118 results about "Sample entropy" patented technology

The invention relates to the technical field of crane line fault diagnosis, in particular to a crane line fault diagnosis system and method based on multi-source data fusion, which comprises a data acquisition and processing unit, a mechanical and electrical coupling characteristic unit and a characteristic fusion and fault quantification unit, the three-axis vibration acceleration and the three-phase current waveform of the track are obtained through the data collecting and processing unit, the mechanical and electrical coupling characteristic unit conducts three-dimensional vector synthesis and wavelet packet decomposition on vibration data, and a time-space incidence matrix of harmonic distortion and vibration is constructed. And the feature fusion and fault quantification unit outputs coupling factors by using a bidirectional long-short-term memory network and an attention mechanism, and outputs a fault probability value through a dynamic time warping matching algorithm after time-frequency domain analysis and sample entropy judgment, so that time-space correlation modeling and dynamic fault matching of multi-source data are realized. And the fault positioning precision and the diagnosis accuracy are improved.

The invention discloses an intelligent power prediction method considering dynamic load change, and relates to the technical field of power grid load prediction, and the method comprises the steps: collecting original load data, carrying out the preprocessing, constructing a VMD constraint optimization model, carrying out the four-stage improvement of optimization parameters through employing an improved dung beetle optimization algorithm, so as to generate IMF components, reconstructing the IMF component by calculating a sample entropy to obtain a low-frequency component and a high-frequency component; establishing a Kalman filtering state space model based on the low-frequency component, and decomposing the low-frequency component into a residual component and a pseudo trend component through a Kalman filtering recursive algorithm; external features are obtained, the high-frequency component, the residual component and the pseudo trend component are aligned and spliced with the external features, multi-component collaborative prediction is carried out through a local-global interactive attention mechanism, and a final load prediction result is obtained; and generating a power demand visualization chart based on the final load prediction result. And reliable decision support is provided for power dispatching and energy management.

The invention relates to the technical field of machine learning, and discloses a load prediction method based on dual feature processing and error correction. The method comprises the following steps: acquiring multi-source time sequence data; decomposing the historical load data into a plurality of modal components by adopting a variational modal decomposition algorithm; classifying each modal component into different frequency levels according to the size of the sample entropy; performing phase-space reconstruction according to the modal component of each frequency level and the corresponding external influence factor data, and generating a multivariable phase-space data set of each frequency level; respectively inputting the multivariable phase space data set of each frequency level into the corresponding load prediction sub-model, generating prediction output results, and superposing the prediction output results; constructing a residual sequence based on the historical load data and the initial load prediction result; inputting the residual error sequence into a residual error prediction model to obtain a load residual error prediction value; and compensating the initial load prediction result through the load residual prediction value. According to the scheme, the load prediction accuracy can be improved.

The invention discloses a spatio-temporal crime prediction method and system fusing spatio-temporal heterogeneous information. The method comprises the following steps: constructing a spatio-temporal data set; constructing a crime time sequence signal data set; periodically decomposing the time sequence signal into a plurality of intrinsic mode functions (IMF); the sample entropy is used as a fitness function to evaluate the advantages and disadvantages of decomposition results under different parameter combinations so as to determine the optimal input time window length; performing clustering processing on the crime data to identify urban crime hotspot areas, and generating crime spatial distribution data; performing feature analysis in time and space, and further completing contribution evaluation of key features to the crime situation; constructing a space-time prediction model, and segmenting a three-dimensional space-time data set into samples in a sliding window mode according to a time window determined in a self-adaptive mode to serve as input and output of the model; and completing training and tuning of the model. And a more accurate and efficient solution is provided for urban crime situation analysis.

The invention relates to the field of state monitoring and fault diagnosis of power equipment, in particular to a transformer fault diagnosis method based on a chaos evolutionary optimization algorithm, which comprises the following steps of: 1, acquiring a magnetic flux leakage signal during operation of a transformer; 2, optimizing a parameter modal number K and a penalty factor alpha of variational modal decomposition by using a chaos evolutionary optimization algorithm; 3, performing variational mode decomposition on the magnetic flux leakage signal to obtain an intrinsic mode function component; 4, calculating the envelope entropy of the intrinsic mode function component, and obtaining an effective intrinsic mode function component through screening; 5, extracting the energy entropy and the sample entropy of the effective intrinsic mode function component to form a feature vector; and 6, inputting the feature vector into a pre-trained support vector machine classifier, and outputting a fault type diagnosis result of the transformer. According to the method, the CEO algorithm is combined with the ergodicity of chaotic mapping and the global search capability of the evolutionary algorithm, and the problems that VMD parameters K and alpha are sensitive and depend on experience, and a traditional optimization algorithm is prone to local optimum are effectively solved.

The invention discloses a landslide displacement double-layer fusion prediction method and model, and the method comprises the steps: carrying out the decomposition of an original displacement time sequence through employing an ICEEMDAN algorithm, and obtaining a plurality of IMF components; performing feature engineering on each IMF component, representing a displacement trend by adopting a trend slope and a window mean value, representing mutation early warning by adopting kurtosis and a frequency spectrum entropy, representing a period rule by adopting a main frequency and a zero-crossing rate, representing system stability by adopting a sample entropy and a standard deviation, and constructing a three-dimensional feature space fusing a time domain and a frequency domain; data standardization is carried out on the extracted features, the interference effect of dimensions on the model is eliminated, and it is ensured that all feature dimensions are within a unified calculation scale range; a CNN-BiLSTM model is constructed for each IMF component; a CPO algorithm is used to optimize the CNN-BiLSTM model; according to the method, the data acquisition difficulty during model training and use can be reduced, and the usability of the model in actual deployment is enhanced; and the prediction precision and the accuracy of landslide displacement prediction are improved.

The invention discloses a wind power plant power prediction method and system in extreme weather and a medium, and belongs to the technical field of renewable energy power generation prediction. Comprising the following steps: acquiring historical weather characteristic data and corresponding wind power plant power data, and screening out weather characteristics with high correlation with the wind power plant power data through a maximum mutual information coefficient; processing the power data of the wind power plant by adopting a CEEMDAN method and a VDM method in combination with a sample entropy analysis and wavelet packet threshold denoising method to obtain a noise-suppressed and feature-enhanced wind power plant power sequence; based on weather features and a wind power plant power sequence with noise suppression and feature enhancement, a wind power plant power prediction model is constructed according to the BiLSTM network after hyper-parameter combination optimization to predict the power of the wind power plant in extreme weather; performing interval shrinkage reconstruction on the wind power plant power in the future extreme weather through time sequence similarity to obtain a final prediction result of the future wind power plant power; according to the invention, the wind power plant power prediction precision in extreme weather is improved.

The invention relates to the technical field of load prediction, and provides a power load prediction method and system based on time sequence decomposition and an attention mechanism, and the method comprises the steps: carrying out the adaptive time sequence decomposition of an obtained original load sequence, calculating the sample entropy of each decomposed component, and carrying out the clustering; constructing a group of encoder and decoder networks for each piece of clustered data, performing parallel encoding to extract features, performing serial decoding reconstruction on the features from low frequency to high frequency, and outputting prediction data from low frequency to high frequency step by step; the weight is initialized based on the sample entropy, the trained weight is obtained through optimization in the encoder and decoder network training process, and the predicted value of the power load is obtained through weighted fusion. According to the method, adaptive time sequence decomposition, a weight mechanism guided by sample entropy and an attention-enhanced encoder-decoder structure are introduced, so that multi-component collaborative modeling and cross-scale dynamic prediction are realized, and the prediction accuracy and stability in complex load data and small sample scenes are effectively improved.

The invention discloses a distribution hidden danger assessment and prediction method based on a natural time domain and a fuzzy rough set, and belongs to the technical field of operation and maintenance of power grid equipment. According to the method, based on collected multi-source monitoring signals of leakage current, induction current, temperature and the like, adaptive signal preprocessing is carried out by adopting a method of combining empirical mode decomposition (EMD) and sample entropy, effective mode components are effectively extracted, and noise interference is suppressed; natural time domain analysis is introduced on the basis of a traditional time domain, an event sequence is constructed, dynamic features are extracted, and a hidden danger feature data set with time sequence evolution information is formed; further performing unsupervised attribute reduction on the high-dimensional features by using a fuzzy rough set theory, removing redundant information, and retaining key discrimination features; and finally, identification and trend prediction of wiring hidden danger types are realized through a support vector machine (SVM) classifier. According to the method, the accuracy and robustness of hidden danger identification are improved, and effective technical support is provided for intelligent operation and maintenance of power distribution.

The invention provides a sea wave height prediction method and system based on hybrid quadratic decomposition, and relates to the technical field of sea wave height prediction, and the method comprises the steps: obtaining to-be-predicted original wave height data; performing primary decomposition on the original wave height data to obtain an intrinsic mode function and a residual term; calculating a sample entropy of the intrinsic mode function; according to the size of the sample entropy, classifying the intrinsic mode function to obtain a reconstructed mode component; performing secondary decomposition on the high-frequency components through a VMD algorithm to generate sub-sequence components; establishing a periodic sensing time deep learning model; and inputting the subsequence component, the intermediate-frequency component and the low-frequency component into a periodic sensing time deep learning model, and outputting a sea wave height prediction result.

The invention discloses an air duct vibration on-line monitoring system based on multiple sensors and a fault prediction method, and belongs to the technical field of air duct state monitoring and fault prediction.The method specifically comprises the steps that air duct vibration acceleration and noise signals are collected in real time through the sensors, and a time-aligned signal sequence is formed through analog-to-digital conversion and preprocessing; extracting a time-frequency domain feature from the vibration signal sequence, extracting a sound pressure level and a harmonic distortion degree from the noise signal sequence, and generating an air duct feature vector through feature fusion; constructing a Bayesian probability model based on the vector, calculating a fault occurrence probability, judging existence and generating a confidence score; calculating the sliding sample entropy of the vibration signal sequence, and marking an entropy value abnormal interval; and a union set of a fault interval diagnosed by the Bayesian probability model and an entropy abnormal interval is obtained, after time sequence trend analysis, early warning information is generated and pushed to an operation and maintenance platform when a joint judgment condition is met, and online updating of a normal vibration entropy model is triggered, so that the fault detection and early warning precision is improved.

The invention discloses a self-adaptive optimization method and system for online decoding of a motor imagery brain-computer interface, and the method comprises the steps: carrying out the real-time self-adaption of an electroencephalogram data stream of a target user through a teacher-student model framework on the premise that the privacy protection of source domain training data does not need to be accessed; performing batch weight normalization during testing, decoupling normalization statistic updating and parameter optimization by stopping gradient operation, and stabilizing feature representation; a dynamic category specific entropy threshold mechanism is combined with online category frequency and batch confidence to adaptively screen a high-confidence sample for each category; a dynamic online reweighting strategy is designed, and weights are distributed according to the sample entropy and the category frequency to balance the optimization process; and decoupling contrast learning based on a fixed prototype is introduced, and feature space distribution is optimized. According to the method, the problems of statistic drift, poor fixed threshold adaptability, category imbalance sensitivity, insufficient feature optimization and the like are solved, and the adaptability, the stability and the robustness of cross-user motor imagery brain-computer interface online decoding are improved.

The invention relates to a landslide deep deformation monitoring data noise reduction method based on an HBP-VMD combined improvement wavelet threshold. The method comprises the steps of collecting a landslide deep deformation original signal; the sample entropy is used as a fitness function, a badger optimization algorithm is adopted to optimize VMD decomposition parameters, and an optimal combination parameter combination is obtained; substituting the optimal combination parameter into the VMD, and performing VMD decomposition on the original signal to obtain K intrinsic mode components IMF of different frequencies; calculating a variance contribution rate and a correlation coefficient corresponding to each obtained IMF component, and dividing the IMF components into an effective component, a noisy component and a noise component; retaining the obtained effective component, abandoning the noise component, and carrying out noise reduction processing on the noisy component by using an improved wavelet soft threshold; and reconstructing the IMF component after noise reduction and the effective IMF component, and finally realizing signal noise reduction. According to the method, the deformation monitoring signal of the deep part of the landslide can be efficiently stripped from the noisy signal, and the waveform is clearer than that before noise reduction; the SNR of the signal after noise reduction is the highest, the SMES is the lowest, and the excellent noise reduction effect is achieved.

The embodiment of the invention provides a multi-mode collaborative gas pipe network hidden danger identification method and system, and relates to the technical field of gas pipe networks. The identification method comprises the following steps: acquiring data of a pressure sensor, a temperature sensor and a gas concentration sensor at key nodes of the gas pipe network; performing wavelet packet decomposition on the data to obtain sub-band signals of the signals; performing feature extraction on the sub-band signals to obtain an energy entropy, a sample entropy and a permutation entropy; normalizing the energy entropy, the sample entropy and the permutation entropy to obtain a corresponding normalized entropy value; inputting the corresponding normalized entropy value into the intelligent prediction model, and outputting a probability distribution diagram of the pipe network hidden danger position; and leakage point positioning and corrosion degree grading identification are carried out according to the probability distribution diagram. According to the method, data of pressure, temperature and gas concentration sensors in the gas pipe network are effectively integrated, through wavelet packet decomposition and deep feature learning, pipe network leakage, corrosion and other hidden dangers are accurately recognized, and stable operation of a gas supply system is ensured.

The application provides a multi-parameter fusion real-time monitoring and early warning method and system for anesthesia depth, relates to the technical field of medical monitoring, and comprises the following steps: acquiring physiological parameter data, performing multi-scale decomposition and calculating sample entropy values, extracting baseline features and fluctuation features to form an anesthesia feature vector; an initial anesthesia depth index is output by using a deep learning model; a cause-effect correlation matrix is constructed by calculating time-varying transfer entropy based on intrinsic mode function decomposition, dominant cause-effect patterns are extracted, and the anesthesia depth index is corrected; and real-time monitoring and early warning reports are generated. The application can improve anesthesia depth evaluation accuracy and reduce anesthesia risks.

The present application provides a method for bearing fault diagnosis across working conditions by using a minimum entropy optimized prototype contrast network, and relates to the technical field of intelligent fault diagnosis. In the pre-training stage, an auxiliary domain discriminator is constructed to assist DA with the discriminant information of the classifier, the classification difficulty is evaluated by sample entropy, and the performance degradation in the DA process is inhibited. In the training stage, the learning vector quantization method is adopted to find the prototype. Through intra-domain prototype contrast learning, the sample features are closely gathered around the same prototype in the feature space, while being separated from the different prototypes. Then, the intra-class consistency of the features is enhanced, and the inter-class distinguishability is improved, so as to realize the precise alignment in the feature space. In addition, the cross-domain instance-prototype learning aligns the semantic structure in the shared embedding space, alleviates the negative transfer problem through the fine-grained alignment strategy, and improves the generalization ability of the model. Through the pseudo-label generation and the weighted loss function, the generalization performance of the model in the cross-working-condition small sample scene is improved.

The invention provides a feature fault diagnosis method for time domain statistics of a main shaft bearing of an aero-engine, which comprises the following steps of: S1, determining a rotating speed working condition of the aero-engine, determining main parameters and feature frequency of the main shaft bearing, and obtaining a bearing vibration signal through a vibration sensor; s2, optimizing the interception length of the data by adopting sample entropy; s3, calculating the short-time kurtosis time of each rotating speed working condition according to the data interception length, and obtaining a history curve of the short-time kurtosis time; and S4, analyzing the rotating speed working condition and the bearing dangerous working condition of the bearing early-stage fault. According to the method, the fault of the main shaft bearing of the aero-engine is diagnosed through the time domain characteristic statistical index of the short-time kurtosis, the working state of the main shaft bearing under each rotating speed working condition is obtained, and the early fault symptom of the bearing can be found; and the interception length of the data is optimized through the sample entropy, so that the method is suitable for complex and changeable rotating speed working conditions of the aero-engine.

This invention belongs to the field of wireless communication technology, and particularly relates to a method for predictability assessment of electromagnetic communication data. It includes the following steps: Step 1: Collect electromagnetic communication data from multiple receivers and transmitters over a period of time, and calculate the intrinsic predictability of the parameter time series at each point based on true entropy, sample entropy, and multi-scale entropy; Step 2: Select the time scale with the minimum entropy value, preprocess the electromagnetic communication data time series, and construct a spatiotemporal similarity graph to transform the electromagnetic communication data into graph data; Step 3: Locally construct a spatiotemporal graph convolutional network model with node-level attention, perform local prediction, and calculate the prediction accuracy; Step 4: Perform predictability assessment on the electromagnetic communication data. This invention proposes for the first time a comprehensive assessment framework that can both calculate intrinsic predictability and guide the establishment of prediction models, enabling a comprehensive evaluation of the predictability of electromagnetic communication data.

The application discloses a rolling bearing degradation trend prediction method based on linear regression and TCN, characterized in that: firstly, the full life cycle vibration data of the rolling bearing is collected through an acceleration sensor, the sample entropy of the data is calculated and obtained as the performance degradation index of the bearing, the sample entropy is preprocessed smoothly, then the SPT point of the bearing is determined according to the linear regression model and mu+3delta, the influence of different window lengths on the determination of the SPT is analyzed, finally the preprocessed sample entropy value is input into the trained TCN network for degradation trend prediction. The application can effectively and timely determine the SPT point of the bearing, the TCN can well fit the performance degradation trend of the bearing, and the problem that the equipment cannot be safely operated due to the too late discovery of the bearing fault is avoided; compared with the method of adopting the recurrent neural network prediction, the accuracy of the method for the equipment health management is obviously improved, and a new idea is provided for the degradation trend prediction method of the bearing.

The invention relates to the technical field of intelligent driving, in particular to a risk early warning method for a heart rate scatter diagram in the driving process. The real-time heart rate scatter diagram is calculated and analyzed in real time through the improved box counting method to obtain the fractal dimension, the dimension and entropy combined feature space is constructed through the improved sample entropy algorithm and the fractal dimension, the primary early warning information is generated through primary early warning judgment, the problems that a traditional method cannot recognize instantaneous risks and early warning is delayed are solved, and the early warning accuracy is improved. The real-time performance and the accuracy of risk identification are improved; the method comprises the following steps: acquiring a heart rate signal, analyzing the heart rate signal through wavelet packet decomposition to obtain a time-frequency feature vector, acquiring a steering wheel grip parameter, carrying out fusion analysis on the steering wheel grip parameter and the time-frequency feature vector to obtain early warning verification information, judging and outputting a risk early warning level through a risk level, and carrying out adaptability early warning through an individual heart rate baseline. The false report and missing report rate is reduced, and an accurate decision basis is provided for an intelligent driving system.

The application discloses a load prediction method and system based on big data, and relates to the technical field of power grid load prediction.The method comprises the following steps: based on a least squares support vector machine, pre-processing load historical data; through decomposing the load historical data with reduced complexity, obtaining a zero-crossing rate and sample entropy, and determining the multi-frequency components of the load data; through a preset hybrid algorithm and in combination with the multi-frequency components of the load data, training a multi-factor weighted combination analysis model; based on the multi-factor weighted combination analysis model and according to an improved grey wolf algorithm, determining the weight of the load prediction result of each prediction factor module; and weighting and combining the load prediction results of each prediction factor module to obtain a final load prediction result.The application improves the robustness of the model, dynamically updates the model output result, optimizes the weight proportion among the factor modules, and improves the adaptability of load prediction in a complex environment.

The invention discloses a carbon emission prediction method based on deep learning. The method comprises the following steps: acquiring input data required by enterprise carbon emission research and judgment through massive multi-source fusion data; then multi-dimensional enterprise carbon emission feature extraction is carried out by combining cross aggregation feature derivation, time sequence sample entropy and a convolutional neural network automatic encoder, finally research and judgment are realized through fusion prediction of a plurality of machine learning models, if enterprise samples are extremely few, a pseudo-label semi-supervised learning algorithm is introduced to process large-scale unlabeled data, and finally, the enterprise carbon emission feature extraction is carried out. The enterprise carbon emission heat map is constructed according to the enterprise carbon emission historical data, the enterprise carbon emission heat map is used as the input of the convolutional neural network, the carbon emission of the enterprise in the next 24 hours is used as the output, and the enterprise carbon emission prediction with complex and changeable energy consumption characteristics is realized.

The invention relates to the technical field of milling cutter state monitoring, and provides a self-adaptive denoising method suitable for a milling cutter state monitoring signal, and the method comprises the steps: decomposing an original complex signal into a plurality of IMF components in a milling process through employing a CEEMDAN algorithm, calculating the sample entropy of each IMF component, and carrying out the classification and recombination, obtaining a reconstructed signal and a high-complexity component; a sample entropy is used as a fitness function for measuring signal complexity, and an RIME algorithm is guided to automatically find a parameter combination capable of enabling the VMD decomposition effect to be optimal; performing VMD secondary decomposition on the high-complexity component by using the optimized parameter combination, calculating PCC of the decomposed IMF component and the original signal, and selecting an effective IMF component according to a correlation coefficient weight; and carrying out superposition reconstruction on the screened effective IMF components to generate a de-noised signal with a high signal-to-noise ratio, and taking the de-noised signal as a feature input signal for milling cutter state monitoring. The optimal denoising state is automatically adjusted according to the tool working condition, and the working condition adaptability of the system is improved.

The invention belongs to the technical field of aviation cable fault positioning, and particularly relates to a cable fault positioning method based on time-frequency domain analysis, which comprises the following steps: firstly, collecting a cable fault signal, adopting a GWO-VMD algorithm, taking a minimum sample entropy as a fitness function, and optimizing to find an optimal penalty factor and decomposition layer number, so as to obtain a fault positioning result; removing a high noise component obtained by VMD decomposition according to the current sample entropy, and obtaining a reflected signal after noise reduction; performing cross term suppression on the reflected signal after noise reduction by using smooth pseudo Wigner distribution to obtain discrete SPWVD; and performing blind area elimination on the discrete SPWVD based on a blind area elimination algorithm of hypothesis verification to obtain a fault distance. A grey wolf optimization algorithm is applied to a variational mode decomposition algorithm, and parameters of decomposition layers and penalty factors of the VMD are optimized, so that the problem of interference of noise on reflection signal extraction is solved; secondly, the smooth pseudo Wigner distribution is adopted to solve the problem of cross term interference among a plurality of linear components in the traditional WVD distribution, and the positioning precision is improved.

The application discloses a kind of millimeter wave sign detection methods based on swarm intelligence and improved wavelet threshold value, wherein the method includes: obtaining the original vital sign signal collected by millimeter wave radar;Construct the multi-objective fitness function of fusion sample entropy, pearson correlation coefficient and kurtosis;The multi-objective fitness function is optimized using the improved Harris eagle optimization algorithm, and the optimal parameter combination of CEEMDAN decomposition algorithm is adaptively solved;Based on the optimal parameter combination, the original vital sign signal is decomposed by CEEMDAN, and a plurality of intrinsic mode function components are obtained.The application constructs the cascade processing framework of "parameter optimization-signal separation-noise suppression", solves the problem that CEEMDAN parameter depends on artificial experience, respiratory and heartbeat signal band aliasing and the problem of insufficient performance of traditional wavelet threshold denoising.

The invention discloses an electroencephalogram signal reconstruction method and system based on artifact removal. The method comprises the steps that multi-channel electroencephalogram signals are obtained and subjected to filtering preprocessing; blind source separation is carried out through independent component analysis, and statistical independent signal source components and a hybrid matrix are obtained through decomposition; constructing an equivalent current dipole model to calculate a space artifact index, evaluating signal complexity in combination with a time sequence sample entropy, and judging and identifying an artifact component and a neural activity component through weighted fusion and a self-adaptive threshold value; neural activity components are reserved, artifact components are zeroed, and pure electroencephalogram signals are reconstructed through mixed matrix inverse transformation. According to the method, accurate identification and effective removal of the electroencephalogram artifacts are realized, and high-quality data are provided for electroencephalogram analysis.

The application discloses a street lamp line ground fault intelligent diagnosis system and method, belongs to the technical field of power equipment operation monitoring, and comprises a data acquisition and preprocessing unit, a feature extraction unit and a fault diagnosis unit. The data acquisition and preprocessing unit collects original residual current data of a street lamp line through a residual current transformer and an oscilloscope. The application adopts a parrot optimization algorithm to adaptively determine optimal decomposition parameters of VMD, overcomes the limitations of artificial parameter setting, ensures the quality of data decomposition, lays a foundation for extracting high-quality fault features, solves the problems of lack of reliability of existing fault protection, great influence of human factors, and low accuracy of existing street lamp ground fault diagnosis systems and methods, calculates sample entropy values, constructs a multi-dimensional feature vector set representing line states, inputs the multi-dimensional feature vector set into a fault diagnosis unit, and the fault diagnosis unit trains a kernel extreme learning machine classifier by using the feature vector set.

This invention claims protection for a short-term power load forecasting method combining CDC-VMD and echo state networks. First, the Local Directional Centrality (CDC) clustering algorithm is used to accurately cluster the raw load data. The Davidson-Bolding Index (DBI) is then used to compare with other clustering algorithms, revealing that the CDC algorithm has the best clustering effect. Next, Variational Mode Decomposition (VMD) is used to process the load data, decomposing the load sequence into multiple Intrinsic Mode Functions (IMFs) to reduce the complexity of the input data for the neural network prediction model. The sample entropy (SE) of each IMF component is calculated, and components with similar sample entropy values ​​are merged into a new sequence to reduce the number of models required for training. Finally, a Deep Echo State Network (DESN) is used to fit the relationship between historical data and predicted data for each sequence, and the prediction results of each model are superimposed to obtain the final predicted value. This invention offers higher accuracy compared to traditional prediction methods.

The invention discloses a window optimization-based intra-day prospective dispatching method and device for a power grid, and relates to the technical field of power dispatching, and the method comprises the steps: obtaining a power grid data feature prediction model through employing historical power grid data features and an adaptive differential evolution algorithm; obtaining predicted values of intra-day power grid data features based on a power grid data feature prediction model, thereby generating a plurality of operation scenes; calculating corresponding net load power based on each operation scene, determining permutation entropy, sample entropy and fuzzy entropy based on all the net load power, and establishing a combined entropy calculation formula by using a multi-target equalization method; constructing a look-ahead window optimization model based on a combined entropy calculation formula, and solving the look-ahead window optimization model by adopting a carbon black stingless bee optimization algorithm to obtain a window optimization result; based on a window optimization result, establishing an intra-day look-ahead optimization scheduling model; and solving the intra-day look-ahead optimization scheduling model by adopting a state optimization algorithm to obtain a power grid intra-day look-ahead scheduling plan. The flexibility of the power grid dispatching plan is improved.

The invention discloses a sparse sample entropy criterion-based physical characteristic parameter estimation method, which comprises the following steps of: determining a pneumatic system fault factor and a physical constraint thereof, and converting the physical constraint into an integral form; solving an integral form based on a Lagrange direct construction method to obtain a probability density function, and ensuring that the probability density function has physical significance; screening target distribution conforming to physical constraints from a preset general probability density function library, and extracting functions conforming to basic distribution shape requirements in the target distribution to form a function group; dividing the function group into a single-parameter group and a double-parameter group based on the parameter scale, respectively calculating to obtain corresponding entropy information, sorting differential entropy results, and taking a maximum value to obtain most unbiased probability density distribution conforming to the physical constraint. According to the method, high-confidence parameter estimation under the condition that the sample size is scarce is realized, the limitation of data shortage is broken through, and the probability density function can be stably generated only by utilizing physical characteristics such as extreme values and peak positions.