CN115545345A - Power grid load prediction method based on GRU-TCN model - Google Patents
Power grid load prediction method based on GRU-TCN model Download PDFInfo
- Publication number
- CN115545345A CN115545345A CN202211389573.9A CN202211389573A CN115545345A CN 115545345 A CN115545345 A CN 115545345A CN 202211389573 A CN202211389573 A CN 202211389573A CN 115545345 A CN115545345 A CN 115545345A
- Authority
- CN
- China
- Prior art keywords
- load
- power grid
- value
- gru
- data
- 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
- 238000000034 method Methods 0.000 title claims abstract description 29
- 239000011159 matrix material Substances 0.000 claims description 29
- 238000004088 simulation Methods 0.000 claims description 18
- 238000010606 normalization Methods 0.000 claims description 13
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 claims description 3
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 claims description 3
- 238000013461 design Methods 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 27
- 230000001364 causal effect Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 101150049349 setA gene Proteins 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
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
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Primary Health Care (AREA)
- Life Sciences & Earth Sciences (AREA)
- Power Engineering (AREA)
- Molecular Biology (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 invention discloses a power grid load prediction method based on a GRU-TCN model, which comprises the following steps: s1: load data of each device in the power grid on the forecast day are collected to form a load data group with time as a sequence
S2: normalizing the collected load data to enlarge the load data group
Obtaining a normalized load prediction data set; s3: the load prediction data set is input into the GRU-TCN model,outputting a load predicted value of each time interval of the power grid; s4: establishing a digital twin power grid, inputting a load predicted value into the digital twin power grid, and outputting a real load value; and correcting the control parameters of the digital twin power grid. The invention integrates a Time Convolution Network (TCN) and a Gated Recursion Unit (GRU) network to design a GRU-TCN prediction model, and high-dimensional data characteristics extracted by the TCN network model are input into the GRU network for prediction, so that a high-precision prediction result can be obtained.
Description
Technical Field
The invention relates to the field of digital construction of power grids, in particular to a power grid load prediction method based on a GRU-TCN model.
Background
At present, a Power Grid (PG) company takes digital transformation as important deployment of PG construction, and digital technology is utilized to promote coordination and interaction of PG loads and energize the PG, so that the power supply service level is improved, and the satisfaction degree of power supply customers is improved. The digital power grid can effectively integrate various resources, promote data sharing and integrate power dispatching, thereby improving the service level of each PG business and realizing the automatic metering of power loads. In order to monitor and perceive the operation indexes of the PG, a digital power grid model is provided and applied to load prediction.
Currently, there is no formal definition of digital power grid in the world, but there are a lot of research results related thereto. Meanwhile, the development of the digital power grid enriches the research content of load prediction, and attracts more experts and scholars to discuss ahead. And the prediction data with different time scales and different precisions are utilized to carry out multi-stage coordination, so that a cold, hot and electricity combined debugging and scheduling scheme is gradually perfected. However, most of the prior art focuses on optimal scheduling, and no deep research is performed on load prediction. For example, the power load is predicted by comprehensively using a K-means clustering model, a K-nearest neighbor classification model and a differential integral moving average autoregressive model, but the processing efficiency needs to be further improved. And as the prior art also provides a power grid load prediction method based on hyper-parameter grid search and support vector regression, which effectively improves the accuracy of power grid load prediction. In view of the unique network structure and the variable form, the traditional load prediction method cannot meet the requirements of the digital power grid.
Disclosure of Invention
In order to solve the problems, the invention provides a power grid load prediction method based on a GRU-TCN model, which can effectively reduce power grid load prediction errors.
In order to achieve the purpose of the invention, the technical scheme adopted by the invention is as follows:
a grid load prediction method based on a GRU-TCN model comprises the following steps:
s1: load data of each device in the power grid on the forecast day are collected to form a load data group with time as a sequence
Wherein n is the type of the load, and t is the acquisition time;
s2: for collected load dataNormalization processing is carried out, load data of each device in the power grid on similar days are selected according to the relevant value model and input into the load data group
In expanding the load data set
Obtaining a normalized load prediction data set;
s3: inputting the load prediction data group into a GRU-TCN model, and outputting a load prediction value of each time period of the power grid;
s4: establishing a digital twin power grid, inputting a load predicted value into the digital twin power grid, and outputting a real load value; and evaluating the predicted load value and the real load value by utilizing the root mean square error RMSE, and correcting the control parameters of the digital twin power grid.
Further, step S2 comprises the following sub-steps:
s21: load data set
Is normalized and mapped to the interval [0,1 ]]In order to obtain a normalized value
S22: computing load data set
All normalized values in (1)
Average value of (2)
S23: acquiring load data groups of devices in power grid on similar days
Where m = n, and using the normalization method in step S21, for the load data set
Load data X in (1) m Normalization is carried out to obtain a normalized value x m ′；
S24: using normalized value x m ' and average value
Calculating the degree of correlation R X ；
S25: similarity R X With a similarity threshold value R X threshold value By comparison, if R X ≥R X threshold value Then the load data X is added m Adding to load data set
The preparation method comprises the following steps of (1) performing; if R is X ＜R X threshold If the load data X is proved m Deleting the load data X for irrelevant variables m ；
S26: traversing load data X of all the devices of the power grid collected in similar days m Executing steps S23-S25 to obtain a load prediction data set with time as a sequence after normalization
s is the increased load data amount.
Further, the normalization method in step S21 is:
wherein, X (t) For load data sets
Any of the load data of (1) is loaded,
are respectively load data set
Maximum and minimum values of (a).
Further, the normalized value x is utilized in step S24 m ' and average value
Calculating the degree of correlation R X The method comprises the following steps:
further, step S3 comprises the following sub-steps:
s31: grouping load prediction data
As an input matrix;
s32: inputting the input matrix into the TCN network model, outputting the output matrix
A feature matrix F;
s33: inputting an output matrix F into the fully-connected layer, and outputting a prediction matrix Y = [ Y ] composed of predicted values 1 ，Y 2 ,…,Y v ]V is the number of predicted values in the prediction matrix;
s34: inputting the prediction matrix Y into a GRU network model, and outputting a state value c corresponding to the time t (t) ；
S34: state value c to be output (t) As the load prediction value of the power grid.
Further onIn step S32, on
The feature matrix F of (1) is:
i＝[1，2，···，l]，k＝[1，2，···，K]，K≥1
wherein, W k V is the amount of data in the output feature matrix F,
and f (-) is the residual error function of the TCN network model, K is a time sequence, K is 2K times of the initial time, and K is a hyperparameter.
Further, the state value c corresponding to the time t in step S34 (t) Comprises the following steps:
ψ r ＝sigmoid(ω rc c (t-1) +ω rx x (t) +b r )
ψ u ＝sigmoid(ω uc c (t-1) +ω ux x (t) +b u )
wherein psi r Is a correlationGate function, ω rc Is the weight, ω, of the output state value at the last moment in the correlation gate function rx As a weight of the predicted value of the input at time t in the correlation gate function, b r An offset input that is a function of the correlation gate; psi u To update the gate function, ω uc To update the weight, ω, of the output state value in the gate function for the last moment in time ux To update the weight of the predicted value of the input at time t in the gate function, b u An offset input that is a function of the correlation gate;
as a candidate function, ω cc As a weight, ω, of the output state value at the previous time in the candidate function cx As a weight of the predicted value of the input at time t in the candidate function, b c Is the offset input to the candidate function.
Further, step S4 comprises the following sub-steps:
s41: establishing an intelligent simulation platform according to the operation equipment, the operation data and the management data of the physical power grid, and setting the parameters of the operation equipment in the intelligent simulation platform according to the operation parameters of the physical power grid;
s42: inputting the load predicted value into the intelligent simulation platform, observing the running state and preset running parameters of each running device in the intelligent simulation platform, inputting the preset running parameters into the entity power grid, and collecting the real load value in the entity power grid once every a period of time t to obtain a real load value data set (y) 1 ，y 2 ，···，y z )；
S43: and evaluating the load predicted value and the real load value by utilizing the root mean square error RMSE, the average absolute error MAE and the average absolute percentage error MAPE.
Further, the method for calculating the root mean square error in step S43 is:
wherein, y l For one of the real load value data setsAnd (4) a real load value.
Further, in step S43, the root mean square error RMSE is compared with the corresponding threshold, and if RMSE is less than or equal to RMSE Threshold value If so, the intelligent simulation platform is accurately built; otherwise, the operation parameters of the operation equipment in the intelligent simulation platform are corrected, and the step S42 is returned.
The invention has the beneficial effects that:
the GRU-TCN prediction model is designed by combining a Time Convolution Network (TCN) and a Gated Recursion Unit (GRU) network, high-dimensional data features extracted by the TCN network model are input into the GRU network for prediction to obtain a high-precision prediction result, and the built digital twin power grid is corrected by the prediction result, so that a more accurate service level is provided for the digital power grid. The method provided by the invention designs a digital power grid architecture by utilizing digital twins, and realizes the digitization of the system through the mapping between the physical entity PG and the digital twins power grid, so that the operation effect of the system is more obvious. In order to improve the accuracy of load prediction, a Time Convolution Network (TCN) is adopted to extract high-dimensional characteristics of the load, and a Gated Recursive Unit (GRU) network with a simple structure and strong learning capacity is used for load prediction, so that the prediction error is further reduced.
Drawings
FIG. 1 is a block diagram of the GRU-TCN model.
Detailed Description
In order to more clearly understand the technical features, objects, and effects of the present invention, specific embodiments of the present invention will now be described. It should be understood that the detailed description and specific examples, while indicating embodiments of the invention, are given by way of illustration only, not by way of limitation, i.e., the embodiments described are intended as a selection of the best mode contemplated for carrying out the invention, not as a full mode. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment provides a power grid load prediction method based on a GRU-TCN model, which comprises the following steps:
s1: collecting load of each device in power grid of forecast dayData forming a time-series payload data set
Wherein n is the type of the load, and t is the acquisition time;
s2: normalizing the collected load data, selecting the load data of each device in the power grid on similar days according to the correlation value model, and inputting the load data into a load data group
In expanding the load data set
Obtaining a normalized load prediction data set;
s3: inputting the load prediction data group into a GRU-TCN model, and outputting a load prediction value of each time period of the power grid;
s4: establishing a digital twin power grid, inputting a load predicted value into the digital twin power grid, and outputting a real load value; and evaluating the predicted load value and the real load value by utilizing the root mean square error RMSE, and correcting the control parameters of the digital twin power grid.
Preferably, step S2 comprises the following sub-steps:
s21: load data group
Is normalized and mapped to the interval [0,1 ]]In the interior, a normalized value is obtained
Wherein X (t) For load data sets
Any of the load data of (a) is loaded,
are load data sets respectively
Maximum and minimum values of;
s22: computing load data set
All normalized values in (1)
Average value of (2)
S23: load data set for collecting devices in power grid on similar days
Where m = n, and using the normalization method in step S21, for the load data set
Load data X in m Normalization is carried out to obtain a normalized value x m ′；
S24: in order to avoid increasing the prediction time of the model due to excessive data and influence of irrelevant variables on the prediction accuracy, the normalization value x is utilized m ' and mean value
Calculating the degree of correlation R X ：
S25: similarity R X Similarity threshold R X threshold By comparison, if R X ≥R X threshold Then the load data X is added m Adding to load data set
The preparation method comprises the following steps of (1) performing; if R is X ＜R X threshold Then prove the load data X m Deleting the load data X for irrelevant variables m ；
S26: traversing load data X of all the devices of the power grid collected in similar days m Executing steps S23-S25 to obtain a load prediction data set with time as a sequence after normalization
s is the amount of increased load data.
Preferably, step S3 comprises the following sub-steps:
s31: load prediction data set
As an input matrix;
s32: inputting the input matrix into the TCN network model, outputting the output matrix
The feature matrix of (F):
i＝[1，2，···，l]，k＝[1，2，···，K]，K≥1
wherein, W k V is the amount of data in the output feature matrix F,
for the output of the i-th layer residual convolution, f (-) is a residual function of the TCN network model, K is a time sequence, K is 2K times of the initial time, and K is a hyper-parameter;
s33: inputting an output matrix F into the fully-connected layer, and outputting a prediction matrix Y = [ Y ] composed of predicted values 1 ，Y 2 ,…,Y v ]V is the number of predicted values in the prediction matrix;
s34: inputting the prediction matrix Y into a GRU network model, and outputting a state value c corresponding to the t moment (t) ：
ψ r ＝sigmoid(ω rc c (t-1) +ω rx x (t) +b r )
ψ u ＝sigmoid(ω uc c (t-1) +ω ux x (t) +b u )
Wherein psi r As a function of the correlation gate, ω rc As a weight, ω, of the output state value at the last instant in the correlation gate function rx As a weight of the predicted value of the input at time t in the correlation gate function, b r An offset input that is a function of the correlation gate; psi u To update the gate function, ω uc To update the weight, ω, of the output state value in the gate function for the last moment in time ux To update the weights of the predicted values input at time t in the gate function,b u an offset input that is a function of the correlation gate;
as a candidate function, ω cc As a weight, ω, of the output state value at the previous time instant in the candidate function cx As a weight of the predicted value of the input at time t in the candidate function, b c Is the offset input to the candidate function.
As shown in FIG. 1, since the feature learning of the GRU network model is not comprehensive enough, the TCN network model is used to predict the data set from the load first
And extracting high-dimensional features to strengthen the data learning capability of the GRU network model. The TCN is a neural network model incorporating extended causal convolution and residual concatenation for sequence modeling or data. The application of causal convolution in a TCN network model can guarantee the integrity of data. The application of the extended causal convolution may allow the TCN network model to have a larger field of view with a smaller number of layers, and thus may receive longer historical data. Where the extended causal convolution uses the ReLU activation function and performs weight renormalization and Dropout regularization. The residual connection jumps from the input to the output through the connection, which ensures the stability of the TCN in more layers.
S34: state value c to be output (t) As the load prediction value of the power grid.
Preferably, step S4 comprises the following sub-steps:
s41: establishing an intelligent simulation platform according to the operation equipment, the operation data and the management data of the physical power grid, and setting the parameters of the operation equipment in the intelligent simulation platform according to the operation parameters of the physical power grid;
s42: inputting the predicted load value into an intelligent simulation platform, observing the running state and preset running parameters of each running device in the intelligent simulation platform, inputting the preset running parameters into an entity power grid, and acquiring the real load value in the entity power grid once every a period of time t to obtain a real load value data set (y) 1 ，y 2 ，···，y z )；
S43: and (3) evaluating the load predicted value and the real load value by utilizing the root mean square error RMSE, the average absolute error MAE and the average absolute percentage error MAPE:
wherein, y l One of the real load values in the set of real load value data.
Comparing the root mean square error RMSE with a corresponding threshold, if RMSE is less than or equal to RMSE Threshold value If so, indicating that the intelligent simulation platform is accurately built; otherwise, the operation parameters of the operation equipment in the intelligent simulation platform are corrected, and the step S42 is returned.
According to the method, a Time Convolution Network (TCN) and a Gated Recursion Unit (GRU) network are fused to design a GRU-TCN prediction model, high-dimensional data characteristics extracted by the TCN network model are input into the GRU network for prediction, a high-precision prediction result is obtained, the built digital twin power grid is corrected by the prediction result, and a more precise business service level is provided for the digital power grid. The method provided by the invention designs a digital power grid architecture by utilizing digital twins, realizes the digitization of the system through the mapping between the physical entity PG and the digital twins power grid, and makes the operation effect of the system more obvious. In order to improve the accuracy of load prediction, a Time Convolution Network (TCN) is adopted to extract high-dimensional characteristics of the load, and a Gated Recursion Unit (GRU) network with a simple structure and strong learning capacity is used for load prediction, so that the prediction error is further reduced.
The digital power grid can optimize and match power transmission according to the load prediction result, and the presented power grid data has strong significance. The prediction result of the method is ideal, and the operation requirement of the digital power grid can be met. The precision and the speed of the data-driven load prediction method are further improved, the load composition of the digital power grid is diversified, and the characteristics of the load composition are strongly changed randomly. Follow-up work will study the combination strategy of data driving and mechanism analysis, and propose a high-performance load prediction method based on an advanced data preprocessing technology.
It should be noted that the foregoing method embodiments are described as a series of acts or combinations for simplicity in description, but it should be understood by those skilled in the art that the present application is not limited by the order of acts described, as some steps may occur in other orders or concurrently depending on the application. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required in this application.
Claims (10)
1. A power grid load prediction method based on a GRU-TCN model is characterized by comprising the following steps:
s1: load data of each device in the power grid on the forecast day are collected to form a load data group with time as a sequence
Wherein n is the type of the load, and t is the acquisition time;
s2: normalizing the collected load data, selecting the load data of each device in the power grid on similar days according to the correlation value model, and inputting the load data into a load data group
In expanding the load data group
Obtaining a normalized load prediction data set;
s3: inputting the load prediction data group into a GRU-TCN model, and outputting a load prediction value of each time interval of the power grid;
s4: establishing a digital twin power grid, inputting the load predicted value into the digital twin power grid, and outputting a real load value; and evaluating the load predicted value and the real load value by utilizing the root mean square error RMSE, and correcting the control parameters of the digital twin power grid.
2. The GRU-TCN model-based grid load prediction method according to claim 1, wherein step S2 comprises the following sub-steps:
s21: load data group
Is normalized and mapped to the interval [0,1 ]]In order to obtain a normalized value
S22: computing load data set
All normalized values in (1)
Average value of (2)
S23: acquiring load data groups of devices in power grid on similar days
Where m = n, and using the normalization method in step S21, for the load data set
Load data X in (1) m Normalization is carried out to obtain a normalized value x m ′；
S24: using normalized value x m ' and average value
Calculating the degree of correlation R X ；
S25: similarity R X Similarity threshold R X threshold By comparison, if R X ≥R X threshold Then the load data X is added m Adding to load data set
Performing the following steps; if R is X ＜R X threshold Then prove the load data X m Deleting the load data X for irrelevant variables m ；
S26: traversing load data X of all the devices of the power grid collected in similar days m Executing steps S23-S25 to obtain a load prediction data set with time as a sequence after normalization
s is the increased load data amount.
3. The GRU-TCN model-based power grid load prediction method according to claim 2, wherein the normalization method in step S21 is:
wherein X (t) For load data sets
Any of the load data of (1) is loaded,
are load data sets respectively
Maximum and minimum values of (a).
4. The GRU-TCN model-based power grid load prediction method of claim 2, wherein the normalization value x is used in step S24 m ' and average value
Calculating the degree of correlation R X The method comprises the following steps:
5. the GRU-TCN model-based grid load prediction method according to claim 2, characterized in that step S3 comprises the following sub-steps:
s31: grouping load prediction data
As an input matrix;
s32: inputting the input matrix into the TCN network model, outputting the output matrix
A feature matrix F;
s33: inputting an output matrix F into the fully-connected layer, and outputting a prediction matrix Y = [ Y ] composed of prediction values 1 ，Y 2 ,…,Y v ]V is the number of predicted values in the prediction matrix;
s34: inputting the prediction matrix Y into a GRU network model, and outputting a state value c corresponding to the time t (t) ；
S34: state value c to be output (t) As the load prediction value of the power grid.
6. The GRU-TCN model-based power grid load prediction method of claim 5, characterized in thatIn step S32
The feature matrix F of (a) is:
i＝[1，2，···，l]，k＝[1，2，···，K]，K≥1
wherein, W k V is the amount of data in the output feature matrix F,
and f (-) is the residual error function of the TCN network model, K is a time sequence, K is 2K times of the initial time, and K is a hyperparameter.
7. The GRU-TCN model-based power grid load prediction method of claim 6, wherein the state value c corresponding to the time t in the step S34 (t) Comprises the following steps:
ψ r ＝sigmoid(ω rc c (t-1) +ω rx x (t) +b r )
ψ u ＝sigmoid(ω uc c (t-1) +ω ux x (t) +b u )
wherein psi r As a function of the correlation gate, ω rc Is the weight, ω, of the output state value at the last moment in the correlation gate function rx For the weight of the predicted value of the input at time t in the correlation gate function, b r An offset input that is a function of the correlation gate; psi u To update the gate function, ω uc To update the weight, ω, of the output state value in the gate function for the last moment in time ux To update the weight of the predicted value of the input at time t in the gate function, b u An offset input that is a function of the correlation gate;
as a candidate function, ω cc As a weight, ω, of the output state value at the previous time in the candidate function cx As a weight of the predicted value of the input at time t in the candidate function, b c Is the offset input to the candidate function.
8. The GRU-TCN model-based grid load prediction method according to claim 7, wherein step S4 comprises the sub-steps of:
s41: establishing an intelligent simulation platform according to the operation equipment, the operation data and the management data of the physical power grid, and setting the parameters of the operation equipment in the intelligent simulation platform according to the operation parameters of the physical power grid;
s42: inputting the load predicted value into the intelligent simulation platform, observing the running state and preset running parameters of each running device in the intelligent simulation platform, inputting the preset running parameters into the entity power grid, and collecting the real load value in the entity power grid once every a period of time t to obtain a real load value data set (y) 1 ，y 2 ，···，y z )；
S43: and evaluating the load predicted value and the real load value by utilizing the root mean square error RMSE, the average absolute error MAE and the average absolute percentage error MAPE.
9. The method for predicting the grid load based on the GRU-TCN model according to claim 8, wherein the root mean square error in step S43 is calculated by:
wherein, y l One of the real load values in the set of real load value data.
10. The GRU-TCN model-based power grid load prediction method of claim 8, wherein the Root Mean Square Error (RMSE) is compared with a corresponding threshold in step S43, and if RMSE is less than or equal to RMSE Threshold value If so, indicating that the intelligent simulation platform is accurately built; otherwise, the operation parameters of the operation equipment in the intelligent simulation platform are corrected, and the step S42 is returned.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211389573.9A CN115545345A (en) | 2022-11-08 | 2022-11-08 | Power grid load prediction method based on GRU-TCN model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211389573.9A CN115545345A (en) | 2022-11-08 | 2022-11-08 | Power grid load prediction method based on GRU-TCN model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115545345A true CN115545345A (en) | 2022-12-30 |
Family
ID=84719999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211389573.9A Pending CN115545345A (en) | 2022-11-08 | 2022-11-08 | Power grid load prediction method based on GRU-TCN model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115545345A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116707141A (en) * | 2023-07-31 | 2023-09-05 | 国网山东省电力公司阳信县供电公司 | Power operation data analysis method and system |
| CN117391310A (en) * | 2023-12-04 | 2024-01-12 | 南京瀚元科技有限公司 | Power grid equipment operation state prediction and optimization method based on digital twin technology |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106600463A (en) * | 2016-12-21 | 2017-04-26 | 广东电网有限责任公司电力调度控制中心 | Local shape similarity ultra short-period load prediction method and apparatus |
| CN112116144A (en) * | 2020-09-15 | 2020-12-22 | 山东科技大学 | Regional power distribution network short-term load prediction method |
| CN112926770A (en) * | 2021-02-22 | 2021-06-08 | 胡书恺 | Unified data metering and collecting system based on digital twins |
| CN114611738A (en) * | 2020-12-08 | 2022-06-10 | 南京工程学院 | Load prediction method based on user electricity consumption behavior analysis |
-
2022
- 2022-11-08 CN CN202211389573.9A patent/CN115545345A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106600463A (en) * | 2016-12-21 | 2017-04-26 | 广东电网有限责任公司电力调度控制中心 | Local shape similarity ultra short-period load prediction method and apparatus |
| CN112116144A (en) * | 2020-09-15 | 2020-12-22 | 山东科技大学 | Regional power distribution network short-term load prediction method |
| CN114611738A (en) * | 2020-12-08 | 2022-06-10 | 南京工程学院 | Load prediction method based on user electricity consumption behavior analysis |
| CN112926770A (en) * | 2021-02-22 | 2021-06-08 | 胡书恺 | Unified data metering and collecting system based on digital twins |
Non-Patent Citations (3)
| Title |
|---|
| 王俊;李强;: "基于相似日的96点短期负荷预测实用方法及其应用", 上海电力, no. 03, pages 1 - 10 * |
| 郑豪丰等: "基于多负荷特征和 TCN-GRU 神经网络的负荷预测", 《中国电力》, vol. 55, no. 11, pages 1 - 10 * |
| 郭玲等: "基于 TCN-GＲU 模型的短期负荷预测方法", 《电力工程技术》, vol. 40, no. 3, pages 1 - 10 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116707141A (en) * | 2023-07-31 | 2023-09-05 | 国网山东省电力公司阳信县供电公司 | Power operation data analysis method and system |
| CN116707141B (en) * | 2023-07-31 | 2023-11-17 | 国网山东省电力公司阳信县供电公司 | Power operation data analysis method and system |
| CN117391310A (en) * | 2023-12-04 | 2024-01-12 | 南京瀚元科技有限公司 | Power grid equipment operation state prediction and optimization method based on digital twin technology |
| CN117391310B (en) * | 2023-12-04 | 2024-03-08 | 南京瀚元科技有限公司 | Power grid equipment operation state prediction and optimization method based on digital twin technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Wind speed forecasting based on quantile regression minimal gated memory network and kernel density estimation | |
| Shen et al. | Wind speed prediction of unmanned sailboat based on CNN and LSTM hybrid neural network | |
| Xuan et al. | Multi-model fusion short-term load forecasting based on random forest feature selection and hybrid neural network | |
| Yuan et al. | Short-term wind power prediction based on LSSVM–GSA model | |
| CN115545345A (en) | Power grid load prediction method based on GRU-TCN model | |
| CN105303262A (en) | Short period load prediction method based on kernel principle component analysis and random forest | |
| CN110969290B (en) | Runoff probability prediction method and system based on deep learning | |
| CN111860982A (en) | Wind power plant short-term wind power prediction method based on VMD-FCM-GRU | |
| Huang et al. | An integrated computational intelligence approach to product concept generation and evaluation | |
| Poczęta et al. | Learning fuzzy cognitive maps using structure optimization genetic algorithm | |
| CN112434848A (en) | Nonlinear weighted combination wind power prediction method based on deep belief network | |
| CN113836783B (en) | Digital regression model modeling method for main beam temperature-induced deflection monitoring reference value of cable-stayed bridge | |
| CN112834927A (en) | Lithium battery residual life prediction method, system, device and medium | |
| CN115860177A (en) | Photovoltaic power generation power prediction method based on combined machine learning model and application thereof | |
| CN112803398A (en) | Load prediction method and system based on empirical mode decomposition and deep neural network | |
| CN114118569A (en) | Wind power multi-step prediction method based on multi-mode multi-task Transformer network | |
| CN115759415A (en) | Power consumption demand prediction method based on LSTM-SVR | |
| CN108537581B (en) | Energy consumption time series prediction method and device based on GMDH selective combination | |
| Bian et al. | Research on short-term load forecasting based on accumulated temperature effect and improved temporal convolutional network | |
| CN112232570A (en) | Forward active total electric quantity prediction method and device and readable storage medium | |
| CN116885699A (en) | Power load prediction method based on dual-attention mechanism | |
| CN115481788B (en) | Phase change energy storage system load prediction method and system | |
| CN116865232A (en) | Wind speed error correction-based medium-and-long-term wind power prediction method and system | |
| CN115130662A (en) | Power distribution network time-varying topological state estimation method based on transfer learning | |
| CN113962431A (en) | Bus load prediction method for two-stage feature processing |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20221230 |
|
| RJ01 | Rejection of invention patent application after publication |