180 results about "Long-term prediction" patented technology

A method for multivariate time-series long-term forecasting based on multi-scale temporal feature enhancements, includes a time-series forcasting model TFEformer. The model utilizes a multi-branch structure and a patch-series attention mechanism to extract global and local time-series features at multiple temporal scales, and designs an adaptive feature fusion mechanism to achieve adaptive fusion of multi-scale temporal features. It employs an variate-wise attention mechanism and a redesigned gated feedforward network to perform feature fusion among multivariate variables and within the time-series, respectively. The time-series forcasting model TFEformer proposed by the present invention significantly improves the prediction of long-term trends in time-series and enhances the fitting ability for short-term local fluctuations, comprehensively increasing prediction accuracy across different prediction time lengths in multivariate time-series forcasting tasks.

A multi-energy load forecasting method, a multi-energy load forecasting system, an electronic device, a program, and a storage medium are provided that realize accurate long-term forecasting of multi-energy loads in a target integrated energy system under conditions where no historical load data is available. [Solution] A multi-energy load forecasting method for an integrated energy system without historical data involves obtaining the meteorological characteristics of a target complex and the cooling, heating, electricity, and gas historical data of a source domain group complex, preprocessing the obtained data, performing cross-correlation and generalization ability analysis of the complex on the preprocessed cooling, heating, electricity, and gas historical data of the source domain group complex, determining appropriate source domain data, constructing a multi-energy load forecasting model, training the model based on the source domain data according to the Metas training policy, obtaining a trained forecasting model, and inputting the preprocessed meteorological characteristics of the target complex into the forecasting model to obtain a forecast result.

The invention belongs to the technical field of tunnel safety monitoring, and particularly relates to a rock stratum stress-deformation coupled tunnel safety real-time dynamic modeling method, which comprises the following steps of: embedding a rock mass mechanical relationship into a neural network model as a hard constraint, and establishing explicit mapping of monitoring data and a physical field, a model parameter field is dynamically optimized by adopting ensemble Kalman filtering real-time general field detection data, and a data driving and physical mechanism collaborative deduction mechanism is designed, so that long-term prediction error accumulation is effectively inhibited; a stress field, a deformation field and other physical quantities deduced and output by the model are directly utilized to calculate disaster risk indexes with clear mechanical significance, and a physically interpretable early warning decision is realized by fusing fuzzy reasoning and multi-source risks; the technical problems of physical misalignment, weak long-term generalization and poor early warning interpretability of a data-driven model are solved.

The invention discloses a digital twin-driven bridge full life cycle damage prediction and evaluation method and system, and belongs to the field of bridge structure health monitoring, and the method comprises the steps: obtaining a monitoring data set of a bridge structure; an initial digital twinborn model embedded with a micro physical layer is constructed and trained, the micro physical layer constructs a damage evolution model applied with monotonic physical constraint based on multi-source monitoring data, and a damage evolution trajectory of the bridge in a future time period is predicted through the damage evolution model based on the physical parameter vector; in combination with an uncertainty quantification method, generating a time-varying reliability index of the bridge in a future time period; and based on the time-varying reliability index, constructing and solving a maintenance decision optimization model to generate a maintenance decision of the bridge. According to the invention, the physical authenticity and reliability of the long-term prediction result are ensured.

The embodiment of the invention discloses a power grid load prediction method and device based on fractal analysis, and the method comprises the steps: obtaining historical load time series data of a power grid, and carrying out the preprocessing, and generating a standardized load data set; fractal features of the load data set are extracted by using a fractal analysis algorithm; judging dynamic behavior characteristics of the load time sequence based on fractal characteristics, adaptively selecting short-term prediction or long-term prediction model parameters, and dynamically adjusting the time scale of a prediction model to adapt to load fluctuation characteristics; analyzing an abnormal mode in the historical load time sequence based on fractal features, and generating an abnormal detection result; constructing a load prediction model based on the fractal features, the anomaly detection result, the model parameters and the external features; and predicting the power grid load in the future time period by using the load prediction model to obtain a prediction result. According to the power grid load prediction method provided by the invention, the prediction accuracy and stability are improved.

The invention relates to the technical field of power load prediction, in particular to a long-term power system load prediction method and system based on multi-scale decomposition fusion. The method comprises the steps of performing data preprocessing based on time sequence data; performing multi-scale decomposition and feature embedding on the preprocessed data to obtain a multi-scale load feature vector set; performing gating adaptive filtering and attention double-path fusion under the multi-scale load characteristics based on the multi-scale load characteristic vector set; performing independent prediction and prediction fusion on a fusion result based on a space-time attention gating mechanism; and evaluating a result after prediction fusion. According to the multi-scale prediction result space-time attention fusion mechanism provided by the invention, prediction information on different scales can be adaptively integrated, deviation caused by a single scale is avoided, the comprehensive performance of long-term prediction is further improved, and the method is suitable for various power system planning and operation scenes.

The invention relates to the technical field of geological early warning, in particular to a dangerous rock mass instability analysis method, system and equipment based on a space-time diagram neural network, by fusing unmanned aerial vehicle LiDAR, multispectral data, meteorological radar data and the space-time diagram neural network (ST-GNN), the system realizes sub-meter spatial resolution and minute-level time response, and the stability of dangerous rock mass instability analysis is improved. The four-dimensional (time and space) analysis result of the instability probability of the dangerous rock mass is obtained through high-precision space-time modeling, the problems that a traditional geological disaster early warning system is low in resolution ratio, slow in response and high in misinformation are solved, the comprehensiveness, accuracy and reliability of instability prediction of the dangerous rock mass are improved, and the early warning effect is good. And full-chain intelligent closed-loop management of real-time data acquisition-dynamic prediction-early warning push-feedback optimization is supported, the emergency decision time is shortened by real-time rainfall superposition risk thermodynamic diagrams, and the attenuation rate of long-term prediction precision is reduced by dynamically fusing newly added geological data and instability events through incremental learning.

The invention belongs to the technical field of new energy electric power, and discloses a new energy bearing intelligent evaluation and regulation system and method, the system is composed of a data acquisition module, an intelligent evaluation module, a prediction module, a regulation strategy module and an execution and feedback module, and the data acquisition module acquires meteorological, historical operation, load and power grid state data in real time; the time sequence of the prediction module and a deep learning algorithm are combined, short-term to medium-and-long-term prediction is carried out on new energy output and load change, output change is captured in real time, regulation and control lag is avoided, power grid risks are quantified through a multi-model fusion algorithm, and node accessible capacity and a risk assessment result are output; and the regulation and control strategy module generates a regulation and control instruction by utilizing optimization algorithms such as a genetic algorithm based on the evaluation result and the prediction data, and realizes closed-loop control through the execution and feedback module, so that the evaluation data directly guides regulation and control actions.

The invention particularly relates to a hydrogeological analysis system and method based on big data, and relates to the technical field of hydrogeological analysis, and the method comprises the steps: placing a new model and an old model in an A / B test stage, and carrying out the parallel processing of real-time data; performing automatic judgment according to a preset performance index, and deciding whether to upgrade the candidate model to a new production model; and applying the model selected after decision making to hydrogeological analysis to solve the problem of long-term prediction precision under dynamic change of the hydrogeological system. According to the method, dynamic threshold monitoring, K-S outlier sample examination, sub-model optimization and A / B test smooth iteration processes are constructed, and a mechanism and machine learning coupling model and parallel calculation are combined, so that the model drift identification delay is reduced, the prediction precision is improved, scenes such as underground water over-mining prevention and control and pollution emergency are effectively supported, and the prediction accuracy is improved. And an accurate and efficient scientific decision basis is provided for water resource management.

The invention relates to the field of power grid real-time monitoring and coordination control, in particular to a distributed photovoltaic grid-connected regional power grid real-time monitoring and coordination control system which comprises a sensing acquisition module, a multi-time scale prediction module, a self-evolution calibration module and a distributed coordination control module. The sensing acquisition module acquires wide-frequency-domain electrical quantity and multi-dimensional meteorological quantity through a double-layer sensing network, and restoration data is obtained through anomaly detection; the multi-time-scale prediction module integrates ultra-short-term, short-term and medium-and-long-term prediction, and realizes cross-scale collaboration based on a combined state space model and the like; the self-evolution calibration module performs online compensation on the multi-scale prediction residual error based on deep reinforcement learning; the distributed coordination control module is combined with an improved alternating direction multiplier method and a non-dominated sorting genetic algorithm to obtain an optimal power instruction meeting voltage, frequency and harmonic ternary constraints; according to the invention, the stability of power grid operation in a high-permeability photovoltaic grid-connected scene is improved.

The invention discloses an intelligent power grid load prediction and dynamic response coordinated scheduling method, and relates to the technical field of power system automation, and the method comprises the steps: accessing intelligent ammeters, distributed power controllers and other devices of Modbus, IEC61850 and DL / T645 protocols through a multi-protocol adaptive gateway, and achieving data standardization; time stamps are calibrated by means of Beidou time service and an IEEE1588PTP protocol, and it is ensured that multi-source data synchronization errors are controllable; deploying an edge computing node cluster, distributing high-priority tasks to low-load nodes through an edge coordinator in combination with a load fluctuation level and a greedy algorithm, and ensuring real-time processing efficiency; the edge nodes generate short-term load prediction, and the cloud platform outputs medium and long-term prediction based on a historical data training model; and finally, the coordinated scheduling decision module fuses the two types of prediction results and the real-time parameters of the power grid, and generates a dynamic instruction to control the output of the adjustable load and the distributed power supply.

The invention provides an electric quantity prediction method and system fusing physical constraint factors, and relates to the technical field of electric quantity prediction. Historical load, weather, electricity price and calendar data are collected, and a key feature set is constructed through preprocessing and feature selection; a prediction model with the physical information neural network as the core is constructed, the prediction model comprises a recursion sub-module used for short-term prediction and a trend sub-module used for long-term prediction, and a physical constraint loss item based on a physical rule is introduced into model training so as to enhance the generalization ability; a multi-time granularity modeling framework is adopted, uncertainty quantization is achieved through a Monte Carlo Dropout or Bayesian neural network, and a confidence interval of a predicted value is output; and finally, causal reasoning is carried out through a Shapley value algorithm and anti-fact simulation, and key influence factors are identified. According to the method, the precision, stability and interpretability of electric quantity prediction are effectively improved, and reliable support is provided for power grid dispatching and decision making.

The invention relates to a mine gas emission quantity prediction method based on deep learning. The deep learning-based mine gas emission quantity prediction method comprises the following steps of (1) preparing a data set, (2) performing correlation analysis on the data set to determine main control elements of mine gas emission, and (3) constructing a VMD-WTC-PatchTST combined model according to time sequence data of the main control elements of mine gas emission. According to the VMD-WTC-PatchTST model prediction performance evaluation method, data and the model are improved through error analysis, and the final VMD-WTC-PatchTST model prediction performance evaluation result is that R2 reaches 0.894, RMSE is 1.80, and MAE is 1.38. Compared with an original PatchTST model, the fitting degree of the PatchTST model is improved by 2.6%, and due to the fact that the PatchTST is also suitable for long-term prediction, the performance of the model is expected to be further improved theoretically along with continuous supplementation of data.

The invention relates to the technical field of hydraulic engineering monitoring and control, in particular to a sluice pump station operation state and water regimen linkage monitoring system. The decision control unit is used for generating long-term prediction information and short-term prediction information based on hydrological observation data and generating control instructions of the opening degree of a water gate and the rotating speed of a water pump. The control instruction execution unit is used for executing the corrected control instruction of the water gate opening degree and the water pump rotating speed; the linkage safety interlocking unit is used for transmitting a standby pump starting instruction to the control instruction execution unit when the first condition and the second condition are met at the same time, sending a blocking signal to the tactical optimization module, and sending a release signal to the tactical optimization module after the standby pump starting instruction is executed. According to the system, low efficiency or safety risk caused by a fixed strategy and execution of wrong operation under the condition that the equipment has a fault or has a fault are avoided, and secondary damage to the equipment and interruption of system operation caused by improper control instructions are effectively prevented.

The invention discloses a deep learning photovoltaic power generation long-term prediction method fusing domain knowledge, and the method comprises the following steps: constructing a meteorological feature fine classification module, carrying out the classification processing of original meteorological data, and generating discrete weather category labels; carrying out embedded representation on discrete weather category labels, and splicing the discrete weather category labels with original continuous meteorological features to form an enhanced input sequence; the enhanced input sequence is input into a Fusionform model, and photovoltaic power generation power prediction is carried out; introducing a theoretical power modeling module based on a photovoltaic physical mechanism, and calculating theoretical power generation power according to photovoltaic system parameters and meteorological data; performing weighted fusion on the output of the Fusionformer model and the theoretical generated power to obtain a final prediction result; according to the method, the physical consistency and interpretability are improved while the precision is ensured, and high-precision long-term prediction support is provided for power grid dispatching.

The invention discloses a control method of a mixed electrolytic cell cluster and a related device, relates to the technical field of intelligent control, and determines the initial starting number of AEL electrolytic cells by utilizing prediction data based on a long-term prediction period. And based on the prediction data of the first short-term prediction period, the initial starting number of the PEMEL electrolytic cell is determined. And based on the prediction data of the target short-term prediction period, predicting a wind-light power prediction value of the target short-term prediction period, and determining the operation state of the mixed electrolytic cell cluster according to the wind-light power prediction value of the target short-term prediction period, the initial starting number of the AEL electrolytic cells and the initial starting number of the PEMEL electrolytic cells. According to the method, long-term prediction and short-term prediction are combined, coordinated operation of the PEMEL electrolytic cell and the AEL electrolytic cell in the mixed electrolytic cell cluster is controlled, real-time performance and flexibility of electrolytic cell control are achieved, and therefore the control effect of the mixed electrolytic cell cluster is improved.

The invention discloses an intelligent analysis system for the running state of a 3C vehicle-mounted contact network, and belongs to the technical field of crossing of intelligent operation and maintenance of rail transit and industrial artificial intelligence. And the system associates the multi-source heterogeneous observation data to a specific equipment unit through an equipment centralized data binding module. And the multi-modal feature extraction and state management unit processes the data and maintains a multi-dimensional state vector of the equipment by using the Kalman filtering updating unit. And the physical-data hybrid decision maker fuses the data driving rule and the simplified physical model to output a diagnosis result. The topology analyzer performs global verification based on a mechanical transfer rule. The system further comprises a long-term health state prediction and feedback module, early failure risks are predicted through a hidden Markov model, Kalman filtering process noise is dynamically fed back and adjusted, and cooperation of long-term prediction and short-term estimation is achieved. According to the method, the technical problems of multi-source data splitting, lack of physical basis in diagnosis and incapability of predictive maintenance are solved.

The invention discloses a global fracture conductivity ultra-long-term prediction method and system, belongs to the technical field of crossing of oil and gas reservoir engineering and artificial intelligence, and is used for solving the technical problem that the calculation cost is high when an existing physical model is used for predicting fracture ultra-long-term conductivity. The method comprises the following steps: generating a global fracture conductivity evolution data set covering multiple working conditions through a numerical simulation system; constructing a multi-channel input tensor containing dynamic fracture conductivity and static proppant concentration distribution; using the data set to train a plurality of Fourier neural operator models for different prediction time scales; and based on a hierarchical time sequence aggregation strategy, iteratively calling trained models of different time scales, and realizing rapid prediction of fracture conductivity in a few years to tens of years in the future. According to the method, on the premise of ensuring the prediction precision, the calculation time consumption of ultra-long-term prediction can be remarkably reduced, and decision support is provided for crack design and production strategy optimization.

The invention discloses a surface temperature medium and long term forecasting method based on a multi-scale TransFuse pyramid network. The surface temperature medium and long term forecasting method comprises the steps of S1, acquiring original LST time sequence data and performing preprocessing; s2, constructing a three-dimensional time delay embedded feature tensor for the data preprocessed in the S1 by adopting a stepped cross-node tensor feature reconstruction method; s3, on the basis of the three-dimensional time delay embedded feature tensor constructed in the S2, meteorological features of different time scales are extracted in parallel through a dynamic multi-resolution convolution kernel group; s4, performing five-stage cascade operation on the meteorological features of different time scales extracted in the S3 to realize deep fusion of cross-scale features, and performing high-level semantic information integration through a global feature pyramid; and S5, finally outputting a multi-step prediction result through a regression prediction framework based on the fusion features obtained in the S4. According to the invention, an extensible technical normal form is effectively provided for high-precision meteorological prediction in a limited computing power scene, and important application potential is shown in actual business deployment.

The invention provides an eye health state prediction system and a prediction method thereof. The prediction method comprises the following steps: S1, data acquisition; s2, multi-source data preprocessing and feature engineering; s3, based on dynamic feature weighting of a scene and an attention mechanism, generating a fusion feature vector used for model input; s4, performing short-term channel and long-term channel double-model prediction and state evaluation, and respectively performing short-term state detection and long-term risk prediction; s5, comprehensively displaying, evaluating and calculating the total score; s6, early warning triggering based on the scene and the severity level, and triggering early warning of different levels according to the total score calculated in the step S4 and in combination with scores of all subitems. User data can be collected in multiple dimensions, the limitation of a single dimension is solved, multi-source data preprocessing and feature processing are conducted on the user data in a multi-data fusion mode, and therefore the accuracy of the user data is improved. And short-term timely early warning and long-term prediction of the user are realized according to a data processing result in cooperation with an early warning module, and eye health detection and fatigue state early warning of the user are realized.

The invention discloses a long time sequence prediction method based on time-varying period coding and hierarchical channel fusion. Comprising the following steps: decomposing an input multivariate time sequence into a trend component and a season component; performing mapping processing on the trend component to obtain trend output; time-varying periodic coding processing and hierarchical channel fusion processing are carried out on the seasonal components to obtain seasonal output, the time-varying periodic coding processing is used for adaptively capturing time-varying periodic features in the seasonal components, and the hierarchical channel fusion processing is used for dynamically capturing correlation differences among different channels in the seasonal components; and generating a prediction result of the future time step based on the trend output and the seasonal output. According to the method, the time-varying periodic characteristics can be modeled in a self-adaptive manner, the strength correlation between the channels can be dynamically captured, and the method has excellent long-term prediction performance and generalization.

The invention discloses a multi-source time sequence decomposition and fusion hydraulic structure behavior prediction method, which comprises the steps of 1, acquiring historical time sequence monitoring data of a hydraulic structure, including behavior parameters and environmental quantity parameters, the behavior parameters including dam body settlement and seepage flow, and the environmental quantity parameters including dam body water level and rainfall; 2, preprocessing the acquired monitoring data, and constructing a hydraulic structure behavior model input data set; 3, determining an optimal decomposition period of an input data set of the hydraulic structure performance model; 4, according to the determined optimal decomposition period, decomposing the preprocessed hydraulic structure performance model input data set into three component parameters according to an original time sequence, wherein the three component parameters are respectively a trend component parameter, a periodic component parameter and a residual component parameter; and 5, superposing the predicted values of the three component parameters according to time points to obtain a final performance predicted value of the hydraulic structure. According to the method, the prediction precision is remarkably improved, and the stability and reliability of long-term prediction are remarkably improved.

The invention provides a non-stationary hydrological time sequence prediction method based on WA-GA-BP, and the method comprises the steps: carrying out the recognition of a non-stationary hydrological time sequence through wavelet analysis, separating a deterministic component from a random component, building a genetic BP neural network for sub-components, carrying out the prediction, and superposing the prediction results of the sub-components, and obtaining a final prediction result. The method is applied to the prediction of four groups of hydrological time sequences, and compared with other two common hydrological time sequence prediction models AR (p) and GM (1, 1), the result shows that the wavelet analysis and genetic algorithm coupling improved BP neural network model can effectively improve the prediction precision and stability, the prediction result of the WA-GA-BP model meets the long-term prediction precision requirement of water regimen, and the prediction result of the WA-GA-BP model meets the long-term prediction precision requirement of the water regimen. Compared with AR (p) and GM (1, 1) models, the method has higher accuracy. An example verifies that the WA-GA-BP model not only can solve the problem that the prediction precision is influenced by the non-stationarity, but also can solve the problem that the prediction stability is influenced by the model parameter selection randomness, and has reasonability and feasibility when being applied to the long-term prediction of the non-stationary hydrological time series.

The invention provides a multi-energy-flow real-time updating method based on a building integrated energy system, and belongs to the technical field of energy scheduling, and the method comprises the steps: obtaining an equipment set and a typical state vector, and constructing a building integrated energy mechanism model based on mechanism data and balance constraint data; collecting a plurality of historical operation data, and determining a plurality of period-mode historical data and a plurality of environment clustering data; determining a plurality of response clustering data, and constructing a data driving model of each environment label of each period-mode label; acquiring equipment parameter data, and determining fitting residual data of each environment label of each period-mode label; performing first optimization on each data driving model; and constructing a mechanism-data hybrid drive model and realizing adaptive optimization. The method can accurately adapt to a subdivided scene, improve deviation fitting precision, give consideration to explanatory and prediction precision, dynamically adapt to system time-varying characteristics, improve the adaptability of the model in a complex scene, and guarantee long-term prediction reliability.

The invention relates to the field of life prediction models and data analysis, and discloses a submarine cable operation life prediction method and device based on dynamic temperature characteristics and a product. The method comprises the following steps: when a target offshore wind field is a to-be-built wind field, acquiring a long-term prediction sequence of the output power of the target offshore wind field, and analyzing the temperature characteristics of a to-be-predicted submarine cable in combination with a submarine cable key parameter set; evaluating the aging speed of the to-be-predicted submarine cable by using the submarine cable dynamic aging model, and determining a first long-term attenuation rate sequence of the performance index of the to-be-predicted submarine cable; and based on a preset life end point, according to the first long-term attenuation rate sequence and the plurality of initial performance index values of the to-be-predicted submarine cable, predicting the operation life of the to-be-predicted submarine cable, and obtaining an operation life prediction value of the to-be-predicted submarine cable. According to the method, the dynamic temperature characteristics of the submarine cable are considered, and the precision of a submarine cable service life prediction technology is improved.

In order to solve the key problems of difficulty in multi-source information fusion, insufficient spatial topology modeling, attenuation of long-term prediction precision and the like in an existing aircraft ground taxiing trajectory prediction method, a parallel processing architecture of a historical trajectory encoder and a pavement path encoder is designed, and trajectory time sequence features are extracted by using a long-short-term memory network; modeling a spatial topological relation of a control path by adopting a graph attention network, and realizing effective integration of heterogeneous information through a feature fusion layer; a multi-component loss function fusing the position, the speed and the acceleration is provided, and the continuity and the smoothness of a prediction track are constrained; a lightweight data enhancement strategy and an adaptive residual connection mechanism are designed, the model generalization ability is improved, and long-term prediction error accumulation is relieved. According to the method, multi-source trajectory information can be effectively fused, and the accuracy and stability of aircraft ground taxiing trajectory prediction are remarkably improved.

The invention relates to the technical field of engineering monitoring, and discloses a soft soil foundation high-speed rail station roadbed safety construction monitoring method and system, and the method comprises the steps: constructing a finite element model and an LSTM model for bidirectional data feedback, and carrying out the monitoring of the safety construction of a soft soil foundation high-speed rail station roadbed according to construction time sequence data and construction real-time data; the method comprises the following steps: accurately performing short-term prediction and long-term prediction on monitoring data of a soft soil foundation high-speed railway station roadbed construction project, outputting a construction suggestion, and after the construction suggestion is implemented, further optimizing a finite element model and an LSTM model for data bidirectional feedback according to newly collected data, so as to obtain a prediction result; short-term and long-term prediction on soft soil foundation high-speed railway station roadbed construction projects is realized at the same time, so that safe construction monitoring is performed, and the problem of insufficient real-time performance of construction monitoring is avoided.

The invention discloses a method for predicting power consumption load of industrial and commercial users, and particularly relates to the field of data analysis, which comprises data preprocessing, seasonal decomposition, independent component prediction and total quantity reconstruction. According to the method, firstly, the limitation of traditional regional prediction is broken through, single-household-granularity accurate load prediction is provided, and refined operation such as electricity transaction and demand response of an electricity selling company and a virtual power plant is directly supported; through separation of trend, season and irregular components and targeted selection of an SARIMA / SVM model for prediction, the problem of error accumulation and amplification of long-term prediction of a traditional seasonal model is thoroughly solved; and finally, the enterprise yield index and the load elastic coefficient characteristics are fused in trend prediction, so that the prediction result sensitively responds to the actual production and operation fluctuation of industrial and commercial users, and the adaptability of prediction to the actual operation state of the enterprise is remarkably improved.

The invention relates to the technical field of power equipment state monitoring, and discloses a switch cabinet partial discharge detection method, which comprises the following steps: synchronously collecting multi-source heterogeneous signals generated by partial discharge through a multi-sensor array, and carrying out preprocessing and feature extraction to obtain a multi-modal feature vector; inputting the multi-modal feature vector into a recognition and evaluation model for automatic recognition of discharge types, evaluation of discharge severity and sensing of abnormal discharge; real-time analysis and fusion decision making are carried out on an abnormal sensing result, a recognition result and an evaluation result through a collaborative architecture of edge computing nodes and a cloud platform; wherein the edge computing node utilizes a lightweight real-time sensing module to execute real-time abnormal discharge sensing and triggers local early warning when sensing abnormity, and the cloud platform utilizes a depth evaluation module to execute comprehensive state evaluation, recognition and long-term prediction. According to the invention, the limitation of a single detection method is overcome through multi-sensor synchronous acquisition and in combination with a D-S evidence theory information fusion method.