CN111415050A - Short-term load prediction method and short-term load prediction model training method and device - Google Patents
Short-term load prediction method and short-term load prediction model training method and device Download PDFInfo
- Publication number
- CN111415050A CN111415050A CN202010341288.4A CN202010341288A CN111415050A CN 111415050 A CN111415050 A CN 111415050A CN 202010341288 A CN202010341288 A CN 202010341288A CN 111415050 A CN111415050 A CN 111415050A
- Authority
- CN
- China
- Prior art keywords
- time sequence
- data
- load
- load time
- 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
- 238000000034 method Methods 0.000 title claims abstract description 45
- 230000015654 memory Effects 0.000 claims abstract description 85
- 230000006403 short-term memory Effects 0.000 claims description 35
- 230000003442 weekly effect Effects 0.000 claims description 16
- 238000013277 forecasting method Methods 0.000 abstract description 4
- 230000007787 long-term memory Effects 0.000 description 28
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 239000002184 metal Substances 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000001537 neural effect Effects 0.000 description 3
- 238000003062 neural network model Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000004590 computer program Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000013439 planning Methods 0.000 description 1
- 230000000750 progressive effect 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- 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
-
- 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
-
- 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)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Marketing (AREA)
- Software Systems (AREA)
- Tourism & Hospitality (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Operations Research (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a short-term load forecasting method, a short-term load forecasting model training method, a device, a readable medium and electronic equipment, wherein the method comprises the following steps: acquiring data to be predicted, wherein the data to be predicted comprises load time sequence data and influence factor data; acquiring a first load time sequence characteristic based on the load time sequence data and a long-short term memory network in a trained prediction model; acquiring influence factor characteristics based on the influence factor data and a fully connected network in the trained prediction model; and acquiring a short-term load prediction result based on the first load time sequence characteristic, the influence factor characteristic and an output network in the trained prediction model. According to the technical scheme provided by the invention, multisource data, namely load time sequence data and influence factor data, are introduced, and multisource characteristics, namely first load time sequence characteristics and influence factor characteristics, are accurately learned by using a long-short-term memory network and a full-connection network, so that the accuracy of a short-term load prediction result obtained based on the multisource characteristics is higher.
Description
Technical Field
The invention relates to the field of energy, in particular to a short-term load forecasting method, a short-term load forecasting model training method and a short-term load forecasting model training device.
Background
Load forecasting is an important basis of energy planning, economic operation and energy management, and generally comprises long-term load forecasting, medium-term load forecasting and short-term load forecasting, wherein the short-term load forecasting generally refers to forecasting the load of a forecasting object in one day or one week in the future, and the short-term load is characterized by being greatly influenced by factors such as weather, equipment conditions, important social activities and the like, so that the short-term load is accurately forecasted with great difficulty.
At present, when a short-term load is predicted, time series models such as ARMA (autoregressive moving average) are often used, and a short-term load prediction result obtained based on the time series models is often obtained according to load time series data in a relatively short time interval, so that the influence of a time factor on the short-term load prediction result is only considered, the influence factor is considered to be too single, and the accuracy of the obtained short-term prediction result is low.
Disclosure of Invention
The invention provides a short-term load forecasting method, a short-term load forecasting model training device, a readable medium and electronic equipment.
In a first aspect, the present invention provides a short-term load prediction method, including:
acquiring data to be predicted, wherein the data to be predicted comprises load time sequence data and influence factor data;
acquiring a first load time sequence characteristic based on the load time sequence data and a long-short term memory network in a trained prediction model;
acquiring influence factor characteristics based on the influence factor data and a fully connected network in the trained prediction model;
and acquiring a short-term load prediction result corresponding to the data to be predicted based on the first load time sequence characteristic, the influence factor characteristic and an output network in the trained prediction model.
Preferably, the first and second electrodes are formed of a metal,
the acquiring the first load timing characteristics, if the long-short term memory network comprises a first long-short term memory network layer and a second long-short term memory network layer, comprises:
inputting the load time sequence data into the first long-short term memory network layer to obtain a second load time sequence characteristic;
and inputting the second load time sequence characteristic into the second long-short term memory network layer to obtain a first load time sequence characteristic.
Preferably, the first and second electrodes are formed of a metal,
the load time sequence data comprises time load time sequence data, daily load time sequence data and weekly load time sequence data.
Preferably, the first and second electrodes are formed of a metal,
the long-short term memory network comprises a first long-short term memory network, a second long-short term memory network and a third long-short term memory network, and the acquiring the first load timing characteristic comprises:
inputting the time load time sequence data into the first long-short term memory network in the trained prediction model to obtain time load time sequence characteristics;
inputting the daily load time sequence data into the second long-short term memory network in the trained prediction model to obtain daily load time sequence characteristics;
inputting the weekly load time sequence data into the third long-short term memory network in the trained prediction model to obtain the weekly load time sequence characteristic;
and acquiring a first load time sequence characteristic based on the moment load time sequence characteristic, the daily load time sequence characteristic and the weekly load time sequence characteristic.
Preferably, the first and second electrodes are formed of a metal,
the fully connected network comprises a first fully connected network layer and a second fully connected network layer, then the obtaining influencing factor characteristics comprises:
inputting the influence factor data into the first fully-connected network layer in the trained prediction model to obtain coding factor characteristics;
and inputting the coding factor characteristics into the second fully-connected network layer in the trained prediction model to obtain the influence factor characteristics.
In a second aspect, the present invention provides a short-term load prediction model training method, including:
acquiring load time sequence training data and influence factor training data;
training a long-short term memory network in a prediction model according to the load time sequence training data, wherein the long-short term memory network is used for acquiring load time sequence characteristics;
training a fully-connected network in the prediction model according to the influence factor training data, wherein the fully-connected network is used for acquiring influence factor characteristics;
and training an output network in the prediction model according to the load time sequence characteristics, the influence factor characteristics and the load time sequence training data, wherein the output network is used for outputting a short-term load prediction result.
In a third aspect, the present invention provides a short-term load prediction apparatus, including:
the data acquisition module is used for acquiring data to be predicted, and the data to be predicted comprises load time sequence data and influence factor data;
the time sequence characteristic acquisition module is used for acquiring a first load time sequence characteristic based on the load time sequence data and a long-short term memory network in a trained prediction model;
the factor characteristic acquisition model is used for acquiring the influence factor characteristics based on the influence factor data and a fully-connected network in the trained prediction model;
and the prediction result obtaining module is used for obtaining a short-term load prediction result corresponding to the data to be predicted based on the first load time sequence characteristic, the influence factor characteristic and an output network in the trained prediction model.
In a fourth aspect, the present invention provides a training apparatus for a short-term load prediction model, including:
the training data acquisition module is used for acquiring load time sequence training data and influence factor training data;
the memory network training module is used for training a long-short term memory network in a prediction model according to the load time sequence training data, and the long-short term memory network is used for acquiring load time sequence characteristics;
the connection network training module is used for training a full connection network in the prediction model according to the influence factor training data, and the full connection network is used for acquiring influence factor characteristics;
and the output network training module is used for training an output network in the prediction model according to the load time sequence characteristics, the influence factor characteristics and the load time sequence training data, and the output network is used for outputting a short-term load prediction result.
In a fifth aspect, the invention provides a readable medium comprising executable instructions which, when executed by a processor of an electronic device, cause the electronic device to perform the method according to any one of the first or second aspects.
In a sixth aspect, the present invention provides an electronic device, comprising a processor and a memory storing execution instructions, wherein when the processor executes the execution instructions stored in the memory, the processor performs the method according to any one of the first aspect or the second aspect.
The invention provides a short-term load forecasting method, a short-term load forecasting model training method, a device, a readable medium and electronic equipment.A forecasting model comprising a long-term and short-term memory network, a full-connection network and an output network is trained in advance, and the forecasting model takes load time sequence data and influence factor data as data to be forecasted, so that multi-source data can be introduced into an input layer; then, accurately learning the load time sequence characteristics of the load time sequence data by using a long and short term memory network in the prediction model, and accurately learning the influence factor characteristics of the influence factor data by using a fully connected network in the prediction model, namely accurately acquiring the multi-source characteristics by using the long and short term memory network and the fully connected network; and further fusing the first load time sequence characteristics and the influence factor characteristics in an output network of the prediction model, and outputting a short-term load prediction result, wherein the short-term load prediction result is obtained by fusing the multi-source characteristics, and has higher accuracy.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed for the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
Fig. 1 is a schematic flow chart of a short-term load prediction method according to an embodiment of the present invention;
FIG. 2 is a flow chart illustrating another short term load prediction method according to an embodiment of the present invention;
FIG. 3 is a flowchart illustrating a short term load prediction method according to another embodiment of the present invention;
FIG. 4 is a flowchart illustrating a short-term load prediction method according to another embodiment of the present invention;
FIG. 5 is a schematic flow chart illustrating a short-term load prediction model training method according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a short-term load prediction apparatus according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of a short-term load prediction model training apparatus according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of an electronic device provided in an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail and completely with reference to the following embodiments and accompanying drawings. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In a first aspect, as shown in fig. 1, an embodiment of the present invention provides a short-term load prediction method, including:
102, acquiring a first load time sequence characteristic based on the load time sequence data and a long-short term memory network in a trained prediction model;
103, acquiring influence factor characteristics based on the influence factor data and a fully-connected network in the trained prediction model;
and 104, acquiring a short-term load prediction result corresponding to the data to be predicted based on the first load time sequence characteristic, the influence factor characteristic and an output network in the trained prediction model.
As shown in the embodiment of fig. 1, a prediction model comprising a long-term and short-term memory network, a fully-connected network and an output network is trained in advance, and the prediction model takes load time series data and influence factor data as data to be predicted, so that multi-source data can be introduced into an input layer; then, accurately learning the load time sequence characteristics of the load time sequence data by using a long and short term memory network in the prediction model, and accurately learning the influence factor characteristics of the influence factor data by using a fully connected network in the prediction model, namely accurately acquiring the multi-source characteristics by using the long and short term memory network and the fully connected network; and further fusing the first load time sequence characteristics and the influence factor characteristics in an output network of the prediction model, and outputting a short-term load prediction result, wherein the short-term load prediction result is obtained by fusing the multi-source characteristics, and has higher accuracy.
Specifically, the load time series data may be composed of load data corresponding to different historical times, and the influence factor data may be composed of data corresponding to factors that influence short-term loads, such as weather factors and equipment condition factors.
It should be noted that the trained prediction model mentioned in this embodiment is substantially a dual-branch neural network model, one is a long-short term memory network, and the other is a fully-connected network, and then features learned by the long-short term memory network and the fully-connected network are fused in the output network, and a short-term load prediction result is output. Because the data to be predicted comprises the load time sequence data and the influence factor data, the load time sequence data and the influence factor data can be respectively input to the double-branch neural network model, namely the load time sequence data is directly used as the input data of the long-term and short-term memory network, and the influence factor data is used as the input data of the full-connection network. Certainly, the data to be predicted can also be used as unified input data of the double-branch neural network model, then the input long-term and short-term memory network only learns the load time sequence data through parameter setting, the fully-connected network only learns the influence factor data, and a specific data input user can set the data according to an actual service scene.
As shown in fig. 2, in an embodiment of the present invention, the long-short term memory network includes a first long-short term memory network layer and a second long-short term memory network layer, and step 102 includes:
step 1021: inputting the load time sequence data into the first long-short term memory network layer to obtain a second load time sequence characteristic;
step 1022: and inputting the second load time sequence characteristic into the second long-short term memory network layer to obtain a first load time sequence characteristic.
In the above embodiment, the long-short term memory network has a two-layer network structure, i.e., a first long-short term memory network layer and a second long-short term memory network layer, wherein the first long-short term memory network layer learns the load timing data, the output result of the first long-short term memory network layer is the second load timing characteristic, then the second load timing characteristic is used as the input data of the second long-short term memory network layer, and the output result of the second long-short term memory network layer is the first load timing characteristic by learning the second load timing characteristic. Therefore, the load time sequence data can be effectively learned through the two layers of long and short term memory networks, accurate load time sequence characteristics can be obtained, and the accuracy of the finally obtained short term load prediction result can be further ensured.
In one embodiment of the invention, the load timing data includes time of day load timing data, daily load timing data and weekly load timing data.
In the above embodiment, the load data corresponding to different historical times are obtained to ensure the multi-source of the data in the time dimension, that is, not only the load time sequence data of a relatively short time interval but also the load time sequence data of adjacent days and adjacent weeks are considered. The time load time sequence data consists of load data adjacent to the time to be predicted, the day load time sequence data consists of load data at the same time as the adjacent day and the time to be predicted, the week load time sequence data consists of load data at the same time as the adjacent week and the time to be predicted, if the short-term load from 0 point of Wednesday to 24 points is to be predicted, the week load time sequence data consists of short-term load from 0 point of Wednesday to 24 points, and the short-term load can be influenced by social important activities, and the social important activities are usually held regularly, so that the load data of the adjacent week is introduced, and the load data of the adjacent month can also be introduced, namely the load time sequence data can also comprise month load time sequence data. Historical load data can be fully utilized by introducing daily load time sequence data and weekly load time sequence data, and accurate short-term load prediction results can be obtained beneficially.
As shown in fig. 3, in an embodiment of the present invention, when the load time series data includes time load time series data, daily load time series data and weekly load time series data, the long-short term memory network includes a first long-short term memory network, a second long-short term memory network and a third long-short term memory network, step 102 includes:
In the above embodiment, different long and short term memory networks are used for learning for different load time sequence data, so that time sequence characteristics corresponding to different load time sequence data can be effectively obtained. Specifically, a first long and short term memory network is used for learning time load time sequence characteristics corresponding to time load time sequence data, a second long and short term memory network is used for learning daily load time sequence characteristics corresponding to daily load time sequence data, and a third long and short term memory network is used for learning weekly load time sequence characteristics corresponding to weekly load time sequence data, wherein the first long and short term memory network, the second long and short term memory network and the third long and short term memory network can be two layers of long and short term memory networks, and can be long and short term memory networks with other network structures; the network structures of the first long-short term memory network, the second long-short term memory network and the third long-short term memory network can be the same or different, and when the network structures are different, the time load time sequence feature, the day load time sequence feature and the week load time sequence feature need to be fused subsequently, so that the time load time sequence feature, the day load time sequence feature and the week load time sequence feature need to be ensured to be obtained with the same dimension. The network structure users of the first long-short term memory network, the second long-short term memory network and the third long-short term memory network can be set up according to actual business scenes.
The first load timing characteristic is composed of a time load timing characteristic, a day load timing characteristic and a week load timing characteristic. In a possible implementation manner, the first load time sequence feature is obtained by linearly fusing the load time sequence feature, the daily load time sequence feature and the weekly load time sequence feature, that is, the moment load time sequence feature, the daily load time sequence feature and the weekly load time sequence feature are input to a neural unit provided with a linear activation function to be linearly fused, so that the first load time sequence feature is obtained, and the first load time sequence feature is input to the output network. In another possible implementation manner, the time load timing sequence feature, the day load timing sequence feature and the week load timing sequence feature are components of the first load timing sequence feature, that is, the time load timing sequence feature, the day load timing sequence feature and the week load timing sequence feature are determined, that is, the first load timing sequence feature is determined, and the first load timing sequence feature is input to and output from the network, that is, the time load timing sequence feature, the day load timing sequence feature and the week load timing sequence feature are respectively input to and output from the network.
As shown in fig. 4, in an embodiment of the present invention, where the fully connected network includes a first fully connected network layer and a second fully connected network layer, step 103 includes:
In the above embodiment, the fully-connected network has a two-layer network structure, that is, the fully-connected network includes a first fully-connected network layer and a second fully-connected network layer, where the first fully-connected network layer may serve as an embedded layer to solve the problem of excessive parameters caused by the excessively high dimensionality of the input data, and a small number of parameters may increase the training speed and the learning speed. Because the influencing factor features obtained after passing through the fully-connected layer need to be fused with the first load time sequence features, in order to ensure that the dimensionalities of the influencing factor features and the first load time sequence features are consistent, the second fully-connected network layer needs to perform dimensionality adjustment on the coding factor features obtained through the first fully-connected network layer.
Specifically, the Re L u function can be used as the activation function of the neural unit in the first fully-connected network layer and the second fully-connected network layer, so that the convergence is faster when the prediction model is trained, and the obtained prediction model has higher prediction accuracy.
In an embodiment of the present invention, obtaining a short-term load prediction result corresponding to the data to be predicted based on the first load timing characteristic, the influencing factor characteristic, and an output network in the trained prediction model includes: and inputting the first load time sequence characteristic and the influence factor characteristic into an output network in the trained prediction model, wherein the activation function of a neural unit in the output network is a linear function, namely, the first load time sequence characteristic and the influence factor characteristic are linearly fused in the output network, and a short-term load prediction result is output. Specifically, when the first load time sequence characteristics include moment load time sequence characteristics, day load time sequence characteristics and week load time sequence characteristics, the moment load time sequence characteristics, the day load time sequence characteristics, the week load time sequence characteristics and the influence factor characteristics are linearly fused and then a short-term load prediction result is output, so that the short-term load prediction result is obtained by fusing the multi-source characteristics, and the accuracy is high.
In a second aspect, as shown in fig. 5, an embodiment of the present invention provides a short-term load prediction model training method, including:
and step 504, training an output network in the prediction model according to the load time sequence characteristics, the influence factor characteristics and the load time sequence training data, wherein the output network is used for outputting a short-term load prediction result.
In the above embodiment, the short-term load prediction model is composed of three parts, namely, a long-term and short-term memory network, a fully-connected network and an output network, and the training data is composed of two parts, namely, load time sequence training data and influencing factor training data. The method comprises the steps of training a long-term and short-term memory network by using load time sequence training data, training a full-connection network by using influencing factor training data, enabling the long-term and short-term memory network to more effectively learn load time sequence characteristics through respective training modes, enabling the full-connection network to more effectively acquire the influencing factor characteristics, then taking the load time sequence characteristics acquired by the long-term and short-term memory network and the influencing factor characteristics acquired by the full-connection network as input data of an output network after the long-term and short-term memory network and the full-connection network are trained, and training the output network according to load time sequence training data corresponding to load true values. It should be noted that the load timing training data and the influencing factor training data correspond to the same historical time, that is, have the same load true value, so that the output network may also be trained according to the load true value corresponding to the influencing factor training data.
Specifically, when the long-short term memory network in the prediction model is trained by using the load time sequence training data, the long-short term memory network is firstly provided with an output layer, then a loss function is constructed according to the output result of the output layer and the load true value, and the parameters of the long-short term memory network are continuously adjusted according to the loss function, so that the long-short term memory network which meets the preset precision is obtained. In this embodiment, the load timing characteristics need to be acquired by the long and short term memory network, so the output layer of the long and short term memory network needs to be removed, and the load timing characteristics acquired before the output layer are directly input into the output network of the prediction model. When the fully-connected network is trained by using the same influencing factor training data, an output layer can be arranged on the fully-connected network, after the fully-connected network which meets the preset precision is obtained, the output layer is removed, and the influencing factor characteristics obtained before the output layer are input into the output network of the prediction model.
It should be noted that in the above embodiment, separate training is performed before the output network, but it is also possible to train the short-term load prediction model by using a unified training method, that is, training data is obtained, the training data includes load time sequence training data and influencing factor training data, a short-term prediction model including a long-term and short-term memory network, a full-connection network and an output network is initialized, the long-term and short-term memory network in the prediction model learns the load time sequence characteristics in the load time sequence training data through parameter setting of the prediction model, the full-connection network learns the influencing factor characteristics in the influencing factor training data, then the load time sequence characteristics and the influencing factor characteristics are fused in the output network to obtain a short-term load prediction result, a loss function is constructed according to the short-term load prediction result and the load true value, and parameters in the initialized prediction model are continuously adjusted through the loss function, so as to obtain a short-term load prediction model which accords with preset precision.
In a third aspect, as shown in fig. 6, an embodiment of the present invention provides a short-term load prediction apparatus, including:
the data acquiring module 601 is configured to acquire data to be predicted, where the data to be predicted includes load time sequence data and influence factor data;
a timing characteristic obtaining module 602, configured to obtain a first load timing characteristic based on the load timing data and a long-term and short-term memory network in a trained prediction model;
a factor feature obtaining model 603, configured to obtain a factor feature based on the factor-affecting data and a fully connected network in the trained prediction model;
a prediction result obtaining module 604, configured to obtain a short-term load prediction result corresponding to the data to be predicted based on the first load timing characteristic, the influencing factor characteristic, and an output network in the trained prediction model.
In a fourth aspect, as shown in fig. 7, an embodiment of the present invention provides a training apparatus for a short-term load prediction model, including:
a training data acquisition module 701, configured to acquire load timing training data and influence factor training data;
a memory network training module 702, configured to train a long-short term memory network in a prediction model according to the load timing training data, where the long-short term memory network is used to obtain load timing characteristics;
a connection network training module 703, configured to train a full connection network in the prediction model according to the influence factor training data, where the full connection network is used to obtain influence factor characteristics;
and the output network training module 704 is configured to train an output network in the prediction model according to the load timing characteristics, the influence factor characteristics, and the load timing training data, where the output network is configured to output a short-term load prediction result.
Fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. On the hardware level, the electronic device includes a processor 801 and a memory 802 storing execution instructions, and optionally further includes an internal bus 803 and a network interface 804. The memory 802 may include a memory 8021, such as a Random-access memory (RAM), and may further include a non-volatile memory 8022 (e.g., at least 1 disk memory); the processor 801, the network interface 804, and the memory 802 may be connected to each other by an internal bus 803, and the internal bus 803 may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, an EISA (extended Industry Standard Architecture) bus, or the like; the internal bus 803 may be divided into an address bus, a data bus, a control bus, etc., which are indicated by only one double-headed arrow in fig. 8 for convenience of illustration, but do not indicate only one bus or one type of bus. Of course, the electronic device may also include hardware required for other services. When the processor 801 executes execution instructions stored by the memory 802, the processor 801 performs the method of any of the embodiments of the present invention and at least is used to perform the method as shown in fig. 1, fig. 2, fig. 3, fig. 4, fig. 5.
In a possible implementation manner, the processor reads corresponding execution instructions from the nonvolatile memory into the memory and then executes the execution instructions, and corresponding execution instructions can also be obtained from other equipment, so as to form a short-term load prediction device or a short-term load prediction model training device on a logic level. The processor executes the execution instructions stored in the memory, so that the executed execution instructions realize a short-term load prediction method or a short-term load prediction model training method provided by any embodiment of the invention.
The processor may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in a processor or instructions in the form of software. The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but also Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
Embodiments of the present invention further provide a computer-readable storage medium, which includes an execution instruction, and when a processor of an electronic device executes the execution instruction, the processor executes a method provided in any one of the embodiments of the present invention. The electronic device may specifically be the electronic device shown in fig. 8; the execution instruction is a computer program corresponding to a short-term load prediction device or a short-term load prediction model training device.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects.
The embodiments of the present invention are described in a progressive manner, and the same and similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, as for the apparatus embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or boiler that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or boiler. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or boiler that comprises the element.
The above description is only an example of the present invention, and is not intended to limit the present invention. Various modifications and alterations to this invention will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.
Claims (10)
1. A method for short-term load prediction, comprising:
acquiring data to be predicted, wherein the data to be predicted comprises load time sequence data and influence factor data;
acquiring a first load time sequence characteristic based on the load time sequence data and a long-short term memory network in a trained prediction model;
acquiring influence factor characteristics based on the influence factor data and a fully connected network in the trained prediction model;
and acquiring a short-term load prediction result corresponding to the data to be predicted based on the first load time sequence characteristic, the influence factor characteristic and an output network in the trained prediction model.
2. The short term load prediction method as claimed in claim 1, wherein the long short term memory network comprises a first long short term memory network layer and a second long short term memory network layer, and the obtaining the first load timing characteristic comprises:
inputting the load time sequence data into the first long-short term memory network layer to obtain a second load time sequence characteristic;
and inputting the second load time sequence characteristic into the second long-short term memory network layer to obtain a first load time sequence characteristic.
3. The method of short term load prediction according to claim 1, wherein the load time series data comprises time of day load time series data, day of day load time series data and week of day load time series data.
4. The short term load prediction method as claimed in claim 3, wherein the long short term memory network comprises a first long short term memory network, a second long short term memory network and a third long short term memory network, and the obtaining the first load timing characteristic comprises:
inputting the time load time sequence data into the first long-short term memory network in the trained prediction model to obtain time load time sequence characteristics;
inputting the daily load time sequence data into the second long-short term memory network in the trained prediction model to obtain daily load time sequence characteristics;
inputting the weekly load time sequence data into the third long-short term memory network in the trained prediction model to obtain the weekly load time sequence characteristic;
and acquiring a first load time sequence characteristic based on the moment load time sequence characteristic, the daily load time sequence characteristic and the weekly load time sequence characteristic.
5. The method of short term load prediction according to claim 1, wherein the fully connected network comprises a first fully connected network layer and a second fully connected network layer, and the obtaining the impact factor characteristics comprises:
inputting the influence factor data into the first fully-connected network layer in the trained prediction model to obtain coding factor characteristics;
and inputting the coding factor characteristics into the second fully-connected network layer in the trained prediction model to obtain the influence factor characteristics.
6. A short-term load prediction model training method is characterized in that,
acquiring load time sequence training data and influence factor training data;
training a long-short term memory network in a prediction model according to the load time sequence training data, wherein the long-short term memory network is used for acquiring load time sequence characteristics;
training a fully-connected network in the prediction model according to the influence factor training data, wherein the fully-connected network is used for acquiring influence factor characteristics;
and training an output network in the prediction model according to the load time sequence characteristics, the influence factor characteristics and the load time sequence training data, wherein the output network is used for outputting a short-term load prediction result.
7. A short-term load prediction apparatus, comprising:
the data acquisition module is used for acquiring data to be predicted, and the data to be predicted comprises load time sequence data and influence factor data;
the time sequence characteristic acquisition module is used for acquiring a first load time sequence characteristic based on the load time sequence data and a long-short term memory network in a trained prediction model;
the factor characteristic acquisition model is used for acquiring the influence factor characteristics based on the influence factor data and a fully-connected network in the trained prediction model;
and the prediction result obtaining module is used for obtaining a short-term load prediction result corresponding to the data to be predicted based on the first load time sequence characteristic, the influence factor characteristic and an output network in the trained prediction model.
8. A training apparatus for a short-term load prediction model, comprising:
the training data acquisition module is used for acquiring load time sequence training data and influence factor training data;
the memory network training module is used for training a long-short term memory network in a prediction model according to the load time sequence training data, and the long-short term memory network is used for acquiring load time sequence characteristics;
the connection network training module is used for training a full connection network in the prediction model according to the influence factor training data, and the full connection network is used for acquiring influence factor characteristics;
and the output network training module is used for training an output network in the prediction model according to the load time sequence characteristics, the influence factor characteristics and the load time sequence training data, and the output network is used for outputting a short-term load prediction result.
9. A readable medium comprising executable instructions which, when executed by a processor of an electronic device, cause the electronic device to perform the method of any of claims 1 to 6.
10. An electronic device comprising a processor and a memory storing execution instructions, the processor performing the method of any of claims 1-6 when the processor executes the execution instructions stored by the memory.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010341288.4A CN111415050B (en) | 2020-04-27 | 2020-04-27 | Short-term load prediction method, short-term load prediction model training method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010341288.4A CN111415050B (en) | 2020-04-27 | 2020-04-27 | Short-term load prediction method, short-term load prediction model training method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111415050A true CN111415050A (en) | 2020-07-14 |
| CN111415050B CN111415050B (en) | 2023-12-05 |
Family
ID=71492057
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010341288.4A Active CN111415050B (en) | 2020-04-27 | 2020-04-27 | Short-term load prediction method, short-term load prediction model training method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111415050B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112348284A (en) * | 2020-11-25 | 2021-02-09 | 新智数字科技有限公司 | Power load prediction method and device, readable medium and electronic equipment |
| CN112668776A (en) * | 2020-12-25 | 2021-04-16 | 国网黑龙江省电力有限公司电力科学研究院 | Long-short term memory network model power load prediction method suitable for alpine region |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018082523A1 (en) * | 2016-11-02 | 2018-05-11 | 中国电力科学研究院有限公司 | Load cycle mode identification method |
| CN108288109A (en) * | 2018-01-11 | 2018-07-17 | 安徽优思天成智能科技有限公司 | Motor-vehicle tail-gas concentration prediction method based on LSTM depth space-time residual error networks |
| CN109214610A (en) * | 2018-11-16 | 2019-01-15 | 国网江苏省电力有限公司经济技术研究院 | A kind of saturation Methods of electric load forecasting based on shot and long term Memory Neural Networks |
| CN109409561A (en) * | 2018-08-22 | 2019-03-01 | 国网天津市电力公司 | The construction method of Multiple Time Scales time series collaborative forecasting model |
| CN109472404A (en) * | 2018-10-31 | 2019-03-15 | 山东大学 | A kind of Short-Term Load Forecasting of Electric Power System, model, apparatus and system |
| CN109919370A (en) * | 2019-02-26 | 2019-06-21 | 国网冀北电力有限公司运营监测（控）中心 | A kind of Methods of electric load forecasting and prediction meanss |
| CN109978261A (en) * | 2019-03-27 | 2019-07-05 | 新奥数能科技有限公司 | Determine method, apparatus, readable medium and the electronic equipment of load forecasting model |
| CN110163421A (en) * | 2019-04-29 | 2019-08-23 | 广东电网有限责任公司电网规划研究中心 | Long-medium term power load forecasting method |
| CN110264083A (en) * | 2019-06-21 | 2019-09-20 | Oppo广东移动通信有限公司 | A kind of agenda managing method, device and computer readable storage medium |
| CN110782096A (en) * | 2019-10-29 | 2020-02-11 | 山东科技大学 | Forex time series prediction method |
| CN110991761A (en) * | 2019-12-13 | 2020-04-10 | 新奥数能科技有限公司 | Heat supply load prediction method and device |
| CN113837503A (en) * | 2021-11-29 | 2021-12-24 | 新风光电子科技股份有限公司 | Power grid load prediction method and system based on artificial intelligence |
-
2020
- 2020-04-27 CN CN202010341288.4A patent/CN111415050B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018082523A1 (en) * | 2016-11-02 | 2018-05-11 | 中国电力科学研究院有限公司 | Load cycle mode identification method |
| CN108288109A (en) * | 2018-01-11 | 2018-07-17 | 安徽优思天成智能科技有限公司 | Motor-vehicle tail-gas concentration prediction method based on LSTM depth space-time residual error networks |
| CN109409561A (en) * | 2018-08-22 | 2019-03-01 | 国网天津市电力公司 | The construction method of Multiple Time Scales time series collaborative forecasting model |
| CN109472404A (en) * | 2018-10-31 | 2019-03-15 | 山东大学 | A kind of Short-Term Load Forecasting of Electric Power System, model, apparatus and system |
| CN109214610A (en) * | 2018-11-16 | 2019-01-15 | 国网江苏省电力有限公司经济技术研究院 | A kind of saturation Methods of electric load forecasting based on shot and long term Memory Neural Networks |
| CN109919370A (en) * | 2019-02-26 | 2019-06-21 | 国网冀北电力有限公司运营监测（控）中心 | A kind of Methods of electric load forecasting and prediction meanss |
| CN109978261A (en) * | 2019-03-27 | 2019-07-05 | 新奥数能科技有限公司 | Determine method, apparatus, readable medium and the electronic equipment of load forecasting model |
| CN110163421A (en) * | 2019-04-29 | 2019-08-23 | 广东电网有限责任公司电网规划研究中心 | Long-medium term power load forecasting method |
| CN110264083A (en) * | 2019-06-21 | 2019-09-20 | Oppo广东移动通信有限公司 | A kind of agenda managing method, device and computer readable storage medium |
| CN110782096A (en) * | 2019-10-29 | 2020-02-11 | 山东科技大学 | Forex time series prediction method |
| CN110991761A (en) * | 2019-12-13 | 2020-04-10 | 新奥数能科技有限公司 | Heat supply load prediction method and device |
| CN113837503A (en) * | 2021-11-29 | 2021-12-24 | 新风光电子科技股份有限公司 | Power grid load prediction method and system based on artificial intelligence |
Non-Patent Citations (4)
| Title |
|---|
| 孔祥玉等: "基于经验模态分解与特征相关分析的短期负荷预测方法", 《电力系统自动化》 * |
| 庞昊等: "基于多神经网络融合的短期负荷预测方法", 《电力自动化设备》, no. 06 * |
| 罗澍忻等: "考虑相关因素的长短时记忆网络短期负荷预测方法", 《机电工程技术》, no. 12 * |
| 许言路等: "基于多模型融合神经网络的短期负荷预测", 《控制工程》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112348284A (en) * | 2020-11-25 | 2021-02-09 | 新智数字科技有限公司 | Power load prediction method and device, readable medium and electronic equipment |
| CN112668776A (en) * | 2020-12-25 | 2021-04-16 | 国网黑龙江省电力有限公司电力科学研究院 | Long-short term memory network model power load prediction method suitable for alpine region |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111415050B (en) | 2023-12-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021238262A1 (en) | Vehicle recognition method and apparatus, device, and storage medium | |
| JP6384065B2 (en) | Information processing apparatus, learning method, and program | |
| CN111210093B (en) | Daily water consumption prediction method based on big data | |
| CN111415050A (en) | Short-term load prediction method and short-term load prediction model training method and device | |
| CN110688620A (en) | Short-term load prediction method and device | |
| CN111797320A (en) | Data processing method, device, equipment and storage medium | |
| CN110784507A (en) | Fusion method and system of population information data | |
| CN110796240A (en) | Training method, feature extraction method, device and electronic equipment | |
| CN114726745B (en) | Network traffic prediction method, device and computer readable storage medium | |
| CN111563829A (en) | Power price prediction method and device and power price prediction model training method and device | |
| CN111404835B (en) | Flow control method, device, equipment and storage medium | |
| EP4307172A1 (en) | Data interpolation method and apparatus and storage medium | |
| CN110766231A (en) | Crime prediction method and system based on multi-head neural network | |
| CN113239272B (en) | Intention prediction method and intention prediction device of network management and control system | |
| CN116029422A (en) | Behavior sequence prediction method and device, electronic equipment and medium | |
| CN111178498A (en) | Stock fluctuation prediction method and device | |
| CN115935822A (en) | Method, system, device and medium for predicting long-term degradation trend of computer parameters | |
| CN115309638A (en) | Method and device for assisting model optimization | |
| CN111767934B (en) | Image recognition method and device and electronic equipment | |
| CN112348284A (en) | Power load prediction method and device, readable medium and electronic equipment | |
| CN112949882A (en) | Load prediction method and device and electronic equipment | |
| CN112926608A (en) | Image classification method and device, electronic equipment and storage medium | |
| CN112733454A (en) | Equipment predictive maintenance method and device based on joint learning | |
| CN112733453A (en) | Equipment predictive maintenance method and device based on joint learning | |
| CN111753598A (en) | Face detection method and device |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20231031 Address after: 065001 China (Hebei) Pilot Free Trade Zone Daxing Airport Area Langfang Airport Economic Zone Hangyidao Free Trade Zone Science and Technology Innovation Base 2101, Langfang City, Hebei Province Applicant after: Xinao Xinzhi Technology Co.,Ltd. Address before: 100102 10th floor, 1 Wangjing East Road, Chaoyang District, Beijing Applicant before: ENNEW DIGITAL TECHNOLOGY Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |