826 results about "Data Noise" patented technology

The invention discloses a solar radiation space-time prediction method and system based on physical information constraint and a neural network, and the method comprises the steps: collecting multi-dimensional time sequence meteorological data, extracting high-dimensional time sequence dynamic features, converting geographic space data into a fuzzy set, and carrying out the defuzzification of the fuzzy set through an inference rule, thereby obtaining geographic space features; and a gating mechanism is adopted to realize deep fusion of the space-time features to generate high-dimensional space-time fusion features. In a model training stage, an energy conservation equation is introduced into an optimization process, a physical residual error is constructed by calculating a time derivative and a space derivative of a predicted value, a physical constraint total loss function is formed in combination with a data loss item, and model parameters are updated by using a gradient descent method. According to the method, the accuracy and reliability of a prediction result are remarkably improved while the calculation efficiency is ensured, and the method is particularly suitable for solar radiation prediction under complex meteorological conditions; according to the method, abnormal prediction caused by data noise can be effectively corrected, and a solution with physical rationality and data adaptability is provided for the fields of solar resource evaluation, photovoltaic power generation power prediction and the like.

The invention discloses a multi-dimensional information integrated transformer health state monitoring method, which relates to the technical field of power systems, and comprises the following steps: deploying a multi-source sensor, collecting various signals of a transformer, and carrying out signal conditioning and analog-to-digital conversion on analog signals collected by the sensor; performing timestamp alignment on the received multi-source heterogeneous data, and eliminating data noise by adopting a filtering algorithm; key features which are sensitive to the health state of the transformer and complement each other are selected from the extracted feature values to form a health state evaluation index set; setting a sliding time window, obtaining historical data in the window, performing normalization processing on the data, calculating the information entropy and the variation coefficient of each index, and fusing the information entropy and the variation coefficient to obtain the dynamic weight of each index; obtaining a health state index of the transformer according to the normalized value and the dynamic weight; and displaying the health state index and the change trend thereof in real time, and setting an early warning threshold value and an alarm threshold value of the health state index.

The invention discloses a laboratory detection data processing method and system based on a machine learning technology, and relates to the technical field of data processing. According to the method, the signal is decomposed through discrete wavelet transform, the noise standard deviation is calculated based on the median of the highest frequency coefficient, the high signal-to-noise ratio data is reconstructed through the dynamic threshold and the soft threshold function, and the signal quality is improved; a sliding window is used for extracting time sequence signal statistics and FFT frequency domain features, spatial features are extracted in combination with an SIFT algorithm, high-correlation features are reserved through mutual information screening, and redundancy is reduced; constructing a graph convolutional network anomaly detection model and an XGBoost-LightGBM weighted regression model, and eliminating pollution data through an anomaly probability threshold to obtain a normal regression predicted value; predicting a compensation amount according to the environmental parameters by using an LSTM network, and obtaining calibration laboratory data according to the normal regression predicted value and the predicted compensation amount; according to the method, the problems of noise sensitivity, feature splitting, model isolation and environment drifting of multi-source data are solved, and the laboratory analysis precision and robustness are remarkably improved.

The invention relates to the technical field of transformer substation intelligent management, in particular to a modular prefabricated cabin transformer substation intelligent comprehensive management system with a multi-source data integration and collaborative management function. Comprising a multi-source data acquisition unit; the edge intelligent processing unit is used for carrying out local preprocessing, feature extraction and abnormal early warning on the acquired multi-source heterogeneous data of the transformer substation, and realizing data noise reduction, abnormal recognition and equipment health degree evaluation through a lightweight AI algorithm and a hybrid communication protocol carried by the edge computing terminal module; and a data integration collaboration unit. According to the invention, through an adaptive sampling strategy and a multi-modal feature fusion mechanism of the multi-source data acquisition unit, multi-dimensional data such as electric power parameters, power environment, fire protection and security protection and the like are incorporated into a monitoring system, feature layer association analysis is realized by depending on a Jousseme distance and a Dempster synthesis rule, the assessment limitation of single equipment and single-dimensional data in a traditional scheme is broken through, and the method has the advantages of high reliability and high reliability. And a more comprehensive state basis is provided for equipment health degree research and judgment.

The invention provides a construction site safety information supervision system based on the Internet of Things, and relates to the technical field of data processing, the risk assessment accuracy is remarkably improved and the data noise interference is inhibited through a deep reinforcement learning dynamic optimization grading early warning strategy and in combination with a multi-source data compression and federated learning framework; a risk coupling model is constructed based on a complex network theory, a conduction path and a linkage effect between regions are analyzed, a global early warning scheme is optimized, and the adaptability of the system to a complex construction environment is enhanced; the model is continuously optimized through a dynamic weight updating mechanism, high adaptability is kept under the environment change of a construction site, meanwhile, a resource dynamic allocation strategy is adopted, the problem of low high-concurrency processing efficiency is effectively solved, and the overall accident early warning accuracy is improved.

The invention discloses an underwater acoustic signal denoising method based on a time-frequency adaptive dual-path Conformer network, and the method comprises the steps: carrying out the preprocessing of real marine environment noise and underwater acoustic target signals, and generating a multi-signal-to-noise-ratio noisy data set; performing short-time Fourier transform on the noisy data, and extracting real part and imaginary part features to form a feature tensor; extracting features by using a feature encoder, and generating intermediate feature representation; respectively extracting a time path feature and a frequency path feature through a time-frequency adaptive dual-path Conformer network; performing multi-scale convolution processing and weighted fusion on the extracted time-frequency features by adopting a multi-scale fusion dynamic gating network; performing nonlinear mapping on the fused features by using a feature decoder to generate a mask matrix; and restoring the complex frequency spectrum based on the mask matrix, and restoring the denoised underwater acoustic time domain signal. According to the method, time-frequency information is fully mined in combination with underwater sound noise characteristics, and the underwater sound signal denoising effect is improved.

The invention discloses a physics-based method for discovering a control equation from scarce and noise data. According to the method, a physical information neural network and a sparse regression method are combined, and a partial differential control equation of a dynamic system is found from scarce and noisy data. Firstly, the size of a candidate function library is effectively reduced through a dimension verification method. And then modeling a physical system and calculating a candidate function by using the strong nonlinear fitting capability and automatic differential characteristic of the deep neural network. And finally, the form of a control equation and the coefficient of the equation can be obtained through sparse regression, and the coefficient is finely adjusted through DNN. The method not only can discover the control equation from the data, but also can obtain the network model to realize the response prediction of the dynamic system. The method is easy to implement, efficient, high in precision and high in universality. The method can be widely applied to physical knowledge mining, modeling and reasoning of a complex dynamic system.

The invention discloses a transcriptomics spatial domain identification method, and belongs to the technical field of transcriptomics. The objective of the invention is to solve the problems of low data noise reduction precision and poor recognition effect of an existing spatial transcriptional spatial domain recognition method. The method comprises the following steps: firstly, obtaining an undirected neighborhood graph according to a gene expression matrix, obtaining embedded representation of the gene expression matrix by utilizing an encoder, obtaining a corresponding reconstruction matrix by utilizing a decoder, and further determining reconstruction loss; meanwhile, a ZINB model is used for fitting a reconstruction matrix, and a ZINB loss function is obtained; then, an augmented graph is constructed based on the undirected neighborhood graph, respective embedded matrixes are obtained through an encoder, the comparison loss of the undirected neighborhood graph and the comparison loss of the augmented graph are obtained through a comparison representation learning mechanism, and then the neighbor comparison loss is obtained; total target loss is obtained based on all losses, a joint optimization strategy is adopted for training, and after training of the whole model is completed, dimensionality reduction and spatial domain recognition are carried out on a generated reconstruction matrix.

The invention discloses an intelligent health management system and method based on edge calculation and reinforcement learning. The system comprises an old people data acquisition module, an edge calculation module, a multi-modal fusion module, a reinforcement learning intervention module and a block chain evidence storage module. Physiological and environmental data are collected through a wearable device, a millimeter-wave radar, a pressure sensor and the like, data noise reduction and anomaly detection are realized at an edge end, a cloud end adopts a CNN-LSTM neural network to fuse multi-modal features, reinforcement learning is combined to dynamically generate personalized intervention actions, intervention behaviors and results are stored through a block chain technology, and the intervention effect is improved. And safe and credible strategy optimization is realized. According to the invention, continuous monitoring and intelligent intervention of the health state of the old people can be realized, and intelligence and credibility of community health management are improved.

The invention provides a sensor anomaly detection method and device based on a long short-term memory network, electronic equipment and a medium, which are applied to the technical field of fault detection, and effectively reduce data noise and improve the input quality of a model by performing preprocessing of null value removal and feature screening on acquired sensor data. Inputting the preprocessed sensor data into an improved attention mechanism-long and short-term memory network model, capturing a long-term dependency relationship of a time sequence by utilizing a long and short-term memory network, and dynamically distributing weights and focusing on key information by virtue of an attention mechanism, so that the model prediction precision is remarkably improved; and finally, the prediction result is compared with historical data and real-time data, so that the sensor abnormality is accurately judged, the defects of a traditional method in processing complex time sequence data are effectively overcome, the accuracy and reliability of abnormality detection are remarkably improved, and meanwhile, the method has prediction capability and adaptivity and is suitable for popularization and application. And characteristic differences of different sensors can be adapted.

The embodiment of the invention provides a hail identification and prediction method based on multi-source meteorological data fusion and an attention mechanism. The method is applied to the technical field of meteorology and artificial intelligence processing, and comprises the following steps: collecting dual-polarization radar data, satellite remote sensing data and ground meteorological observation data, and carrying out time sequence alignment, data standardization and feature splicing processing; extracting physical mechanism features and statistical texture features of hail clouds from the dual-polarization radar data, the satellite remote sensing data and the ground meteorological observation data; extracting multiple spatio-temporal features of the hail cloud by using a spatio-temporal feature extraction network; and inputting real-time observation data into the trained spatial-temporal feature extraction network, and outputting a hail occurrence probability, a nuclear region position and intensity grade distribution. In this way, the technical problems that in the prior art, multi-modal data noise, inconsistency and insufficient time-space feature capture are caused, and the precision and real-time performance of hail recognition are limited can be solved.

The invention discloses an information transmission attack event intelligent prediction model training optimization method, which comprises the steps of a data preparation stage, a feature engineering stage, a model selection and construction stage, a model training stage and a model evaluation and optimization stage, and is characterized in that data are collected from multiple channels, noise is cleaned and removed, missing values and abnormal values are processed, and accurate labeling is carried out; then key features are extracted, redundant irrelevant features are removed through selection, standardized transformation is carried out, a proper model is selected based on data features, an integrated learning combination model is adopted, then a training set, a verification set and a test set are divided, a proper optimizer is selected, parameters are adjusted to be over, training is carried out through the training set, and over-fitting is prevented through the verification set; and finally, using index evaluation of accuracy and recall rate, and adjusting hyper-parameters and continuous optimization of the model structure through cross validation. According to the method, data collection, data imbalance processing and data noise interference processing are adopted, data diversity is improved, and interference of data noise is reduced.

The invention discloses an AI model training data protection method and device based on local differential privacy. The method comprises the steps of noise scale determination, wherein the noise scale is determined through a Laplace differential privacy mechanism according to a preset privacy budget value and obtained input data; a data noise adding step: generating Laplacian noise according to the noise scale, and adding the Laplacian noise to the input data; a gradient noise adding step: determining the gradient of each data point in the input data after the Laplacian noise is added according to a preset initial model training parameter, and adding the Laplacian noise to the gradient of each data point to obtain a noise-added gradient value; and a parameter determination step: determining a prediction model training parameter according to the noise-added gradient value. According to the invention, the accuracy and efficiency of model training can be improved under the condition of effectively avoiding privacy leakage of training data.

The invention discloses a water quality monitoring system and method, and relates to the technical field of water quality monitoring, the system comprises a water quality data acquisition unit used for acquiring water quality parameter information; the data noise reduction unit is used for constructing a characteristic component matrix group according to the water quality parameter information and generating water quality characteristic sequence data through threshold filtering and matrix reconstruction; and the abnormal data monitoring unit is used for acquiring a water quality linkage sensitive factor and a water quality unconventional fluctuation index based on the water quality characteristic sequence data, calculating a water quality reference weight of each water quality characteristic sequence, and analyzing a water quality abnormal degree in combination with an abnormal score. Based on fusion analysis of the water quality reference weight and the anomaly detection score, the efficiency and reliability of water quality monitoring are improved.

Embodiments of the present application provide a video generation method and apparatus, a device, and a medium. The method can be applied to the technical field of video content generation, and is used for improving the video generation quality. The method comprises: acquiring a sample video frame sequence from a sample video, determining a first step count and sample original noise, and performing data noise addition processing on the sample video frame sequence to obtain video input data; inputting the video input data, a sample text encoded feature corresponding to sample description text, and first embedding information corresponding to the first step count into an initial generation model; and performing noise prediction on the sample video frame sequence by means of M spatiotemporal residual components and M spatiotemporal attention components in the initial generation model to obtain sample predicted noise, correcting a network parameter in the initial generation model on the basis of the sample original noise and the sample predicted noise, and determining the initial generation model comprising the corrected network parameter as a video generation model.

The invention proposes a hard negative sampling comparison recommendation method based on a graph generation diffusion model, and the method comprises the following steps: coding a user-article interaction graph through a graph convolutional neural network, obtaining a node embedded vector, and carrying out the inner product of the node embedded vector, and obtaining a recommendation probability; the model parameters of the graph convolutional neural network are obtained through the following training steps: S1, generating a comparison view 1 and a comparison view 2 from the node embedding vector through an inline diffusion model view generator and a parameter topological noise reduction view generator; the inline diffusion model view generator comprises an inline diffusion model, and the parameter topological noise reduction view generator comprises a parameter topological noise reduction network; and S2, carrying out comparative learning on the comparative view 1 and the comparative view 2. According to the method, a signal alignment strategy is combined, a double-view generator comparison learning framework is built, and the problems that the model generation capacity is insufficient and the influence of data noise is large are solved, so that the performance of a recommendation system is improved.

The invention relates to the technical field of man-machine interaction, and discloses an LED display screen interaction method and an LED display screen interaction system.The method comprises the following steps that user interaction data are collected through a multi-sensor module, and the user interaction data comprise visual data, touch data, infrared data and environment data; preprocessing the collected multi-modal user interaction data, including data noise reduction, feature extraction and unified coding; performing time sequence modeling on the multi-modal user interaction data based on the dynamic data fusion model to generate a fusion state; and identifying the user interaction intention based on the user interaction intention prediction model. Through the technical scheme of combining multi-modal data fusion and deep learning, all-around acquisition and comprehensive perception of user interaction information are realized by collecting visual, touch, infrared and environmental data of a user, and the defects of limited data dimension, poor environmental adaptability and insufficient interaction precision in single-modal interaction are overcome.

The invention provides a crop planting environment data acquisition and real-time optimization device integrated with the Internet of Things technology, which comprises a data acquisition module, a data cleaning and feature extraction module, an environment optimization strategy generation module, a decision support and execution module, an intelligent learning and optimization module and a monitoring and data visualization module, relates to the field of communication and data processing. The environment data is collected in real time through the distributed Internet of Things equipment, the planting area is fully covered, the problem of insufficient data acquisition in the prior art is solved, and a reliable data source is provided for subsequent optimization. The real-time performance and comprehensiveness of data acquisition are improved, and the accuracy of basic data optimization is ensured. An adaptive signal enhancement model is introduced, complex interference is effectively identified and eliminated, the intensity of weak signals is enhanced, and the problem that the optimization effect is reduced due to data noise in the prior art is solved. The precision and efficiency of data processing are improved, and high-quality input is provided for an optimization model.

The invention relates to the field of knowledge base construction, and discloses a knowledge base construction method for assisting digital employee question answering, and the method comprises the following steps: collecting internal multi-modal data of an enterprise, including document data, voice data and metadata; carrying out information amount analysis on the collected data, and screening high-value data through an information entropy theory; performing interactive verification on the screened multi-modal data to analyze data consistency; dynamically correcting the verified data, and reducing data noise in combination with a prediction model and an observation model; and performing semantic extraction and clustering optimization on the corrected data to generate knowledge entries. Through the closed-loop process of multi-modal data collection, dynamic correction, semantic extraction, distributed updating and intelligent question and answer, the data processing efficiency, the knowledge base quality and the question and answer accuracy are remarkably improved, and the problems that in the prior art, unstructured data processing efficiency is low, updating is lagged, and dynamic optimization capacity is lacked are solved.

The invention provides an aircraft fault early warning method and system based on flight data, and relates to the technical field of aircraft fault early warning. The method comprises the following steps: firstly, mining out an aircraft external data semantic vector and an aircraft internal data semantic vector corresponding to flight data to be processed; secondly, performing semantic enhancement operation on a pre-generated flight data noise vector based on an aircraft external data semantic vector, and outputting a flight data enhancement vector; then, according to the aircraft internal data semantic vector, mining an aircraft internal depth vector; and finally, according to the flight data reinforcement vector, carrying out association mining operation on the aircraft internal depth vector, outputting a target flight data semantic vector, and according to the target flight data semantic vector, analyzing aircraft fault early warning information. On the basis of the content, the problem that in the prior art, the reliability of aircraft fault early warning is relatively low can be solved.

The invention discloses a high-precision real-time positioning system and method for positioning personnel based on multi-mode fusion. The system comprises a multi-source data acquisition module, an intelligent data processing module, a fusion positioning model construction module and a positioning result application module, provides visual monitoring, trajectory analysis and scene customization functions, and supports flexible expansion through modular design. According to the method, satellite positioning, Bluetooth beacon and inertial navigation technologies are integrated, the weight is dynamically adjusted in combination with an adaptive weighted fusion algorithm, and a fusion positioning model supports outdoor (satellite dominant + EKF calibration, the precision being 2-5 meters) and indoor (beacon dominant + fingerprint matching, the precision being lt); according to the invention, intelligent switching between three modes (inertial navigation + ZUPT correction) and signal loss (1 meter) is realized, data noise is optimized by using Kalman filtering, and high-precision real-time positioning in indoor and outdoor complex scenes is realized. According to the invention, the problems of insufficient precision, poor environmental adaptability and weak expansibility of a single positioning technology are solved, and the comprehensive performance of the positioning system is significantly improved.

The invention relates to the technical field of air leakage big data analysis, in particular to a compressed air pipeline air leakage intelligent analysis system and method based on LLM. A plurality of monitoring nodes in a compressed air pipeline are analyzed, and data normal range parameters and adaptive acquisition frequency are preset for each node in the pipeline operation stage; the clear definition standard avoids misjudgment of original data due to fuzzy threshold values, then data noise is effectively reduced by analyzing the data synchronization degree of each monitoring node and intelligently repairing the problems of missing, deviation and the like, the repaired data better fits the actual operation state of the compressed air pipeline, the missed detection and mispositioning risks are reduced, and the reliability of the compressed air pipeline is improved. And finally fusing the data of the gas leakage node and the adjacent node to supplement multi-dimensional information, and optimizing the fusion data quality. The LLM model depends on high-quality fusion data, deep analysis is carried out in combination with compressed air pipeline multi-monitoring-node correlation characteristics and historical experience, deviation caused by data defects is reduced, and the accuracy of an intelligent air leakage analysis result is greatly improved.

The invention discloses a dynamic identification method for unusual cells before young tumors based on multi-omics data, and relates to the technical field of cell unusual identification. A dynamic correlation intensity matrix and a cumulative effect contribution matrix are constructed, a differentiation screening strategy is implemented according to individual response characteristics, and the unusual cells before young tumors are identified. And the abnormal dynamic high-fidelity identification of the young tumor pre-cells is realized. And aiming at individuals of different response types, an instant path, a long-term path or a double-path fusion strategy is respectively adopted, key behavior data is accurately screened, and the input quality is improved. According to the method, redundant interference is effectively eliminated, the simulation capability of the model on key processes such as immunosuppression and DNA damage accumulation is enhanced, the biological rationality and prediction precision of a cell state evolution sequence are remarkably improved, and the problems of model response lag, low calculation efficiency and output distortion caused by data noise in the prior art are solved; and a reliable technical support is provided for early warning and individualized intervention of precancerous lesions.

The invention discloses an unmanned aerial vehicle target identification method based on multi-modal information fusion, and mainly relates to application research of an artificial intelligence technology in the field of unmanned aerial vehicle target identification. The method is based on a multi-modal large model, radar, optical, radio and other multi-modal unmanned aerial vehicle detection data are input at the same time for co-training, processing and reasoning, few-sample reasoning of downstream tasks is realized by using rich priori knowledge, and a small unmanned aerial vehicle target identification result is returned. According to the method, the features of the unmanned aerial vehicle are comprehensively described from different dimensions by utilizing mutual complementation and mutual promotion of different modal information, so that the problem of multi-modal data missing can be solved, high-quality multi-modal data are integrated, data noise is better coped with, and the accuracy, availability and robustness of reasoning of an unmanned aerial vehicle target recognition algorithm are improved.

The invention relates to the related technical field of power network control, in particular to a multi-mode fusion power network operation state evaluation method and device, and the method comprises the steps: receiving power network switch control and operation state monitoring parameters, obtaining prediction parameters through the analysis of a primary network and a secondary network, adding the same attributes, comparing the obtained prediction parameters with the monitoring parameters, and obtaining a prediction result; the abnormal circuit is fed back to the decision-making layer, switch disconnection control is executed, and the technical problems that in a complex power network, influences of operation environment fluctuation and data noise are likely to happen, the operation state of the power network is difficult to respond and adjust in time, and abnormal conditions and sudden faults cannot be effectively handled are solved; the method has the technical effects of improving the accuracy of operation state prediction parameters, accurately isolating and disconnecting abnormal circuits, analyzing prediction deviation of a main network and a secondary network layer by layer, establishing a dynamic adjustment and grading evaluation mechanism, efficiently processing redundancy and abnormal conditions of multi-modal data in the power network, and improving the operation efficiency and safety of the power network.

The present invention is a parameter estimation method for compartment model based on physics-informed neural networks. Starting from a physical model, the method extracts information from an AIF and a small amount of measurement data to obtain kinetic parameters, thereby greatly improving the scanning efficiency of a measuring instrument, and reducing occurrence of inaccurate estimation results due to patient movement. In addition, the present invention has the robustness to AIF noise and measurement data noise, and can flexibly arrange the time of data acquisition, reduce an error of inaccurate estimation caused by long time 10 acquisition and the patient movement, and improve the efficiency of data acquisition of the instrument. Experimental results show that the present invention is more stable and has less errors. Meanwhile, the present invention does not require the setup of training datasets, and is superior to an end-to-end supervised reconstruction method U-net network with fewer samples.

The invention belongs to the technical field of wastewater treatment, and provides a wastewater treatment equipment remote control system based on the Internet of Things, which aims at solving the problems that the traditional fixed DO concentration control cannot adapt to water inlet load fluctuation, the effluent is easy to be substandard or the aeration energy consumption is high, the monitoring data noise is large, and the time sequence is misplaced. According to the scheme, a multi-dimensional real-time sensing network is constructed, flow, COD, BOD, ammonia nitrogen, water temperature, distributed DO and sludge activity sensors are deployed, discrete wavelet transform is adopted for noise reduction, and a data time sequence is aligned; a three-layer load-DO-energy consumption dynamic correlation model is designed, a basic layer predicts a load trend through LSTM, a middle layer quantifies a DO demand through a microbial metabolism model, and an optimization layer outputs an optimal DO set value through a PPO algorithm; and developing a dynamic decision-making system, adaptively generating a DO interval according to a load state, and adjusting fan parameters and number in a linkage manner. According to the invention, load dynamic adaptation is realized, effluent ammonia nitrogen is guaranteed to reach the standard, aeration energy consumption is reduced, and remote control precision and equipment operation efficiency are improved.

The invention relates to a user operation risk dynamic monitoring method and device based on data consanguinity and a medium, and the method comprises the following steps: obtaining operation behaviors of all users in real time, and extracting a segment of continuous operation behaviors of each user as a group of behavior fragments with semantic relevance; a plurality of continuous behavior segments of each user are matched with a preset behavior template, so that a plurality of intermediate behavior blocks are obtained through dimension reduction; according to a time sequence, traversing and extracting a predetermined number of intermediate behavior blocks from the plurality of intermediate behavior blocks, and respectively combining to generate a behavior fragment sequence; and respectively inputting the behavior fragment sequences into a pre-trained neural network model to obtain a risk classification result or a risk probability distribution result corresponding to each behavior fragment sequence. The method has the advantages that each operation behavior can be subjected to structured expression, the potential behavior chain can be reconstructed through the combination strategy, and the high-risk behavior path hidden in the data noise is recognized by means of the deep model.

The invention discloses a photovoltaic power generation power prediction method based on deep learning and multi-source data fusion, and the method comprises the steps: carrying out the preprocessing of the real-time power generation data, environment parameter data and weather forecast data of a photovoltaic power station, reducing the data noise through Gaussian filtering and bilateral filtering, and then carrying out the prediction of the real-time power generation data, environment parameter data and weather forecast data; according to the method, feature extraction is carried out on data based on a model of an LSTM and an Attention mechanism, the LSTM is used for capturing a long and short term dependency relationship in a time sequence, the Attention mechanism focuses on key environment features through a weighting mode, the feature extraction precision and the model interpretation are improved, finally, multi-source data are fused into a final power prediction result by using a full-connection layer, and the power prediction accuracy is improved. And further smoothing processing is carried out through median filtering, so that the stability and the anti-interference performance of a prediction result are improved. According to the method, the real-time performance and accuracy of photovoltaic power prediction can be improved under the complex environment condition, and the adaptability and robustness of the system are remarkably enhanced.

The invention discloses an uncertainty quantification method based on an embedded physical information neural network. The problems that data uncertainty cannot be eliminated and a model uncertainty method is difficult to select and popularize are solved. According to the scheme, the method comprises the following steps: randomly selecting an input-output sample containing data noise, establishing and training a data-driven neural network model, and establishing and training a neural network model embedded with physical information in combination with a physical equation satisfied by an input-output variable; based on the neural network model driven by the data and the neural network model embedded with the physical information, predicting the same sample, and comparing prediction results of the two neural network models; based on a neural network model embedded with physical information, a Bayesian method, a deep integration method and an MC Dropout method are used to improve the neural network model; and on the basis of the improved neural network model embedded with the physical information, predicting the same data, comparing the variance and the deviation of prediction results, and providing guidance for the selection of a model uncertainty quantification method.