CN114625831A - Classification evaluation feedback method for load identification of smart power grid - Google Patents

Classification evaluation feedback method for load identification of smart power grid Download PDF

Info

Publication number
CN114625831A
CN114625831A CN202011416235.0A CN202011416235A CN114625831A CN 114625831 A CN114625831 A CN 114625831A CN 202011416235 A CN202011416235 A CN 202011416235A CN 114625831 A CN114625831 A CN 114625831A
Authority
CN
China
Prior art keywords
evaluation
algorithm
classification
feedback
text
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011416235.0A
Other languages
Chinese (zh)
Inventor
张翼英
周保先
尚静
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University of Science and Technology
Original Assignee
Tianjin University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tianjin University of Science and Technology filed Critical Tianjin University of Science and Technology
Priority to CN202011416235.0A priority Critical patent/CN114625831A/en
Publication of CN114625831A publication Critical patent/CN114625831A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/3332—Query translation
    • G06F16/3335—Syntactic pre-processing, e.g. stopword elimination, stemming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • Economics (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Machine Translation (AREA)

Abstract

The invention relates to a classification evaluation feedback method for load identification of a smart power grid, which analyzes the evaluation content of a user power consumption list, establishes a classification evaluation feedback model by combining a deep learning algorithm, classifies and analyzes evaluation texts, and finally establishes an evaluation feedback mechanism according to actual analysis results to provide adjustment feedback for load identification. According to simulation experiment result analysis, the algorithm can accurately distinguish user experience, can provide adjustment suggestions, reduces manual participation in an evaluation process, improves the efficiency of finding problems, realizes communication with users as soon as possible, has certain value and positive effect on non-invasive load identification algorithm research, and has positive effect and instructive suggestion on non-invasive load identification of a power grid. The scheme of the invention aims to research corpus data processing, classification evaluation analysis and evaluation feedback mechanisms and construct a classification evaluation feedback system for intelligent power grid load identification.

Description

Classification evaluation feedback method for load identification of smart power grid
Technical Field
The invention relates to a classification evaluation feedback method for load identification of a smart power grid, which analyzes the evaluation content of a user power consumption list, establishes a classification evaluation feedback model by combining a deep learning algorithm, classifies and analyzes evaluation texts, and finally establishes an evaluation feedback mechanism according to actual analysis results to provide adjustment feedback for load identification. According to simulation experiment result analysis, the algorithm can accurately distinguish user experience, can provide adjustment suggestions, reduces manual participation in an evaluation process, improves the efficiency of finding problems, realizes communication with users as soon as possible, has certain value and positive effect on non-invasive load identification algorithm research, and has positive effect and instructive suggestion on non-invasive load identification of a power grid.
Background
With continuous progress and development of current electric power technology, enterprises pay more attention to the daily life electricity utilization experience of users. Meanwhile, with the introduction of concepts of smart home and smart life, accurate analysis of power consumption behaviors of users and detailed power consumption expenditure are important in realizing power consumption safety and improving user satisfaction. The non-invasive charge identification algorithm is just oriented to the requirement, and on the basis of the existing household and civilian circuits, the analysis on the household power consumption behavior of the user is realized by means of the existing electric energy meters and other electric equipment.
In recent years, a non-invasive load identification technology has been widely applied to the fields of building energy conservation, smart cities, smart power grids and the like, and along with the continuous development of the field of artificial intelligence and the continuous maturity of big data analysis and mining technology, the technology can monitor the power consumption behavior through equipment such as an electric energy meter and the like, and protect the privacy of users to a certain degree. For the user, the power consumption details of the equipment can be timely fed back through load monitoring, and the reasonable power consumption of the user is guided; for an electric power company, non-invasive load monitoring can realize fine-grained perception of each component of a load on the premise of not obviously improving investment, the load prediction accuracy of an electric power system is improved, the safety and the economy of a power grid are improved, modeling on user behaviors can be facilitated more accurately, and differentiation and accurate service for users is realized. However, the existing non-invasive load monitoring products in the market are limited by computing power, power consumption and cost, and the algorithm of the production equipment is single and fixed, so that the classification parameters or the matching library are continuously updated, and the recognition achievement rate of various types of products can be improved. The problems of data link occupation and poor self-adaptive capability caused by regional similarity of the household appliances and updating of classification parameters or matching libraries are caused. On the basis of the research of the existing non-invasive load identification algorithm, the evaluation data information of the user is fully utilized, the deep learning algorithm is combined to classify the evaluation texts and perform emotion analysis, a classified evaluation feedback model is established, and the feedback adjustment of the power grid load identification is realized according to the analysis result.
Disclosure of Invention
The invention firstly processes the whole evaluation data in the user evaluation data. The method realizes the steps of deleting special symbols, filtering common words, sentence segmentation and the like, processes the whole text into a single-sentence load evaluation short text form, removes redundant characters in the text and reduces model operation noise. Sending the preprocessed data to a classification evaluation analysis part for text classification and analysis; and then modeling and analyzing the data processing data, performing model training according to the existing and manually labeled load evaluation texts of various types of equipment, classifying the processed load evaluation text input models to obtain type feedback, and analyzing the load evaluation texts obtained after the linguistic data preprocessing by using an evaluation feedback algorithm to obtain the identification error feedback condition of each equipment. Sending the two feedback results to an evaluation feedback mechanism part for processing; and finally, respectively processing according to the two feedback results. And judging whether the classification result belongs to the novel equipment category according to a feedback mechanism, and updating the equipment library and the training parameters. And for the two effects of the analysis result, feeding back the identification effect of each type respectively, and providing a new algorithm adjustment suggestion.
In order to achieve the purpose, the technical steps adopted by the invention are as follows:
step A: design algorithm framework
According to the whole algorithm design of the classified evaluation feedback, the whole algorithm design is divided into three parts, and the three parts respectively comprise a corpus data processing mechanism, a classified evaluation analysis mechanism and an evaluation feedback mechanism. As shown in figure 1 of the drawings, in which,
(1) and a corpus preprocessing module. Designing a data reading processing program, taking a user evaluation list and the equipment type as algorithm input, generating a regularized short text after processing, and storing the regularized short text in a database.
(2) And a classification evaluation analysis module. And the classification evaluation feedback model respectively reads and trains the text, realizes text classification and evaluation analysis, generates a class vector and an evaluation analysis result, and stores the class vector and the evaluation analysis result in a database.
(3) The feedback mechanism is evaluated. The algorithm processes the category vector data and the evaluation analysis result respectively. And (3) reading and analyzing data of the two processing results, analyzing and processing according to a set threshold, finally predicting the problems of the existing non-invasive load identification algorithm, and giving a reasonable and standardized suggestion.
And B: corpus pre-processing
The overall evaluation data of the user is processed according to the steps shown in fig. 2, and since the user evaluation data is a paragraph and a sentence with a complex structure, the evaluation text is first sentence-divided, and the overall text X of each user is processednSplitting into (x)1，x2，...，xn) The sentence phrase of (2) is stored.
And common tone words, special symbols and the like are deleted according to the common word dictionary, so that the operation burden of the model is reduced, null data are screened and removed, the effectiveness of the data is ensured, and the test and training conditions of the model are not influenced. Finally, the sentence sequence corresponding to each user is obtained.
And C: evaluation classification algorithm based on LSTM
The user evaluation text is generated into a sentence text sequence through corpus processing, and due to the fact that the text is short and the dependency of a classification result on the context is strong, the LSTM neural network algorithm is used for processing the problems of time vectors and long-term dependency. The defects of the conventional RNN information loss in the process of processing the long-term dependence context problem are avoided, and the accuracy of text classification is effectively improved. The structural units of the neurons are shown in fig. 3.
Where long-term state C stores long-term memory information so that the feature vectors above can be saved and passed as long-term states.
Step D: Bi-LSTM-based evaluation analysis
The Bi-LSTM neural network is formed by combining forward LSTM and backward LSTM, two LSTM with inverted directions are given out in each operation, and the output result is determined by two single LSTM. The structure can keep more input information, and is convenient for emotion analysis. Fig. 5 shows a structure of the neural network.
Step E: design evaluation classification model
The sequence text x obtained in the above stepsnThe selection part carries out manual category marking according to the equipment category of the evaluation text and marks the equipment category as data _ train to obtain a model training set (x _ train, x _ test) and a test set (y _ train, y _ test); then the training set and the test set are sequenced to obtain a text xnPerforming word segmentation processing to x(s)1，s2，…，sn) The word vector patterns are input to the model for training and testing.
After the model training is finished, performing word segmentation processing on the rest user data according to the same steps, inputting the rest user data into the model, and finally outputting a text category vector Y represented by corresponding user evaluation datan(y1，y2，…，yn) And transmitted as input to the evaluation feedback mechanism.
Step F: design evaluation feedback model
And establishing an evaluation feedback mechanism according to the obtained user evaluation text classification and text emotion analysis, analyzing an output result by using a category analysis and evaluation algorithm, and judging whether the load identifies the occurrence of new equipment, the type of the equipment is required to be adjusted in time, and whether an error condition exists and an equipment parameter library matching model is required to be updated in time. Fig. 6 shows a diagram of the evaluation feedback algorithm.
Compared with the prior art, the invention has the advantages and positive effects that:
the algorithm is based on user power consumption list evaluation content, a classification evaluation feedback model is established by combining a deep learning algorithm, evaluation texts are classified and subjected to emotion analysis, and then an evaluation feedback mechanism is established according to actual analysis results to provide adjustment feedback for load identification. The method and the device realize accurate differentiation of user experience, provide adjustment suggestions in a targeted manner, and have positive effects and instructive suggestions on non-invasive load identification of the power grid.
Drawings
FIG. 1 is a block diagram of an algorithm design framework
FIG. 2 is a corpus pre-processing flow chart
FIG. 3 is a diagram of the structure of the LSTM neural network
FIG. 4 is a diagram of a Bi-LSTM neural network architecture
FIG. 5 is a flowchart of an evaluation classification algorithm
FIG. 6 is a diagram of an evaluation feedback algorithm
FIG. 7 is a diagram of an evaluation text classification training process
FIG. 8 is a diagram illustrating the load recognition of the device
Detailed Description
Step A: corpus pre-processing
The experiment processes the whole text into a single-sentence load evaluation short text form, so that model operation noise is reduced; and processing the overall evaluation data in the user evaluation data. The method realizes the steps of deleting special symbols, filtering common words, sentence segmentation and the like, processes the whole text into a single-sentence load evaluation short text form, and simultaneously removes redundant words in the text. Because the user evaluation data is paragraph-like and has complicated structure, the evaluation text is firstly processed in sentence division manner, and the whole text X of each user is processednSplitting into (x)1，x2，…，xn) The sentence phrase storage of (1).
And B: LSTM-based rating classification
The invention is based on the evaluation classification of LSTM, forgets the input information x and the output information h of the previous unit of the gate floor(t-1)Through ftCalculating and judging whether to transmit to C, and determining the discarding of the information according to the judgment, wherein W is a weight value, b is an offset:
ft＝σ(Wf·[h(t-1)，xt]+bf) (1)
the update layer comprises an input gate layer itDetermining values to be updated and a vector of candidate values
Figure BSA0000226871330000051
Where tanh is the activation function:
it＝σ(Wi·[h(t-1)，xt]+bi) (2)
Figure BSA0000226871330000052
then updating the long-term status to C(t-1)Is updated to CtOld state C(t-1)And ftThe multiplication determines the discard information and then adds the new state to update the delivery change for each state:
Figure BSA0000226871330000053
and finally, outputting the finally determined output part, namely the classification result, through the processing of an output layer:
ot＝σ(Wi*[h(t-1)，xt]+b0) (5)
ht＝ot*tanh(Ct) (6)
and C: evaluation analysis based on Bi-LSTM
X(s) in the structure diagram of Bi-LSTM neural network1，s2，…，sn)，x′(s′1，s′2，…，s′n) For input, the following operations are obtained through a forward operation and a reverse operation:
hLn＝f(Wst+VhL-1) (7)
hRn＝f(Wst′+VhL-1) (8)
where f is the LSTM neuron activation function, hLnAnd hRnThe calculated output information vectors are respectively, W is a weight matrix from the hidden layer to the output layer, and V is a weight matrix of the hidden layer calculated from the previous moment to the next moment. Finally h isLnAnd hRnThe method comprises the following steps of obtaining through the activation function operation of an output layer:
On＝g(W′hLn+W′hRn) (9)
where g is the output layer activation function and W' is the output layer weight matrix. And finally, calculating to obtain a vector representing the emotion analysis of the user.
Step D: evaluation analysis and feedback
Rating analysis by combining user text xnProcessing into s(s) according to word segmentation form1，s2，…，sn) Selecting partial data to be manually marked as a training set to train an evaluation analysis model, loading the model, inputting the model and the text to be analyzed into a Bi-LSTM neural network for processing, and finally obtaining an analysis result and using OnAnd respectively representing the text analysis result of each user and serving as the input of the evaluation feedback mechanism.
step 1: traversing the user evaluation category information to judge the user XiText category vector y ofiWhether or not to satisfy
Figure BSA0000226871330000061
Wherein
Figure BSA0000226871330000062
Is yiThe maximum value in the user data is delta t, the threshold value is set for the category, if the threshold value is exceeded, the user equipment is considered to be updated, the user load identification model is adjusted in time, and category feedback response is made.
step 2: for evaluation result OiUsing evaluation algorithm analysis, first traverse user XiAll y corresponding to the Chinese text categoryiExtracting y from the sequence tag ofiCorresponding emotion analysis vector oiAnd storing the value according to the label.
step 3: calculate each yiO corresponding to label typeiMean value, i.e. solving
Figure BSA0000226871330000063
And determining the value of the evaluation threshold value and delta l, if the value is smaller than the evaluation threshold value, feeding back that the load identification effect of the type equipment is poor, and adjusting the load identification effect, otherwise, the load identification effect is normal.
The invention uses pycharm software to build LSTM and Bi-LSTM neural networks by using a TensorFlow2.0 framework under the environment based on python3.7, and realizes a classified evaluation feedback mechanism, 120 groups of user text data are used in the experiment, 800 groups of equipment sample sets are obtained by sentence division processing, a training set is set to be 70%, namely 560 groups, and the specific experiment result is shown in figure 7.
The evaluation analysis effect of the classified evaluation text is as described above, and the actual load recognition is as shown in fig. 8. The test result can be obtained by performing text test on the actual operation result of the six randomly selected devices, and the evaluation feedback mechanism can realize the feedback of abnormity and deficiency and realize the detection of the load identification system by analyzing the text evaluation result and the classification effect of the evaluation text.
The above description is only a preferred embodiment of the present invention, and all changes and modifications that come within the scope of the invention as defined by the appended claims fall within the scope of the invention.

Claims (2)

1. A classification evaluation feedback algorithm for power grid load identification is characterized by comprising the following steps:
step 1: and (5) processing the user corpus data. The integral text is processed into a single-sentence load evaluation short text form, so that model operation noise is reduced;
step 2: and modeling the corpus data by classification evaluation analysis. Building an evaluation classification model, performing model training according to existing data and realizing load evaluation text classification; establishing an evaluation analysis model to carry out error analysis on the corpus according to an evaluation feedback algorithm;
and step 3: and establishing a classified evaluation feedback mechanism according to the two feedback results. And judging the load identification problem type according to a feedback mechanism on the result, and realizing the parameter updating of the equipment library or the non-invasive load identification algorithm.
2. The grid load identification-oriented classification evaluation feedback algorithm according to claim 1, characterized in that: the algorithm is based on user power consumption list evaluation content, a classification evaluation feedback model is established by combining a deep learning algorithm, evaluation texts are classified and subjected to emotion analysis, and then an evaluation feedback mechanism is established according to actual analysis results to provide adjustment feedback for load identification. The method and the device realize accurate differentiation of user experience, provide adjustment suggestions in a targeted manner, and have positive effects and instructive suggestions on non-invasive load identification of the power grid.
The main functions realized are as follows:
first, the overall evaluation data in the user evaluation data is processed. And (3) carrying out processing such as special symbol deletion, common word filtering, sentence division and the like on the text, and processing the whole text into a form of a load evaluation short text of a single sentence, thereby reducing model operation noise. Then the preprocessed data is sent to a classification evaluation analysis part for text classification and analysis.
And secondly, respectively building an evaluation classification model based on the LSTM and an evaluation analysis algorithm based on the Bi-LSTM to perform text processing. Performing model training according to the existing and manually labeled load evaluation texts of various types of equipment, and then performing model classification on the processed user evaluation texts to be classified to obtain class vectors; and simultaneously, analyzing the load evaluation text obtained after the corpus is preprocessed by using an evaluation analysis algorithm to obtain the identification error feedback condition of each device. Then the two feedback results are sent to an evaluation feedback mechanism part for processing.
And finally, establishing an evaluation feedback mechanism for judging the type of the problem caused by equipment identification by the non-invasive load identification algorithm. And judging whether the algorithm identification problem belongs to the defect of the novel equipment category or not according to the two operation results or whether the existing load identification algorithm model parameters need to be updated. For the two effects of the analysis result, feedback is performed respectively.
(1) And a corpus preprocessing module. Designing a data reading processing program, taking a user evaluation list and the equipment type as algorithm input, generating a regularized short text after processing, and storing the regularized short text in a database.
(2) And a classification evaluation analysis module. And the classification evaluation feedback model respectively reads and trains the text, realizes text classification and evaluation analysis, generates a class vector and an evaluation analysis result, and stores the class vector and the evaluation analysis result in a database.
(3) The feedback mechanism is evaluated. The algorithm processes the category vector data and the evaluation analysis result respectively. And (4) reading and analyzing the data of the two processing results, analyzing and processing according to a set threshold, finally predicting the problems of the existing non-invasive load identification algorithm, and giving a reasonable and standardized suggestion.
CN202011416235.0A 2020-12-08 2020-12-08 Classification evaluation feedback method for load identification of smart power grid Pending CN114625831A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011416235.0A CN114625831A (en) 2020-12-08 2020-12-08 Classification evaluation feedback method for load identification of smart power grid

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011416235.0A CN114625831A (en) 2020-12-08 2020-12-08 Classification evaluation feedback method for load identification of smart power grid

Publications (1)

Publication Number Publication Date
CN114625831A true CN114625831A (en) 2022-06-14

Family

ID=81895794

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011416235.0A Pending CN114625831A (en) 2020-12-08 2020-12-08 Classification evaluation feedback method for load identification of smart power grid

Country Status (1)

Country Link
CN (1) CN114625831A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117828030A (en) * 2024-03-01 2024-04-05 微网优联科技(成都)有限公司 User analysis method based on big data and electronic equipment

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117828030A (en) * 2024-03-01 2024-04-05 微网优联科技(成都)有限公司 User analysis method based on big data and electronic equipment
CN117828030B (en) * 2024-03-01 2024-05-07 微网优联科技(成都)有限公司 User analysis method based on big data and electronic equipment

Similar Documents

Publication Publication Date Title
Shen et al. Wind speed prediction of unmanned sailboat based on CNN and LSTM hybrid neural network
CN111832647A (en) Abnormal flow detection system and method
CN113283027B (en) Mechanical fault diagnosis method based on knowledge graph and graph neural network
CN110751318B (en) Ultra-short-term power load prediction method based on IPSO-LSTM
CN111651593A (en) Text emotion analysis method based on word vector and word vector mixed model
CN111814956B (en) Multi-task learning air quality prediction method based on multi-dimensional secondary feature extraction
CN110415071B (en) Automobile competitive product comparison method based on viewpoint mining analysis
CN111309909B (en) Text emotion classification method based on hybrid model
CN111237988A (en) Control method and system for subway vehicle-mounted air conditioning unit
CN112732921B (en) False user comment detection method and system
CN109214444B (en) Game anti-addiction determination system and method based on twin neural network and GMM
Wu et al. Optimized deep learning framework for water distribution data-driven modeling
Chu et al. Co-training based on semi-supervised ensemble classification approach for multi-label data stream
CN113298318A (en) Novel overload prediction method for distribution transformer
CN112101473A (en) Smoke detection algorithm based on small sample learning
CN115564194A (en) Method and system for constructing metering abnormality diagnosis information generation model of smart power grid
CN114692623A (en) Emotion analysis method for environment network public sentiment
CN114625831A (en) Classification evaluation feedback method for load identification of smart power grid
CN111723127A (en) Stock trend prediction method and system based on text abstract emotion mining
Zhang et al. Credit risk control algorithm based on stacking ensemble learning
CN113762591B (en) Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning
CN115759343A (en) E-LSTM-based user electric quantity prediction method and device
CN115577111A (en) Text classification method based on self-attention mechanism
CN112598186B (en) Improved LSTM-MLP-based small generator fault prediction method
CN112348275A (en) Regional ecological environment change prediction method based on online incremental learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication