CN109255505A - A kind of short-term load forecasting method of multi-model fused neural network - Google Patents
A kind of short-term load forecasting method of multi-model fused neural network Download PDFInfo
- Publication number
- CN109255505A CN109255505A CN201811385455.4A CN201811385455A CN109255505A CN 109255505 A CN109255505 A CN 109255505A CN 201811385455 A CN201811385455 A CN 201811385455A CN 109255505 A CN109255505 A CN 109255505A
- Authority
- CN
- China
- Prior art keywords
- model
- training
- data
- neural network
- short
- 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.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 23
- 238000013277 forecasting method Methods 0.000 title claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 65
- 230000004927 fusion Effects 0.000 claims abstract description 12
- 239000013598 vector Substances 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims abstract description 6
- 230000011218 segmentation Effects 0.000 claims abstract description 5
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000001373 regressive effect Effects 0.000 claims description 3
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- 238000010276 construction Methods 0.000 abstract description 3
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 6
- 238000000034 method Methods 0.000 description 6
- 238000004880 explosion Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 230000000875 corresponding effect Effects 0.000 description 3
- 230000001351 cycling effect Effects 0.000 description 3
- 230000007774 longterm Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 238000013179 statistical model Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 1
- 240000002853 Nelumbo nucifera Species 0.000 description 1
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000013604 expression vector Substances 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- Marketing (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The present invention discloses a kind of short-term load forecasting method of multi-model fused neural network, step are as follows: data acquisition and pretreatment obtain data set；It is training set and verifying collection by Segmentation of Data Set；Individually a variety of models of the construction based on back-propagation algorithm, use the input vector sequence training pattern of training set；Use the output of a variety of independent models as input one top layer DNN model of training；By individually trained model and top layer DNN Model Fusion, accurate adjustment is carried out；Trained model is verified with verifying collection, by the precision and error of contrast test collection and verifying collection, adjusts model parameter, by repeatedly training, the model for selecting verifying collection to behave oneself best is as training result；When relatively large deviation occur in predicted value and actual value training set training pattern again is added in latest data by moving model in the actual environment.The present invention reduces the complexity of deployment by the way of independent training, combination tuning, and precision is better than single model, has good practical value.
Description
Technical field
The present invention relates to a kind of Load Prediction In Power Systems technology, specially a kind of multi-model fused neural network it is short-term
Load forecasting method.
Background technique
Current power industry needs accurate short-term load forecasting under open sale of electricity environment, to realize the rule of electric load
Draw the implementation with management strategy.In order to improve precision of prediction, various new power loads have been proposed in the past few decades
Lotus Predicting Technique.Since electric load and its non-linear and randomness of influence factor, load forecast have become electric power city
One of the most challenging task that field entity faces.
Model for load forecast can substantially be divided into four classes: (1) statistical model；(2) Knowledge based engineering expert
System；(3) mixed model and the model of (4) based on artificial intelligence.
Conventional statistics model, such as autoregression model have studied the qualitative relationships between electric load and its influence factor,
And it is easily achieved.However, these statistical models be largely dependent upon it is related between current loads and historical load
Property, it faces very big difficult and calculates at high cost when selecting nonlinear function appropriate.Because of these limitations, researcher
It is increasingly interested in artificial intelligence.
In all available prediction models, short-term load forecasting side is being solved the problems, such as based on the model of artificial neural network
Face attraction is maximum.Even if neural network is also well-known with its learning ability in complicated non-linear environment.It is current common
Neural network model includes DNN, LSTM and ResNet etc., but the back-propagation algorithm of DNN needs to search for huge function space,
It is easily trapped into non-optimum local minimum, so that the precision of model is adversely affected；LSTM can solve long-term continuous time
Dependence Problem between step, however when overlong time, model can still lose the key message in long term data；ResNet
Model depth can be increased, with the intensification of depth, the more advanced feature of model learning, but boundary effect is still remained, it lacks
Few perception to overall data.
Summary of the invention
Exist for Load Prediction In Power Systems model in the prior art in boundary effect, lacks the sense to overall data
Know, the deficiencies such as precision of prediction is low, the object of the present invention is to provide one kind can effectively promote power-system short-term load forecasting precision
Multi-model fused neural network short-term load forecasting method.
In order to solve the above technical problems, the technical solution adopted by the present invention is that:
A kind of short-term load forecasting method of multi-model fused neural network of the present invention, comprising the following steps:
1) data acquisition and pretreatment, obtain data set；
2) be training set and verifying collection two parts by Segmentation of Data Set: training set is used to training pattern, and verifying collection is used to test
Demonstrate,prove training result；
3) a variety of models based on back-propagation algorithm are individually constructed, the input vector sequence training mould of training set is used
Type；
4) use the output of a variety of independent models as input one top layer DNN model of training；
5) by individually trained model and top layer DNN Model Fusion, and accurate adjustment is carried out；
6) trained model is verified with verifying collection, by the precision and error of contrast test collection and verifying collection, adjusts mould
Shape parameter, by repeatedly training, the model for selecting verifying collection to behave oneself best is as training result；
7) latest data is added when relatively large deviation occur in predicted value and actual value for moving model in the actual environment
Training set training pattern again.
In step 1), data acquisition and pretreatment are as follows:
101) power system load data are obtained, while acquiring corresponding weather, festivals or holidays data；
102) above-mentioned Various types of data is pre-processed, including but not limited to: missing values mean value interpolation, data normalization,
One-hot coding, timestamp conversion；
103) pretreated data form input vector sequence according to chronological order；
104) using the system loading in any time following a period of time as prediction target.
In step 3), different models individually train and record prediction result in a variety of models, and meet:
The model structure that A is selected is based on back-propagation algorithm；
With the difference in structure between the model that B is selected, there is different characteristics；
In step 4), top layer DNN model has three-decker, including one layer of input layer, one layer of hidden layer and one layer of input layer,
It is input with the training set output of individually three models of training, using the actual value of training set as regressive object, is trained.
In step 5), by individually trained model and top layer DNN Model Fusion, and accurate adjustment is carried out are as follows:
By trained independent model parallel arranged, using the output node of each independent model as top layer DNN model
Input be connected to DNN hidden layer, finally reduce Fusion Model learning rate carry out accurate adjustment.
The invention has the following beneficial effects and advantage:
1. the invention proposes a kind of short-term load forecasting method based on multi-model fused neural network, by a variety of differences
Neural network individually train；Again using the output of independent model as input, one top layer DNN of training；It finally will be trained
Model Fusion, and accurate adjustment training is carried out, the experimental results showed that, the precision of this method is better than single model.Illustrate that this method has
Good practical value.
2. the present invention is reduced the complexity of deployment, is facilitated engineering construction by the way of independent training, combination tuning.
Detailed description of the invention
Fig. 1 is a kind of short-term load forecasting method flow chart of multi-model fused neural network of embodiment of the present invention；
Fig. 2 is that embodiment of the present invention multi-model merges schematic diagram.
Specific embodiment
The present invention is further elaborated with reference to the accompanying drawings of the specification.
As shown in Figure 1, a kind of short-term load forecasting method of multi-model fused neural network of the present invention, following steps:
1) data acquisition and pretreatment, obtain data set；
2) be training set and verifying collection two parts by Segmentation of Data Set: training set is used to training pattern, and verifying collection is used to test
Demonstrate,prove training result；
3) a variety of models based on back-propagation algorithm are individually constructed, the input vector sequence training mould of training set is used
Type；
4) use the output of a variety of independent models as input one top layer DNN model of training；
5) by individually trained model and top layer DNN Model Fusion, and accurate adjustment is carried out；
6) trained model is verified with verifying collection, by the precision and error of contrast test collection and verifying collection, adjusts mould
Shape parameter, by repeatedly training, the model for selecting verifying collection to behave oneself best is as training result；
7) latest data is added when relatively large deviation occur in predicted value and actual value for moving model in the actual environment
Training set training pattern again.
In step 2), data acquisition and pretreatment are as follows:
101) power system load data are obtained, while acquiring corresponding weather, festivals or holidays data；
In the present embodiment, using somewhere network load data, precision is statistics hourly.It is every that this area has been collected simultaneously
Daily maximum temperature and every daily minimal tcmperature and festivals or holidays data.
102) above-mentioned Various types of data is pre-processed, including but not limited to: missing values mean value interpolation, data normalization,
One-hot coding, timestamp conversion；
In the present embodiment, time data are converted to numeric type data, are standardized to all data, to missing values
Mean value interpolation processing is carried out, festivals or holidays data are converted to two-value data.
103) pretreated data form input vector sequence according to chronological order；
The present embodiment forms input vector sequence according to chronological order, i.e., each hour data are a vector, presses
Chronological order arrangement.
104) using the system loading in any time following a period of time as prediction target；
The present embodiment prediction target is any time following 6 hours power system loads.
Be two parts by Segmentation of Data Set in step 2): training set is used to training pattern, and verifying collection is used to verify training
As a result.Present case is used as training set using preceding the 80% of data set, and rear 20% as verifying collection.
In step 3), different models individually train and record prediction result in a variety of models, and meet:
The model structure that A is selected is based on back-propagation algorithm；
With the difference in structure between the model that B is selected, there is different characteristics.
In the present embodiment, step 301) individually constructs tri- kinds of models of DNN, LSTM and ResNet and training；Then construction is single
Only 3 layers of DNN model, and use the input vector sequence training pattern of training set.
Single DNN model includes one layer of hidden layer, input layer and output layer, using full connection structure and tanh activation primitive,
Correlation between available any input node.Theoretically, as long as hidden neuron is enough, the DNN of 3-tier architecture can also
To be fitted arbitrarily complicated function.It is herein that DNN model is made more to obtain initial data using the main reason for single hidden layer configuration
Whole information, and each neuron of single hidden layer is all connected to each node of input.
In step 302), independent LSTM model is constructed, and use the input vector sequence training pattern of training set；
LSTM model is one kind of Recognition with Recurrent Neural Network.Recognition with Recurrent Neural Network is highly effective in sequence data modeling, because
The information in past input can be remembered for it.But when sequence length is longer, simple RNN structure can encounter gradient
The problem of explosion or gradient disappear.LSTM very good solution with thresholding cycling element this problem.LSTM can be effective
It avoids the problem that gradient explosion or gradient disappear, therefore obtains in terms of modeling with the time series relied on for a long time and answer extensively
With.
RNN is by list entries { x1,x2,…,xnInput shaped like with the cycling element of minor function:
ht=f (ht-1,xt)
Here, xtIt is the input of t moment, htIt is hidden state, htIt can be considered as until t moment, before all inputs
Expression vector.In order to solve the problems, such as that gradient explosion or gradient disappear, LSTM introduces a variety of doors in cycling element,
Its calculating is as follows:
it=σ (Wi·[ht-1, xt]+bi)
ft=σ (Wf·[ht-1, xt]+bf)
ot=σ (Wo·[ht-1, xt]+bo)
ht=ot*tanh(Ct)
Here it、ftAnd otIt is input gate respectively, forgets door and out gate.CtIt is location mode, C%tRepresent candidate list
First state, itDecide whether to be updated location mode CtIn.Location mode helps model preferably to transmit gradient information.Wi、
Wf、Wo、WCAnd bi、bf、bo、bCIt is input gate, the parameter and biasing for forgeing door, out gate and location mode transmitting respectively.Pass through
By list entries { x1,x2,…,xnBe input in LSTM, obtain the corresponding hidden state { h of a column1,h2,…,hn}.These are hidden
Character representation of the hiding state as sequence, for generating output, while the input as next circulation.
LSTM can solve the Dependence Problem between long-term continuous time step, and achieve good result.But
LSTM may still lose the key message in long history when facing long time sequence.This is facing LSTM
When very long list entries, still it is difficult to obtain the overall recognition of historical data, to affect the promotion of precision.
Here for the precision of prediction of lift scheme, present invention employs the structures of 3 layers of LSTM element stack.
Step 303) constructs independent ResNet model, and uses the input vector sequence training pattern of training set；
In order to obtain more information in historical data, periodicity, trend including data etc., Power system load data it is defeated
Enter that sequence is generally longer, this requires the top layer neurons of convolution model should obtain bigger receptive field.Top layer neuron
Receptive field is influenced by two factors: the size of depth and convolution kernel, and the size of receptive field and the two factors are linear
It is positively correlated.One be generally accepted the fact is that, lesser convolution kernel and deeper network can bring higher precision, therefore
ResNet model with bigger depth just becomes natural selection.
The it is proposed of ResNet is a revolution of depth network, it starts deep neural network model toward depth development.
In deep neural network structure, if the depth for merely increasing network can be brought, gradient disappears or explosion is degenerated with accuracy rate
The problem of.
It disappears for gradient or explosion issues, ResNet is addressed by standardization.For accuracy rate problem, ResNet
It is solved by residual error study.Residual block is an a series of branch comprising transformation F, and residual error connection is by the output of residual block
It is activated with the output phase adduction of an identical mapping:
Y=F (x, { Wi})+x
X and y is the tensor that outputs and inputs an of residual block, function F (x, { Wi) represent the residual error that learns is needed to reflect
It penetrates.Residual error connection allows layer effectively to learn the modification to identical mapping, rather than learns entire transformation, this has been proved to pair
It is highly useful in depth network.
ResNet is originally designed for solving problem of image recognition, but wherein use residual error study and it is standardized
Thought can be applied in arbitrary model, similar for solving the problems, such as.Due to the presence of identical mapping, gradient information can
With as needed bypass any number of residual block be directly delivered to lower layer, this be equivalent to model and meanwhile simulate one it is non-
The network of constant width can extract the feature of any depth and width.ResNet has been demonstrated effectively to extract in input data
Feature, including many advanced features.However there are boundary effects for convolution, can not obtain the overall recognition of list entries.
Here, ResNet has used the basic structure of 32 residual blocks, in a residual block of ResNet, comprising one-dimensional
Convolutional layer, standardization and activation primitive, activation primitive is using rectification linear unit (rectifier linear unit, ReLU).
In addition, in order to prevent over-fitting and increase model anti-noise ability, joined one layer of Dropout.The convolution kernel size that model uses
It is 8, because excessive convolution kernel can reduce the resolution ratio of details, and too small convolution kernel will limit advanced spy under one-dimensional environment
The scope of sign, both of which will affect the promotion of precision.
In step 4), top layer DNN model has three-decker, including one layer of input layer, one layer of hidden layer and one layer of input layer,
It is input with the training set output of individually three models of training, using the actual value of training set as regressive object, is trained.
In step 5), by individually trained model and top layer DNN Model Fusion, and accurate adjustment is carried out are as follows:
By trained independent model parallel arranged, using the output node of each independent model as top layer DNN model
Input be connected to DNN hidden layer, finally reduce Fusion Model learning rate carry out accurate adjustment, as shown in Figure 2.
In order to verify effectiveness of the invention, the essence of multi-model fusion method and independent model compared in the present embodiment
Degree, result is as shown in the table, the results show that the short-term load forecasting method precision of prediction of multi-model fused neural network
Highest in several models illustrates that the present invention can effectively promote the precision of short-term load forecasting.
Short-term load forecasting plays key effect in for economic, the reliable and safe operation reserve of electric system formulation.
In order to improve precision of prediction, the invention proposes a kind of short-term load forecasting methods based on multi-model fused neural network.It is first
First a variety of different neural networks are individually trained；Again using the output of independent model as input, one top layer DNN of training；Most
Afterwards by trained Model Fusion, and carry out accurate adjustment training.Short-term load forecasting the experimental results showed that, the precision of this method is excellent
In single model, illustrate that this method has good practical value.
Claims (5)
1. a kind of short-term load forecasting method of multi-model fused neural network, it is characterised in that the following steps are included:
1) data acquisition and pretreatment, obtain data set；
2) be training set and verifying collection two parts by Segmentation of Data Set: training set is used to training pattern, and verifying collection is used to verify instruction
Practice result；
3) a variety of models based on back-propagation algorithm are individually constructed, the input vector sequence training pattern of training set is used；
4) use the output of a variety of independent models as input one top layer DNN model of training；
5) by individually trained model and top layer DNN Model Fusion, and accurate adjustment is carried out；
6) trained model is verified with verifying collection, passes through the precision and error of contrast test collection and verifying collection, adjustment model ginseng
Number, by repeatedly training, the model for selecting verifying collection to behave oneself best is as training result；
7) latest data is added and trains when relatively large deviation occur in predicted value and actual value by moving model in the actual environment
Collect training pattern again.
2. the short-term load forecasting method of multi-model fused neural network according to claim 1, it is characterised in that: step
1) in, data acquisition and pretreatment are as follows:
101) power system load data are obtained, while acquiring corresponding weather, festivals or holidays data；
102) above-mentioned Various types of data is pre-processed, including but not limited to: missing values mean value interpolation, data normalization, solely heat
Coding, timestamp conversion；
103) pretreated data form input vector sequence according to chronological order；
104) using the system loading in any time following a period of time as prediction target.
3. the short-term load forecasting method of multi-model fused neural network according to claim 1, it is characterised in that: step
3) in, different models individually train and record prediction result in a variety of models, and meet:
The model structure that A is selected is based on back-propagation algorithm；
With the difference in structure between the model that B is selected, there is different characteristics.
4. the short-term load forecasting method of multi-model fused neural network according to claim 1, it is characterised in that: step
4) in, top layer DNN model has three-decker, including one layer of input layer, one layer of hidden layer and one layer of input layer, with individually training
The training set output of three models is input, using the actual value of training set as regressive object, is trained.
5. the short-term load forecasting method of multi-model fused neural network according to claim 1, it is characterised in that: step
5) in, by individually trained model and top layer DNN Model Fusion, and accurate adjustment is carried out are as follows:
By trained independent model parallel arranged, using the output node of each independent model as the defeated of top layer DNN model
Enter to be connected to DNN hidden layer, the learning rate for reducing Fusion Model carries out accurate adjustment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811385455.4A CN109255505B (en) | 2018-11-20 | 2018-11-20 | Short-term load prediction method of multi-model fusion neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811385455.4A CN109255505B (en) | 2018-11-20 | 2018-11-20 | Short-term load prediction method of multi-model fusion neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109255505A true CN109255505A (en) | 2019-01-22 |
| CN109255505B CN109255505B (en) | 2021-09-24 |
Family
ID=65043724
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811385455.4A Active CN109255505B (en) | 2018-11-20 | 2018-11-20 | Short-term load prediction method of multi-model fusion neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109255505B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109947088A (en) * | 2019-04-17 | 2019-06-28 | 北京天泽智云科技有限公司 | Equipment fault early-warning system based on model lifecycle management |
| CN110046005A (en) * | 2019-04-15 | 2019-07-23 | 珠海格力电器股份有限公司 | Model training method for product big data capable of reducing transmission flow |
| CN110543942A (en) * | 2019-08-28 | 2019-12-06 | 广西大学 | Multi-space-time long and short memory depth network accurate prediction method |
| CN110837934A (en) * | 2019-11-11 | 2020-02-25 | 四川大学 | Smart grid short-term residential load prediction method based on deep learning |
| CN110927584A (en) * | 2019-12-09 | 2020-03-27 | 天津市捷威动力工业有限公司 | Neural network-based battery life extension prediction method |
| CN111160659A (en) * | 2019-12-31 | 2020-05-15 | 国家电网公司西南分部 | Power load prediction method considering temperature fuzzification |
| CN111160619A (en) * | 2019-12-06 | 2020-05-15 | 北京国电通网络技术有限公司 | Power load prediction method based on data derivation |
| CN111222687A (en) * | 2019-12-06 | 2020-06-02 | 北京国电通网络技术有限公司 | Power load probability prediction method and device based on heterogeneous neural network |
| CN111507507A (en) * | 2020-03-24 | 2020-08-07 | 重庆森鑫炬科技有限公司 | Big data-based monthly water consumption prediction method |
| CN112101653A (en) * | 2020-09-10 | 2020-12-18 | 湘潭大学 | Novel electrified railway traction load prediction method |
| CN112102004A (en) * | 2020-09-18 | 2020-12-18 | 合肥工业大学 | Click rate prediction fusion method based on residual error learning |
| CN112098714A (en) * | 2020-08-12 | 2020-12-18 | 国网江苏省电力有限公司南京供电分公司 | ResNet-LSTM-based electricity stealing detection method and system |
| CN112163715A (en) * | 2020-10-14 | 2021-01-01 | 腾讯科技（深圳）有限公司 | Training method and device of generative countermeasure network and power load prediction method |
| CN112215426A (en) * | 2020-10-16 | 2021-01-12 | 国网山东省电力公司信息通信公司 | Short-term power load prediction method |
| CN112488404A (en) * | 2020-12-07 | 2021-03-12 | 广西电网有限责任公司电力科学研究院 | Multithreading efficient prediction method and system for large-scale power load of power distribution network |
| CN113239029A (en) * | 2021-05-18 | 2021-08-10 | 国网江苏省电力有限公司镇江供电分公司 | Completion method for missing daily freezing data of electric energy meter |
| CN114091724A (en) * | 2021-10-15 | 2022-02-25 | 国网浙江省电力有限公司 | Power supply equipment load and service life prediction model construction method |
| WO2022112895A1 (en) * | 2020-11-30 | 2022-06-02 | International Business Machines Corporation | Automated deep learning architecture selection for time series prediction with user interaction |
| CN114936704A (en) * | 2022-06-02 | 2022-08-23 | 南京东唯电子科技有限公司 | Source station load prediction method based on multi-sliding time window MSD-LSTM neural network |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679287A (en) * | 2013-12-05 | 2014-03-26 | 王海燕 | Combined type power load forecasting method |
| CN104794538A (en) * | 2015-04-20 | 2015-07-22 | 浙江大学 | Power system short-term load prediction method based on IEEMD-BPNN (improved ensemble empirical mode decomposition-back propagation neutral network) |
| CN106022954A (en) * | 2016-05-16 | 2016-10-12 | 四川大学 | Multiple BP neural network load prediction method based on grey correlation degree |
| CN106127360A (en) * | 2016-06-06 | 2016-11-16 | 国网天津市电力公司 | A kind of multi-model load forecasting method analyzed based on user personality |
| CN106952181A (en) * | 2017-03-08 | 2017-07-14 | 深圳市景程信息科技有限公司 | Electric Load Prediction System based on long Memory Neural Networks in short-term |
| CN107392364A (en) * | 2017-07-12 | 2017-11-24 | 河海大学 | The short-term load forecasting method of variation mode decomposition and depth belief network |
| CN107704970A (en) * | 2017-10-24 | 2018-02-16 | 上海电器科学研究所（集团）有限公司 | A kind of Demand-side load forecasting method based on Spark |
| CN108256697A (en) * | 2018-03-26 | 2018-07-06 | 电子科技大学 | A kind of Forecasting Methodology for power-system short-term load |
| CN108510006A (en) * | 2018-04-08 | 2018-09-07 | 重庆邮电大学 | A kind of analysis of business electrical amount and prediction technique based on data mining |
| CN108830487A (en) * | 2018-06-21 | 2018-11-16 | 王芊霖 | Methods of electric load forecasting based on long neural network in short-term |
-
2018
- 2018-11-20 CN CN201811385455.4A patent/CN109255505B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679287A (en) * | 2013-12-05 | 2014-03-26 | 王海燕 | Combined type power load forecasting method |
| CN104794538A (en) * | 2015-04-20 | 2015-07-22 | 浙江大学 | Power system short-term load prediction method based on IEEMD-BPNN (improved ensemble empirical mode decomposition-back propagation neutral network) |
| CN106022954A (en) * | 2016-05-16 | 2016-10-12 | 四川大学 | Multiple BP neural network load prediction method based on grey correlation degree |
| CN106127360A (en) * | 2016-06-06 | 2016-11-16 | 国网天津市电力公司 | A kind of multi-model load forecasting method analyzed based on user personality |
| CN106952181A (en) * | 2017-03-08 | 2017-07-14 | 深圳市景程信息科技有限公司 | Electric Load Prediction System based on long Memory Neural Networks in short-term |
| CN107392364A (en) * | 2017-07-12 | 2017-11-24 | 河海大学 | The short-term load forecasting method of variation mode decomposition and depth belief network |
| CN107704970A (en) * | 2017-10-24 | 2018-02-16 | 上海电器科学研究所（集团）有限公司 | A kind of Demand-side load forecasting method based on Spark |
| CN108256697A (en) * | 2018-03-26 | 2018-07-06 | 电子科技大学 | A kind of Forecasting Methodology for power-system short-term load |
| CN108510006A (en) * | 2018-04-08 | 2018-09-07 | 重庆邮电大学 | A kind of analysis of business electrical amount and prediction technique based on data mining |
| CN108830487A (en) * | 2018-06-21 | 2018-11-16 | 王芊霖 | Methods of electric load forecasting based on long neural network in short-term |
Non-Patent Citations (3)
| Title |
|---|
| JURADO S等: "Hybrid methodologies for electricity load forecasting: Entropy-based feature selection with machine learning and soft computing techniques", 《ENERGY》 * |
| KHWAJA A S: "Boostedneuralnetworksforimprovedshort-termelectricloadforecasting", 《ELECTRIC POWER SYSTEMS RESEARCH》 * |
| 李冬辉等: "基于MFOA-GRNN模型的年电力负荷预测", 《电网技术》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110046005A (en) * | 2019-04-15 | 2019-07-23 | 珠海格力电器股份有限公司 | Model training method for product big data capable of reducing transmission flow |
| CN109947088A (en) * | 2019-04-17 | 2019-06-28 | 北京天泽智云科技有限公司 | Equipment fault early-warning system based on model lifecycle management |
| CN110543942A (en) * | 2019-08-28 | 2019-12-06 | 广西大学 | Multi-space-time long and short memory depth network accurate prediction method |
| CN110837934A (en) * | 2019-11-11 | 2020-02-25 | 四川大学 | Smart grid short-term residential load prediction method based on deep learning |
| CN111160619A (en) * | 2019-12-06 | 2020-05-15 | 北京国电通网络技术有限公司 | Power load prediction method based on data derivation |
| CN111222687A (en) * | 2019-12-06 | 2020-06-02 | 北京国电通网络技术有限公司 | Power load probability prediction method and device based on heterogeneous neural network |
| CN110927584B (en) * | 2019-12-09 | 2022-05-10 | 天津市捷威动力工业有限公司 | Neural network-based battery life extension prediction method |
| CN110927584A (en) * | 2019-12-09 | 2020-03-27 | 天津市捷威动力工业有限公司 | Neural network-based battery life extension prediction method |
| CN111160659A (en) * | 2019-12-31 | 2020-05-15 | 国家电网公司西南分部 | Power load prediction method considering temperature fuzzification |
| CN111507507A (en) * | 2020-03-24 | 2020-08-07 | 重庆森鑫炬科技有限公司 | Big data-based monthly water consumption prediction method |
| CN112098714A (en) * | 2020-08-12 | 2020-12-18 | 国网江苏省电力有限公司南京供电分公司 | ResNet-LSTM-based electricity stealing detection method and system |
| CN112098714B (en) * | 2020-08-12 | 2023-04-18 | 国网江苏省电力有限公司南京供电分公司 | Electricity stealing detection method and system based on ResNet-LSTM |
| CN112101653A (en) * | 2020-09-10 | 2020-12-18 | 湘潭大学 | Novel electrified railway traction load prediction method |
| CN112102004A (en) * | 2020-09-18 | 2020-12-18 | 合肥工业大学 | Click rate prediction fusion method based on residual error learning |
| CN112102004B (en) * | 2020-09-18 | 2022-08-30 | 合肥工业大学 | Click rate prediction fusion method based on residual error learning |
| CN112163715A (en) * | 2020-10-14 | 2021-01-01 | 腾讯科技（深圳）有限公司 | Training method and device of generative countermeasure network and power load prediction method |
| CN112215426A (en) * | 2020-10-16 | 2021-01-12 | 国网山东省电力公司信息通信公司 | Short-term power load prediction method |
| WO2022112895A1 (en) * | 2020-11-30 | 2022-06-02 | International Business Machines Corporation | Automated deep learning architecture selection for time series prediction with user interaction |
| GB2616770A (en) * | 2020-11-30 | 2023-09-20 | Ibm | Automated deep learning architecture selection for time series prediction with user interaction |
| CN112488404A (en) * | 2020-12-07 | 2021-03-12 | 广西电网有限责任公司电力科学研究院 | Multithreading efficient prediction method and system for large-scale power load of power distribution network |
| CN113239029A (en) * | 2021-05-18 | 2021-08-10 | 国网江苏省电力有限公司镇江供电分公司 | Completion method for missing daily freezing data of electric energy meter |
| CN114091724A (en) * | 2021-10-15 | 2022-02-25 | 国网浙江省电力有限公司 | Power supply equipment load and service life prediction model construction method |
| CN114936704A (en) * | 2022-06-02 | 2022-08-23 | 南京东唯电子科技有限公司 | Source station load prediction method based on multi-sliding time window MSD-LSTM neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109255505B (en) | 2021-09-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109255505A (en) | A kind of short-term load forecasting method of multi-model fused neural network | |
| Qin et al. | A dual-stage attention-based recurrent neural network for time series prediction | |
| CN109543901A (en) | Short-Term Load Forecasting Method based on information fusion convolutional neural networks model | |
| Yu et al. | Prediction of highway tunnel pavement performance based on digital twin and multiple time series stacking | |
| CN109829587A (en) | Zonule grade ultra-short term and method for visualizing based on depth LSTM network | |
| CN109614981A (en) | The Power System Intelligent fault detection method and system of convolutional neural networks based on Spearman rank correlation | |
| Azadeh et al. | Performance assessment of electric power generations using an adaptive neural network algorithm | |
| Luan et al. | Out-of-distribution detection for deep neural networks with isolation forest and local outlier factor | |
| CN114565187B (en) | Traffic network data prediction method based on graph space-time self-coding network | |
| Yan et al. | Small watershed stream-flow forecasting based on LSTM | |
| Kumar et al. | Wind speed prediction using deep learning-LSTM and GRU | |
| CN114611617A (en) | Depth field self-adaptive image classification method based on prototype network | |
| CN116843083A (en) | Carbon emission prediction system and method based on hybrid neural network model | |
| CN115391553A (en) | Method for automatically searching time sequence knowledge graph complement model | |
| Xu et al. | Short‐term traffic flow prediction based on whale optimization algorithm optimized BiLSTM_Attention | |
| Hamad et al. | Deep learning-based load forecasting considering data reshaping using MATLAB\Simulink | |
| Liu et al. | A CNN-LSTM-based domain adaptation model for remaining useful life prediction | |
| Hu et al. | Self-supervised pre-training for robust and generic spatial-temporal representations | |
| Zhu et al. | An efficient algorithm for the incremental broad learning system by inverse Cholesky factorization of a partitioned matrix | |
| CN115048873B (en) | Residual service life prediction system for aircraft engine | |
| Yang et al. | Guest Editorial: Industrial Artificial Intelligence for Smart Manufacturing | |
| CN114401135B (en) | Internal threat detection method based on LSTM-Attention user and entity behavior analysis technology | |
| CN112256858B (en) | Double-convolution knowledge tracking method and system fusing question mode and answer result | |
| CN115293249A (en) | Power system typical scene probability prediction method based on dynamic time sequence prediction | |
| Ogawa et al. | PV output forecasting by deep Boltzmann machines with SS‐PPBSO |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |