402 results about "Multi variable" patented technology

The invention belongs to the technical field of power system load prediction, and relates to a long-range multivariable load prediction method and system based on time-frequency domain collaboration, and the system carries out the normalization and stabilization of a multivariate load time sequence through a data preprocessing module; the feature embedding module performs linear embedding on the block sequence to construct high-dimensional feature representation; the state space coding module extracts long-range dependency features and generates depth time sequence representation; the decoding prediction module maps the coding features into a preliminary prediction sequence; the time sequence alignment module identifies a leading-lagging relation among multiple variables and aligns a time sequence; the frequency domain optimization module realizes frequency domain component fusion based on adaptive filtering; and the model training optimization module is used for performing training and optimization through a signal attenuation loss function. The method can effectively improve the precision and robustness of long-range multivariable load prediction, and especially has obvious advantages in the aspects of processing complex dependency relationships and dynamic time delay.

The invention provides an automatic data management method and system based on a multi-modal large model, and the method comprises the steps: collecting multi-source heterogeneous industrial data, and carrying out the standardization processing, and forming standardized multivariable time series data; constructing a process knowledge base, and performing semantic embedding coding on a process knowledge text and storing the process knowledge text; constructing and finely tuning a KTSF multi-modal large model, and fusing process knowledge semantics and multivariable time sequence data through a cross-modal attention mechanism to generate joint semantic representation; based on prediction of a KTSF multi-mode large model, outputting a residual error with actual data, and dynamically identifying abnormal data; performing attribution analysis; based on an attribution result, calling a KTSF multi-mode large model to generate a repair value, and performing intelligent correction on the abnormal data; the design quality evaluation and feedback learning module is used for calculating a data quality score and driving incremental updating of the model; and the design rule self-learning module is used for automatically extracting the governance rule through clustering analysis and updating the knowledge base.

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 key parameter long-time-sequence prediction method for a complex process industry, and the method comprises the steps: collecting multivariable sensor data in the process industry, and constructing a high-dimensional long-time-sequence prediction data set; constructing a PatchConvRNN prediction model by combining time slice embedding, dimension decoupling convolution, depth separable convolution and a recurrent neural network based on a sequence-to-sequence normal form; a point value-statistical mixed loss function is adopted, the point prediction precision, the sequence mean value and the standard deviation consistency are optimized at the same time, a prediction model is trained in combination with an optimization algorithm, and network model parameters are adjusted; and comprehensively evaluating the prediction model through a root-mean-square error, an average absolute percentage error and a standard deviation average absolute error. According to the method, high-precision prediction and fluctuation maintenance of the key time sequence variables under the complex working condition of the industrial process are achieved, and powerful support is provided for quality control and predictive maintenance of the production process.

The invention belongs to the technical field of multivariable measurement and data analysis, and relates to a multi-source data joint measurement and analysis method based on a physical consistency relationship, which comprises the following steps of: 1, acquiring and preprocessing multi-source data; 2, modeling physical characteristics; 3, constructing a physical consistency relationship; 4, multi-source joint feature extraction; 5, carrying out consistency evaluation and abnormity identification; and 6, result fusion and dynamic updating. According to the method, collaborative analysis and consistency feature extraction of different types of measurement signals are realized by establishing a physical constraint model among multi-source sensor data, so that the reliability and precision of a multivariable measurement system are improved, the robustness of the system to measurement errors, noise and abnormity under complex working conditions is effectively enhanced, and the measurement accuracy of the system is improved. The method is suitable for the fields of intelligent driving, industrial monitoring, multi-sensor fusion measurement and the like.

The invention discloses a hybrid sequential network power consumption prediction method for reinforcement learning dynamic calibration, and the method comprises the steps: collecting multivariable power consumption time sequence data, completing the data preprocessing through resampling, feature engineering, normalization and sliding window technologies, generating a supervised learning sample set, and dividing the supervised learning sample set into a training set, a verification set and a test set; constructing a hybrid prediction model comprising a dynamic capture module, a long-term dependence modeling module, a regression prediction module and a reinforcement learning dynamic fine tuning module; training and optimizing by adopting a staged training strategy to obtain a hybrid prediction model; multivariable power consumption time sequence data are collected in real time and preprocessed, the preprocessed data serve as input, real-time prediction of future total consumption is conducted through the mixed prediction model, and a final prediction result after dynamic fine adjustment is output. According to the method, accurate and efficient prediction of the power grid load can be realized, and a reliable technical solution can be provided for power system scheduling optimization, demand side management, market transaction and other scenes.

The invention discloses a shale gas well fracturing parameter optimization design method and system based on a depth Q network, and relates to the technical field of shale gas development, and the method comprises the following steps: S1, collecting geological parameters, fracturing construction parameters and productivity data of a shale gas well in advance; s2, constructing an agent model and performing environment simulation; and S3, constructing a DQN model. According to the method, a LightGBM algorithm is adopted to construct a data-driven proxy model to simulate a real fracturing environment, an optimization model is constructed based on a deep Q network (DQN), state, action, reward and epsilon-greedy strategies are defined, and technologies such as a variable step size search mechanism, experience playback and target network soft update are combined, so that the real fracturing environment is simulated. The problems that a traditional optimization method is weak in local search capability and low in convergence speed are effectively solved, multivariable synchronous optimization of the unit perforation length proppant dosage and the fracturing fluid dosage is achieved, and a global optimal parameter combination can be rapidly explored.

The invention discloses an intelligent tea withering control system based on multi-modal feature fusion and time sequence prediction. The system is composed of a multi-modal feature extraction module, a time sequence modeling module, a transfer learning module and an intelligent decision control module, a Transform attention mechanism is adopted to construct a cross-modal fusion framework, RGB images, hyperspectrum and time sequence information are fused, and accurate recognition and trend prediction of the withering state are achieved. According to the system, an adaptive attention fusion network is designed, optimal fusion of images, spectrums and grade information is realized through dynamic distribution of modal weights, and the recognition accuracy and stability are remarkably improved. The classification accuracy of 180 verification samples reaches 93.33%, and is improved by 15.93%-34.83% compared with that of a single-mode method. And cross-environment self-adaption is realized through fusion transfer learning, and the performance is improved by 6.7%-14.3%. The intelligent decision engine optimizes temperature and humidity parameters based on a multivariable coupling control theory, the control precision reaches + / -0.5 DEG C and + / -2.0%, and the response time is 2.5 seconds.

The invention provides a multivariable time series prediction method and system based on structural representation learning, and belongs to the technical field of multivariable time series prediction.The method comprises the steps that a similarity matrix between variables is constructed according to multivariable time series data, and variable community tags are obtained through community detection; carrying out weighted binarization to obtain a binary community sensing adjacency matrix; inputting the initial variable feature matrix and the adjacent matrix into a graph auto-encoder to obtain a variable structure embedded matrix; variable value embedding, time step position embedding and structure embedding are fused to obtain a unified input matrix, a time sequence prediction model based on a structure information guiding attention mechanism is input for modeling, and a prediction result is obtained; wherein structure embedding serves as an auxiliary feature to participate in modeling, and explicit structure constraint is not applied. According to the method, variable association is learned and mined through structural representation, and the accuracy and generalization ability of multivariable time sequence prediction are improved.

The invention discloses a load prediction method considering data enhancement in extreme weather, which relates to the technical field of power system load prediction and comprises the following steps: acquiring meteorological data and historical industry load data of a target area according to a fixed sampling interval; calculating the correlation between a temperature sequence and a load sequence in the multivariable feature matrix, and screening a temperature-sensitive target industry load set by combining the correlation strength and stability in an extreme scene; calculating daily extreme temperature of the temperature sequence by day, and judging and identifying different extreme weather days and corresponding similar day sets according to a temperature threshold value; performing sample expansion and elimination according to the similar day set and the representative test day, and establishing a basic training set and an enhanced training set; and constructing a reference model through double-stage training. Compared with a traditional method, all load data are simply used, redundant features are remarkably reduced, and model training efficiency and prediction accuracy are improved.

The invention relates to the technical field of boiler control, in particular to a multivariable coupling thermal process regulation and control system and method for carbon pollution governing, and the method comprises the steps: collecting multi-source data such as acoustic emission, temperature, humidity and spectrum, and constructing a feature sequence through time mark alignment and wavelet packet enhancement; a heat value characterization quantity is predicted by using a Shenchang differential equation model fused with dynamic gating, a partition equivalent thermal network model is driven on this basis, and accurate prediction of a future time domain temperature field is realized by dynamically correcting thermal resistance and thermal capacity; based on the prediction result, a control instruction is solved through multi-objective optimization under the condition that the active temperature constraint is met; and in combination with heat flow density feedback, a layered reinforcement learning controller is adopted for online compensation of a pre-feedback instruction, and stable and efficient regulation and control of the boiler are achieved.

The invention relates to the technical field of new energy power generation and control, and discloses a new energy station operation decision support system based on multi-objective optimization, which comprises a multi-source state space reconstruction module, a Riemannian manifold geometry engine module, a self-adaptive inertia Hamiltonian evolution module and a symplectic geometric integral and instruction mapping module. The system collects station data to construct a dimensionless state space, constructs a Riemannian metric tensor according to physical constraints to reconstruct a Riemannian manifold space, and calculates a geometric connection strength factor; the factor is used to adaptively modulate a virtual inertia matrix, and a dissipative Hamiltonian kinetic model is constructed; and finally, solving the steady-state generalized coordinates through a pungent-preserving numerical integration algorithm, and decoding the steady-state generalized coordinates into an equipment control instruction. According to the method, physical constraints are converted into geometric measurements, and a self-adaptive inertia mechanism is introduced, so that the problem of optimization convergence under multivariable strong constraints is solved, and the safety and accuracy of a control instruction are ensured.

The invention provides a drought and flood sudden change prediction method and system based on artificial intelligence, and the method comprises the steps: collecting the in-out reservoir runoff observation data of a reservoir in a target region, obtaining a restored natural reservoir runoff series based on a water balance method, calibrating a long-short-term memory model through a minimum batch gradient descent method, and carrying out the prediction of the drought and flood sudden change. Based on the natural reservoir runoff, the actual reservoir runoff and meteorological data, constructing a long-short-term memory model to simulate the influence of water conservancy project regulation and storage on the runoff; based on an earth system mode set and a multivariable deviation correction method, performing spatial downscaling on output data of earth system modes to obtain corrected meteorological variables; driving a multi-member set of an earth system mode, and separating by adopting a detection attribution technology to obtain a contribution rate of man-made climate compulsion to drought and flood sudden turning event evolution; and predicting a daily runoff process under future climate change, and predicting a future drought and flood sudden turning event by adopting deep learning and superposition of regulation and storage influence of a water conservancy project.

The invention provides a wind power prediction method and system based on multi-channel multi-scale decomposition, and relates to wind power prediction, and the method comprises the steps: obtaining historical time sequence data of a prediction region; extracting the historical time sequence data to obtain a segmented data set; the segmented data set comprises a plurality of segmented data; the segmented data set comprises trend item data, periodic data, stepped data and fluctuation item data; constructing a multi-modal time sequence prediction model; and on the basis of the multi-modal time sequence prediction model, combining historical time sequence data to obtain a wind power prediction result of the prediction area. According to the method, the information of the multi-channel time sequence can be effectively extracted and fused, the correlation among multi-channel multiple variables is modeled, the deep semantic features strongly related to the target task are automatically extracted, and accurate power prediction is realized.

The invention discloses a multivariable time sequence feature extraction and grade prediction method and system for a flotation process and a storage medium. The method comprises the steps that S1, original data are input and coded; carrying out coding and structured input on time sequence data formed by various process variables; s2, extracting dynamic characteristics; s3, modeling based on condition guidance coding; s4, enhancing the significance of the key variable and the important time slice; a multi-head attention mechanism is adopted, and a query vector based on target guidance is matched with a key value pair generated by multi-scale features; s5, outputting a prediction module; and carrying out feature fusion and nonlinear mapping on an attention mechanism output result, and outputting a concentrate grade and recovery rate prediction result at a future moment. The system and the storage medium are both realized based on the method. The method has the advantages of being higher in intelligent degree, better in controllability, capable of improving prediction accuracy and model adaptability of key indexes in the flotation process and the like.

The invention relates to the crossing field of geological disaster monitoring and early warning, artificial intelligence and edge computing technology, and discloses a slope early warning method based on unsupervised time series prediction and dynamic fusion, and the method comprises the steps: obtaining a standardized multivariable time series data matrix; performing hyper-parameter configuration of the stacked self-attention encoder model of the new slope; outputting a prediction sequence of a plurality of time steps in the future; generating a slope comprehensive health index; generating a dynamically adjusted multi-level early warning threshold adjustment instruction and a resource preset suggestion; and performing real-time judgment on the comprehensive health index of the side slope and a multi-level early warning threshold value, determining an early warning level, and executing a linkage response corresponding to the early warning level. According to the method, the industrial core problem that a supervised model cannot be trained due to extremely few unstable samples in the geological disaster early warning field is fundamentally solved, and the technical breakthrough of high-precision advanced prediction under the label-free condition is realized.

The invention relates to an SCR ammonia spraying multivariable decoupling control method based on partition concentration field real-time feedback, and belongs to the technical field of flue gas denitration automatic control. The method comprises the following steps: constructing a total cross-section concentration field vector, and establishing and dynamically updating a concentration field coupling relation matrix; building a multivariable prediction model based on the matrix, predicting the coupling influence of the valve action on the full-field concentration, and generating a compensation signal through self-adaptive decoupling logic; performing credibility grading and reconstruction on the concentration data, cooperatively resolving a multi-partition control instruction set in combination with a prediction model, and configuring a differentiated execution step length according to data quality; and with uniform distribution of full-field concentration as a target, decoupling parameters and calibration weights are iteratively optimized through deviation feedback, and a parameter-working condition mapping library is established, so that self-adaptive rolling optimization of control parameters is realized. According to the invention, the problem of strong coupling of multi-partition ammonia spraying is effectively solved, and rapid, accurate and stable leveling of a concentration field is realized.

According to the time series prediction method based on the multi-modal contrast learning technology, original multivariable time series data are converted into structured visual representation and language representation, multi-modal representation with consistent inner performance can be constructed without depending on external natural language or real image data, and the time series prediction method is high in practicability. The deep semantic understanding capability of the model on the complex operation state of the rail transit is effectively enhanced; a multi-modal contrast learning mechanism is introduced, visual and text modal representation is aligned in a shared embedding space, positive sample consistency is maximized through InfoNCE loss, negative sample interference is suppressed, and the robustness and generalization ability of time sequence features are remarkably improved; and the importance of each variable on a prediction task is dynamically evaluated by using the aligned multi-modal representation, and key variables are automatically screened, so that the redundant information interference is reduced, and the prediction precision and the calculation efficiency of the model in a high-dimensional multi-variable scene are also improved.

The invention discloses an ice-coated blade aeroelastic flutter and complete machine mistuning identification method based on multi-dimensional dynamic feature engineering, and belongs to the technical field of wind power. Comprising the steps of collecting original multivariable time series data, preprocessing the original multivariable time series data, performing feature extraction on the preprocessed original multivariable time series data from three dimensions of time domain, time-frequency domain wavelet energy and control system domain to obtain basic physical features, performing feature screening on the basic physical features based on a Gini index, and obtaining the basic physical features. The method comprises the following steps: obtaining a fault seed feature, constructing an interactive feature and an adaptive feature based on the fault seed feature, fusing the basic physical feature, the interactive feature and the adaptive feature into a high-dimensional feature set, and inputting the high-dimensional feature set into a weighted ensemble learning classifier to realize the identification of the ice-coated blade health, aeroelastic flutter and complete machine detuning state. According to the method, the problems of poor fault identification robustness and low precision of a traditional method are solved, and reliable guarantee is provided for safe and stable operation of the wind turbine generator under the icing working condition.

The invention relates to the technical field of process industry process modeling and data-driven prediction, and discloses a process industry key index prediction method and device based on time-delay distribution learning. The method and the device are used for solving the problems of unreasonable historical information alignment and insufficient key index prediction precision and stability caused by difficulty in measurement of time-delay distribution attributes in process industry multivariable data. According to the method, a current window, a candidate lag historical window and a future target window are constructed by adopting a sliding window, time-lag weight distribution is learned on a candidate lag axis based on a state gating cross attention mechanism, time-lag distribution learning and a future window prediction task are jointly optimized through a time-lag modeling auto-encoder, and finally, time-lag prediction is performed on the candidate lag historical window and the future target window. And outputting a key index prediction value under the guidance of time-delay distribution. The method is suitable for online prediction of key indexes such as quality, yield and energy consumption of the process industry, and can be used for process monitoring and optimization control.

The invention provides a gas concentration time sequence distribution prediction method based on multivariable data driving. Based on the theoretical relationship between the mean square displacement and the mean scattering angle cosine of neutrons in an infinite homogeneous medium, the mean square displacement is directly counted through Monte Carlo simulation, the mean scattering angle cosine is reversely deduced, and a first-order scattering matrix meeting the mean square displacement conservation is constructed accordingly. In order to be suitable for a finite geometric model, a correction factor is further introduced to correct an average scattering angle cosine obtained by a traditional method. A verification result shows that the method remarkably reduces the calculation deviation of effective proliferation factors, improves the neutron flux distribution precision, and is particularly suitable for high-precision multi-group calculation of high-anisotropy fast spectrum reactor cores such as pebble-bed high-temperature gas cooled reactors and sodium cooled fast reactors. According to the method, the time sequence probability distribution prediction of the gas concentration can be realized, the prediction uncertainty is quantified, the risk indexes such as the over-limit probability are directly output, and the scientificity and the reliability of coal mine safety early warning are improved.

The invention discloses an underground water level change prediction method and system based on machine learning, and relates to the technical field of hydrogeology, and the method comprises the steps: carrying out the standardization processing of multi-source time series data, obtaining a standardized multivariable time series data matrix, and constructing a supervised learning sample set; inputting the underground water level sequence in the supervised learning sample set into a physical information guided variational mode decomposition network; decomposing the underground water level sequence into K intrinsic mode component sequences and residual term sequences through a loss function of physical driving consistency constraint; and for each of the K intrinsic mode component sequences, dynamically assembling a differentiable simulator from the differentiable simplified physical simulator basic library, and carrying out cooperative training by taking approximation to each intrinsic mode component sequence and overall reconstruction of the original water level sequence as targets to obtain K completely trained assembled differentiable simulators. According to the method, a reliable visual prediction result is generated through multi-simulator collaborative deduction and uncertainty quantization.

A new energy thermal power bundling delivery system control method based on multivariable adaptive decision comprises the following steps: monitoring the total output of a new energy generator group and the total output of a thermal power generating unit in real time, and calculating the real-time bundling proportion; calculating a system vulnerability degree V based on fuzzy comprehensive evaluation: constructing an evaluation system taking the bundling proportion, the system inertia and the voltage sensitivity as indexes, and outputting the system comprehensive vulnerability degree V; carrying out fault event identification and severity S quantification; determining a control mode according to V and S, and calculating the total thermal power output increasing amount and the direct current power decreasing amount of each loop; safety verification and iterative correction of the control instruction: calculating the stability margin of the controlled system based on a transient energy function method, and if the requirement is not met, feeding back to carry out control quantity iterative correction; and issuing the control instruction passing the verification to an execution unit. The method has the advantages that the problem of decoupling of a traditional control strategy and an operation state is solved, and self-adaptive stable control based on system real-time vulnerability and fault severity is realized.

The invention provides a soil moisture sensor data quality inspection and interpolation method, which comprises the following steps: screening multivariable soil moisture time sequence data based on a preset core physical feature list, carrying out abnormal value detection through physical rule constraint and an isolation forest algorithm, and carrying out data labeling by creating a complete time axis; generating a training sample from the preprocessed data through sliding window sampling, and performing deep feature learning by using a denoising network based on a space-time diffusion probability model; performing interpolation on missing values in the original data by using the trained space-time diffusion probability model, generating a noisy data sample through a forward noise adding process, and performing conditional data interpolation based on a known observation value and a mask matrix in a reverse denoising process to generate a preliminary interpolation result; and performing post-processing correction on the preliminary interpolation result, wherein the post-processing correction comprises clamping correction based on a monthly historical range and correction based on interlayer physical logic.

The invention discloses a regional extreme drought and flood event prediction system based on a climate model, and relates to the technical field of climate model prediction and disaster early warning, and the system comprises a data input module which is used for receiving regional multivariable prediction data processed by downscaling of the climate model, and the multivariable prediction data comprises rainfall, soil humidity and evapotranspiration. According to the regional extreme drought and flood event prediction system based on the climate model, equation strong constraint and variational optimization are carried out on key variables such as rainfall, soil humidity and evapotranspiration through the physical constraint dynamic coupler, the physical law of water circulation mass conservation and surface energy balance is forcibly met, the distortion phenomenon is eliminated, and the prediction accuracy is improved. According to the method, physical correction residual errors are converted into weight factors, Copula function parameters are dynamically adjusted, it is ensured that joint probability distribution of extreme drought and flood events strictly follows a physical mechanism, the path of misinformation physical impossible events is blocked from the source, and the problem of prediction distortion caused by multivariate physical inconsistency in downscaling output of a climate model is solved.

The invention discloses a vancomycin antibacterial effect prediction method based on multivariate analysis, and relates to the technical field of medical data analysis and prediction.The vancomycin antibacterial effect prediction method comprises the steps that S1, prediction data of a target patient is collected, the prediction data comprises static baseline data acquired at the beginning of treatment and dynamic time sequence data acquired at a preset time interval in the treatment process; and S2, preprocessing the prediction data, wherein the preprocessing comprises performing data cleaning on missing values and abnormal values in the static baseline data and the dynamic time sequence data. According to the method, the static baseline data at the beginning of the treatment and the dynamic time sequence data in the treatment process are fused, and the time sequence evolution rule of the dynamic index is captured by using the long and short-term memory network, so that the treatment failure probability and the kidney injury risk probability are dynamically predicted. Meanwhile, supervised learning samples are generated through a sliding window, targeted data preprocessing and risk early warning and trend chart visualization are carried out, the prediction result can be updated in real time, and the prediction precision is improved.

The invention discloses a multi-dimensional fusion sensor data anomaly detection method based on a hydrogen-based shaft furnace, and belongs to the technical field of metallurgical equipment monitoring. The method comprises the steps of data preprocessing, single-variable anomaly detection, multivariable consistency detection, anomaly type judgment and alarm output. The single variable detection adopts a rolling median absolute deviation method, drift detection, continuous tiny variable threshold and border crossing detection; the multivariate detection calculates reconstruction residuals of each sensor through principal component analysis (PCA) and identifies overall consistency anomalies. And the system judges the fault type according to the comprehensive score of the multiple detection results and outputs a detailed alarm record and a statistical report. According to the method, multi-algorithm fusion analysis is carried out on time sequence data collected by a plurality of sensors in the operation process of the shaft furnace, so that the abnormal states of the sensors are accurately recognized. The method can be widely applied to real-time monitoring of the hydrogen-based shaft furnace smelting process, the accuracy and robustness of anomaly detection are improved, and false alarms and missing alarms are reduced.

The invention relates to a multi-causal model integrated abnormal root cause tracing method and device and computer equipment. The method is applied to industrial equipment, and comprises the following steps: acquiring multivariable time sequence data of the industrial equipment in an abnormal state; based on the multivariable time series data, training at least three causal inference models adopting different prediction functions; determining a unified root cause score vector based on the root cause score vector of each causal inference model after training; and determining a root dependent variable based on the unified root cause score vector. By adopting the method, the root cause identification efficiency and accuracy can be improved.

A system and method that generates and stores a multivariate time series of the user's conditions and activities, based on the interpretation of the user's natural language input through trained AI including large language models. This time series serves as a comprehensive record of the user's well-being and provides the foundation for the system's ability to identify the factors that most significantly impact the user's desired outcome. By tracking and analyzing the relationships between the various conditions and activities, the system can determine the key drivers of the outcome of interest to the user and provide insights into how to improve and maintain overall well-being.

The invention belongs to the technical field of electric vehicle charging load prediction, and discloses a method for predicting the load of an electric vehicle charging station on a highway, and the method can accurately sense the sudden change trend of charging demands during holidays and festivals through introducing holiday category codes and a model, and improves the prediction precision during the peak period of holidays and festivals. A multi-granularity frequency-time sequence fusion module is adopted, long-term periodicity and short-time disturbance characteristics are effectively integrated, and the modeling capacity of the model for a non-stationary load sequence is enhanced; the designed dynamic Tanh normalization module has an adaptive learning capability, and can improve the numerical stability and nonlinear expression capability of the model in a severe load fluctuation scene; the parallel TCN-MLP structure realizes collaborative modeling of nonlinear interaction and time dependency relationship between variables, and enhances the multivariable prediction performance. The method is accurate in prediction and good in stability, has expandability and engineering landing potential, and can provide efficient and reliable technical support for intelligent scheduling, power resource configuration and holiday and festival energy management.