adaptability and insufficient traceability of evaluation indexes in the prior art, improves the equipment health evaluation accuracy, reduces the delay control, and is suitable for the full-life-cycle health management of various industrial equipment.

The present application relates to the technical field of artificial intelligence driven complex dynamic system model representation and state prediction, in particular to a flue gas desulfurization system state prediction method based on an adaptive radial basis function neural network, comprising: obtaining real-time working parameters of a target flue gas desulfurization system; inputting the real-time working parameters into a preset state prediction model to predict the real-time flue gas desulfurization state of the target flue gas desulfurization system, wherein the state prediction model is obtained based on training of a training set, the training set comprises historical data of different desulfurization processes of the flue gas desulfurization system, and the state prediction model is constructed by introducing a weight memory mechanism and an adaptive hyperparameter into a radial basis function neural network. The present application efficiently extracts multi-step information by using a model weight information memory mechanism, a learning rate and a momentum factor adaptive update mechanism, improves the perception and prediction accuracy of the prediction model for the SO2 concentration state of the outlet gas, and reduces the model calculation complexity.

The invention provides a new energy control parameter adaptive identification method based on double-layer particle swarm optimization, and belongs to the technical field of power system simulation analysis. The typical control strategies of the heterogeneous new energy unit are integrated, and a strategy database is constructed and verified; acquiring test data for processing and feature extraction to obtain high and low pass threshold values, a current initial state and test feature data; and constructing a fitness function based on the strategy library, screening an optimal strategy combination from the strategy library by using a discrete particle swarm algorithm in a global optimization layer, and continuously optimizing a parameter set of a selected strategy in combination with a momentum factor and a learning rate in a local optimization layer. According to the method, the characteristic value of the actual unit test data is extracted as input of the double-layer particle swarm algorithm, the optimal strategy-parameter pair is judged and output through total error index weighting, the accuracy of the strategy-parameter pair is verified, and the efficiency and precision of heterogeneous new energy equipment modeling are effectively improved.