CN116881811A - Platform region heavy overload prediction method, model training method, device and computer equipment - Google Patents
Platform region heavy overload prediction method, model training method, device and computer equipment Download PDFInfo
- Publication number
- CN116881811A CN116881811A CN202310595201.XA CN202310595201A CN116881811A CN 116881811 A CN116881811 A CN 116881811A CN 202310595201 A CN202310595201 A CN 202310595201A CN 116881811 A CN116881811 A CN 116881811A
- Authority
- CN
- China
- Prior art keywords
- data set
- training
- platform region
- heavy overload
- feature
- 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
Links
- 238000012549 training Methods 0.000 title claims abstract description 208
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000012360 testing method Methods 0.000 claims abstract description 100
- 230000002159 abnormal effect Effects 0.000 claims abstract description 52
- 238000004458 analytical method Methods 0.000 claims abstract description 32
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 32
- 238000000605 extraction Methods 0.000 claims abstract description 29
- 238000012937 correction Methods 0.000 claims abstract description 16
- 230000003014 reinforcing effect Effects 0.000 claims abstract description 6
- 238000004590 computer program Methods 0.000 claims description 15
- 230000002708 enhancing effect Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 238000005516 engineering process Methods 0.000 description 10
- 230000008859 change Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 5
- 239000004973 liquid crystal related substance Substances 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000005611 electricity Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000013210 evaluation model Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000005856 abnormality Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012946 outsourcing Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 230000003442 weekly effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The application relates to a method for predicting heavy overload of a platform area, a model training method, a device, computer equipment and a storage medium. The method comprises the following steps: acquiring a reference data set of a station area; the area reference data set comprises a limit data set and an influence factor data set of the area in a specified time period; abnormal data identification and correction are carried out on the reference data set of the platform region by adopting an isolated forest algorithm, and a training data set and a testing data set are obtained; performing feature analysis and feature extraction on the training data set and the test data set to obtain a training feature data set and a test feature data set, and adding an influence factor data set into the training feature data set and the test feature data set; wherein the training feature dataset comprises negative samples; and reinforcing the negative samples in the training characteristic data set, and constructing a platform region heavy overload prediction model according to the reinforced training characteristic data set. By adopting the method, the accuracy of the heavy overload prediction of the platform region can be improved.
Description
Technical Field
The present application relates to the field of a platform region heavy overload prediction technology, and in particular, to a platform region heavy overload prediction method, a model training method, a device, a computer device, and a storage medium.
Background
With the continuous expansion of the power grid scale, the rapid growth of social economy and the continuous improvement of the living standard of residents, the power consumption demands of people are increasing. The distribution transformer area is used as the terminal power supply unit facing the low-voltage user, and the running state of the power supply equipment in the distribution transformer area directly influences the power supply quality of the distribution transformer area. One of the main causes of a fault outage is the occurrence of heavy overload operation of the equipment. The heavy overload of the transformer can greatly influence the electricity utilization safety, accelerate the equipment loss, reduce the service life of the equipment and bring greater risk to the safety of the power distribution network; especially summer, winter and holidays, often overload the circuit.
The distribution network is an important link in the power transmission and transformation and distribution links, and overload operation of a circuit can cause overload and even faults of a distribution network transformer, so that more serious accidents are caused. The complaint rate of the electricity customers is high, and huge economic losses can be caused. Therefore, the reason for heavy overload of the distribution network transformer needs to be found through a scientific early warning method, the probability of heavy overload of the distribution network transformer is predicted, and the distribution conveying system is adjusted in advance, so that the ever-increasing power demands of people are met.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a method, a model training method, an apparatus, a computer device, and a storage medium for predicting heavy overload of a region, which can improve the accuracy of predicting heavy overload of the region.
In a first aspect, the present application provides a training method for a heavy overload prediction model of a platform. The method comprises the following steps:
acquiring a reference data set of a station area; the area reference data set comprises a limit data set and an influence factor data set of the area in a specified time period;
abnormal data identification and correction are carried out on the reference data set of the platform region by adopting an isolated forest algorithm, and a training data set and a testing data set are obtained;
performing feature analysis and feature extraction on the training data set and the test data set to obtain a training feature data set and a test feature data set, and adding an influence factor data set into the training feature data set and the test feature data set; wherein the training feature dataset comprises negative samples;
and reinforcing the negative samples in the training characteristic data set, and constructing a platform region heavy overload prediction model according to the reinforced training characteristic data set.
In one embodiment, performing abnormal data identification and correction on the reference data set of the platform area by adopting an isolated forest algorithm, and acquiring a training data set and a test data set comprises:
Randomly selecting a data sample from the platform region reference data set, acquiring a platform region reference data subset, and placing the platform region reference data subset into a root node of an isolated tree;
randomly selecting target features from the reference data subsets of the platform region, and randomly selecting a threshold under the data value of the features of the current node;
dividing the current node data space into two subspaces according to a threshold value, placing a point smaller than the threshold value under the current characteristic in the left branch of the current node, and placing a point larger than or equal to the threshold value in the right branch of the current node;
repeating the steps to continuously construct new leaf nodes, and generating a target isolated tree when only one data or isolated tree on the leaf nodes has grown to a specified height;
repeating the step of generating the target isolated tree to obtain a target isolated tree set; obtaining an isolated tree anomaly score according to a target isolated tree set;
judging abnormal data according to the abnormal score, and acquiring an abnormal data set;
and correcting the abnormal data set, dividing the corrected reference data set of the district according to a specified proportion, and obtaining a training data set and a test data set.
In one embodiment, performing feature analysis and feature extraction on the training data set and the test data set, obtaining the training feature data set and the test feature data set, and adding the influencing factor data set to the training feature data set and the test feature data set includes:
And respectively carrying out feature analysis and feature extraction on the maximum load of the area in the training data set and the testing data set by utilizing a decomposable time sequence model NeuralProphet to obtain a training feature data set and a testing feature data set, and adding an influence factor data set into the training feature data set and the testing feature data set, wherein the influence factor data set comprises the highest daily temperature.
In one embodiment, enhancing the negative samples in the training feature data set and constructing the platform region heavy overload prediction model according to the enhanced training feature data set includes:
utilizing Gaussian noise technology to enhance negative samples in the training characteristic data set;
and training a target algorithm according to the enhanced training characteristic data set, and generating a platform region heavy overload prediction model.
In one embodiment, generating the platform region heavy overload prediction model according to the training target algorithm of the enhanced training feature data set comprises:
initializing weight distribution of the enhanced training feature data set;
the target algorithm learns by using a training feature data set with weight distribution to obtain a basic classifier;
acquiring a classification error rate and a coefficient of a basic classifier on a training feature data set with weight distribution, and updating the weight distribution of the training feature data set;
And linearly combining the basic classifiers to obtain a platform region heavy overload prediction model.
In a second aspect, the present application also provides a method for predicting a heavy overload of a region, the method using the model for predicting heavy overload of a region as provided in the first aspect, the method comprising:
acquiring a test characteristic data set and calling a platform region heavy overload prediction model;
and inputting the test characteristic data set into a platform region heavy overload prediction model to obtain a platform region heavy overload prediction value.
In a third aspect, the application further provides a training device of the platform region heavy overload prediction model. The device comprises:
the platform region reference data set acquisition module is used for acquiring a platform region reference data set; the area reference data set comprises a limit data set and an influence factor data set of the area in a specified time period;
the abnormal data identification and correction module is used for carrying out abnormal data identification and correction on the reference data set of the platform region by adopting an isolated forest algorithm to obtain a training data set and a test data set;
the feature analysis and feature extraction module is used for carrying out feature analysis and feature extraction on the training data set and the test data set, obtaining the training feature data set and the test feature data set, and adding an influence factor data set into the training feature data set and the test feature data set; wherein the training feature dataset comprises negative samples;
The platform region heavy overload prediction model construction module is used for reinforcing negative samples in the training characteristic data set and constructing a platform region heavy overload prediction model according to the reinforced training characteristic data set.
In a fourth aspect, the application also provides a device for predicting the heavy overload of the station area. The device comprises:
the platform region heavy overload prediction model calling module is used for acquiring the test characteristic data set and calling the platform region heavy overload prediction model;
the platform region heavy overload prediction value acquisition module is used for inputting the test characteristic data set into the platform region heavy overload prediction model to obtain the platform region heavy overload prediction value.
In a fifth aspect, the present application also provides a computer device. The computer equipment comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the step of the training method of the platform region heavy overload prediction model when executing the computer program; or, the step of implementing the above-mentioned method for predicting the heavy overload of the area.
In a sixth aspect, the present application also provides a computer readable storage medium. The computer readable storage medium is stored with a computer program, and the computer program realizes the steps of the training method of the platform region heavy overload prediction model when being executed by a processor; or, the step of implementing the above-mentioned method for predicting the heavy overload of the area.
The method, the device, the computer equipment and the storage medium for predicting the heavy overload of the area are characterized by acquiring an area reference data set, wherein the area reference data set comprises a limit data set and an influence factor data set of the area in a specified time period; abnormal data identification and correction are carried out on the reference data set of the platform region by adopting an isolated forest algorithm, and a training data set and a testing data set are obtained; and then carrying out feature analysis and feature extraction on the training data set and the test data set to obtain the training feature data set and the test feature data set, and adding an influence factor data set into the training feature data set and the test feature data set, wherein the training feature data set comprises a negative sample. The method can extract the change characteristics of the load from the data, so that the characteristics of the data are fully mined. And the feature extraction mode is more objective, and the influence of priori experience on analysis modeling is reduced. The method further enhances the negative samples in the training characteristic data set, and builds a platform region heavy overload prediction model according to the enhanced training characteristic data set. And considering the property of the data, selecting a time sequence modeling mode to be used, and conforming to the characteristics of the data. And acquiring a test characteristic data set, calling a platform region heavy overload prediction model, and inputting the test characteristic data set into the platform region heavy overload prediction model to obtain a platform region heavy overload prediction value. The selected index considers the actual condition requirement, and penalizes the condition of insufficient heavy overload early warning.
Drawings
FIG. 1 is a flow chart of a training method of a region heavy overload prediction model in one embodiment;
FIG. 2 is a block diagram illustrating a method for predicting heavy overload of a cell in one embodiment;
fig. 3 is a flowchart of another embodiment of a method for predicting heavy overload of a cell;
FIG. 4 is a schematic diagram of an apparatus for training a region heavy overload prediction model in one embodiment;
fig. 5 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
With the continuous expansion of the power grid scale, the rapid growth of social economy and the continuous improvement of the living standard of residents, the power consumption demands of people are increasing. As a distribution transformer area of the endmost power supply unit facing the low-voltage user, the running state of the power supply equipment in the distribution transformer area directly influences the power supply quality of the distribution transformer area. One of the main causes of a fault outage is the occurrence of heavy overload operation of the equipment. The heavy overload of the transformer can greatly influence the electricity utilization safety, accelerate the equipment loss, reduce the service life of the equipment, bring great risks to the safety of the power distribution network, and particularly lead to overload operation of the circuit in summer, winter and holidays.
The distribution network is an important link in the power transmission and transformation and distribution links, overload operation of a circuit can cause overload and even faults of a distribution network transformer, and then serious accidents are caused, so that the complaint rate of electricity customers is high, and huge economic loss is caused. Therefore, how to find out the reason of heavy overload of the distribution network transformer through a scientific early warning method, predict the probability of heavy overload of the distribution network transformer, adjust the distribution transmission system in advance, and meet the ever-increasing power demands of people has extremely important significance.
However, the conventional area heavy overload prediction technology has the following problems:
firstly, feature selection is subjective and does not start from data. In the traditional technology, the influence factors of the heavy overload of the transformer area are selected by historical experience or expert analysis to obtain a plurality of influence factors of the running data of the transformer area, wherein the influence factors comprise six types of equipment ledgers, meteorological data, low-voltage topology, holiday data, distribution transformer transformation records and user files. And using an analysis method of nonlinear correlation to select an influence factor with a larger correlation coefficient as a heavy overload influence factor of the platform region. However, the data of the platform region operation has periodicity and data characteristics of years, weeks and days, and the service-based characteristic selection method does not fully mine the characteristics of the data.
And secondly, modeling of the time sequence is not characterized by the time sequence. Predicting whether heavy overload will occur in the future using historical data is itself a time series problem. That is to say, a certain development rule is analyzed from the historical data. BP (Back Propagation) neural network is used as a multi-layer feedforward neural network, has strong nonlinear mapping capability and flexible network structure, and can be used for modeling time series data. However, the conventional technology only refers to the influence factor of the heavy overload of the area when modeling by using the neural network, but does not consider the area history data of the first n points of the predicted data as input, and does not embody the characteristics of time series data.
And finally, selecting the evaluation index which does not accord with the objective condition. Whether heavy overload occurs is a 0-1 classification problem, but for whether heavy overload occurs, the damage of a platform area which is not predicted to occur heavy overload is far greater than that of a platform area which is not predicted to occur heavy overload, so that the prediction accuracy rate is not in accordance with the actual condition requirement when the evaluation index selection is performed.
The application provides a method for predicting heavy overload of a platform area, a model training method, a device, computer equipment and a storage medium, which can improve the accuracy of the heavy overload prediction of the platform area.
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
In one embodiment, as shown in fig. 1, a training method of a platform heavy overload prediction model is provided, and this embodiment is applied to a terminal for illustration by using the method, it can be understood that the method can also be applied to a server, and can also be applied to a system including the terminal and the server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the steps of:
102, obtaining a reference data set of a platform area; the zone reference data set includes a limit data set and an influence factor data set of the zone for a specified period of time.
The limit data set comprises operation data such as daily maximum load, maximum load rate, occurrence time and the like, and the influence factor data set comprises climate data such as temperature, humidity, wind speed and the like of corresponding days.
Optionally, operational data such as daily maximum load, maximum load rate, occurrence time and the like of a certain area in the past five years are obtained, and climate data such as temperature, humidity, wind speed and the like of a corresponding day are obtained.
And 104, carrying out abnormal data identification and correction on the reference data set of the platform region by adopting an isolated forest algorithm to obtain a training data set and a test data set.
In one embodiment, isolated forests are utilized to identify data outliers in the reference dataset of the region. Specifically, the basic idea of an isolated forest is to cut the data space continuously using a random hyperplane, each time a cut results in two subspaces, and the subspaces are continuously circularly cut until each subspace contains only one data point. It will be appreciated that for denser areas, cutting a data point into separate subspaces requires a greater number of cuts; for relatively sparse regions, cutting one data point into separate subspaces requires a relatively small number of cuts.
Step 106, carrying out feature analysis and feature extraction on the training data set and the test data set to obtain a training feature data set and a test feature data set, and adding an influence factor data set into the training feature data set and the test feature data set; wherein the training feature data set comprises a negative sample.
In one embodiment, neuralpopset is selected for feature analysis and extraction. Neuralpopset is a neural network-based time series model, the core concept being the combinability of its modules. The model is composed of a plurality of modules of trend terms, period terms, autoregressive terms, etc., each contributing an additional component to the prediction, and most of the components can also be configured to scale with trends to obtain multiplicative effects. Each module has independent inputs and modeling processes, and each module must produce h outputs, where h defines the number of steps in one prediction future.
Specifically, the mathematical model of the neuroalpephet prediction is shown in formula (1):
wherein, the liquid crystal display device comprises a liquid crystal display device,is a predicted value of the time series; t (T) represents the trend of the time series; s (t) represents that the time series is affected by seasons; e (t) represents that the time series is affected by holidays and other times; f (t) represents the regression effect of the time series on future known outsourcing variables; a (t) represents an autoregressive effect of the time series based on past observations; l (t) represents the regression effect of the time series exogenous variable hysteresis observation. All of the model component modules described above may be individually configured and combined to form a model. By default, the neuroalpyphet activates only the trend and seasonal modules.
In one embodiment, the training feature data set includes positive and negative samples. Specifically, the training feature data set of the date on which heavy overload occurs is set as a negative sample, and the training feature data set of the date on which normal operation occurs is set as a positive sample.
And step 108, enhancing the negative samples in the training characteristic data set, and constructing a platform region heavy overload prediction model according to the enhanced training characteristic data set.
In one embodiment, the negative samples in the training set are enhanced using gaussian noise techniques to obtain an enhanced training feature data set. Further, based on a classification frame corresponding to an Adaboost (Adaptive boosting, adaptive enhancement) algorithm trained by positive samples and negative samples in the training feature data set after sample enhancement, a state evaluation model of whether heavy overload occurs in the platform region is obtained, namely a platform region heavy overload prediction model.
According to the training method of the platform region heavy overload prediction model, the platform region reference data set is obtained, the platform region reference data set comprises the limit data set and the influence factor data set of the platform region in a specified time period, the abnormal data identification and correction are carried out on the platform region reference data set by adopting an isolated forest algorithm, and the training data set and the test data set are obtained. And carrying out feature analysis and feature extraction on the training data set and the test data set to obtain the training feature data set and the test feature data set, and adding the influencing factor data set into the training feature data set and the test feature data set, wherein the training feature data set comprises a negative sample. And extracting the change characteristics of the load from the data, and fully mining the characteristics of the data. The feature extraction mode is more objective, and the influence of priori experience on analysis modeling is reduced. And reinforcing the negative samples in the training characteristic data set, and constructing a platform region heavy overload prediction model according to the reinforced training characteristic data set. And considering the property of the data, selecting a time sequence modeling mode to be used, and conforming to the characteristics of the data.
In one embodiment, performing abnormal data identification and correction on the reference data set of the platform area by adopting an isolated forest algorithm, and acquiring a training data set and a test data set comprises:
randomly selecting a data sample from the platform region reference data set, acquiring a platform region reference data subset, and placing the platform region reference data subset into a root node of an isolated tree; randomly selecting target features from the reference data subsets of the platform region, and randomly selecting a threshold under the data value of the features of the current node; dividing the current node data space into two subspaces according to a threshold value, placing a point smaller than the threshold value under the current characteristic in the left branch of the current node, and placing a point larger than or equal to the threshold value in the right branch of the current node; repeating the steps to continuously construct new leaf nodes, and generating a target isolated tree when only one data or isolated tree on the leaf nodes has grown to a specified height; repeating the step of generating the target isolated tree to obtain a target isolated tree set; obtaining an isolated tree anomaly score according to a target isolated tree set; judging abnormal data according to the abnormal score, and acquiring an abnormal data set; and correcting the abnormal data set, dividing the corrected reference data set of the district according to a specified proportion, and obtaining a training data set and a test data set.
In one embodiment, the step of constructing an isolated forest is generally divided into two steps of constructing t isolated trees and constructing the t isolated trees into an isolated forest, and the step of constructing a single tree, namely, an isolated tree, includes:
n data samples are randomly selected from the region reference data set to serve as a region reference data subset, and then the n data samples are placed into a root node of an isolated tree. And randomly selecting a certain characteristic from the given data, namely the target characteristic, and randomly selecting a threshold value p under the data value of the characteristic of the current node. And dividing the current node data space into two subspaces according to the value of the threshold value p, placing a point smaller than p under the current characteristic in the left branch of the current node, and placing a point larger than or equal to p in the right branch of the current node. And each branch node repeatedly constructs new leaf nodes according to the steps until only one data or isolated tree grows to a set height on the leaf nodes, namely, a target isolated tree is generated, wherein the set height is the designated height.
Further, repeating the step of generating the target isolated tree to generate t isolated trees, namely the target isolated tree set. The degree of abnormality, i.e., the statistical abnormality score S, was evaluated with the generated t isolated trees. The results of the individual isolated trees are comprehensively counted for each sample x, and an anomaly score is estimated, wherein the estimated anomaly score is shown in a formula (2) and a formula (3):
Where E [ h (X) ] is the sample's path length h (X) expected at each orphan tree; h (X) is the path length of sample X in each tree; c (m) is the height value normalization of the tree, where H (k) =nk+δ, δ is the euler constant.
Further, abnormal data is judged according to the abnormal score, and an abnormal data set is obtained. Specifically, if the anomaly score is close to 1, it indicates that the anomaly suspicion is large; if the anomaly scores of all points are about 0.5, the anomaly points possibly do not exist in the data set; if the anomaly score is close to 0, no anomaly suspicion is indicated.
Further, correction is performed on data outliers and missing values in the outlier dataset identified by the isolated forest. Specifically, by filling the abnormal data set with the average value for correction, a specific filling formula is shown as formula (4):
where y (t, d) is the data of the d-th class data at the time t, q represents the time when the value is null or abnormal, p is the time when the value is not null after continuous deletion, and when only 1 value is continuously deleted, p=q+1.
Further, the corrected site reference data set is divided into a training data set and a test data set according to a ratio of 8:2.
In this embodiment, the data sample is randomly selected from the area reference data set, so as to obtain the area reference data subset, and the area reference data subset is put into the root node of an isolated tree. Randomly selecting target features from the reference data subsets of the platform region, randomly selecting a threshold value under the data value of the feature where the current node is located, dividing the current node data space into two subspaces according to the threshold value, placing the point smaller than the threshold value under the current feature on the left branch of the current node, and placing the point larger than or equal to the threshold value on the right branch of the current node. And repeating the steps to continuously construct new leaf nodes, and generating a target isolated tree when only one data or isolated tree on the leaf nodes has grown to a specified height. And repeating the step of generating the target isolated tree to obtain a target isolated tree set. Obtaining an isolated tree anomaly score according to a target isolated tree set; and judging the abnormal data according to the abnormal score, and acquiring an abnormal data set. And correcting the abnormal data set, dividing the corrected reference data set of the district according to a specified proportion, and obtaining a training data set and a test data set. Abnormal data is detected through an isolated forest algorithm, so that the abnormal data can be screened out more simply and efficiently, and the efficiency of constructing the platform region heavy overload prediction model is improved. And filling and correcting the abnormal data set through the average value to generate a training data set and a test data set, so as to prepare for constructing the area heavy overload prediction model and calling the area heavy overload prediction model to implement the area heavy overload prediction.
In one embodiment, performing feature analysis and feature extraction on the training data set and the test data set, obtaining the training feature data set and the test feature data set, and adding the influencing factor data set to the training feature data set and the test feature data set includes:
and respectively carrying out feature analysis and feature extraction on the maximum load of the area in the training data set and the testing data set by utilizing a decomposable time sequence model NeuralProphet to obtain a training feature data set and a testing feature data set, and adding an influence factor data set into the training feature data set and the testing feature data set, wherein the influence factor data set comprises the highest daily temperature.
Specifically, feature analysis and extraction are respectively carried out on the maximum load of the platform region in the training data set and the testing data set by utilizing the NeuralProphet, the features comprise overall trend, annual seasonality, weekly seasonality, holidays and the like, a training feature data set A and a testing feature data set B are obtained, and influence factors such as the highest daily temperature and the like are added.
In this embodiment, the feature analysis and feature extraction are performed by using the resolvable time series model neuroalpyphet on the maximum load of the area in the training data set and the test data set respectively, so that the feature extraction mode is more objective, and the influence of the priori experience on analysis modeling is reduced. A training feature data set and a test feature data set are obtained, and an influence factor data set is added into the training feature data set and the test feature data set, wherein the influence factor data set comprises the highest daily temperature. The load change characteristics are extracted from the data, so that the characteristics of the data can be fully mined.
In one embodiment, enhancing the negative samples in the training feature data set and constructing the platform region heavy overload prediction model according to the enhanced training feature data set includes:
utilizing Gaussian noise technology to enhance negative samples in the training characteristic data set; and training a target algorithm according to the enhanced training characteristic data set, and generating a platform region heavy overload prediction model.
In one embodiment, the negative samples in the training feature data set a are enhanced by using a gaussian noise technique, so as to obtain an enhanced training feature data set a', and the enhancement method is as shown in formula (5):
f aug (x)＝λ*max(f(x))*random() (5)
wherein f aug () Representing the enhanced negative samples; f (x) represents the negative sample that needs enhancement; maxf () represents the maximum value of the negative sample; λ represents the proportion of sample enhancement (e.g., negative samples are to be enhanced 5-fold, 7-fold); random () represents a random number generation function in which random numbers obey a gaussian distribution of 0-1.
Further, training a classification frame corresponding to an Adaboost algorithm based on positive samples and negative samples in the sample enhanced training feature data set A' to obtain a state evaluation model of whether heavy overload occurs in the platform region, namely a platform region heavy overload prediction model.
In this embodiment, a negative sample in the training feature data set is enhanced by using a gaussian noise technique, and a platform region heavy overload prediction model is generated according to a training target algorithm of the enhanced training feature data set. Considering that the heavy overload condition of the platform area is generally less than that of normal operation, the number of negative samples is relatively less, and more abundant information in the negative samples can be obtained after the training characteristic data set is enhanced by the Gaussian noise technology.
In one embodiment, generating the platform region heavy overload prediction model according to the training target algorithm of the enhanced training feature data set comprises:
initializing weight distribution of the enhanced training feature data set; the target algorithm learns by using a training feature data set with weight distribution to obtain a basic classifier; acquiring a classification error rate and a coefficient of a basic classifier on a training feature data set with weight distribution, and updating the weight distribution of the training feature data set; and linearly combining the basic classifiers to obtain a platform region heavy overload prediction model.
In one embodiment, training the classification frames corresponding to the Adaboost algorithm based on positive and negative samples in the sample-enhanced training feature dataset to obtain a state evaluation model of whether heavy overload occurs in the area. Specifically, firstly, initializing the weight distribution of the enhanced training feature data set, wherein the weight distribution of the initialized training feature data set is shown in a formula (6):
wherein i=1, 2, …, N is the number of samples, ω 1i The samples are given weights.
Further, for m=1, 2, …, M, the following steps are cycled:
using a distribution D with weights m Is learned by the training feature data set to obtain a basic classifier G m () Recalculating G m () Classification error rate e on training feature data set m The calculation method is shown in the formula (7):
wherein omega mi The samples are given weights.
Further, a basic classifier G is calculated m () Coefficient alpha of (2) m The calculation method is shown in the formula (8):
wherein e m Is the classification error rate of the basic classifier on the training feature data set.
Further, the weight distribution of the training feature data set is updated according to the classification error rate and the coefficient, and the updating method is as shown in the formula (9) and the formula (10):
D m+1 ＝(ω m+1,1 ,…,ω m+1,i ,…ω m+1,N ) (9)
wherein, the liquid crystal display device comprises a liquid crystal display device,ω mi weights, alpha, assigned to samples m Is a coefficient of the basic classifier.
Further, constructing a linear combination of the basic classifier, wherein the construction method is shown in a formula (11):
wherein f (x) is a linear combination function of the basic classifier, G m () As a basic classifier, alpha m Is a coefficient of the basic classifier.
Further, a final classifier G (x) is obtained, and the final classifier G (x) is shown in formula (12):
G(x)＝sign(f(x)) (12)
specifically, the final classifier G (x) is a region heavy overload prediction model.
In this embodiment, the weight distribution of the enhanced training feature data set is initialized. The target algorithm learns using the training feature data set with the weight distribution to obtain the basic classifier. And obtaining the classification error rate and the coefficient of the basic classifier on the training characteristic data set with the weight distribution, and updating the weight distribution of the training characteristic data set. And linearly combining the basic classifiers to obtain a platform region heavy overload prediction model. The characteristics of the enhanced training characteristic data set and the Adaboost algorithm are fully utilized, and a model capable of improving the accuracy of the heavy overload prediction of the platform region is constructed.
In one embodiment, a method for predicting a heavy overload of a region is provided, the method using a heavy overload prediction model of the region, the method comprising:
acquiring a test characteristic data set and calling a platform region heavy overload prediction model; and inputting the test characteristic data set into a platform region heavy overload prediction model to obtain a platform region heavy overload prediction value.
In one embodiment, the test feature data set B is input into a region heavy overload prediction model to predict, and output predicts whether heavy overload will occur for a period of time in the future for the region. Specifically, the recall rate is selected as an evaluation index, that is, the correct prediction is positive, and the actual proportion of the recall rate to the total recall rate is positive, and the specific calculation mode is shown in a formula (13):
in this embodiment, a test feature dataset is obtained, and a region heavy overload prediction model is invoked. And inputting the test characteristic data set into a platform region heavy overload prediction model to obtain a platform region heavy overload prediction value. Since in practical situations the hazard of underestimating heavy overload is larger than overestimating heavy overload, in this embodiment the recall is chosen as an evaluation index, a high recall means that there may be more false detections, but it is best to find every object that should be found. The selected index in the embodiment considers the actual situation requirement, and can punish the insufficient heavy overload early warning condition.
In one embodiment, as shown in fig. 2, a method for predicting a heavy overload of a cell is provided. Firstly, the data of the platform area is obtained as a platform area reference data set, wherein the data of the platform area comprise daily maximum load, maximum load rate, maximum load occurrence time and influencing factors, and the influencing factors comprise temperature, humidity, wind speed and the like. And detecting data abnormal points in the reference data set of the platform region by using the isolated forest, correcting the missing values and the data abnormal points, and dividing the training data set and the test data set according to the proportion of 8:2. And then, adopting a Neuralpopset model to perform feature analysis extraction on the training data set and the test data set to obtain a training feature data set A and a test feature data set B, and adding influencing factors such as temperature. The date on which heavy overload occurred was set as negative sample and normal operation was set as positive sample. Considering that the heavy overload condition of the platform area is generally less than that of normal operation, the number of negative samples is relatively small, so that the negative samples in the training characteristic data set are enhanced by using the Gaussian noise technology, and the enhanced training characteristic data set A' is obtained. And (3) putting the training characteristic data set A' into an Adaboost model for classification training to obtain a platform region heavy overload prediction model. And predicting the test characteristic data set by using the model to obtain an output result. And finally, selecting the recall rate as an evaluation index to judge the quality of the model.
In another embodiment, as shown in fig. 3, there is provided a method for predicting a heavy overload of a station, the method including the steps of:
step 302, obtaining a reference data set of a platform area; the zone reference data set includes a limit data set and an influence factor data set of the zone for a specified period of time.
Step 304, randomly selecting data samples from the platform region reference data set, obtaining a platform region reference data subset, and placing the platform region reference data subset into a root node of an isolated tree.
And 306, randomly selecting target features from the reference data subset of the area, and randomly selecting a threshold value under the data value of the features where the current node is located.
Step 308, according to the threshold value, the current node data space is divided into two subspaces, the point with the characteristic smaller than the threshold value is placed on the left branch of the current node, and the point with the characteristic larger than or equal to the threshold value is placed on the right branch of the current node.
Step 310, repeating the above steps, continuously constructing new leaf nodes, and generating a target orphan tree when only one data or orphan tree on the leaf nodes has grown to a specified height.
Step 312, repeating the step of generating the target isolated tree to obtain a target isolated tree set; and obtaining the isolated tree anomaly score according to the target isolated tree set.
And step 314, judging the abnormal data according to the abnormal score, and acquiring an abnormal data set.
And step 316, correcting the abnormal data set, dividing the corrected reference data set of the district according to a specified proportion, and obtaining a training data set and a test data set.
Step 318, performing feature analysis and feature extraction on the maximum load of the platform region in the training data set and the test data set by using a decomposable time sequence model NeuralProphet respectively to obtain a training feature data set and a test feature data set, and adding an influence factor data set into the training feature data set and the test feature data set, wherein the influence factor data set comprises the highest daily temperature; wherein the training feature data set comprises a negative sample.
Step 320, negative samples in the training feature dataset are enhanced using gaussian noise techniques.
Step 322, initializing the weight distribution of the enhanced training feature data set.
In step 324, the target algorithm learns using the training feature data set with the weight distribution to obtain the base classifier.
Step 326, obtaining the classification error rate and coefficient of the basic classifier on the training feature data set with weight distribution, and updating the weight distribution of the training feature data set.
And 328, linearly combining the basic classifiers to obtain a platform region heavy overload prediction model.
Step 330, a test feature data set is acquired and a platform region heavy overload prediction model is invoked.
And step 332, inputting the test characteristic data set into a platform region heavy overload prediction model to obtain a platform region heavy overload prediction value.
It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.
Based on the same inventive concept, the embodiment of the application also provides a training device for realizing the training method of the district heavy overload prediction model. The implementation scheme of the solution provided by the device is similar to the implementation scheme described in the above method, so the specific limitation in the embodiments of the training device for one or more area heavy overload prediction models provided below may be referred to the limitation of the training method for the area heavy overload prediction model hereinabove, and will not be repeated herein.
In one embodiment, as shown in fig. 4, there is provided a training apparatus of a region heavy overload prediction model, including: a region reference data set acquisition module 402, an abnormal data identification correction module 404, a feature analysis and feature extraction module 406, and a region heavy overload prediction model construction module 408, wherein:
a region reference data set acquisition module 402, configured to acquire a region reference data set; the area reference data set comprises a limit data set and an influence factor data set of the area in a specified time period;
the abnormal data identification and correction module 404 is configured to identify and correct abnormal data of the reference dataset of the platform area by using an isolated forest algorithm, and obtain a training dataset and a test dataset;
The feature analysis and feature extraction module 406 is configured to perform feature analysis and feature extraction on the training data set and the test data set, obtain the training feature data set and the test feature data set, and add an influence factor data set into the training feature data set and the test feature data set; wherein the training feature dataset comprises negative samples;
the platform region heavy overload prediction model construction module 408 is configured to enhance the negative samples in the training feature data set, and construct a platform region heavy overload prediction model according to the enhanced training feature data set.
In one embodiment, the apparatus further comprises:
the system comprises a platform region reference data subset acquisition module, a platform region reference data set acquisition module and a data processing module, wherein the platform region reference data subset acquisition module is used for randomly selecting data samples from a platform region reference data set to acquire the platform region reference data subset, and placing the platform region reference data subset into a root node of an isolated tree;
the threshold selection module is used for randomly selecting target features from the reference data subset of the area and randomly selecting a threshold under the data value of the feature where the current node is located;
the subspace segmentation module is used for segmenting the current node data space into two subspaces according to a threshold value, placing a point smaller than the threshold value under the characteristic of the current node in the left branch of the current node, and placing a point larger than or equal to the threshold value in the right branch of the current node;
The target isolated tree generation module is used for repeating the steps to continuously construct new leaf nodes, and generating a target isolated tree when only one data or isolated tree grows to a specified height on the leaf nodes;
the target isolated tree set acquisition module is used for repeatedly generating the target isolated tree to acquire a target isolated tree set; obtaining an isolated tree anomaly score according to a target isolated tree set;
the abnormal data set acquisition module is used for judging abnormal data according to the abnormal score to acquire an abnormal data set;
the training data set and test data set acquisition module is used for correcting the abnormal data set, dividing the corrected reference data set of the area according to the specified proportion, and acquiring the training data set and the test data set.
In one embodiment, the apparatus further comprises:
and the time sequence model feature analysis and extraction module is used for carrying out feature analysis and feature extraction on the maximum load of the platform region in the training data set and the test data set by utilizing the decomposable time sequence model NeuralProphet respectively to obtain the training feature data set and the test feature data set, and adding an influence factor data set into the training feature data set and the test feature data set, wherein the influence factor data set comprises the highest daily temperature.
In one embodiment, the apparatus further comprises:
the negative sample enhancement module is used for enhancing the negative samples in the training characteristic data set by using Gaussian noise technology;
and the target algorithm training module is used for training a target algorithm according to the enhanced training characteristic data set and generating a platform region heavy overload prediction model.
In one embodiment, the apparatus further comprises:
the weight distribution initializing module is used for initializing the weight distribution of the enhanced training characteristic data set;
the basic classifier acquisition module is used for learning by using the training characteristic data set with weight distribution by a target algorithm to acquire a basic classifier;
the weight distribution updating module is used for acquiring the classification error rate and the coefficient of the basic classifier on the training characteristic data set with weight distribution and updating the weight distribution of the training characteristic data set;
and the basic classifier combination module is used for linearly combining the basic classifiers to obtain a platform region heavy overload prediction model.
All or part of each module in the training device of the platform region heavy overload prediction model can be realized by software, hardware and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, there is provided a district heavy overload prediction apparatus, including:
the platform region heavy overload prediction model calling module is used for acquiring the test characteristic data set and calling the platform region heavy overload prediction model;
the platform region heavy overload prediction value acquisition module is used for inputting the test characteristic data set into the platform region heavy overload prediction model to obtain the platform region heavy overload prediction value.
The above-mentioned various modules in the platform region heavy overload prediction device may be implemented in whole or in part by software, hardware, and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a terminal, and the internal structure of which may be as shown in fig. 5. The computer device includes a processor, a memory, an input/output interface, a communication interface, a display unit, and an input means. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless mode can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program, when executed by a processor, implements a training method of a region heavy overload prediction model and a region heavy overload prediction method. The display unit of the computer device is used for forming a visual picture, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be a key, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by those skilled in the art that the structure shown in FIG. 5 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements may be applied, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.
In an embodiment, there is also provided a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the method embodiments described above when the computer program is executed.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, carries out the steps of the method embodiments described above.
It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related country and region.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples illustrate only a few embodiments of the application and are described in detail herein without thereby limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.
Claims (10)
1. A method for training a region heavy overload prediction model, the method comprising:
acquiring a reference data set of a station area; the platform region reference data set comprises a limit data set and an influence factor data set of the platform region in a specified time period;
abnormal data identification and correction are carried out on the reference data set of the platform region by adopting an isolated forest algorithm, and a training data set and a testing data set are obtained;
Performing feature analysis and feature extraction on the training data set and the test data set to obtain a training feature data set and a test feature data set, and adding an influence factor data set into the training feature data set and the test feature data set; wherein the training feature data set comprises a negative sample;
and reinforcing the negative sample in the training characteristic data set, and constructing a platform region heavy overload prediction model according to the reinforced training characteristic data set.
2. The method of claim 1, wherein the employing an orphan forest algorithm to identify and correct abnormal data for the site reference data set, the obtaining training data set and test data set comprises:
randomly selecting a data sample from a platform region reference data set, acquiring a platform region reference data subset, and placing the platform region reference data subset into a root node of an isolated tree;
randomly selecting target features from the reference data subsets of the platform region, and randomly selecting a threshold value under the data value of the features where the current node is located;
dividing the current node data space into two subspaces according to the threshold value, placing a point smaller than the threshold value under the characteristic of the current node in the left branch of the current node, and placing a point larger than or equal to the threshold value in the right branch of the current node;
Repeating the steps to continuously construct new leaf nodes, and generating a target isolated tree when only one data or isolated tree on the leaf nodes has grown to a specified height;
repeating the step of generating the target isolated tree to obtain a target isolated tree set; obtaining an isolated tree abnormal score according to the target isolated tree set;
judging abnormal data according to the abnormal score, and acquiring an abnormal data set;
and correcting the abnormal data set, dividing the corrected reference data set of the district according to a specified proportion, and obtaining a training data set and a test data set.
3. The method of claim 1, wherein performing feature analysis and feature extraction on the training data set and the test data set to obtain a training feature data set and a test feature data set, and adding an influencing factor data set to the training feature data set and the test feature data set comprises:
and respectively carrying out feature analysis and feature extraction on the maximum load of the platform region in the training data set and the test data set by utilizing a decomposable time sequence model NeuralProphet to obtain a training feature data set and a test feature data set, and adding an influence factor data set into the training feature data set and the test feature data set, wherein the influence factor data set comprises the highest daily temperature.
4. The method of claim 1, wherein the enhancing the negative samples in the training feature data set and constructing a region heavy overload prediction model from the enhanced training feature data set comprises:
enhancing negative samples in the training feature dataset using gaussian noise techniques;
and training a target algorithm according to the enhanced training characteristic data set, and generating a platform region heavy overload prediction model.
5. The method of claim 4, wherein generating the region heavy overload prediction model based on the training target algorithm for the enhanced training feature data set comprises:
initializing weight distribution of the enhanced training feature data set;
the target algorithm learns by using a training feature data set with weight distribution to obtain a basic classifier;
acquiring a classification error rate and a coefficient of the basic classifier on the training feature data set with weight distribution, and updating the weight distribution of the training feature data set;
and linearly combining the basic classifiers to obtain a platform region heavy overload prediction model.
6. A method for predicting a heavy overload of a region, the method using the region heavy overload prediction model of claim 1, the method comprising:
Acquiring a test characteristic data set and calling a platform region heavy overload prediction model;
and inputting the test characteristic data set into a platform region heavy overload prediction model to obtain a platform region heavy overload prediction value.
7. A training device for a heavy overload prediction model of a platform, the device comprising:
the platform region reference data set acquisition module is used for acquiring a platform region reference data set; the platform region reference data set comprises a limit data set and an influence factor data set of the platform region in a specified time period;
the abnormal data identification and correction module is used for carrying out abnormal data identification and correction on the platform region reference data set by adopting an isolated forest algorithm to obtain a training data set and a test data set;
the feature analysis and feature extraction module is used for carrying out feature analysis and feature extraction on the training data set and the test data set to obtain a training feature data set and a test feature data set, and adding an influence factor data set into the training feature data set and the test feature data set; wherein the training feature data set comprises a negative sample;
and the platform region heavy overload prediction model construction module is used for reinforcing the negative sample in the training characteristic data set and constructing a platform region heavy overload prediction model according to the reinforced training characteristic data set.
8. A station heavy overload prediction apparatus, characterized in that the apparatus comprises:
the platform region heavy overload prediction model calling module is used for acquiring the test characteristic data set and calling the platform region heavy overload prediction model;
and the platform region heavy overload prediction value acquisition module is used for inputting the test characteristic data set into a platform region heavy overload prediction model to obtain a platform region heavy overload prediction value.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 5 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310595201.XA CN116881811A (en) | 2023-05-24 | 2023-05-24 | Platform region heavy overload prediction method, model training method, device and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310595201.XA CN116881811A (en) | 2023-05-24 | 2023-05-24 | Platform region heavy overload prediction method, model training method, device and computer equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116881811A true CN116881811A (en) | 2023-10-13 |
Family
ID=88266860
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310595201.XA Pending CN116881811A (en) | 2023-05-24 | 2023-05-24 | Platform region heavy overload prediction method, model training method, device and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116881811A (en) |
-
2023
- 2023-05-24 CN CN202310595201.XA patent/CN116881811A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110516910A (en) | Declaration form core based on big data protects model training method and core protects methods of risk assessment | |
| US20230169230A1 (en) | Probabilistic wind speed forecasting method and system based on multi-scale information | |
| CN114493052B (en) | Multi-model fusion self-adaptive new energy power prediction method and system | |
| CN111738522B (en) | Photovoltaic power generation power prediction method, storage medium and terminal equipment | |
| CN114266421B (en) | New energy power prediction method based on composite meteorological feature construction and selection | |
| CN113435122B (en) | Real-time flow data processing method, device, computer equipment and storage medium | |
| CN110751317A (en) | Power load prediction system and prediction method | |
| Lopes et al. | A comparative approach of methods to estimate machine productivity in wood cutting | |
| CN116933037A (en) | Photovoltaic output prediction method based on multi-model fusion and related device | |
| CN116523001A (en) | Method, device and computer equipment for constructing weak line identification model of power grid | |
| CN116881811A (en) | Platform region heavy overload prediction method, model training method, device and computer equipment | |
| Sharma et al. | Deep Learning Based Prediction Of Weather Using Hybrid_stacked Bi-Long Short Term Memory | |
| CN113537607B (en) | Power failure prediction method | |
| CN114648406A (en) | User credit integral prediction method and device based on random forest | |
| CN115860273B (en) | Method, apparatus, computer device and storage medium for predicting electric load | |
| Premachandra et al. | Ensemble Methods based Machine Learning Approach for Weather Prediction for Precision Agriculture | |
| CN115577857B (en) | Method and device for predicting output data of energy system and computer equipment | |
| CN114676167B (en) | User persistence model training method, user persistence prediction method and device | |
| Shaji et al. | Weather Prediction Using Machine Learning Algorithms | |
| Gao et al. | Risk Analysis and Prediction of Enterprise Operations Based on Random Forest Algorithm and Power Big Data | |
| CN117852968A (en) | Evaluation model determination method, apparatus, device, storage medium, and program product | |
| CN116780515A (en) | Power consumption prediction method and device, computer equipment and storage medium | |
| Wang et al. | Platform Merchant Demand Prediction Based on Decision Tree and Multi-Layer Perceptron Models | |
| CN117807374A (en) | Spare part abnormal leading data identification method, device and computer equipment | |
| CN117829623A (en) | Airport terminal cold load prediction method and system based on RF-IGOA-LSTM model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |