CN114707426A - Line dynamic capacity increasing method based on multi-source data fusion - Google Patents
Line dynamic capacity increasing method based on multi-source data fusion Download PDFInfo
- Publication number
- CN114707426A CN114707426A CN202210471739.5A CN202210471739A CN114707426A CN 114707426 A CN114707426 A CN 114707426A CN 202210471739 A CN202210471739 A CN 202210471739A CN 114707426 A CN114707426 A CN 114707426A
- Authority
- CN
- China
- Prior art keywords
- capacity
- line
- model
- power transmission
- target
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/04—Power grid distribution networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a line dynamic capacity increasing method based on multi-source data fusion, which has the effects that in a neural network structure model based on a probability sparse self-attention mechanism, a neural network embedded layer is established by fusing multi-source information such as characteristic sequence information, hierarchical time sequence information and the like, modeling is carried out by utilizing variables to extract the periodic rule of the capacity change of a power transmission line, long-time sequence prediction is realized based on a historical sequence, and reasonable prejudgment is carried out on the change trend of the future working condition. The method has the advantages that the prediction judgment in a period of time in the future can be realized, the comprehensive change condition of a certain line or a plurality of lines in the period of time in the future can be known through the judgment in the period of time, and the method is favorable for better and reasonable allocation of the rated load capacity of the power transmission line in advance.
Description
Technical Field
The invention belongs to the technical field of power systems, and relates to a dynamic capacity increasing method for a high-voltage line, in particular to a method for modeling based on dynamic capacity increasing deep learning of the line during multi-source data fusion.
Background
The dynamic capacity increasing technology of the existing power transmission line is relatively mature, the maximum allowable current carrying capacity of a lead is calculated according to a Mogan equal-load flow model on the premise of not breaking through the existing technical regulation, the hidden capacity objectively existing in the line is fully utilized, and the actual transmission capacity of the power transmission line is improved. The parameters monitored in real time comprise the parameter characteristics of the temperature, the tension, the sag, the temperature, the sunshine, the customs and the like of the wire, the maximum allowable current-carrying capacity of each power transmission line is timely changed according to the real-time parameter characteristics, the performance of each line is fully exerted, larger electric load is conveniently transmitted, and the compensation capacity of the transmission capacity of the power transmission lines is improved.
With the rise of new energy and the progress of science and technology, more new energy power plants including solar energy and wind energy are built in western regions of China. The transmission of electricity from these new energy sources is characterized by their renewable nature, but it also brings about problems. Because of the long distance and high energy, although the transmission line can be additionally increased, the existing transmission line is also used for carrying, so whether the new energy electric energy with high energy can bear the load capacity of the existing transmission line or not is considered, and how to more efficiently utilize and allocate the transmission capacity of the existing transmission line is also considered.
The prior art mostly adopts single-working-condition prediction in the process of dynamic capacity increase of a power transmission line, for example, patent with publication number CN110321601A, which discloses a method and a system for predicting dynamic current-carrying capacity of an overhead line in advance, wherein the method and the system are predicted by a recurrent neural network with attention mechanism, and such a result can only predict one point or a plurality of points in a future period of time, for example, predict a line point 15 minutes in the future, the required capacity of the line at the time node, and the prediction mode cannot perform iterative learning. For the times of multi-source power data fusion, only such prediction obviously cannot achieve the prediction purpose. For example, the rated load on multiple lines in the next hour period needs to be predicted so as to make reasonable adjustments in high-voltage, peak-hour large-scale transmission.
Disclosure of Invention
The invention aims to enable a plurality of power transmission lines to simultaneously predict load capacity in a period of time in the future so as to realize the technical problem of dynamic capacity increase effect in a multi-source power data fusion state. The training process comprises two processes of pre-training and super-parameter fine-tuning, firstly, pre-training of a neural network model is carried out based on historical operation data of a plurality of lines with the same voltage level of a regional power grid, fuzzy association relations among input characteristic variables, between input characteristics and prediction targets and between the prediction targets are extracted, then, the pre-training model is used as an initial parameter, and fine-tuning is carried out on the super-parameter of the network model based on the historical operation data of the target lines. In the forward prediction stage of the model, long-time sequence prediction is realized based on a historical sequence, and the change trend of the future working condition is reasonably predicted.
The technical scheme of the invention is that the line dynamic capacity increasing method based on multi-source data fusion predicts the capacity of a future power transmission line after a prediction model is constructed, and the key is that the line dynamic capacity increasing method comprises the following specific steps:
a. constructing a neural network structure model of a probability sparse self-attention mechanism, adding a model embedding layer consisting of hierarchical time sequence information, characteristic sequence information and sequence position information, and taking the model embedding layer as an input sample set of a pre-training model to form the pre-training model;
b. extracting historical operation data of a plurality of lines with the same voltage level, and pre-training the lines by using a pre-training model to form an incidence relation between a model embedding layer and the capacity of the power transmission line;
c. when the target power transmission lines of the team are predicted, training the target power transmission lines by using a pre-training model according to the hierarchical time sequence information of the target power transmission lines and the relevant information operation data of the target variables to obtain a final target line dynamic capacity-increasing prediction model;
d. and performing long sequence prediction on the capacity of the power transmission line at multiple moments in the future by adopting the trained line dynamic capacity-increasing prediction model.
In the step a, the hierarchical time sequence information includes information of season, month, week, day and time.
In the step a, the characteristic sequence information includes temperature, ambient humidity, wind speed, wind direction, rainfall, radiation intensity and historical line capacity.
In the step b, the incidence relation between the model embedded layer and the capacity of the power transmission line comprises incidence relations among all input characteristic variables and between the input characteristics and the prediction target.
In the step c, the hierarchical time sequence information of the target transmission line and the relevant information operation data of the target variable comprise season, month, week, day and time information, temperature, environment humidity, wind speed, wind direction, rainfall, radiation intensity and historical line capacity.
The beneficial effect of the invention is that,
1. the neural network model is pre-trained on the basis of historical operation data of a plurality of lines with the same voltage class of a regional power grid, so that when one line needs to be evaluated, the parameter of the line is input, the pre-trained model can optimize the line, the pre-trained model can continuously perform iterative training due to the fact that the pre-trained record exists, deep learning can be continuously performed through the model, the near-perfect pre-judgment processing capacity is achieved, and finally the pre-judgment result is better and accurate.
2. The invention can realize the prediction judgment in a period of time in the future, can know the comprehensive change condition of a certain line or a plurality of lines in a period of time in the future through the judgment in a period of time, and is favorable for better and reasonably allocating the rated load capacity of the power transmission line in advance.
3. In a neural network structure model based on a probability sparse self-attention mechanism, a neural network embedded layer is established by fusing multi-source information such as characteristic sequence information and hierarchical time sequence information, modeling is carried out by utilizing variables to extract the periodic rule of the capacity change of the power transmission line, long-time sequence prediction is realized based on a historical sequence, and the change trend of the future working condition is reasonably predicted.
Drawings
FIG. 1 is a line dynamic compatibilization model embedded layer structure of multi-source data fusion in the invention.
Detailed Description
The detailed steps of the line dynamic capacity increasing method of the present invention will be described with reference to the accompanying drawings and specific embodiments.
The invention focuses on the selection of models and the improvement of models. The basic model adopts a neural network structure based on a probability sparse self-attention mechanism, and a neural network embedding layer is established to perform a model pre-training process by fusing multi-source information such as characteristic sequence information, layer time sequence information and sequence position information. The addition of the embedded layer enables the prediction model to be analyzed better and more comprehensively, and the variable factor relation influencing the line capacity can be analyzed in a panoramic way. Wherein the hierarchical timing information includes: season, month, week, day, time information, the characteristic sequence information includes: temperature, ambient humidity, wind speed, wind direction, rainfall, radiation intensity, historical line capacity. The method is used for extracting the periodic rule of the capacity change of the power transmission line.
In the model pre-training stage, pre-training a neural network model based on historical operation data of a plurality of lines with the same voltage level in each regional power grid, and extracting fuzzy association relations among input characteristic variables, between input characteristics and a predicted target and between predicted targets;
aiming at a target variable training stage in a certain line, taking a pre-trained model as an initial parameter, and performing model fine adjustment according to the level time sequence information of a target power transmission line and the operation data of the target variable related information;
in the forward prediction stage of the model, long-time sequence prediction is realized based on a historical sequence, and the change trend of the future working condition is reasonably predicted.
The base model is a neural network structure based on a probabilistic sparse self-attention mechanism. Wherein the input data comprises: season, month, week, day, time information as well as temperature, ambient humidity, wind speed, wind direction, rainfall, radiation intensity, historical line capacity; the output data is: transmission line capacity. The specific steps of the probability sparseness self-attention mechanism of the basic model are as follows:
step 1, input determination: q ∈ Rm×d，K∈Rn×d，V∈Rn×dWherein m and n are the lengths of input sequences, and d is the dimension of model input;
step 2, setting an influence factor c, calculating a sampling dimension U of Q to be c. lnm, and randomly selecting a sampling dimension K to be c. lnn;
step 3, randomly selecting U keys from K to form
Calculating dot product scores
Step 4, calculating the measurement indexes according to rows
Extracting top-u queries from Q based on M to form
Step 5, calculate the dot product of u queries and all keys, that is
Step 6, calculating S0＝mean(V)，S＝{S1,S0The probability is the output of probability sparseness self-attention.
In order to analyze the variable factor relationship affecting the line capacity in a panoramic way, a model embedding layer construction method for multi-source data fusion of characteristic sequence information, characteristic sequence position information and hierarchical time sequence information is provided, wherein the model embedding layer comprises hierarchical time sequence information, sequence information and sequence position information, and is shown in figure 1.
By curve observation of line capacity and characteristic influence factors thereof, the change rule of the line capacity and the characteristic influence factors thereof is found to have extremely strong time periodicity, the level time sequence information of seasons, months, weeks, days, hours and the like is extracted, the level time sequence information is added to a model embedded layer after being coded, and the influence of the time period on the line capacity is mined by the level time sequence information extracted through training of an Informer model. The characteristic sequence information comprises temperature, ambient humidity, wind speed, wind direction, rainfall, radiation intensity and historical line capacity, and corresponding to u0-u6 in the figure, the sequence position information E0-E6 represents the relative positions of the variables in the encoding process, and the intrinsic relation between the characteristic variables and the line capacity is deepened. The multi-source data fusion expression of the feature sequence information, the feature sequence position information and the hierarchical time sequence information is as follows:
wherein:
representing multi-source data fusion embedding layer variables;
representing a sequence feature variable; pe represents sequence feature variable position information; se represents hierarchical timing information, pos represents sequence position, i represents model dimension position, t represents sample working condition, and LxRepresenting the input sample length. Wherein the characteristic variable bitThe coding sequence formula is as follows:
pe(pos,2i)＝sin(pos/100002i/d)
pe(pos,2i+1)＝cos(pos/100002i/d)
where pos represents the sequence position and i represents the model dimension position.
And in the model pre-training process, selecting the power transmission lines with certain similarity in operation conditions, and training according to multi-source information.
And in the model fine adjustment process, on the basis of the result of the pre-training model, the model is continuously trained on the historical data of the target line, so that the model can be more suitable for the operation condition of the target line.
In the long-time sequence prediction process, the change trend prediction of the future long-time period of the line capacity can be realized by setting the super parameters such as the historical sequence length, the prediction target prediction length and the like. The model decomposes the input of the decoding layer into two parts, namely a part of history sequence and a target placeholder sequence, and realizes the one-step prediction of the long sequence, and the formula is as follows:
wherein the history sequence
LtokenIs the length of the history sequence; target placeholder sequence
LyFor the predicted target sequence length, its sequence data is set to 0. Sparse self-attention in the decoding layer employs a masking mechanism with the dot product of the mask set to-inf, which prevents each location from focusing on future locations, thus avoiding autoregressive.
Claims (5)
1. A line dynamic capacity increasing method based on multi-source data fusion is a process for predicting the capacity of a future power transmission line after a prediction model is built, and is characterized in that: the method for dynamically increasing the capacity of the line comprises the following specific steps:
a. constructing a neural network structure model of a probability sparse self-attention mechanism, adding a model embedding layer consisting of hierarchical time sequence information, characteristic sequence information and sequence position information, and taking the model embedding layer as an input sample set of a pre-training model to form the pre-training model;
b. extracting historical operation data of a plurality of lines with the same voltage level, and pre-training the lines by using a pre-training model to form an incidence relation between a model embedding layer and the capacity of the power transmission line;
c. when the target power transmission lines of the team are predicted, training the target power transmission lines by using a pre-training model according to the hierarchical time sequence information of the target power transmission lines and the relevant information operation data of the target variables to obtain a final target line dynamic capacity-increasing prediction model;
d. and performing long sequence prediction on the capacity of the power transmission line at multiple moments in the future by adopting the trained line dynamic capacity-increasing prediction model.
2. The line dynamic capacity increasing method based on multi-source data fusion, according to claim 1, is characterized in that: in the step a, the hierarchical time sequence information includes information of season, month, week, day and time.
3. The line dynamic capacity increasing method based on multi-source data fusion, according to claim 1, is characterized in that: in the step a, the characteristic sequence information includes temperature, ambient humidity, wind speed, wind direction, rainfall, radiation intensity and historical line capacity.
4. The line dynamic capacity increasing method based on multi-source data fusion, according to claim 1, is characterized in that: in the step b, the incidence relation between the model embedded layer and the capacity of the power transmission line comprises incidence relations among all input characteristic variables and between the input characteristics and the prediction target.
5. The line dynamic capacity increasing method based on multi-source data fusion, according to claim 1, is characterized in that: in the step c, the hierarchical time sequence information of the target transmission line and the relevant information operation data of the target variable comprise season, month, week, day and time information, temperature, environment humidity, wind speed, wind direction, rainfall, radiation intensity and historical line capacity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210471739.5A CN114707426A (en) | 2022-04-29 | 2022-04-29 | Line dynamic capacity increasing method based on multi-source data fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210471739.5A CN114707426A (en) | 2022-04-29 | 2022-04-29 | Line dynamic capacity increasing method based on multi-source data fusion |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114707426A true CN114707426A (en) | 2022-07-05 |
Family
ID=82176871
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210471739.5A Pending CN114707426A (en) | 2022-04-29 | 2022-04-29 | Line dynamic capacity increasing method based on multi-source data fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114707426A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115496002A (en) * | 2022-11-16 | 2022-12-20 | 国网湖北省电力有限公司信息通信公司 | Multi-dimensional feature interactive line dynamic capacity increasing method, system and medium |
| CN115579887A (en) * | 2022-12-07 | 2023-01-06 | 国网湖北省电力有限公司信息通信公司 | Dynamic capacity increasing method, system and medium for transmission line |
| CN116896167A (en) * | 2023-09-11 | 2023-10-17 | 山东和兑智能科技有限公司 | Power transmission line dynamic capacity-increasing monitoring and early warning method based on artificial intelligence |
-
2022
- 2022-04-29 CN CN202210471739.5A patent/CN114707426A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115496002A (en) * | 2022-11-16 | 2022-12-20 | 国网湖北省电力有限公司信息通信公司 | Multi-dimensional feature interactive line dynamic capacity increasing method, system and medium |
| CN115496002B (en) * | 2022-11-16 | 2023-02-24 | 国网湖北省电力有限公司信息通信公司 | Multi-dimensional feature interactive line dynamic capacity increasing method, system and medium |
| CN115579887A (en) * | 2022-12-07 | 2023-01-06 | 国网湖北省电力有限公司信息通信公司 | Dynamic capacity increasing method, system and medium for transmission line |
| CN116896167A (en) * | 2023-09-11 | 2023-10-17 | 山东和兑智能科技有限公司 | Power transmission line dynamic capacity-increasing monitoring and early warning method based on artificial intelligence |
| CN116896167B (en) * | 2023-09-11 | 2023-12-15 | 山东和兑智能科技有限公司 | Power transmission line dynamic capacity-increasing monitoring and early warning method based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114707426A (en) | Line dynamic capacity increasing method based on multi-source data fusion | |
| Lin et al. | An improved moth-flame optimization algorithm for support vector machine prediction of photovoltaic power generation | |
| CN110288136B (en) | Wind power multi-step prediction model establishment method | |
| CN102930358B (en) | A kind of neural net prediction method of photovoltaic power station power generation power | |
| CN105631558A (en) | BP neural network photovoltaic power generation system power prediction method based on similar day | |
| CN111832812A (en) | Wind power short-term prediction method based on deep learning | |
| Kharlova et al. | Forecasting photovoltaic power production using a deep learning sequence to sequence model with attention | |
| CN114792156A (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
| Park et al. | Multi-layer RNN-based short-term photovoltaic power forecasting using IoT dataset | |
| CN105426989A (en) | EEMD and combined kernel RVM-based photovoltaic power short-term prediction method | |
| Liao et al. | Ultra-short-term interval prediction of wind power based on graph neural network and improved bootstrap technique | |
| CN108694475B (en) | Short-time-scale photovoltaic cell power generation capacity prediction method based on hybrid model | |
| CN110852492A (en) | Photovoltaic power ultra-short-term prediction method for finding similarity based on Mahalanobis distance | |
| Fan et al. | Spatial-temporal genetic-based attention networks for short-term photovoltaic power forecasting | |
| CN114611757A (en) | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network | |
| Buonanno et al. | Comprehensive method for modeling uncertainties of solar irradiance for PV power generation in smart grids | |
| CN110909994A (en) | Small hydropower station power generation amount prediction method based on big data drive | |
| CN112348236B (en) | Abnormal daily load demand prediction system and method for intelligent electricity consumption monitoring terminal | |
| Zhou et al. | Short-term wind power prediction based on EMD-LSTM | |
| Zhang et al. | Joint forecasting of regional wind and solar power based on attention neural network | |
| CN115660893A (en) | Transformer substation bus load prediction method based on load characteristics | |
| Hiba et al. | Stacked Denoising Autoencoder network for short-term prediction of electrical Algerian load | |
| Zhang et al. | Short-term wind power interval prediction based on gd-lstm and bootstrap techniques | |
| Succetti et al. | Multivariate prediction of energy time series by autoencoded LSTM networks | |
| Wang et al. | GWTSP: A multi-state prediction method for short-term wind turbines based on GAT and GL |
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 |