20 results about "Additive model" patented technology

The invention discloses a driving mechanism analysis method of reservoir group scheduling for flood non-consistent evolution, which comprises the following steps: acquiring a smooth reservoir entering flood sequence by utilizing a topologically coupled physical manifold constraint denoising model, and constructing dynamic topological characteristics representing propagation time lag and sensitivity through a hydraulic propagation mechanism-based Figure ordinary differential equation; constructing a structural causal model comprising a scheduling capability index, ectogenic hydrological driving and topological characteristics, and fitting a nonlinear dependency relationship by using a causal generalized additive model; executing anti-fact inference, and calculating a local causal driving index and a global causal cumulative effect index through a dry budget; and generating an optimal scheduling strategy for suppressing variation based on the causal index. According to the method, systematic analysis from data physical restoration to causal mechanism decoupling can be realized, the driving contribution of a scheduling behavior to flood inconsistency is accurately quantified, and decision support is provided for scientific flood control of a drainage basin.

The invention relates to the technical field of ecological assessment, and discloses a coastal wetland ecosystem condition assessment method and device, a medium and a product, different smooth regression functions are obtained through double generalized additive model processing, and the influence of an environmental single factor on vegetation and the cooperative influence of the environment and the vegetation on benthic organisms are accurately quantified. Furthermore, the index value is determined by taking historical data as a reference and combining a nonlinear rule of a smooth function, so that subjective weighting deviation is avoided. Furthermore, the weight is calculated based on the association between the real-time index value and the biological index, so that human subjective intervention is avoided, and the weight is ensured to reflect the actual contribution of the index to the ecological system. Meanwhile, the influence of single data deviation is reduced by using the multi-tree integration characteristic of the random forest model, so that the weight result is more stable. Furthermore, the complex ecosystem condition of the coastal wetland can be comprehensively evaluated in combination with the index value and the weight, the biodiversity is protected, and the stability of the ecosystem is maintained.

The invention discloses an automatic station soil humidity observation data layer-by-layer correction method based on meteorological environment factors and multi-model integration, which considers that the influence of evaporation and rainfall on soil humidity has a hysteresis effect and the change of humidity of different layers of soil is influenced by the change of humidity of an upper layer at the same time. The method comprises the following steps of: performing layer-by-layer deviation correction on automatic station soil humidity observation data by taking manually observed soil humidity as a reference, combining meteorological and environmental factors such as air temperature, wind speed, rainfall and vegetation index and comprehensively adopting a generalized additive model (GAM) to integrate various machine learning algorithms such as Cubist, random forest, XGBoost and CatBoost; the observation precision of the corrected soil humidity data is remarkably improved, particularly, more obvious improvement is shown in deep soil, construction of a high-precision and long-term stable soil humidity observation network is facilitated, and a reliable data basis is provided for verification of remote sensing and numerical simulation soil humidity products.

The invention relates to an intelligent decomposition and optimization method for an annual power generation plan of a power plant, belongs to the technical field of power plan scheduling, and solves the problems of low granularity and low evaluation dimension of an existing power plan scheduling model. The method comprises the following steps: dividing a target year into a plurality of basic optimization time periods based on the operation mode difference between a heat supply season and a non-heat supply season; taking the minimum annual total gas consumption predicted by the generalized additive model as an optimization target, and coupling a smooth penalty term to construct a basic optimization target function; generating a feasible operation interval of each basic optimization period; generating a plurality of feasible initial solutions in the feasible operation interval, performing iterative solution on each feasible initial solution by adopting a constraint optimization algorithm to obtain an optimal solution, and selecting a solution with the lowest basic optimization objective function value from all the optimal solutions as a final optimal solution; integer correction is carried out on the power generation amount and the heat supply amount in the final optimal solution, and a prediction interval of the gas consumption amount is generated based on the corrected power generation amount and heat supply amount.

The invention relates to the technical field of data analysis, and discloses a multi-dimensional feature-based connection passenger flow prediction method and device, and the method comprises the steps: obtaining user riding data corresponding to a public transport network and passenger flow influence data associated with the user riding data; preprocessing the user riding data to obtain target riding data; counting historical connection passenger flow data based on a connection passenger flow prediction formula and according to the target riding data; extracting features of the passenger flow influence data and the historical connection passenger flow data based on a multi-dimensional feature processing model to obtain multi-dimensional features associated with the historical connection passenger flow data; and on the basis of a dynamic generalized additive model, according to the multi-dimensional features, calculating the predicted connection passenger flow of any connection station. Therefore, the dynamic prediction of the bus and subway connection passenger flow can be realized on the basis of improving the analysis comprehensiveness of the multi-dimensional characteristics of the bus and subway connection, so that the prediction accuracy of the bus and subway connection passenger flow is improved.

The invention relates to a landslide displacement prediction method and system, and the method comprises the following steps: firstly obtaining landslide monitoring data, and carrying out the preprocessing of the data; decomposing the accumulated displacement into trend term displacement and periodic term displacement by adopting a time sequence addition model; analyzing and extracting influence factors of the displacement of the trend term and the periodic term, predicting the displacement of the trend term by using an extreme learning machine (ELM) model, and comparing a prediction result with a cubic polynomial prediction result based on a least square method; the influence of rainfall and reservoir level change is considered, a grid search support vector regression (GS-SVR) model is selected to predict periodic term displacement, and a prediction result is compared with an ELM prediction result; and finally, adding the displacement predicted values of the trend term and the periodic term to obtain a total displacement predicted value, and performing prediction accuracy evaluation according to the goodness of fit Rand the root mean square error RMSE.

This invention proposes an interpretable load forecasting method based on a neural additive model and a gradient boosting tree, belonging to the field of load forecasting technology. The method includes: constructing a basic sample set from historical load, meteorological sequences, calendar time, and resource allocation data, and dividing it into a main effect feature set and a complex interaction feature set; inputting the main effect feature set into an improved neural additive model to obtain first-stage features and fusion contribution values; calculating the first-stage prediction residual based on the first-stage features and actual load values; constructing a second-stage enhancement input based on the fusion contribution values ​​and the complex interaction feature set, inputting it into a gradient boosting tree, and deriving a corrected residual with the first-stage prediction residual as the target; obtaining the load forecasting result based on the first-stage features and the corrected residual, and performing hierarchical interpretable analysis based on shape function, feature attribution, and the load forecasting result. This achieves load forecasting with both high accuracy and high interpretability.

The invention discloses a neural imaging data processing method based on spatial effect detection. The method comprises the following steps: S1, establishing a regression relationship between high-dimensional function type data and response variables by adopting a function additive model; s2, space coordinates of function data are introduced; s3, introducing a regression coefficient, and identifying the boundary of sudden disappearance of the effect; and determining a brain region which has an influence on the cognitive function. According to the method, the space-time continuity of the internal effect of the boundary and the jumping property of the effect on the boundary are fully considered, and the non-standard space change mode of the effect is effectively identified, so that the actual data characteristics are described more accurately.

The invention discloses an electric power meteorological evaluation method based on a hybrid strategy generalized additive model. The method comprises the following steps: obtaining transformer area distribution transformer historical load data and meteorological data; on the basis of the preprocessed data, feature analysis is carried out on meteorological factors through a maximum mutual information coefficient method, a maximum mutual information coefficient matrix is formed, and key meteorological factors having significant influence on loads are screened out; establishing a generalized additive model, and fitting the load data and the key meteorological factors to obtain a nonlinear fitting curve between the power load and the meteorological factors; introducing a hybrid strategy evaluation index, and identifying a critical meteorological index corresponding to a load sudden change point through a maximum curvature method or a maximum slope method; and fine evaluation and regional risk early warning of the power load are realized by using the critical meteorological indexes. The method can quantify the influence of different meteorological factors on the load, automatically identifies the load sensitive critical point, improves the accuracy and scientificity of power grid risk early warning, and is suitable for load management and intelligent scheduling of each transformer area.

The invention discloses an industrial and commercial energy storage intelligent scheduling method based on data decomposition and ensemble learning, and the method comprises the following steps: S1, data preprocessing: calculating actual load data according to ammeter data, and carrying out missing value filling and filtering; s2, seasonal decomposition: decomposing the load data by using an addition model; and S3, abnormal value detection: using an abnormal value detection algorithm for the load data. S4, model training and testing: based on results of S3 and S2, sequentially performing feature engineering, parameter adjustment, cross fusion training and testing to obtain a load prediction value; and S5, photovoltaic prediction: carrying out model training and testing according to weather data and photovoltaic data. And S6, scheduling algorithm: based on the data in S4 and S5, adding energy storage operation parameters, constructing a revenue maximization model, solving and correcting by using an optimizer, and obtaining an energy storage charging and discharging strategy. Based on a data decomposition technology and load prediction, photovoltaic prediction and scheduling algorithms, an efficient charging and discharging strategy is made for industrial and commercial energy storage.

The application discloses a kind of reservoir group scheduling to the driving mechanism analysis method of non-consistent evolution of flood, comprising: using topological coupling physical manifold constraint denoising model to obtain smooth reservoir flood sequence, and the dynamic topological feature representing propagation time lag and sensitivity is constructed by graph neural ordinary differential equation based on hydraulic propagation mechanism;Structural causal model containing scheduling ability index, exogenous hydrology driving and topological feature is constructed, and nonlinear dependence relationship is fitted using causal generalized additive model;Perform counterfactual inference, calculate local causal driving index and global causal cumulative effect index by intervention operator;Based on causal index, the optimization scheduling strategy of inhibiting variation is generated.The application can realize the systematic analysis from data physics reduction to causal mechanism decoupling, accurately quantify the driving contribution of scheduling behavior to non-consistency of flood, and provide decision support for scientific flood control of river basin.

The invention provides an accurate regulation and control method and system for a water treatment biological membrane flora, and belongs to the technical field of water treatment biological membranes. The method comprises the following steps: a data acquisition step: acquiring microbial community data and environment operation parameters from a plurality of water treatment biological membrane systems; an index calculation step: calculating microbial diversity and functional characteristic indexes by using a 16S rRNA high-throughput sequencing result; a model construction step: constructing a neural network based on a structure interaction generalized additive model; an influence analysis step: obtaining the influence rate of each input variable and the variable pair on the microbial diversity and functional characteristic indexes through interpretability analysis; and a regulation and control output step: outputting a visual regulation and control suggestion based on a modeling result. The system comprises a data acquisition module, an index calculation module, an interpretation modeling module, an influence analysis module and a parameter regulation and control module. The method has the advantages of being high in interpretability, good in generalization performance, high in engineering applicability and the like, and a scientific basis and a universal method are provided for intelligent regulation and control of flora.

The present application belongs to the technical field of data processing, and particularly relates to a forest microwave remote sensing ground surface dead combustible moisture content monitoring method and system, which comprises the following steps: constructing a soil background scattering prediction model by using historical time series data, and obtaining the soil background scattering coefficient under the current environment; constructing a dielectric and energy coupling balance equation considering the two-way attenuation effect, and introducing a water impedance attenuation function and a water gain scattering function; constructing a residual target function according to the total backscattering coefficient observed by a satellite and the soil background scattering coefficient, and using a constrained interval search algorithm to invert the ground surface dead combustible moisture content. The present application decouples the soil background and the dead combustible signal, solves the problem that the traditional linear additive model is invalid in a humid environment, and realizes the accurate monitoring of the ground surface dead combustible moisture content under complex water and heat conditions.

The invention relates to a power load prediction method and device based on dynamic event awareness and electronic equipment, and the method comprises the steps: constructing a power news quantification model driven by a preset large model; based on the electric power news quantitative model, a medium and long term prediction model of time sequence decomposition and dynamic attention fusion is constructed, and the medium and long term prediction model comprises a time sequence prediction framework based on an additive model and a time sequence deep learning model based on an attention mechanism so as to capture a long-term periodicity rule and a short-term dynamic feature of the load; and outputting a predicted load result of the power system based on the dynamic sensing event by using a medium and long term prediction model of time sequence decomposition and dynamic attention fusion. Therefore, the problems that in related technologies, due to the fact that multi-modal fusion lacks an effective cross-modal space-time dynamic association mechanism and an artificial intelligence prediction model lacks interpretability analysis on key social event influence loads, deep fusion of social event semantics is difficult, response to emergencies is lagged, and interpretability is poor are solved.

This disclosure provides a method and system for detecting communication anomalies in a wind turbine pitch control system. First, the original CAN message stream is processed into a time series, extracting multiple inter-frame arrival time sequences. Then, the wind turbine's operating condition is determined in real-time using SCADA data, generating an operating condition time series. Further, an adaptive time series decomposition is used to combine the inter-frame arrival time sequences and the operating condition time series. Before decomposition, a logarithmic transformation is performed on the data, effectively addressing the inherent shortcomings of traditional additive model time series decomposition in handling heteroscedasticity, thus obtaining multiple variance-stable residual sequences. Finally, a pre-trained isolated forest model is used to score these residual sequences for anomalies, and the cumulative evaluation of anomaly scores combined with state judgment outputs the health status of the wind turbine pitch control system. This approach significantly improves the accuracy and robustness of communication anomaly detection in wind turbine pitch control systems.

The application discloses a kind of oilfield production robust prediction method and system based on neural additive model, it is related to petroleum well production prediction technical field, including: based on neural additive model SMART, by obtaining the input data consisting of multidimensional time series data, model training is carried out, and prediction model is constructed;Based on prediction model, using sparse learning strategy, mode-based measurement method and non-convex optimization algorithm, model optimization is carried out, and the production of oilfield is predicted according to the optimized prediction model.The application combines neural network and additive model, introduces mode-based measurement, sparse learning and non-convex optimization algorithm, effectively improves the accuracy, robustness, interpretability and efficiency of oilfield production prediction method, so that it is more suitable for application in actual scene.

The application discloses a land use data downscaling method based on a generalized additive model and related equipment, and the method comprises the following steps: acquiring a LUH2 data set and downscaling requirements, wherein the LUH2 data set comprises data of various land class types; performing fishing net construction on a preset area based on the LUH2 data set, the downscaling requirements and preset driving factor data to obtain a coarse fishing net and a fine fishing net; inputting the coarse fishing net and the fine fishing net into a preset initial generalized additive model for downscaling processing to obtain an intermediate land class area prediction result for the preset area; optimizing the initial generalized additive model based on the intermediate land class area prediction result to obtain a target generalized additive model; and outputting a target land class area prediction result for the preset area based on the target generalized additive model. The problem that the resolution of the LUH2 data set is low is solved.

The invention relates to a pump station dispatching system based on an AI intelligent agent. The system comprises a water load module and a pump set scheduling module, the water load module comprises an enhanced generalized additive model and a DQN reinforcement learning agent; the enhanced generalized additive model comprises a generalized additive model and a gradient boosting tree, the generalized additive model constructs a nonlinear response function of each feature according to historical data, and the gradient boosting tree optimizes the nonlinear response functions; the enhanced generalized additive model calculates the predicted water supply amount according to the nonlinear response function of each feature and the weight of the nonlinear response function; the DQN optimizes the weight according to the supply and demand deviation; after the enhanced generalized additive model obtains the predicted water supply amount, key boundary conditions of water supply demands are generated after confidence interval evaluation; and the pump set scheduling module outputs the optimal scheme of pump set scheduling in each time period under the condition of meeting the key boundary conditions of the water supply requirements in different time periods. The system has the advantages of accurate prediction, good robustness and excellent energy efficiency.

This invention relates to the fields of aerospace vehicle overall design, and discloses a method for predicting the cost and efficiency of solid rocket propulsion based on coordinate descent deep integration. The method includes: constructing a multi-source fusion database for solid rocket engines and training an integrated sparse neural network based on coordinate descent optimization; the integrated sparse neural network includes an additive model, which is constructed from base learners and corresponding weight coefficients; during the training of each base learner in the integrated sparse neural network, an objective function including a mean squared error term and a norm regularization term is constructed; the norm regularization term is used to apply sparse constraints to the weight matrix from the input layer to the hidden layer of the base learners; for a new design scheme of the solid rocket engine, its feature parameters are obtained and input features are constructed, which are then input into the trained integrated sparse neural network to obtain the predicted output response, which is used to estimate the cost or efficiency of the solid rocket engine propulsion system; this invention has fast convergence speed and high accuracy.

The invention relates to the field of aerospace craft overall design and the like, and discloses a solid rocket power cost performance prediction method based on coordinate descent depth integration, which comprises the following steps: constructing a multi-source fusion database of a solid rocket engine, and training an integrated sparse neural network based on coordinate descent optimization; the integrated sparse neural network comprises an additive model, and the additive model is constructed by a base learner and a corresponding weight coefficient; constructing a target function including a mean square error term and a norm regularization term in the training of each base learner of the integrated sparse neural network; the norm regularization item is used for applying row sparse constraint to a weight matrix from an input layer to a hidden layer of the base learner; for a new design scheme of the solid rocket engine, characteristic parameters of the solid rocket engine are obtained, input characteristics are constructed, the input characteristics are input into the trained integrated sparse neural network, and a predicted output response is obtained and used for carrying out cost estimation or efficiency estimation on a power system of the solid rocket engine; the method is high in convergence speed and high in precision.