Short-term load forecasting method and system based on long short-term memory network and terminal

Abstract

The application discloses a short-term load prediction method and system based on a long short-term memory network and a terminal, mainly relates to the technical field of short-term load prediction, and aims to solve the problem that the existing method is difficult to fit the load law. The application comprises the following steps: obtaining an original power data set from a power transformation device, determining an occurrence frequency value, and creating an FP tree; obtaining a sample data set based on the FP tree and a preset node threshold value; and taking the sample data set as an input of an LSTM network algorithm to obtain a prediction result. The application realizes fast and effective obtaining of the prediction result through the above method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of short-term load forecasting, in particular to a short-term load forecasting method, system and terminal based on a long short-term memory network. BACKGROUND

[0002] With the deep integration of energy revolution and digital revolution, big data and artificial intelligence technology are deeply combined and applied in the energy field, promoting energy revolution, promoting the rational allocation of energy resources, and greatly improving the utilization efficiency of energy.

[0003] At present, user electric energy data is mostly collected from electric meters, and data mining, data analysis and deep learning technologies are used for analysis, which can deeply understand the energy demand of enterprises and residents and better develop power generation plans.

[0004] However, the power load data is a time series sequence, which is affected by nonlinear factors such as economic level, weather, population, etc. The data is affected by historical load information characteristics and external dynamic factors, and shows randomness. It is difficult for conventional modeling methods to fit the load law. SUMMARY

[0005] In view of the above shortcomings of the prior art, the present application provides a short-term load forecasting method, system and terminal based on a long short-term memory network to solve the above technical problems.

[0006] In a first aspect, the present application provides a short-term load forecasting method based on a long short-term memory network, the method comprising: obtaining an original power data set from a power transformation device, determining a frequency value to create an FP tree; obtaining a sample data set based on the FP tree and a preset node threshold; taking the sample data set as an input of an LSTM network algorithm to obtain a prediction result.

[0007] Further, the original power data set is obtained from the power transformation device, and the frequency value is determined to create the FP tree, specifically comprising: obtaining the original power data set; wherein the original data set contains a plurality of power data; the power data is cut to obtain sub-data, and then an original sub-data set is obtained; the frequency value of each sub-data in the original sub-data set is determined, and the sub-data lower than the preset threshold is filtered out from the sub-data to create a high-frequency data-frequency value table; based on the high-frequency data-frequency value table and the original sub-data set, the FP tree is established.

[0008] Further, based on the FP tree and the preset node threshold, a training sample data set is obtained, specifically comprising: obtaining the node value of each node in the FP tree, and filtering out the sub-data lower than the preset node threshold again to obtain the training sample data set.

[0009] Further, the sample data set is taken as an input of the LSTM network algorithm to obtain a prediction result, specifically including: dividing the sample data set into a test set and a training set; inputting the test set into the LSTM network algorithm to train the LSTM network algorithm to obtain a trained LSTM network algorithm; inputting the training set into the LSTM network algorithm to obtain the prediction result.

[0010] In a second aspect, the application provides a short-term load prediction system based on a long short-term memory network, which comprises: a creating module configured to obtain an original power data set from a power transformation device, determine occurrence frequency values to create an FP tree; an obtaining module configured to obtain a sample data set based on the FP tree and a preset node threshold; and an obtaining module configured to take the sample data set as an input of an LSTM network algorithm to obtain a prediction result.

[0011] Further, the creating module further comprises a creating unit configured to obtain the original power data set, wherein the original data set comprises a plurality of power data; split the power data to obtain sub-data, and then obtain an original sub-data set; determine occurrence frequency values of each sub-data in the original sub-data set, filter out sub-data lower than a preset threshold from the sub-data to create a high-frequency data-frequency value table; and establish an FP tree based on the high-frequency data-frequency value table and the original sub-data set.

[0012] Further, the obtaining module further comprises an obtaining unit configured to obtain node values of each node in the FP tree, and filter out sub-data lower than a preset node threshold again to obtain a training sample data set.

[0013] Further, the obtaining module further comprises a prediction unit configured to divide the sample data set into a test set and a training set; input the test set into the LSTM network algorithm to train the LSTM network algorithm to obtain a trained LSTM network algorithm; and input the training set into the LSTM network algorithm to obtain the prediction result.

[0014] In a third aspect, the application further provides a short-term load prediction terminal based on a long short-term memory network, which comprises a processor, a memory and an execution instruction stored on the memory, the execution instruction being configured to enable the terminal to perform the short-term load prediction method based on the long short-term memory network of any one of the above aspects when executed by the processor.

[0015] Those skilled in the art can understand that the application has at least the following beneficial effects: the application first uses the FP tree to mine factors (sample data set / strong correlation factors) that have a greater impact on power load, and uses the strong correlation factors to train a load prediction model, thereby achieving fast and effective prediction results, providing a scientific and effective method for power load prediction, and having high application value. BRIEF DESCRIPTION OF DRAWINGS

[0016] Some embodiments of the present disclosure will be described in detail below with reference to the drawings, in which:

[0017] Figure 1 is a schematic diagram of an internal structure of a short-term load prediction system based on a long short-term memory network provided by an embodiment of the present application.

[0018] Figure 2 is a flowchart of a short-term load prediction method based on a long short-term memory network provided by an embodiment of the present application.

[0019] Figure 3 is a schematic diagram of an internal structure of a short-term load prediction terminal based on a long short-term memory network provided by an embodiment of the present application. DETAILED DESCRIPTION

[0020] Those skilled in the art will understand that the embodiments described below are merely preferred embodiments of the present disclosure, and do not represent that the present disclosure can only be implemented by the preferred embodiments. The preferred embodiments are only used to explain the technical principles of the present disclosure, and are not intended to limit the protection scope of the present disclosure. Based on the preferred embodiments provided by the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative labor shall fall within the protection scope of the present disclosure.

[0021] It should also be noted that the terms "comprising", "including", or any other variant thereof are intended to cover non-exclusive inclusion, so that processes, methods, articles or devices that include a series of elements not only include those elements, but also include other elements not explicitly listed, or further include elements inherent in such processes, methods, articles or devices. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or device that includes the element.

[0022] The technical solutions provided by the embodiments of the present application will be described in detail below with reference to the drawings.

[0023] The embodiments of the present application provide a short-term load prediction method based on a long short-term memory network, as shown in Figure 1 The method provided by the embodiments of the present application mainly includes the following steps:

[0024] Step 110, obtaining an original power data set from a power transformation device, determining a frequency value to create an FP tree.

[0025] Specifically, the method can include: obtaining an original power dataset; wherein the original dataset includes a plurality of power data; splitting the power data to obtain sub-data, and then obtaining an original sub-data set; determining the occurrence frequency value of each sub-data in the original sub-data set, filtering out sub-data lower than a preset threshold from the sub-data, and creating a high-frequency data-frequency value table; and establishing an FP tree based on the high-frequency data-frequency value table and the original sub-data set. It should be noted that the method of splitting the power data can be any feasible data splitting algorithm.

[0026] Step 120, obtaining a sample data set based on the FP tree and a preset node threshold.

[0027] Specifically, the method can include: obtaining the node value of each node in the FP tree, filtering out sub-data lower than a preset node threshold again, and obtaining a training sample data set. It should be noted that the preset node threshold can be any feasible numerical value.

[0028] Step 130, inputting the sample data set into the LSTM network algorithm to obtain a prediction result.

[0029] Specifically, the method can include: dividing the sample data set into a test set and a training set; inputting the test set into the LSTM network algorithm to train the LSTM network algorithm to obtain a trained LSTM network algorithm; and inputting the training set into the LSTM network algorithm to obtain a prediction result.

[0030] In addition, Figure 2 A short-term load prediction system based on a long short-term memory network is provided in the embodiments of the present application. As shown in Figure 2 The system provided by the embodiments of the present application mainly includes:

[0031] The creating module 210 is configured to obtain an original power dataset from a power transformation device, and determine an occurrence frequency value to create an FP tree.

[0032] The creating module 210 further includes a creating unit 211. The creating unit 211 is configured to obtain an original power dataset; wherein the original dataset includes a plurality of power data; split the power data to obtain sub-data, and then obtain an original sub-data set; determine the occurrence frequency value of each sub-data in the original sub-data set, filter out sub-data lower than a preset threshold from the sub-data, and create a high-frequency data-frequency value table; and establish an FP tree based on the high-frequency data-frequency value table and the original sub-data set.

[0033] The obtaining module 220 is configured to obtain a sample data set based on the FP tree and a preset node threshold.

[0034] The acquisition module 220 also includes an acquisition unit 221, which is used to acquire the node value of each node in the FP tree, and then filter out sub-data that are lower than the preset node threshold to obtain the training sample dataset.

[0035] The acquisition module 230 is used to take the sample dataset as input to the LSTM network algorithm and obtain the prediction results.

[0036] The acquisition module 230 also includes a prediction unit 231, which is used to divide the sample dataset into a test set and a training set; input the test set into the LSTM network algorithm to train the LSTM network algorithm to obtain a trained LSTM network algorithm; and input the training set into the LSTM network algorithm to obtain the prediction result.

[0037] In addition, this application also provides a residential integrated energy management terminal, such as... Figure 3 As shown, it stores executable instructions, which, when executed, implement a residential integrated energy management method as described above. Specifically, the server sends execution instructions to the memory via a bus. When the memory receives the execution instructions, it sends an execution signal to the processor via the bus to activate the processor.

[0038] It should be noted that the processor is used to obtain the raw power dataset from the power equipment, determine the frequency of occurrence, and create an FP tree; based on the FP tree and a preset node threshold, a sample dataset is obtained; the sample dataset is used as the input to the LSTM network algorithm to obtain the prediction results.

[0039] The technical solutions of this disclosure have been described in conjunction with the preceding embodiments. However, it will be readily understood by those skilled in the art that the scope of protection of this disclosure is not limited to these specific embodiments. Without departing from the technical principles of this disclosure, those skilled in the art can disassemble and combine the technical solutions in the above embodiments, and can also make equivalent changes or substitutions to the relevant technical features. Any changes, equivalent substitutions, improvements, etc., made within the technical concept and / or technical principles of this disclosure will fall within the scope of protection of this disclosure.

Claims

1.A short-term load forecasting method based on a long short-term memory network, characterized in that, The method comprises: obtaining an original power dataset from a power transformation device, determining a frequency value to create an FP tree; obtaining a sample dataset based on the FP tree and a preset node threshold value; inputting the sample dataset into an LSTM network algorithm to obtain a prediction result; wherein the obtaining of the original power dataset from the power transformation device and the determination of the frequency value to create the FP tree specifically comprises: obtaining an original power dataset; wherein the original dataset contains a plurality of power data; segmenting the power data to obtain sub-data, and then obtaining an original sub-data set; determining the frequency value of each sub-data in the original sub-data set, filtering out sub-data lower than a preset threshold value from the sub-data to create a high-frequency data-frequency value table; establishing an FP tree based on the high-frequency data-frequency value table and the original sub-data set; wherein the obtaining of the training sample dataset based on the FP tree and the preset node threshold value specifically comprises: obtaining the node value of each node in the FP tree, and filtering out sub-data lower than a preset node threshold value again to obtain a training sample dataset. 2.The short-term load forecasting method based on long short-term memory network according to claim 1, characterized in that, The method comprises: inputting the sample dataset into an LSTM network algorithm to obtain a prediction result, specifically comprising: dividing the sample dataset into a test set and a training set; inputting the test set into the LSTM network algorithm to train the LSTM network algorithm to obtain a trained LSTM network algorithm; 3.A short-term load forecasting system based on a long short-term memory network, characterized by, inputting the training set into the LSTM network algorithm to obtain a prediction result. The system comprises: a creating module for obtaining an original power dataset from a power transformation device, determining a frequency value to create an FP tree; an obtaining module for obtaining a sample dataset based on the FP tree and a preset node threshold value; an obtaining module for inputting the sample dataset into an LSTM network algorithm to obtain a prediction result; the creating module further comprises a creating unit; the creating unit is configured to obtain an original power dataset; wherein the original dataset contains a plurality of power data; segment the power data to obtain sub-data, and then obtain an original sub-data set; determine the frequency value of each sub-data in the original sub-data set, filter out sub-data lower than a preset threshold value from the sub-data to create a high-frequency data-frequency value table; and establish an FP tree based on the high-frequency data-frequency value table and the original sub-data set; 4.The short-term load forecasting system based on long short-term memory network according to claim 3, wherein, the obtaining module further comprises an obtaining unit; the obtaining unit is configured to obtain the node value of each node in the FP tree, and filter out sub-data lower than a preset node threshold value again to obtain a training sample dataset. The obtaining module further comprises a prediction unit; 5.A short-term load forecasting terminal based on a long short-term memory network, characterized in that, for dividing the sample dataset into a test set and a training set; inputting the test set into the LSTM network algorithm to train the LSTM network algorithm to obtain a trained LSTM network algorithm; and inputting the training set into the LSTM network algorithm to obtain a prediction result. The terminal comprises: a processor; and a memory having executable code stored thereon, which, when executed, causes the processor to perform a short-term load prediction method based on a long short-term memory network as claimed in claims 1-2.

Images