A power grid power consumption prediction method based on data mining and neural network

Abstract

The present application relates to a kind of power grid electric quantity prediction method based on data mining and neural network, comprising the following steps: first data cleaning and recovery, using "3 sigma" principle to eliminate abnormal point, for missing data, using quadratic Lagrange interpolation to restore, avoid its influence on time coherence;Then, from internal and external two factors, construct electric quantity characteristics, for internal factor, consider historical electric quantity data, and combine time coding, for external factor, consider regional population, economic factor, introduce these factors into prediction model;Finally, establish the prediction model based on Prophet model and LSTM model, the component decomposition of power grid electric quantity data is carried out, for different components, respectively establish LSTM network to carry out fitting, and using MLP fuses each LSTM network to obtain final prediction result.

Description

Technical Field

[0001] This invention relates to a power grid power prediction method based on data mining and neural networks, belonging to the field of power data prediction technology. Background Technology

[0002] In modern society, electricity, as one of the most important energy sources, plays a crucial role in various fields. The supply and distribution of electricity require sophisticated planning and management to ensure efficient energy utilization and operational stability. In power systems, power forecasting is an important technology that provides critical decision support for energy dispatch, load balancing, electricity procurement, generation planning, and energy market operation. However, existing power forecasting algorithms typically suffer from the following problems: Outliers in power grid electricity data have a significant impact on algorithm performance. For electricity data at all levels, data exported from multiple platforms contains bad data due to human measurement and recording errors, or due to physical environment, aging monitoring equipment, etc., which manifest as outliers and missing data points during data analysis.

[0003] Regional factors influencing electricity consumption, such as regional economic and demographic information, need to be given more attention by algorithms. Regional information is often related to changes in electricity consumption. For example, a more economically developed region typically consumes more electricity. Cities with larger influxes of people may see their electricity consumption gradually increase over the future.

[0004] Power grid electricity data fluctuates significantly, is voluminous, and requires greater precision, making it difficult for existing algorithms to achieve high accuracy. With the widespread adoption and improved functionality of automatic power grid monitoring devices, more refined electricity curves are being stored in power grid databases. These new electricity curves are characterized by weak regularity, large quantity, and significant differences between each other.

[0005] Therefore, a method is needed to improve the accuracy of power grid data prediction. This method should be able to identify, remove, and supplement abnormal data in power grid data, and combine power generation regional factors with the construction of a specific prediction model to solve the power prediction problem. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention proposes a power grid power prediction method based on data mining and neural networks.

[0007] The technical solution of the present invention is as follows: On the one hand, this invention provides a power grid power prediction method based on data mining and neural networks, comprising the following steps: Collect regional power grid electricity data, clean the electricity data to remove abnormal data, and then use second-order Lagrange interpolation to recover the data that was accidentally deleted during the cleaning process, so as to obtain the processed regional power grid electricity data. New features are extracted based on the processed regional power grid power data and related regional information. The extracted new features are added to the processed regional power grid power data. The processed regional power grid power data is sampled according to a preset length and a power grid power prediction model training set is constructed. A power grid power prediction model is constructed, which consists of a Prophet model and an LSTM model. The Prophet model decomposes the training set into multiple training sets with different components. Each training set is input into a separate LSTM model, and the outputs of all LSTM models are fed into a multilayer perceptron. The multilayer perceptron performs feature fusion on the input data to obtain the prediction result. The parameters of the power grid power prediction model are updated based on the prediction result. The model is trained cyclically until a preset number of iterations are reached to obtain the trained power grid power prediction model. This model is then used to predict the power grid power.

[0008] As a preferred embodiment of the present invention, data cleaning of regional power grid electricity data is performed based on the Laida criterion, and the specific steps are as follows: Calculate the mean of each feature in the regional power grid electricity data. With variance Calculate the normal data range for the corresponding feature based on the Raida criterion. Delete data that is not within this range.

[0009] In a preferred embodiment of the present invention, for the loss of normal power data in the regional power grid due to equipment malfunction or data cleaning, the missing normal data is recovered by using second-order Lagrange interpolation. The specific steps are as follows: Let the function Defined in interval The curve has One point, that is Then there is a function curve that passes through this According to the Lagrange interpolation theorem, there exists a point that does not exceed [a certain value]. polynomials of degree After this There are 10 points that satisfy the following equation:

[0010] in: for of Lagrange interpolation formula; When calculating the time point When taking the value, Approximately ,but The calculation method is as follows: First, Add each point to a set, and define a set of all points. Define another one containing A set of points This includes removal The Middle Remaining after points One point; Secondly, the polynomial is defined as follows:

[0011] For the above formula have That is, only when the value at time is hour, The value is not 0, and has ; Finally, based on polynomials The Lagrange interpolation formula is obtained. :

[0012] Based on the above Lagrange interpolation formula, a second-order Lagrange interpolation polynomial is constructed. This involves taking a data segment of length 3 that includes missing data, and defining the first two terms of the missing data as... , The latter is defined as This leads to the following formula:

[0013] The result of the above calculation is the value of the missing data.

[0014] As a preferred embodiment of the present invention, the extracted new features include the temporal characteristics of regional power grid electricity data and regional population change information, regional economic development information, regional fixed asset investment information, and regional environmental information from regional related information. The steps for extracting the time features of the regional power grid electricity data are as follows: Extract the collection time from any regional power grid electricity data, and project the month, day, and hour from the collection time onto the sine and cosine functions respectively, as shown in the following formula:

[0015]

[0016]

[0017]

[0018]

[0019]

[0020] in: , These represent the projection of the number of days onto the sine and cosine functions, respectively. This indicates which day of the year the day falls on; , These represent the projection of month information onto the sine and cosine functions, respectively. This indicates which month of the year the month falls in; , These represent hourly information projected onto the sine and cosine functions, respectively. This indicates which hour of the day the hour falls within; The month, day, and hour information projected onto the sine and cosine functions are used as the time characteristics of the regional power grid electricity data.

[0021] In a preferred embodiment of the present invention, the extracted new features are normalized by dimensionless processing, and the normalization formula is shown below:

[0022] in: This corresponds to the normalized value of the new feature; It is the maximum value corresponding to the new feature; It is the minimum value corresponding to the new feature.

[0023] In a preferred embodiment of the present invention, the Prophet model performs component decomposition on regional power grid electricity data through the following steps: The Prophet model decomposes regional power grid electricity data into: trend component, seasonal component, holiday component, and uncertain component, as shown in the following formula:

[0024] in: for Regional power grid electricity data at any given time; for Trend components of regional power grid electricity data at a given time; for Seasonal component of regional power grid electricity data at a given time; for Holiday component of regional power grid electricity data at any given time; for Uncertainty in the regional power grid electricity data at a given time; The trend components of the regional power grid electricity data are divided into non-linear growth and linear growth, among which non-linear growth... The saturated nonlinear growth model is used to describe this, as shown in the following equation:

[0025] in: for The capacity extremum; This represents the growth rate of the function curve; This is the bias value; linear growth It is described using a piecewise linear function, as shown in the following equation:

[0026] in: The initial growth rate; vector This represents the change in the rate of increase; The initial bias; vector The change in bias; The seasonal component of the regional power grid electricity data is constructed based on a Fourier series expansion. Let the regional power grid electricity data... by If it is periodic, then expanding it yields the following equation:

[0027] in: , These are the corresponding coefficients of the triangular components; Indicates the number of components; The holiday component of the regional power grid electricity data is expressed by the following formula:

[0028] in: This is the vector of holiday impact factors, representing the degree of impact of different holidays on regional power grid electricity data; The regression vector indicates the holidays.

[0029] On the other hand, the present invention also provides a power grid power prediction system based on data mining and neural networks, including a data collection module, a feature extraction module, and a power grid power prediction module; The data collection module is used to collect regional power grid electricity data, clean the electricity data to remove abnormal data, and then use second-order Lagrange interpolation to recover the data that was accidentally deleted during the cleaning process, so as to obtain the processed regional power grid electricity data. The feature extraction module is used to extract new features based on the processed regional power grid power data and regional related information, add the extracted new features to the processed regional power grid power data, and sample the processed regional power grid power data according to a preset length to build a power grid power prediction model training set. The power grid power prediction module is used to construct a power grid power prediction model, which consists of a Prophet model and an LSTM model. The Prophet model decomposes the training set into multiple training sets with different components. Each training set is then input into a separate LSTM model, and the outputs of all LSTM models are fed into a multilayer perceptron. The multilayer perceptron performs feature fusion on the input data to obtain the prediction result. The parameters of the power grid power prediction model are updated based on the prediction result. After training for a preset number of times, the trained power grid power prediction model is obtained. This module uses this model to predict the power grid power.

[0030] In another aspect, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method as described in any embodiment of the present invention.

[0031] In another aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any embodiment of the present invention.

[0032] The present invention has the following beneficial effects: 1. This invention identifies and cleans outlier data based on the Laida criterion, and completes the outlier data through quadratic Lagrange interpolation, thus solving the problem of inaccurate prediction results caused by the difficulty in handling outliers in traditional methods.

[0033] 2. This invention divides the factors affecting the power grid into internal and external factors. Regional factors affecting the power grid, such as regional economic information and regional population information, are regarded as external factors. By combining the features of internal and external factors into the prediction model, the problem of traditional models ignoring regional factors is solved.

[0034] 3. This invention constructs a prediction model based on the Prophet model and the LSTM model. The Prophet model is used to decompose the power grid data into different trends. Then, the LSTM is used to model each trend separately. Finally, the data is fused and summarized by a multilayer perceptron to obtain the final output prediction result, thereby improving the performance of the prediction model. Attached Figure Description

[0035] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a structural diagram of the regional power grid power prediction model of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0038] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0039] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0040] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0041] Example 1: See Figure 1 A power grid power prediction method based on data mining and neural networks includes the following steps: Collect regional power grid electricity data, clean the electricity data to remove abnormal data, and then use second-order Lagrange interpolation to recover the data that was accidentally deleted during the cleaning process, so as to obtain the processed regional power grid electricity data. New features are extracted based on the processed regional power grid power data and related regional information. The extracted new features are added to the processed regional power grid power data. The processed regional power grid power data is sampled with lengths of 64, 128, and 256 to construct a training set for the power grid power prediction model. A power grid power prediction model is constructed, which consists of a Prophet model and an LSTM model. The Prophet model decomposes the training set into multiple training sets with different components. Each training set is input into a separate LSTM model, and the outputs of all LSTM models are fed into a multilayer perceptron. The multilayer perceptron performs feature fusion on the input data to obtain the prediction result. The parameters of the power grid power prediction model are updated based on the prediction result. The model is trained cyclically until a preset number of iterations are reached to obtain the trained power grid power prediction model. This model is then used to predict the power grid power.

[0042] As a preferred embodiment of this practice, data cleaning of regional power grid electricity data is performed based on the Laida criterion. The specific steps are as follows: Calculate the mean of each feature in the regional power grid electricity data. With variance Calculate the normal data range for the corresponding feature based on the Raida criterion. Delete data that is not within this range.

[0043] The Raida Criterion is also known as "3". The principle is that when the amount of data is large enough, the samples in the data will approximate a normal distribution. And in a normal distribution, 99.73% of the data are distributed within a certain range. The data falls within a specified range. Therefore, data outside this range can be considered outlier. This method is highly efficient and particularly suitable for large-scale data scenarios involving power grid electricity.

[0044] After data cleaning is complete, missing data needs to be recovered. Missing data includes data lost due to equipment malfunctions and data lost due to "3". The missing data is due to the removal of data according to the principle. Since power grid electricity data is time series data, missing data will cause temporal discontinuity, thus having a significant impact on subsequent prediction algorithms. This embodiment uses the Lagrange quadratic interpolation method to recover the missing data.

[0045] In a preferred embodiment of this invention, for the loss of normal power data in the regional power grid due to equipment malfunction or data cleaning, the missing normal data is recovered by using second-order Lagrange interpolation. The specific steps are as follows: Let the function Defined in interval The curve has One point, that is Then there is a function curve that passes through this According to the Lagrange interpolation theorem, there exists a point that does not exceed [a certain value]. polynomials of degree After this There are 10 points that satisfy the following equation:

[0046] in: for of Lagrange interpolation formula; When calculating the time point When taking the value, Approximately ,but The calculation method is as follows: First, Add each point to a set, and define a set of all points. Define another one containing A set of points This includes removal The Middle Remaining after points One point; Secondly, the polynomial is defined as follows:

[0047] For the above formula have That is, only when the value at time is hour, The value is not 0, and has ; Finally, based on polynomials The Lagrange interpolation formula is obtained. :

[0048] Based on the above Lagrange interpolation formula, a second-order Lagrange interpolation polynomial is constructed. This involves taking a data segment of length 3 that includes missing data, and defining the first two terms of the missing data as... , The latter is defined as This leads to the following formula:

[0049] The result of the above calculation is the value of the missing data.

[0050] This embodiment constructs power prediction features from two aspects: internal factors and external factors.

[0051] The internal factors mainly consist of historical power grid consumption data. This invention adds a time feature to this data. Power grid consumption, as a type of data with time characteristics, exhibits a relatively clear time trend. For example, residential areas typically consume less electricity during the day and more at night. In winter and summer, electricity consumption is higher because users frequently use air conditioning for heating and cooling. Therefore, this invention uses time as one of the important features in the prediction algorithm, specifically as follows: First, extract the collection time of a specific power grid data point, for example, "2020-08-30 14:30". Next, extract the month, day, and hour of the year in which this time falls, and project them onto sine and cosine functions respectively, as shown below:

[0052]

[0053]

[0054]

[0055]

[0056]

[0057] in: , These represent the projection of the number of days onto the sine and cosine functions, respectively. This indicates which day of the year the day falls on; , These represent the projection of month information onto the sine and cosine functions, respectively. This indicates which month of the year the month falls in; , These represent hourly information projected onto the sine and cosine functions, respectively. This indicates which hour of the day the hour falls within; By incorporating time features through the above approach, the model can better learn the time information of the data.

[0058] For external factors, information such as regional population changes, regional economic development, regional fixed asset investment, and regional environmental information are considered. These factors can be obtained by consulting government-published documents, which are then coded as part of the input to the predictive model.

[0059] The extracted new features also need to be normalized. Grid electricity data has different specifications and distributions, meaning the data dimensions are inconsistent. If this is not addressed, it may lead to increased weighting of features with larger value ranges during subsequent model training, causing bias in the training process. To solve this problem, the data needs to be dimensionless. For the extracted new features, the following formula is used for normalization:

[0060] in: This corresponds to the normalized value of the new feature; It is the maximum value corresponding to the new feature; It is the minimum value corresponding to the new feature. As a preferred embodiment of this example, for a segment from... arrive Power grid electricity data: The Prophet model decomposes regional power grid electricity data into trend components, seasonal components, holiday components, and uncertain components, as shown in the following formula:

[0061] in: for Regional power grid electricity data at any given time; for Trend components of regional power grid electricity data at a given time; for Seasonal component of regional power grid electricity data at any given time; for Holiday component of regional power grid electricity data at any given time; for Uncertainty in the regional power grid electricity data at a given time; The trend components of the regional power grid electricity data are divided into non-linear growth and linear growth, among which non-linear growth... The saturated nonlinear growth model is used to describe this, as shown in the following equation:

[0062] in: for The capacity extremum; This represents the growth rate of the function curve; This is the bias value; linear growth It is described using a piecewise linear function, as shown in the following equation:

[0063] in: The initial growth rate; vector This represents the change in the rate of increase; The initial bias; vector The change in bias; The seasonal component of the regional power grid electricity data is constructed based on a Fourier series expansion. Let the regional power grid electricity data... by If it is periodic, then expanding it yields the following equation:

[0064] in: , These are the corresponding coefficients of the triangular components; Indicates the number of components; The holiday component of the regional power grid electricity data is expressed by the following formula:

[0065] in: The vector of factors influencing holidays is represented as follows: This represents the degree of impact of different holidays on the regional power grid electricity data; The holiday indicator regression vector represents the time points influencing the impact of holiday factors, denoted as: .

[0066] Next, a power grid power prediction model is constructed based on the LSTM model. LSTM is a type of recurrent neural network (RNN) that can learn long-term dependent information and is well-suited for modeling power grid power prediction tasks. The LSTM consists of an information storage unit and three logic gates that control the state of the stored information. The information storage unit contains sequential historical information, and the three logic gates are the output gate (determining which memories need to be output in the current state), the forget gate (determining which memories need to be forgotten), and the update gate (determining whether to update the storage unit with the current sequence information).

[0067] The structure of the regional power grid power prediction model is as follows: Figure 2 As shown, firstly, four trend decomposition components are obtained through the Prophet model. For each trend decomposition component, an LSTM network is used for alignment and modeling to fit each component. Finally, the outputs of the four LSTM networks are fused and summarized using a multilayer perceptron to obtain the final prediction result.

[0068] Example 2: A power grid power prediction system based on data mining and neural networks includes a data collection module, a feature extraction module, and a power grid power prediction module. The data collection module is used to collect regional power grid electricity data, clean the electricity data to remove abnormal data, and then use second-order Lagrange interpolation to recover the data that was accidentally deleted during the cleaning process, so as to obtain the processed regional power grid electricity data. The feature extraction module is used to extract new features based on the processed regional power grid power data and regional related information, add the extracted new features to the processed regional power grid power data, and sample the processed regional power grid power data according to a preset length to build a power grid power prediction model training set. The power grid power prediction module is used to construct a power grid power prediction model, which consists of a Prophet model and an LSTM model. The Prophet model decomposes the training set into multiple training sets with different components. Each training set is then input into a separate LSTM model, and the outputs of all LSTM models are fed into a multilayer perceptron. The multilayer perceptron performs feature fusion on the input data to obtain the prediction result. The parameters of the power grid power prediction model are updated based on the prediction result. After training for a preset number of times, the trained power grid power prediction model is obtained. This module uses this model to predict the power grid power.

[0069] This embodiment is used to implement the function of Embodiment 1, and will not be described again here.

[0070] Example 3: This embodiment proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the method described in any embodiment of the present invention.

[0071] Example 4: This embodiment proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in any embodiment of the present invention.

[0072] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0073] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0074] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0075] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0076] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A power grid power prediction method based on data mining and neural networks, characterized in that, Includes the following steps: Collect regional power grid electricity data, clean the electricity data to remove abnormal data, and then use second-order Lagrange interpolation to recover the data that was accidentally deleted during the cleaning process, so as to obtain the processed regional power grid electricity data. New features are extracted based on the processed regional power grid power data and related regional information. The extracted new features are added to the processed regional power grid power data. The processed regional power grid power data is sampled according to a preset length and a power grid power prediction model training set is constructed. A power grid power prediction model is constructed, which consists of a Prophet model and an LSTM model. The Prophet model decomposes the training set into multiple training sets with different components. Each training set is input into a separate LSTM model, and the outputs of all LSTM models are fed into a multilayer perceptron. The multilayer perceptron performs feature fusion on the input data to obtain the prediction result. The parameters of the power grid power prediction model are updated based on the prediction result. The model is trained repeatedly until a preset number of iterations are reached to obtain the trained power grid power prediction model. The power grid power is then predicted using this model. Data cleaning of regional power grid electricity data based on the Laida criterion is performed, and the specific steps are as follows: Calculate the mean of each feature in the regional power grid electricity data. With variance Calculate the normal data range for the corresponding feature based on the Raida criterion. Delete data that is not within this range; For the loss of normal power data in the regional power grid due to equipment malfunction or data cleaning, the missing normal data can be recovered by using second-order Lagrange interpolation. The specific steps are as follows: Let the function Defined in interval The curve has One point, that is Then there is a function curve that passes through this According to the Lagrange interpolation theorem, there exists a point that does not exceed [a certain value]. polynomials of degree After this There are 10 points that satisfy the following equation: in: for of Lagrange interpolation formula; When calculating the time point When taking the value, Approximately ,but The calculation method is as follows: First, Add each point to a set, and define a set of all points. Define another one containing A set of points This includes removal The Middle Remaining after 1 point One point; Secondly, the polynomial is defined as follows: For the above formula have That is, only when the value at time is hour, The value is not 0, and has ; Finally, based on polynomials The Lagrange interpolation formula is obtained. : Based on the above Lagrange interpolation formula, a second-order Lagrange interpolation polynomial is constructed. This involves taking a data segment of length 3 that includes missing data, and defining the first two terms of the missing data as... , The latter is defined as This leads to the following formula: The result of the above formula is the value of the missing data; The extracted new features include the temporal characteristics of regional power grid electricity data, as well as regional population change information, regional economic development information, regional fixed asset investment information, and regional environmental information from relevant regional information. The steps for extracting the time features of the regional power grid electricity data are as follows: Extract the collection time from any regional power grid electricity data, and project the month, day, and hour from the collection time onto the sine and cosine functions respectively, as shown in the following formula: in: , These represent the projection of the number of days onto the sine and cosine functions, respectively. This indicates which day of the year the day falls on; , These represent the projection of month information onto the sine and cosine functions, respectively. This indicates which month of the year the month falls in; , These represent hourly information projected onto the sine and cosine functions, respectively. This indicates which hour of the day the hour falls within; The month, day, and hour information projected onto the sine and cosine functions are used as the time characteristics of the regional power grid electricity data.

2. The power grid power prediction method based on data mining and neural networks according to claim 1, characterized in that, The extracted new features are normalized by dimensionless processing, and the normalization formula is shown in the following equation: in: This corresponds to the normalized value of the new feature. It is the maximum value corresponding to the new feature; It is the minimum value corresponding to the new feature.

3. The power grid power prediction method based on data mining and neural networks according to claim 1, characterized in that, The Prophet model performs component decomposition on regional power grid electricity data through the following steps: The Prophet model decomposes regional power grid electricity data into: trend component, seasonal component, holiday component, and uncertain component, as shown in the following formula: in: for Regional power grid electricity data at any given time; for Trend components of regional power grid electricity data at a given time; for Seasonal component of regional power grid electricity data at a given time; for Holiday component of regional power grid electricity data at any given time; for Uncertainty in the regional power grid electricity data at a given time; The trend components of the regional power grid electricity data are divided into non-linear growth and linear growth, among which non-linear growth... The saturated nonlinear growth model is used to describe this, as shown in the following equation: in: for The capacity extremum; This represents the growth rate of the function curve; This is the bias value; linear growth It is described using a piecewise linear function, as shown in the following equation: in: The initial growth rate; vector This represents the change in the rate of increase; The initial bias; vector The change in bias; The seasonal component of the regional power grid electricity data is constructed based on a Fourier series expansion. Let the regional power grid electricity data... by If it is periodic, then expanding it yields the following equation: in: , These are the corresponding coefficients of the triangular components; Indicates the number of components; The holiday component of the regional power grid electricity data is expressed by the following formula: in: This is the vector of holiday impact factors, representing the degree of impact of different holidays on regional power grid electricity data; The regression vector indicates the holidays.

4. A power grid power prediction system based on data mining and neural networks, characterized in that, The method for implementing the method as described in any one of claims 1 to 3 includes a data collection module, a feature extraction module, and a power grid prediction module; The data collection module is used to collect regional power grid power data, clean the power data to remove abnormal data, and then use second-order Lagrange interpolation to recover the data that was accidentally deleted during the cleaning process, so as to obtain the processed regional power grid power data. The feature extraction module is used to extract new features based on the processed regional power grid power data and regional related information, add the extracted new features to the processed regional power grid power data, and sample the processed regional power grid power data according to a preset length to build a power grid power prediction model training set. The power grid power prediction module is used to construct a power grid power prediction model, which consists of a Prophet model and an LSTM model. The Prophet model decomposes the training set into multiple training sets with different components. Each training set is then input into a separate LSTM model, and the outputs of all LSTM models are fed into a multilayer perceptron. The multilayer perceptron performs feature fusion on the input data to obtain the prediction result. The parameters of the power grid power prediction model are updated based on the prediction result. After training for a preset number of times, the trained power grid power prediction model is obtained. This module uses this model to predict the power grid power.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 3.

Images