Electric quantity fluctuation prediction method and system, electronic equipment and storage medium

Abstract

The invention relates to an artificial intelligence technology, in particular to an electric quantity fluctuation prediction method and system, electronic equipment and a storage medium. An analysis method for electric quantity fluctuation related influence factors comprises the following steps: model training: training a meteorological analysis model, a public opinion analysis model and a market analysis model; probability graph generation: utilizing a meteorological analysis model, a public opinion analysis model and a market analysis model to generate a probability distribution graph corresponding to each power consumption fluctuation influence factor and a corresponding real probability value; and model application: utilizing a meteorological analysis model, a public opinion analysis model, a market analysis model and the real probability value corresponding to each power consumption fluctuation influence factor to predict the target power consumption fluctuation condition in the target prediction time period. According to the method, the diversified data set corresponding to the target prediction time period and various prediction analysis models are utilized to accurately predict the power consumption fluctuation condition corresponding to the target prediction time period.

Description

Technical Field

[0001] The present invention relates to artificial intelligence technology, and particularly to a method, a system, an electronic device and a storage medium for predicting power consumption fluctuations. Background Art

[0002] Predicting power consumption fluctuations is one of the important ways for power system state estimation. Accurately predicting power consumption fluctuation data and deeply analyzing the influencing factors of power consumption fluctuations not only contribute to the planning of power systems, but also have far-reaching significance for power demand forecasting. Predicting power consumption fluctuations is of great significance for the government to master the economic operation situation, promote the development and construction of the power market, support power companies to reasonably determine the total power consumption quota, formulate load management plans, and guide the reasonable operation of transmission and distribution networks.

[0003] However, power companies usually only use datasets corresponding to a single influencing factor for predicting power consumption fluctuations at present, without considering the influence of meteorological factors, public opinion information factors, and market factors on power consumption fluctuations. Therefore, the prediction data is relatively single, thus greatly reducing the prediction accuracy.

[0004] Defects of the prior art: Currently, usually only datasets corresponding to a single influencing factor are used to predict power consumption fluctuations, and the prediction data is relatively single, thus greatly reducing the prediction accuracy; currently, the influence of meteorological factors, public opinion information factors, and market factors on power consumption fluctuations is not considered, so the accuracy of predicting power consumption fluctuations is further reduced. Summary of the Invention

[0005] The present invention considers the influence of meteorological factors, public opinion information factors, and market factors on power consumption fluctuations, constructs a meteorological analysis model corresponding to meteorological factors, a public opinion analysis model corresponding to public opinion information factors, and a market analysis model corresponding to market factors, and collects a dataset for predicting power consumption fluctuations with multiple influencing factors to predict power consumption fluctuations, thereby improving the prediction accuracy.

[0006] In a first aspect, the present invention provides a method for predicting power consumption fluctuations, including the following processes:

[0007] S100: Model training; training the meteorological analysis model, the public opinion analysis model, and the market analysis model;

[0008] S200: Probability graph generation; using the meteorological analysis model, the public opinion analysis model, and the market analysis model to generate probability distribution graphs corresponding to each influencing factor of power consumption fluctuations and corresponding true probability values;

[0009] S300: Model Application; Using a meteorological analysis model, a public opinion analysis model, a market analysis model, and the true probability values corresponding to each electricity consumption fluctuation influencing factor, predict the electricity consumption fluctuation situation during the target prediction period.

[0010] Preferably, the S100 includes the following specific processes:

[0011] S110: Obtain sample data corresponding to the electricity consumption fluctuation influencing factors for each first sample period from the power grid platform; among them, the sample data includes a first sample vector corresponding to meteorological factors, a second sample vector corresponding to public opinion information factors, and a third sample vector corresponding to market factors;

[0012] Meteorological factors include: high temperature, low temperature, rainfall, extreme weather;

[0013] Public opinion information factors include: positive incentives and negative impacts;

[0014] Market factors include: raw material price changes, supply chain impacts, enterprise profit changes, inventory backlogs, product price changes, market demand impacts, export trade impacts;

[0015] The first sample period is any number of periods before the target prediction period;

[0016] S120: Obtain the electricity consumption fluctuation data corresponding to each first sample period from the power grid platform; obtain the electricity consumption fluctuation data Q1 corresponding to the first sample period T 1,1 corresponding to the electricity consumption fluctuation data Q2 corresponding to the first sample period T 1,2 corresponding to the electricity consumption fluctuation data Q3 corresponding to the first sample period T 1,n corresponding to the electricity consumption fluctuation data Q n ;

[0017] S130: Use the sample data corresponding to the electricity consumption fluctuation influencing factors for each first sample period as input samples, and use the electricity consumption fluctuation data corresponding to each first sample period as output samples to train the meteorological analysis model, the public opinion analysis model, and the market analysis model respectively.

[0018] Preferably, the S200 includes the following specific processes:

[0019] S210: Obtain sample data corresponding to the electricity consumption fluctuation influencing factors for each second sample period from the power grid platform; the second sample period is different from the first sample period;

[0020] S220: Input the sample data corresponding to each electricity consumption fluctuation influencing factor into the corresponding prediction model respectively, and use the prediction model to output the electricity consumption fluctuation prediction data corresponding to each second sample period; the prediction model includes: a meteorological analysis model, a public opinion analysis model, and a market analysis model;

[0021] S230: Obtain the actual electricity consumption fluctuation data corresponding to each second sample period from the power grid platform, and determine the correlation coefficient between the electricity consumption fluctuation prediction data and the actual electricity consumption fluctuation data according to the electricity consumption fluctuation prediction data and the actual electricity consumption fluctuation data corresponding to the second sample period;

[0022] S240: Generate a corresponding probability distribution diagram based on the correlation coefficient array;

[0023] S250: Integrate each probability distribution curve based on the 2σ principle, and determine the actual probability corresponding to each electricity consumption fluctuation influencing factor.

[0024] Preferably, the S300 includes the following specific processes:

[0025] S310: Receive the data information uploaded by the user corresponding to the target prediction period, and obtain the data corresponding to the meteorological factors, the data corresponding to the public opinion information factors, and the data corresponding to the market factors corresponding to the target prediction period from the power grid platform;

[0026] S320: Input the meteorological prediction vector into the meteorological analysis model, and output the first electricity consumption fluctuation situation Q 1,x ; Input the public opinion information vector into the public opinion analysis model, and output the second electricity consumption fluctuation situation Q 2,x ; Input the market prediction vector into the market analysis model, and output the third electricity consumption fluctuation situation Q 3,x ;

[0027] S330: The prediction module receives the first electricity consumption fluctuation situation Q 1,x , the second electricity consumption fluctuation situation Q 2,x , and the third electricity consumption fluctuation situation Q 3,x respectively sent by the meteorological analysis model, the public opinion analysis model, and the market analysis model, and uses the actual probability P k,1 corresponding to the meteorological factors, the actual probability P k,2 corresponding to the public opinion information factors, and the actual probability P k,3 corresponding to the market factors sent by the probability diagram generation module to determine the target electricity consumption fluctuation situation corresponding to the target prediction period;

[0028] The specific calculation formula is as follows:

[0029] P k = P k,1 × Q 1,x + P k,2 × Q 2,x + P k,3 × Q 3,x 。

[0030] In a second aspect, the present invention provides a power fluctuation prediction system, including:

[0031] A data receiving module, a data acquisition module, a meteorological analysis module, a public opinion analysis module, a market analysis module, a probability graph generation module, a prediction module, and a data output module;

[0032] The data receiving module is connected to the data acquisition module and is used to determine the target prediction period;

[0033] The data acquisition module is respectively connected to the meteorological analysis module, the public opinion analysis module, and the market analysis module; and is used to obtain data corresponding to the influencing factors of the electricity consumption fluctuation in the target prediction period from the power grid platform;

[0034] The meteorological analysis module, the public opinion analysis module, and the market analysis module are respectively connected to the prediction module; the meteorological analysis module is used to analyze the first electricity consumption fluctuation situation predicted based on the meteorological factor data; the public opinion analysis module is used to analyze the second electricity consumption fluctuation situation predicted based on the public opinion information factor data; the market analysis module is used to analyze the third electricity consumption fluctuation situation predicted based on the market factor data;

[0035] The probability graph generation module is used to generate and store the probability distribution graphs corresponding to the respective influencing factors of the electricity consumption fluctuation and the corresponding true probability values; the probability graph generation module is connected to the prediction module;

[0036] The prediction module is used to predict the electricity consumption fluctuation situation corresponding to the respective influencing factors of the electricity consumption; the prediction module is connected to the data output module.

[0037] In a third aspect, the present invention provides an electronic device, specifically including: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute any one of the above prediction methods.

[0038] In a fourth aspect, the present invention provides a computer-readable storage medium, specifically including: storing a computer program, and the computer program causes a computer to execute any one of the above prediction methods.

[0039] Advantages of the present invention over the prior art: The present invention mainly considers the main factors affecting the prediction of electricity consumption fluctuations, such as meteorological factors, public opinion information factors, and market factors, constructs and trains a meteorological analysis model corresponding to meteorological factors, a public opinion analysis model corresponding to public opinion information factors, and a market analysis model corresponding to market factors, and collects a diversified data set corresponding to the target prediction period. Thus, the diversified data set corresponding to the target prediction period and various prediction analysis models can be further used to accurately predict the electricity consumption fluctuation situation corresponding to the target prediction period. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a schematic structural diagram of a power consumption fluctuation prediction system in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] Compare Figure 1 , a power consumption fluctuation prediction system, comprising:

[0042] a data receiving module, a data acquisition module, a meteorological analysis module, a public opinion analysis module, a market analysis module, a probability graph generation module, a prediction module, and a data output module;

[0043] The data receiving module is connected to the data acquisition module and is used to determine the target prediction period;

[0044] The data acquisition module is respectively connected to the meteorological analysis module, the public opinion analysis module, and the market analysis module; and is used to obtain data corresponding to the influencing factors of power consumption fluctuation in the target prediction period from the power grid platform;

[0045] The meteorological analysis module, the public opinion analysis module, and the market analysis module are respectively connected to the prediction module; the meteorological analysis module is used to analyze the first power consumption fluctuation situation predicted based on meteorological factor data; the public opinion analysis module is used to analyze the second power consumption fluctuation situation predicted based on public opinion information factor data; the market analysis module is used to analyze the third power consumption fluctuation situation predicted based on market factor data;

[0046] The probability graph generation module is used to generate and store the probability distribution graph corresponding to each influencing factor of power consumption fluctuation and the corresponding true probability value; the probability graph generation module is connected to the prediction module;

[0047] The prediction module is used to be based on the predicted power consumption fluctuation situations corresponding to each influencing factor of power consumption; the prediction module is connected to the data output module.

[0048] A power consumption fluctuation prediction method implemented on the above system includes the following processes:

[0049] S100: Model training; training the meteorological analysis model, the public opinion analysis model, and the market analysis model;

[0050] S200: Probability graph generation; generating a probability distribution graph and corresponding true probability values corresponding to each influencing factor of power consumption fluctuation by using a meteorological analysis model, a public opinion analysis model, and a market analysis model;

[0051] S300: Model application; predicting the target power consumption fluctuation situation in the target prediction period by using a meteorological analysis model, a public opinion analysis model, and a market analysis model, and the true probability values corresponding to each influencing factor of power consumption fluctuation.

[0052] The said S100 includes the following specific processes:

[0053] S110: The data acquisition module obtains sample data corresponding to the influencing factors of power consumption fluctuation in each first sample period from the power grid platform. Among them, the sample data includes a first sample vector corresponding to meteorological factors, a second sample vector corresponding to public opinion information factors, and a third sample vector corresponding to market factors.

[0054] Among them, the influencing factors of power consumption fluctuation can include, for example, meteorological factors, public opinion information factors, and market factors. And meteorological factors include, for example, high temperature, low temperature, rainfall, extreme weather, etc. Public opinion information factors include, for example, positive incentives and negative impacts, etc. Market factors include, for example, raw material price changes, supply chain impacts, enterprise profit changes, inventory backlogs, product price changes, market demand impacts, export trade impacts, etc. And the first sample period can be, for example, any number of periods before the target prediction period. In this embodiment, the first sample period can be, for example, one day, that is, 24h.

[0055] For example, the data acquisition module obtains the first sample vector corresponding to the first sample period T 1,1 corresponding thereto The first sample vector corresponding to the first sample period T 1,2 corresponding thereto The first sample vector corresponding to the first sample period T 1,n corresponding thereto Thus, the data acquisition module can determine multiple first sample vectors corresponding to the meteorological factors in the first sample period T 1,1 ~T 1,n Among them, for example, the meteorological factors include h secondary influencing factors, such as high temperature, low temperature, rainfall, extreme weather, etc.

[0056] The data acquisition module obtains the second sample vector corresponding to the first sample period T 1,1 corresponding thereto The second sample vector corresponding to the first sample period T 1,2The corresponding second sample vector corresponding to the first sample period T 1,n The corresponding second sample vector Thus, the data acquisition module can determine the multiple second sample vectors corresponding to the public opinion information factors during the first sample period T 1,1 ~T 1,n Among them, for example, the public opinion information factors include h secondary influencing factors, such as positive incentives and negative impacts, etc. Among them, for example, the public opinion information factors include h secondary influencing factors, such as positive incentives and negative impacts, etc.

[0057] The data acquisition module obtains the third sample vector corresponding to the first sample period T from the power grid platform 1,1 The corresponding third sample vector corresponding to the first sample period T 1,2 The corresponding third sample vector corresponding to the first sample period T 1,n The corresponding third sample vector Thus, the data acquisition module can determine the multiple third sample vectors corresponding to the market factors during the first sample period T 1,1 ~T 1,n Among them, for example, the market factors include h secondary influencing factors, such as raw material price changes, supply chain impacts, enterprise profit changes, inventory backlogs, product price changes, market demand impacts, export trade impacts, etc. Among them, for example, the market factors include h secondary influencing factors, such as raw material price changes, supply chain impacts, enterprise profit changes, inventory backlogs, product price changes, market demand impacts, export trade impacts, etc.

[0058] Of course, it should be noted that the number of secondary features corresponding to each power consumption fluctuation influencing factor can also be different, which will not be elaborated here.

[0059] S120: The data acquisition module obtains the power consumption fluctuation data corresponding to each first sample period from the power grid platform.

[0060] For example, the data acquisition module obtains the power consumption fluctuation data corresponding to the first sample period, the power consumption fluctuation data corresponding to the first sample period,..., the power consumption fluctuation data corresponding to the first sample period from the power grid platform. Among them, since the first sample period in this embodiment is one day, the power consumption fluctuation data can be calculated, for example, from 0:00 of the previous day to 0:00 of the next day, which will not be elaborated here.

[0061] S130: The data acquisition module sends the sample data corresponding to the power consumption fluctuation influencing factors of each first sample period, and the power consumption fluctuation data corresponding to each first sample period to the model training module. The model training module uses the sample data corresponding to the power consumption fluctuation influencing factors of each first sample period as input samples, and uses the power consumption fluctuation data corresponding to each first sample period as output samples, and trains the meteorological analysis model, the public opinion analysis model, and the market analysis model respectively.

[0062] For example, the model training module uses multiple first sample vectors as input samples, and uses the power consumption fluctuation data Q1 to Q n as output samples to train a pre-constructed meteorological analysis model. Thus, the trained meteorological analysis model is a model for predicting power consumption fluctuation data based on data corresponding to meteorological factors.

[0063] For another example, the model training module uses multiple second sample vectors as input samples, and uses the power consumption fluctuation data Q1 to Q n as output samples to train a pre-constructed public opinion analysis model. Thus, the trained public opinion analysis model is a model for predicting power consumption fluctuation data based on data corresponding to public opinion information factors.

[0064] For another example, the model training module uses multiple third sample vectors as input samples, and uses the power consumption fluctuation data Q1 to Q n as output samples to train a pre-constructed market analysis model. Thus, the trained market analysis model is a model for predicting power consumption fluctuation data based on data corresponding to public opinion information factors.

[0065] The S200 includes the following specific processes:

[0066] S210: The data acquisition module obtains sample data corresponding to the influencing factors of power consumption fluctuation in each second sample period from the power grid platform. Among them, the second sample period is different from the first sample period. And in this embodiment, the second sample period can be, for example, one day (24h).

[0067] For example, the data acquisition module obtains a fourth sample vector corresponding to the second sample period T 2,1 from the power grid platform The fourth sample vector corresponding to the second sample period T 2,2 The fourth sample vector corresponding to the second sample period T The fourth sample vector corresponding to the second sample period T 2,n The fourth sample vector corresponding to the second sample period T Thus, the data acquisition module can determine multiple fourth sample vectors corresponding to the meteorological factors of the second sample period T 2,1 ~T 2,n Among them, for example, the meteorological factors include h secondary influencing factors, such as high temperature, low temperature, rainfall, extreme weather, etc.

[0068] The data acquisition module obtains a fifth sample vector corresponding to the second sample period T 2,1 from the power grid platform​ corresponding to the second sample period T 2,2 the fifth sample vector corresponding to the second sample period T 2,n the fifth sample vector Thus, the data acquisition module can determine the multiple fifth sample vectors corresponding to the public opinion information factors during the second sample period T 2,1 ~T 2,n For example, the meteorological factors include h secondary influencing factors, such as high temperature, low temperature, rainfall, extreme weather, etc. For example, the public opinion information factors include h secondary influencing factors, such as positive incentives and negative impacts, etc.

[0069] The data acquisition module obtains the sixth sample vector corresponding to the second sample period T from the power grid platform 2,1 corresponding to the second sample period T 2,2 the sixth sample vector corresponding to the second sample period T 2,n the sixth sample vector Thus, the data acquisition module can determine the multiple sixth sample vectors corresponding to the market factors during the second sample period T 2,1 ~T 2,n For example, the market factors include h secondary influencing factors, such as high temperature, low temperature, rainfall, extreme weather, etc. For example, the public opinion information factors include h secondary influencing factors, such as raw material price changes, supply chain impacts, enterprise profit changes, inventory backlogs, product price changes, market demand impacts, export trade impacts, etc.

[0070] S220: The probability distribution graph generation module inputs the sample data corresponding to each power consumption fluctuation influencing factor into the corresponding prediction models (i.e., the meteorological analysis model, the public opinion analysis model, and the market analysis model) respectively, and uses the prediction models to output the power consumption fluctuation prediction data corresponding to each second sample period.

[0071] For example, the probability distribution graph generation module inputs the fourth sample vector into the meteorological analysis model, and uses the meteorological analysis model to output the first power consumption fluctuation prediction data J 2,1 corresponding to the second sample period T 1,1 .

[0072] The probability distribution graph generation module inputs the fourth sample vector into the meteorological analysis model, and uses the meteorological analysis model to output the first power consumption fluctuation prediction data J 2,2 corresponding to the second sample period T​​​1,2 .

[0073] And so on.

[0074] The probability distribution map generation module inputs the fourth sample vector into the meteorological analysis model, and uses the meteorological analysis model to output the first electricity consumption fluctuation prediction data J corresponding to the second sample period T 2,n . 1,n .

[0075] Thus, the meteorological analysis model can output the electricity consumption fluctuation prediction data J corresponding to each second sample period T 2,1 ~T 2,n . 1,1 ~J 1,n .

[0076] Similarly, the probability distribution map generation module inputs each fifth sample vector into the public opinion analysis model respectively, so that the public opinion analysis model can output the second electricity consumption fluctuation prediction data J corresponding to each second sample period T 2,1 ~T 2,n . 2,1 ~J 2,n .

[0077] The probability distribution map generation module inputs each sixth sample vector into the market analysis model respectively, so that the market analysis model can output the second electricity consumption fluctuation prediction data J corresponding to each second sample period T 2,1 ~T 2,n . 3,1 ~J 3,n .

[0078] S230: The data acquisition module obtains the real electricity consumption fluctuation data corresponding to each second sample period from the power grid platform, and the probability distribution map generation module determines multiple groups of correlation coefficients based on the real electricity consumption fluctuation data corresponding to each second sample period (including the first electricity consumption fluctuation prediction data, the second electricity consumption fluctuation prediction data, and the third electricity consumption fluctuation prediction data), and the electricity consumption fluctuation prediction data corresponding to each second sample period.

[0079] Among them, the data acquisition module obtains the real electricity consumption fluctuation data J corresponding to the second sample period T 2,1 , the real electricity consumption fluctuation data J corresponding to the second sample period T x,1 ,..., the real electricity consumption fluctuation data J corresponding to the second sample period T 2,2 . x,2 , the real electricity consumption fluctuation data J corresponding to the second sample period T 2,n . x,n .

[0080] Then, based on the predicted power consumption fluctuation data and the actual power consumption fluctuation data corresponding to the second sample period, the probability distribution graph generation module determines the correlation coefficient between the predicted power consumption fluctuation data and the actual power consumption fluctuation data.

[0081] For example, the probability distribution graph generation module determines the first predicted power consumption fluctuation data J 2,1 corresponding to the second sample period T 1,1 , the second predicted power consumption fluctuation data J 2,1 , and the third predicted power consumption fluctuation data J 3,1 , and determines the actual power consumption fluctuation data J 2,1 corresponding to the second sample period T x,1 . Based on the above data, the coefficients k 1,1 ~k 1,3 corresponding to each predicted power consumption fluctuation data are determined. The specific calculation formula is as follows: J x,1 =k 1,1 ×J 1,1 +k 1,2 ×J 1,2 +k 1,3 ×J 1,3

[0082] Similarly, the probability distribution graph generation module determines the first predicted power consumption fluctuation data J 2,2 corresponding to the second sample period T 1,2 , the second predicted power consumption fluctuation data J 2,2 , and the third predicted power consumption fluctuation data J 3,2 , and determines the actual power consumption fluctuation data J 2,2 corresponding to the second sample period T x,2 . Based on the above data, the coefficients k 2,1 ~k 2,3 corresponding to each predicted power consumption fluctuation data are determined. The specific calculation formula is as follows: J x,2 =k 2,1 ×J 2,1 +k 2,2 ×J 2,2 +k 2,3 ×J 2,3

[0083] And so on.

[0084] The probability distribution graph generation module can determine the coefficients k n,1 ~k n,3 corresponding to each predicted power consumption fluctuation data. The specific calculation formula is as described above and will not be elaborated here.

[0085] Thus, the probability distribution graph generation module can determine multiple correlation arrays corresponding to the electricity consumption fluctuation prediction data of each electricity consumption fluctuation influencing factor. For example, the correlation array k corresponding to the electricity consumption fluctuation prediction data of the meteorological factor 1,1 ~k n,1 。The correlation array k corresponding to the electricity consumption fluctuation prediction data of the public opinion information factor 1,2 ~k n,2 。The correlation array k corresponding to the electricity consumption fluctuation prediction data of the market factor 1,3 ~k n,3 。

[0086] S240: The probability distribution graph generation module generates corresponding probability distribution graphs based on the correlation arrays corresponding to the electricity consumption fluctuation prediction data of each electricity consumption fluctuation influencing factor.

[0087] Specifically, the probability distribution graph generation module fits the correlation array k corresponding to the electricity consumption fluctuation prediction data of the meteorological factor 1,1 ~k n,1 and generates a probability distribution graph related to the meteorological factor. Among them, the probability distribution curve can be, for example, a normal distribution curve H1.

[0088] Similarly, the probability distribution graph generation module fits the correlation array k corresponding to the electricity consumption fluctuation prediction data of the public opinion information factor 1,2 ~k n,2 and generates a probability distribution graph related to the public opinion information factor. Among them, the probability distribution curve can be, for example, a normal distribution curve H2.

[0089] Similarly, the probability distribution graph generation module fits the correlation array k corresponding to the electricity consumption fluctuation prediction data of the market factor 1,3 ~k n,3 and generates a probability distribution graph related to the market factor. Among them, the probability distribution curve can be, for example, a normal distribution curve H3.

[0090] S250: When the probability distribution graph generation module determines the probability distribution curves corresponding to the electricity consumption fluctuation prediction data of each electricity consumption fluctuation influencing factor, further determine the true probability.

[0091] Specifically, based on the 2σ principle, integrate each probability distribution curve and determine the true probability corresponding to each electricity consumption fluctuation influencing factor.

[0092] For example, based on the 2σ principle, integrate the probability distribution curve corresponding to the meteorological factor and calculate the true probability P corresponding to the meteorological factor k,1 。The specific calculation formula is as follows:

[0093]

[0094] Among them, P k,1 represents the true probability corresponding to the meteorological factors, H1(k) represents the normal distribution function corresponding to the meteorological factors, and 95.45% represents the probability that the numerical values are distributed in (k - 2σ, k + 2σ) based on the 2σ principle.

[0095] For another example, based on the 2σ principle, integrate the probability distribution curve corresponding to the public opinion information factors, and calculate the true probability P corresponding to the public opinion information factors k,2 . The specific calculation formula is as follows:

[0096]

[0097] Among them, P k,2 represents the true probability corresponding to the public opinion information factors, H2(k) represents the normal distribution function corresponding to the public opinion information factors, and 95.45% represents the probability that the numerical values are distributed in (k - 2σ, k + 2σ) based on the 2σ principle.

[0098] For another example, based on the 2σ principle, integrate the probability distribution curve corresponding to the market factors, and calculate the true probability P corresponding to the market factors k,3 . The specific calculation formula is as follows:

[0099]

[0100] Among them, P k,3 represents the true probability corresponding to the market factors, H3(k) represents the normal distribution function corresponding to the market factors, and 95.45% represents the probability that the numerical values are distributed in (k - 2σ, k + 2σ) based on the 2σ principle.

[0101] Thus, based on the above method, the probability distribution map generation module can determine the true probability P corresponding to the meteorological factors k,1 , the true probability P corresponding to the public opinion information factors k,2 and the true probability P corresponding to the market factors k,3 .

[0102] The S300 includes the following processes:

[0103] S310: The data receiving module receives the data information corresponding to the target prediction period uploaded by the user through the terminal device, and when the data acquisition module determines the target prediction period, obtains the data corresponding to the meteorological factors of the target prediction period, the data corresponding to the public opinion information factors of the target prediction period, and the data corresponding to the market factors of the target prediction period from the power grid platform.

[0104] For example, the data acquisition module obtains from the power grid platform the meteorological prediction vector x,1 corresponding to the target prediction period T the meteorological prediction vector x,2 corresponding to the target prediction period T the meteorological prediction vector x,n corresponding to the target prediction period T Thus, the data acquisition module can determine multiple meteorological prediction vectors corresponding to the meteorological factors for the target prediction period T x,1 ~T x,n wherein, for example, the meteorological factors include h secondary influencing factors, such as high temperature, low temperature, rainfall, extreme weather, etc.

[0105] The data acquisition module obtains from the power grid platform the public opinion information vector corresponding to the target prediction period T x,1 the public opinion information vector corresponding to the target prediction period T x,2 the public opinion information vector corresponding to the target prediction period T x,n Thus, the data acquisition module can determine multiple public opinion information vectors corresponding to the public opinion information factors for the target prediction period T x,1 ~T x,n wherein, for example, the public opinion information factors include h secondary influencing factors, such as positive incentives and negative impacts, etc.

[0106] The data acquisition module obtains from the power grid platform the market prediction vector corresponding to the target prediction period T x,1 the market prediction vector corresponding to the target prediction period T x,2 the market prediction vector corresponding to the target prediction period T x,n

[0107] Thus, the data acquisition module can determine multiple market prediction vectors corresponding to the market factors for the target prediction period T x,1 ~T x,n wherein, for example, the market factors include h secondary influencing factors, such as raw material price changes, supply chain impacts, enterprise profit changes, inventory backlogs, product price changes, market demand impacts, export trade impacts, etc.

[0107] S320: Then, the data acquisition module inputs the meteorological prediction vector corresponding to the meteorological factors into the meteorological analysis model and outputs the first power consumption fluctuation situation Q 1,x; The public opinion information vector corresponding to the public opinion information factor is input into the public opinion analysis model, and the second power consumption fluctuation situation Q is output 2,x ; The market prediction vector corresponding to the market factor is input into the market analysis model, and the third power consumption fluctuation situation Q is output 3,x .

[0108] S330: The prediction module receives the first power consumption fluctuation situation Q, the second power consumption fluctuation situation Q, and the third power consumption fluctuation situation Q respectively sent by the meteorological analysis model, the public opinion analysis model, and the market analysis model 1,x , the second power consumption fluctuation situation Q 2,x and the third power consumption fluctuation situation Q 3,x , and uses the true probability P corresponding to the meteorological factor sent by the probability graph generation module k,1 , the true probability P corresponding to the public opinion information factor k,2 and the true probability P corresponding to the market factor k,3 to determine the target power consumption fluctuation situation corresponding to the target prediction period

[0109] The specific calculation formula is as follows

[0110] P k =P k,1 ×Q 1,x +P k,2 ×Q 2,x +P k,3 ×Q 3,x

[0111] S340: The prediction module sends the calculated target power consumption fluctuation situation corresponding to the target prediction period to the data output module, and the data output module sends the target power consumption fluctuation situation to the terminal device. Thus, the user can view the target power consumption fluctuation situation corresponding to the target prediction period through the terminal device, so as to further specify the adjustment plan

Claims

1. A method for predicting power fluctuations, characterized in that, It includes the following processes: S100: Model training; training the meteorological analysis model, public opinion analysis model, and market analysis model; S200: Probability graph generation; using the meteorological analysis model, public opinion analysis model, and market analysis model to generate probability distribution graphs and corresponding true probability values corresponding to each electricity consumption fluctuation influencing factor; S300: Model application; using the meteorological analysis model, public opinion analysis model, and market analysis model, and the true probability values corresponding to each electricity consumption fluctuation influencing factor, to predict the electricity consumption fluctuation situation in the target prediction period.

2. The method for predicting power fluctuations according to claim 1, wherein The S100 includes the following specific processes: S110: Obtain sample data corresponding to the electricity consumption fluctuation influencing factors in each first sample period from the power grid platform; among them, the sample data includes a first sample vector corresponding to meteorological factors, a second sample vector corresponding to public opinion information factors, and a third sample vector corresponding to market factors; Meteorological factors include: high temperature, low temperature, rainfall, extreme weather; Public opinion information factors include: positive incentives and negative impacts; Market factors include: raw material price changes, supply chain impacts, enterprise profit changes, inventory backlogs, product price changes, market demand impacts, export trade impacts; The first sample period is any number of periods before the target prediction period; S120: Obtain the power consumption fluctuation data corresponding to each first sample period from the power grid platform; obtain the power consumption fluctuation data Q1 corresponding to the first sample period T 1,1 from the power grid platform, the power consumption fluctuation data Q2 corresponding to the first sample period T 1,2 , and so on, until the power consumption fluctuation data Q 1,n corresponding to the first sample period T n ; S130: Use the sample data corresponding to the electricity consumption fluctuation influencing factors in each first sample period as input samples, and use the electricity consumption fluctuation data corresponding to each first sample period as output samples, and train the meteorological analysis model, public opinion analysis model, and market analysis model respectively.

3. The power consumption fluctuation prediction method according to claim 1, wherein The S200 includes the following specific processes: S210: Obtain sample data corresponding to the electricity consumption fluctuation influencing factors in each second sample period from the power grid platform; the second sample period is different from the first sample period; S220: Input the sample data corresponding to each electricity consumption fluctuation influencing factor into the corresponding prediction model respectively, and use the prediction model to output the electricity consumption fluctuation prediction data corresponding to each second sample period; the prediction models include: meteorological analysis model, public opinion analysis model, and market analysis model; S230: Obtain the true electricity consumption fluctuation data corresponding to each second sample period from the power grid platform, and determine the correlation coefficient between the electricity consumption fluctuation prediction data and the true electricity consumption fluctuation data according to the electricity consumption fluctuation prediction data and the true electricity consumption fluctuation data corresponding to the second sample period; S240: Generate a corresponding probability distribution graph based on the correlation coefficient array; S250: Integrate each probability distribution curve based on the 2σ principle, and determine the true probability corresponding to each electricity consumption fluctuation influencing factor.

4. The power fluctuation prediction method according to claim 1, characterized in that, The S300 includes the following specific processes: S310: Receive the data information uploaded by the user corresponding to the target prediction period, and obtain the data corresponding to the meteorological factors, the data corresponding to the public opinion information factors, and the data corresponding to the market factors in the target prediction period from the power grid platform; S320: Input the meteorological prediction vector into the meteorological analysis model and output the first electricity consumption fluctuation situation Q 1,x ; Input the public opinion information vector into the public opinion analysis model and output the second electricity consumption fluctuation situation Q 2,x ; Input the market prediction vector into the market analysis model and output the third electricity consumption fluctuation situation Q 3,x ; S330: Determine the target power consumption fluctuation corresponding to the target prediction period based on the first power consumption fluctuation Q 1,x , the second power consumption fluctuation Q 2,x , and the third power consumption fluctuation Q 3,x , the true probability P corresponding to the meteorological factor k,1 , the true probability P corresponding to the public opinion information factor k,2 , and the true probability P corresponding to the market factor k,3 . The specific calculation formula is as follows: P k = P k,1 × Q 1,x + P k,2 × Q 2,x + P k,3 × Q 3,x 。 5. The method for predicting power fluctuations according to claim 2, characterized in that In the S110, Obtain the first sample vector corresponding to the first sample period T from the power grid platform 1,1 The first sample vector The first sample vector corresponding to the first sample period T 1,2 The first sample vector The first sample vector corresponding to the first sample period T 1,n The first sample vector Determine multiple first sample vectors corresponding to the meteorological factors from the first sample period T 1,1 ~T 1,n where the meteorological factors include h1 secondary influencing factors; where the meteorological factors include h1 secondary influencing factors; Obtain the second sample vector corresponding to the first sample period T 1,1 from the power grid platform The second sample vector corresponding to the first sample period T 1,2 from the power grid platform The second sample vector corresponding to the first sample period T 1,n from the power grid platform Determine the multiple second sample vectors corresponding to the public opinion information factors from the first sample period T 1,1 to T 1,n where the public opinion information factors include h2 secondary influencing factors; ​ Obtain the third sample vector corresponding to the first sample period T from the power grid platform 1,1 The third sample vector corresponding to the first sample period T The third sample vector corresponding to the first sample period T 1,2 The third sample vector corresponding to the first sample period T The third sample vector corresponding to the first sample period T 1,n The third sample vector corresponding to the first sample period T Determine multiple third sample vectors corresponding to the market factors from the first sample period T 11 ~T 1n The market factors include h3 secondary influencing factors; The market factors include h3 secondary influencing factors; In the S130, during model training, multiple first sample vectors are used as input samples, and the power consumption fluctuation data Q1 to Q n is used as the output sample to train the pre-constructed meteorological analysis model; thus, the trained meteorological analysis model is a model for predicting power consumption fluctuation data based on data corresponding to meteorological factors. Using multiple second sample vectors as input samples, and using power consumption fluctuation data Q1 to Q n as output samples, train a pre-constructed public opinion analysis model; thus, the trained public opinion analysis model is a model for predicting power consumption fluctuation data based on data corresponding to public opinion information factors; Take multiple third sample vectors as input samples, and use the power consumption fluctuation data Q1 to Q n as output samples to train a pre-constructed market analysis model; thus, the trained market analysis model is a model for predicting power consumption fluctuation data based on data corresponding to public opinion information factors.

6. The power consumption fluctuation prediction method according to claim 3, wherein In the S210, Obtain the fourth sample vector corresponding to the second sample period T from the power grid platform 2,1 The fourth sample vector The fourth sample vector corresponding to the second sample period T 2,2 The fourth sample vector The fourth sample vector corresponding to the second sample period T 2,n The fourth sample vector Determine multiple fourth sample vectors corresponding to the meteorological factors during the second sample period T 2,1 ~T 2,n The multiple fourth sample vectors corresponding to the meteorological factors Obtain the fifth sample vector corresponding to the second sample period T from the power grid platform 2,1 The fifth sample vector The fifth sample vector corresponding to the second sample period T 2,2 The fifth sample vector The fifth sample vector corresponding to the second sample period T 2,n The fifth sample vector Determine the multiple fifth sample vectors corresponding to the public opinion information factors from the second sample period T 2,1 ~T 2,n The multiple fifth sample vectors Obtain the sixth sample vector corresponding to the second sample period T from the power grid platform 2,1 The sixth sample vector corresponding to the second sample period T The sixth sample vector corresponding to the second sample period T 2,2 The sixth sample vector corresponding to the second sample period T The sixth sample vector corresponding to the second sample period T 2,n The sixth sample vector corresponding to the second sample period T Determine the multiple sixth sample vectors corresponding to the market factors from the second sample period T 2,1 ~T 2,n The multiple sixth sample vectors corresponding to the market factors In S220, the fourth sample vector is input into the meteorological analysis model, and the meteorological analysis model is used to output the first predicted power consumption fluctuation data corresponding to the second sample period T 2,1 ; Input the fourth sample vector into the meteorological analysis model, and use the meteorological analysis model to output the first predicted power consumption fluctuation data J 2,2 corresponding to the second sample period T 1,2 ; And so on; Input the fourth sample vector into the meteorological analysis model, and use the meteorological analysis model to output the first predicted power consumption fluctuation data J 2,n corresponding to the second sample period T 1,n ; The output of the meteorological analysis model and each second sample period T 2,1 ~T 2,n The corresponding predicted power consumption fluctuation data J 1,1 ~J 1,n ; Input the fifth sample vector into the public opinion analysis model respectively, so that the public opinion analysis model can output the second power consumption fluctuation prediction data J 2,1 ~T 2,n corresponding to each second sample period T 2,1 ~J 2,n ; Input each of the sixth sample vectors into the market analysis model respectively, so that the market analysis model can output second power consumption fluctuation prediction data J 2,1 ~T 2,n corresponding to each second sample period T 3,1 ~J 3,n ; In S230, determine the first predicted power consumption fluctuation data J 2,1 corresponding to the second sample period T 1,1 , the second predicted power consumption fluctuation data J 2,1 and the third predicted power consumption fluctuation data J 3,1 , and determine the true power consumption fluctuation data J 2,1 corresponding to the second sample period T x,1 . Based on the above data, determine the coefficients k 1,1 ~k 1,3 corresponding to each predicted power consumption fluctuation data; The specific calculation formula is as follows: J x,1 = k 1,1 × J 1,1 + k 1,2 × J 1,2 + k 1,3 × J 1,3 ; Determine the first predicted power consumption fluctuation data J corresponding to the second sample period T 2,2 , the second predicted power consumption fluctuation data J 1,2 , and the third predicted power consumption fluctuation data J 2,2 , and determine the true power consumption fluctuation data J corresponding to the second sample period T 3,2 , and based on the above data, determine the coefficients k 2,2 ~k x,2 corresponding to each predicted power consumption fluctuation data 2,1 ; 2,3 ; The specific calculation formula is as follows: J x,2 = k 2,1 × J 2,1 + k 2,2 × J 2,2 + k 2,3 × J 2,3 ; And so on; determine the coefficients k corresponding to each power consumption fluctuation prediction data n,1 ~k n,3 ; The correlation array corresponding to the electricity consumption fluctuation prediction data related to meteorological factors is k 1,1 ~k n,1 ; The correlation array k corresponding to the electricity consumption fluctuation prediction data of public opinion information factors 1,2 ~k n,2 ; The correlation coefficient array k corresponding to the predicted data of power consumption fluctuations due to market factors 1,3 ~k n,3 ; In the step S240, the correlation coefficient arrays k 1,1 ~k n,1 corresponding to the electricity consumption fluctuation prediction data of meteorological factors are fitted, and a probability distribution graph related to meteorological factors is generated; wherein, the probability distribution curve is a normal distribution curve H1; The correlation array k corresponding to the power consumption fluctuation prediction data of public opinion information factors 1,2 ~k n,2 is fitted to generate a probability distribution graph related to public opinion information factors; among them, the probability distribution curve is the normal distribution curve H2; The correlation coefficient array k corresponding to the predicted data of the electricity consumption fluctuation related to market factors 1,3 ~k n,3 is fitted to generate a probability distribution graph related to market factors; wherein, the probability distribution curve is a normal distribution curve H3; In S250, based on the 2σ principle, integrate the probability distribution curve corresponding to the meteorological factor, and calculate the true probability P corresponding to the meteorological factor k,1 ; The specific calculation formula is as follows: Among them, P k,1 represents the true probability corresponding to the meteorological factor, H1(k) represents the normal distribution function corresponding to the meteorological factor, and 95.45% represents the probability that the numerical distribution is in (k - 2σ, k + 2σ) based on the 2σ principle; Based on the 2σ principle, integrate the probability distribution curve corresponding to the public opinion information factor, and calculate the true probability P corresponding to the public opinion information factor k,2 ; The specific calculation formula is as follows: Among them, P k,2 represents the true probability corresponding to the public opinion information factor, H2(k) represents the normal distribution function corresponding to the public opinion information factor, and 95.45% represents the probability that the numerical distribution is in (k - 2σ, k + 2σ) based on the 2σ principle; Based on the 2σ principle, integrate the probability distribution curve corresponding to the market factor and calculate the true probability P corresponding to the market factor k,3 ; The specific calculation formula is as follows: where P k,3 represents the true probability corresponding to the market factor, H3(k) represents the normal distribution function corresponding to the market factor, and 95.45% represents the probability that the numerical distribution is in (k - 2σ, k + 2σ) based on the 2σ principle.

7. The method for predicting power fluctuation according to claim 4, characterized in that In the S310, Obtain the meteorological prediction vector corresponding to the target prediction period T from the power grid platform x,1 The meteorological prediction vector corresponding to the target prediction period T The meteorological prediction vector corresponding to the target prediction period T x,2 The meteorological prediction vector corresponding to the target prediction period T The meteorological prediction vector corresponding to the target prediction period T x,n The meteorological prediction vector corresponding to the target prediction period T Determine the multiple meteorological prediction vectors corresponding to the meteorological factors from T x,1 ~T x,n The meteorological prediction vector corresponding to the meteorological factors from T Obtain the public opinion information vector corresponding to the target prediction period T from the power grid platform x,1 The corresponding public opinion information vector The public opinion information vector corresponding to the target prediction period T x,2 The corresponding public opinion information vector The public opinion information vector corresponding to the target prediction period T x,n The corresponding public opinion information vector Determine the multiple public opinion information vectors corresponding to the public opinion information factors from T x,1 ~T x,n The corresponding public opinion information vectors Obtain the market prediction vector corresponding to the target prediction period T from the power grid platform x,1 The market prediction vector The market prediction vector corresponding to the target prediction period T x,2 The market prediction vector The market prediction vector corresponding to the target prediction period T x,n The market prediction vector Determine the multiple market prediction vectors corresponding to the market factors from T x,1 ~T x,n The market prediction vector 8. A power fluctuation prediction system, characterized in that, It includes: Data receiving module, data acquisition module, meteorological analysis module, public opinion analysis module, market analysis module, probability graph generation module, prediction module, data output module; The data receiving module is connected to the data acquisition module and is used to determine the target prediction period; The data acquisition module is respectively connected to the meteorological analysis module, the public opinion analysis module and the market analysis module; and is used to obtain data corresponding to the influencing factors of the power consumption fluctuation during the target prediction period from the power grid platform; The meteorological analysis module, the public opinion analysis module and the market analysis module are respectively connected to the prediction module; the meteorological analysis module is used to analyze the first power consumption fluctuation situation predicted based on meteorological factor data; the public opinion analysis module is used to analyze the second power consumption fluctuation situation predicted based on public opinion information factor data; the market analysis module is used to analyze the third power consumption fluctuation situation predicted based on market factor data; The probability graph generation module is used to generate and store the probability distribution graph corresponding to each power consumption influencing factor and the corresponding true probability value; the probability graph generation module is connected to the prediction module; The prediction module is used to predict the power consumption fluctuation situation corresponding to each power consumption influencing factor; the prediction module is connected to the data output module.

9. An electronic device, characterized in that, Comprising: A memory storing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory and is used to execute the prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Comprising: A computer program stored, and the computer program causes a computer to execute the prediction method according to any one of claims 1 to 7.

Images