A moving average-based marketized transactional electricity forecasting method and system

Abstract

The application relates to a market transaction electricity prediction method and system based on a sliding average, which comprises the following steps: querying the electricity consumption of all users and market users in each day according to historical daily marketing archives, and calculating the market electricity proportion; a time series model of the market electricity proportion is established, and historical data is cleaned; based on the time series model, the average value of the market electricity proportion is calculated according to a set window width and weight, and is used as the prediction value of the market electricity proportion of the next day; the market electricity proportion of each day in a future period is predicted in a rolling manner with one day as a step; the total electricity consumption of each day in a future period is predicted by using a time series prediction model, and the market transaction electricity is calculated by combining the market electricity proportion. The method and system are beneficial to accurately predicting the market transaction electricity.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power grid power prediction, and particularly relates to a market transaction power prediction method and system based on moving average. BACKGROUND

[0002] Power grid enterprises, power transaction centers and government agencies all need to pay attention to the trend of market transaction power and predict the market transaction power in the future. However, unlike the prediction of power sales or power consumption, the archives of market users can change at any time, and users can exit the market or new users can join the market at any time, resulting in uncertainty of the prediction subject itself, and lack of a clear prediction range (or prediction set), so the traditional method for predicting the power sales or power consumption of a provincial power grid cannot be directly applied to the prediction of market transaction power, and it is necessary to provide a market transaction power prediction method. SUMMARY

[0003] The application aims to provide a market transaction power prediction method and system based on moving average, which is beneficial to accurately predict market transaction power.

[0004] To achieve the above-mentioned purpose, the technical solution adopted by the application is: a market transaction power prediction method based on moving average, comprising:

[0005] According to the historical daily retained marketing archives, the power consumption of all users and market users is queried, and the market power ratio is calculated;

[0006] A time series model of the market power ratio is established, and historical data is cleaned;

[0007] Based on the time series model, the average value of the market power ratio is calculated according to the set window width and weight, as the predicted value of the market power ratio of the next day;

[0008] The market power ratio of each day in the future is predicted in steps of one day;

[0009] The total power consumption of each day in the future is predicted by using a time series prediction model, and the market transaction power is calculated in combination with the market power ratio.

[0010] Further, the market users are industrial and commercial users recorded in the historical daily marketing archives, rather than market users screened according to the latest marketing archives.

[0011] Further, the specific method for cleaning the historical data is:

[0012] The daily total user power consumption Q t is compared with the power consumption Qt-1 , the electric quantity Q of the previous 7 days t-7 , the electric quantity Q of the same period of the new calendar t,new , and the electric quantity Q of the same period of the lunar calendar t,lunar are compared, and the following four rules are set:

[0013] Rule 1: |Q t - Q t-1 | > a x Q t-1

[0014] Rule 2: |Q t - Q t-7 | > b x Q t-7

[0015] Rule 3: |Q t - Q t,new | > c x Q t,new

[0016] Rule 4: |Q t - Q t,lunar | > d x Q t,lunar

[0017] Wherein, a, b, c, d are deviation threshold values, if the four rules are all met, it is considered that the daily electric quantity data is abnormal, and as long as one rule is not met, it is considered that the daily electric quantity data is normal;

[0018] If the total user electric quantity of day t is determined to be abnormal data, the total user electric quantity of day t and the market user electric quantity of day t are simultaneously corrected using the data of the previous 7 days, that is, the total user electric quantity of day t is calculated according to Q t = 0.5 x Q t-7 + 0.5 x (Q t-1 + Q t-2 + Q t-3 + Q t-4 + Q t-5 + Q t-6 + Q t-7 ); wherein, Q t to Q t-7 represent the total user electric quantity of day t to day t-7 respectively; at the same time, the market user electric quantity of day t W t is corrected, that is, the market user electric quantity of day t is calculated according to W t = 0.5 x W t-7 + 0.5 x (W t-1 + W t-2 + W t-3 + W t-4 + W t-5 + W t-6 + W t-7 ); wherein, W t to W t-7respectively represent the marketization user power consumption from day t to day t-7.

[0019] Further, the window width and weight determination method is:

[0020] Let the window width be l, then the window width needs to be a multiple of 7 days, i.e. mod(l,7)=0;

[0021] Let the weight coefficient of day t be a t , and the centralized market entry time specified by the power trading center is day k of each month, then the weight coefficient of day k is 10, i.e. a t =10; the weight coefficient of other dates is 1.

[0022] Further, the specific method of rolling prediction of marketization power consumption ratio per day in a future period of time with a step of one day is:

[0023] Let the marketization user power consumption ratio of day t be p t , then according to the formula , the marketization user power consumption ratio of day t+1 is calculated, and then the calculated value of the marketization user power consumption ratio of day t+1 is taken as the input to predict the marketization user power consumption ratio of day t+2, i.e. , and the marketization user power consumption ratio of a future date is predicted in this way.

[0024] Further, the first aspect of the time series prediction model contains three variables: independent variables, covariates and dependent variables, wherein the independent variables are time d1, d2,…,d t , the time interval is 1 day, i.e. d t -d t-1 =1 day; the dependent variables are power Q1, Q2,…,Q t ; the covariates, also known as characteristic quantities, include daily maximum temperature, daily minimum temperature, daily rainfall, daily humidity, daily weather state, sunrise time, sunset time, time-of-use electricity price and GDP, denoted as z1, z2,…,z n ; the second aspect includes the functional relationship between the independent variables and the covariates on the dependent variables, i.e. Q t =f(t)=g(d1,d2,…,…,d t ,Q1,Q2,…,Q t-1 ,z1,z2,…,z n ); the parameters in the function f(t) are obtained by training historical data, wherein the historical data participating in the training are cleaned data rather than original data; the time series model contains periodic terms and considers annual, monthly and weekly periodicity, i.e. the periodic term s(t) is contained in f(t):

[0025]

[0026] wherein a year and b year are annual periodic term coefficients; a month and b month are monthly periodic term coefficients; a week and b week are weekly periodic term coefficients, all of which are obtained by least square fitting.

[0027] The application also provides a market transaction electricity quantity prediction system based on moving average, which is used to realize the above method, and is characterized in that it comprises:

[0028] a data query module, which is used to export daily electricity quantity data from a power marketing database according to settlement units, electricity categories and marketization attributes;

[0029] a data cleaning module, which is used to judge, clean and correct abnormal data;

[0030] a rolling prediction module, which is used to predict marketization electricity quantity in a future period of time;

[0031] a result display module, which is used to summarize data and output a visual report according to preset rules.

[0032] Further, the data query module decomposes provincial power grid electricity quantity according to power supply units and electricity categories, wherein the electricity categories include large industrial electricity, non-general industrial electricity, commercial electricity and non-residential lighting electricity, and the power supply units are divided according to settlement units, and daily electricity quantity data is output.

[0033] Further, the rolling prediction module predicts marketization user electricity quantity proportion of each power supply unit and electricity category in a future period of time, and predicts daily electricity quantity of all users of each power supply unit and electricity category in a future period of time, so as to calculate marketization user daily electricity quantity in a future period of time, and further summarize monthly electricity quantity.

[0034] Compared with the prior art, the application has the following beneficial effects:

[0035] 1. The application divides marketization transaction electricity quantity into two sub-terms of all user electricity quantity and marketization user electricity quantity proportion for prediction, which not only utilizes the regularity of all user electricity quantity, but also considers the randomness of marketization user electricity quantity proportion, thereby improving the prediction accuracy of marketization transaction electricity quantity.

[0036] 2. When the application cleanses data, it simultaneously takes new calendar same period electricity quantity, lunar calendar same period electricity quantity and recent electricity quantity as references, thereby reducing contingency and avoiding misjudgment of normal data as abnormal data. Meanwhile, the application judges abnormal data by using all user electricity quantity and cleanses marketization user electricity quantity at the same time, thereby ensuring the data quality of marketization user electricity quantity proportion.

[0037] 3、The application can realize daily rolling prediction of marketization user electricity proportion, and considers different situations of user collective marketization and separate marketization by setting different weights, so as to accurately track the change of user profile and realize electricity prediction under the condition of user market attribute change. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is a method implementation flowchart of the embodiment of the application;

[0039] Figure 2 is a historical data cleaning schematic diagram in the embodiment of the application;

[0040] Figure 3 is a schematic diagram for rolling prediction of marketization user electricity proportion by using a moving average method in the embodiment of the application;

[0041] Figure 4 is a system structure schematic diagram of the embodiment of the application;

[0042] Figure 5 is a visualization chart generated by the result display module in the embodiment of the application. DETAILED DESCRIPTION

[0043] The application will be further described below in combination with the drawings and embodiments.

[0044] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0045] It should be noted that the terms used herein are only for the purpose of describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and it should also be understood that when the terms "comprise" and / or "include" are used in the specification, there is a presence of the features, steps, operations, devices, components and / or combinations thereof.

[0046] As shown in Figure 1 The present embodiment provides a marketization transaction electricity prediction method based on a moving average, which comprises:

[0047] 1) According to the historical daily marketing profile retained, the electricity consumption of all users and marketization users is queried, and the marketization electricity proportion is calculated.

[0048] Among them, the marketization user is the industrial and commercial user recorded as marketization in the historical daily marketing profile, rather than the marketization user screened according to the latest marketing profile.

[0049] 2) Establish a time series model of the proportion of market-oriented electricity and clean up historical data.

[0050] As shown in the following table, the specific method for cleaning up historical data is: Figure 2

[0051] Compare the daily total user electricity consumption (assuming Q t ) with the electricity consumption of the previous day (assuming Q t-1 ), the electricity consumption of the previous 7 days (assuming Q t-7 ), the electricity consumption of the same period in the new calendar (assuming Q t,new ), and the electricity consumption of the same period in the lunar calendar (assuming Q t,lunar ), and set the following four rules:

[0052] Rule 1: |Q t -Q t-1 |>a×Q t-1

[0053] Rule 2: |Q t -Q t-7 |>b×Q t-7

[0054] Rule 3: |Q t -Q t,new |>c×Q t,new

[0055] Rule 4: |Q t -Q t,lunar |>d×Q t,lunar

[0056] Where a, b, c, and d are deviation thresholds. If all four rules are met, the daily electricity consumption data is considered abnormal. If only one rule is not met, the daily electricity consumption data is considered normal.

[0057] If the total user electricity consumption on day t is determined to be abnormal data, then the previous 7 days of data are used to correct the total user electricity consumption on day t and the market-oriented user electricity consumption on day t, i.e., Q t =0.5×Q t-7 +0.5×(Q t-1 +Q t-2 +Q t-3 +Q t-4 +Q t-5 +Q t-6 +Q t-7 ). Where Q t to Q t-7 represent the total user electricity consumption from day t to day t-7. At the same time, the market-oriented user electricity consumption on day t (assuming W t ) is corrected, i.e., W​t = 0.5 x W t-7 + 0.5 x (W t-1 + W t-2 + W t-3 + W t-4 + W t-5 + W t-6 + W t-7 ) Calculate the marketization user electricity consumption on day t. Wherein, W t to W t-7 respectively represent the marketization user electricity consumption from t to t-7 days.

[0058] 3) Based on the time series model, the average value of marketization electricity consumption ratio is calculated according to the set window width and weight, which is used as the predicted value of marketization electricity consumption ratio of the next day.

[0059] Wherein, the determination method of the window width and weight is:

[0060] Let the window width be l, then the window width needs to be a multiple of 7 days, that is, mod(l,7) = 0, and the common one is 14 days.

[0061] Let the weight coefficient of day t be a t , and the centralized market entry time specified by the power trading center is k day of each month, then the weight coefficient of k day is 10, that is, a t = 10; The weight coefficient of other dates is 1.

[0062] 4) Roll forward to predict the marketization electricity consumption ratio of each day in the future for a period of time.

[0063] As shown in Figure 3 , the specific method of rolling forward to predict the marketization electricity consumption ratio of each day in the future for a period of time is:

[0064] Let the marketization user electricity consumption ratio of day t be p t , then according to formula , the marketization user electricity consumption ratio of day t+1 is calculated, and then the calculated value of the marketization user electricity consumption ratio of day t+1 is used as the input to predict the marketization user electricity consumption ratio of day t+2, that is , and so on to predict the marketization user electricity consumption ratio of a certain date in the future.

[0065] 5) Use the time series prediction model to predict the total daily electricity consumption in the future for a period of time, and calculate the marketization transaction electricity consumption combined with the marketization electricity consumption ratio.

[0066] Wherein, the time series prediction model first aspect contains three kinds of variables, namely independent variables, dependent variables and dependent variables, wherein the independent variables are time d1, d2, …, d t , and the time interval is 1 day, that is, d t -dt-1 = 1 day; dependent variable is electricity quantity Q1, Q2, …, Q t ; covariate, also known as characteristic variable, the characteristic variables considered in the present application include daily maximum temperature, daily minimum temperature, daily rainfall, daily humidity, daily weather state, sunrise time, sunset time, time-of-use electricity price and GDP, denoted as z1, z2, …, z n ; the second aspect includes the functional relationship of independent variable and covariate to dependent variable, i.e. Q t = f(t) = g(d1, d2, …, d t , Q1, Q2, …, Q t-1 , z1, z2, …, z n ), which can use existing mature frameworks, common ones are AR model (autoregression model), MA model (moving average model), ARMA model (autoregression moving average model) and Prophet model, in the present embodiment, the Prophet model disclosed by Facebook is used to establish, in which there are:

[0067] Q t = f(t) = r(t) + s(t) + h(t) + ξ t

[0068] wherein r(t) is a trend item, representing the change trend of electricity quantity above the non-period, usually represented by a piecewise function r(t) = (k + a(t) T δ)·t + (m + a(t) T γ), k, a(t), δ, T, m and γ are all parameters, which need to be trained from historical data, wherein the data participating in training are all cleaned data rather than original data.

[0069] s(t) is a periodic item, the time series model established in the present application considers annual, monthly and weekly periodicity, i.e.

[0070]

[0071] wherein a year and b year are annual periodic item coefficients; a month and b month are monthly periodic item coefficients; a week and b week are weekly periodic item coefficients, all of which are obtained by least square fitting.

[0072] h(t) is a holiday item, representing the influence of non-fixed period holidays on electricity quantity prediction, i.e.

[0073]

[0074] wherein, L is the number of holidays; D i is the time before and after the holiday; represents the time range of the influence of the holiday on the power; κ i is a parameter subject to normal distribution, obtained by model fitting.

[0075] Finally, ξ t is a white noise error term, satisfying normal distribution.

[0076] As Figure 4 shown, the embodiment also provides a marketization transaction power prediction system based on moving average, for realizing the above method, comprising a data query module 401, a data cleaning module 402, a rolling prediction module 403 and a result display module 404.

[0077] The data query module 401 is used to export daily power data according to settlement units, power categories and marketization attributes from the power marketing database.

[0078] The data cleaning module 402 is used to judge, clean and correct abnormal data.

[0079] The rolling prediction module 403 is used to predict marketization power in the future.

[0080] The result display module 404 is used to aggregate data and output visual reports according to preset rules.

[0081] Among them, the data query module 401 can decompose provincial power grid power according to power supply units and power categories, wherein the power categories include large industrial power, non-general industrial power, commercial power and non-residential lighting power, the power supply units are divided according to settlement units, and daily power data is output.

[0082] The data cleaning module 402 can detect abnormal data in daily power of all users and daily power of marketization users according to the cleaning rules proposed by the present application, and complete data correction according to the correction method proposed by the present application.

[0083] The rolling prediction module 403 can predict the proportion of marketization user power in the future according to the rolling prediction method proposed by the present application, and predict the daily power of all users in the future according to the time series prediction model proposed by the present application, so as to calculate the daily power of marketization users in the future, and further aggregate monthly power.

[0084] Finally, the result display module 404 can generate dynamic visual reports by using the Echarts technology of Baidu, as Figure 5As shown, the report contains the historical and predicted values of the proportion of daily electricity consumption of industrial and commercial customers in the province, and the historical and predicted values of daily electricity consumption of industrial and commercial customers in the province.

[0085] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the above-mentioned system is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be realized in the form of hardware or software. In addition, the specific names of each functional unit and module are only for easy distinction, and do not limit the protection scope of the present application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0086] It should be noted that the method and its detailed examples provided in the above embodiments can be combined into the system provided in the embodiments, and mutual reference will not be repeated.

[0087] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0088] In the embodiments provided in the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the above-mentioned modules or units is only a logical function division, and actual implementation can be performed by another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0089] The above is only a preferred embodiment of the present application, and is not intended to limit the other forms of the present application. Any skilled person in the art can use the disclosed technical content to make changes or modifications to equivalent embodiments. However, any simple modification, equivalent change and modification made according to the technical essence of the present application to the above embodiments, without departing from the technical solution content of the present application, still belongs to the protection scope of the present application.

Claims

1. A method for predicting market-based transaction electricity based on sliding average, characterized in that: include: Query the daily electricity consumption of all users and market users based on the historical marketing archives, and calculate the proportion of market electricity consumption; Establish a time series model for the proportion of market-based electricity consumption and clean up historical data; Based on the time series model, the average value of the market-based electricity proportion is calculated according to the set window width and weight, and used as the predicted value of the market-based electricity proportion for the next day; Rolling forecast of the proportion of daily market electricity consumption in the future with a step of one day; Use the time series forecasting model to predict the total daily electricity consumption in the future, and calculate the market-traded electricity volume based on the market-based electricity ratio; The method for determining the window width and weight is: Let the window width be l, then the window width is a multiple of 7 days, that is, mod(l,7)=0; Assume the weight coefficient of day t is a t The centralized market entry time stipulated by the power trading center is day k of each month, and the weight coefficient of day k is 10, that is, a t =10; the weight coefficient of other dates is 1; The specific method for rolling prediction of the daily market electricity proportion in the future period with a step length of one day is as follows: Assume that the proportion of electricity consumption by market users on day t is p t , then according to the formula Calculate the proportion of electricity consumption of market users on day t+1, and then use the calculated value of the proportion of electricity consumption of market users on day t+1 as input to predict the proportion of electricity consumption of market users on day t+2, that is, This can be used to move forward and predict the proportion of electricity consumption by market users at a certain date in the future.

2. The method for predicting market-based transaction electricity based on sliding average according to claim 1, characterized in that: The market-oriented users are the industrial and commercial users recorded as market-oriented in the historical marketing archives of the day, rather than market-oriented users screened based on the latest marketing archives.

3. The method for predicting market-based transaction electricity based on sliding average according to claim 1, characterized in that: The specific method for cleaning historical data is as follows: The daily electricity consumption of all users Q t Compared with the previous day's electricity Q t-1 、Electricity in the previous 7 days Q t-7 、Electricity Q in the same period of the new calendar t,new The electricity consumption during the same period of the lunar calendar is Q t,lunar For comparison, set the following four rules: Rule 1: |Q t -Q t-1 |>a×Q t-1 Rule 2: |Q t -Q t-7 |>b×Q t-7 Rule 3: |Q t -Q t,new |>c×Q t,new Rule 4: |Q t -Q t,lunar |>d×Q t,lunar Among them, a, b, c, and d are deviation thresholds. If all four rules are met, the daily power data is considered abnormal. As long as one rule is not met, the daily power data is considered normal. If the electricity consumption of all users on day t is judged to be abnormal data, the electricity consumption of all users on day t and the electricity consumption of market users on day t will be corrected using the data from the previous 7 days, that is, according to Q t =0.5×Q t-7 +0.5×(Q t-1 +Q t-2 +Q t-3 +Q t-4 +Q t-5 +Q t-6 +Q t-7 ) calculate the electricity consumption of all users on day t; where Q t To Q t-7 They represent the electricity consumption of all users from t to t-7 respectively; at the same time, they correct the electricity consumption of market users on day t, W t , that is, according to W t =0.5×W t-7 +0.5×(W t-1 +W t-2 +W t-3 +W t-4 +W t-5 +W t-6 +W t-7 ) calculate the market-based electricity consumption of users on day t; where W t To W t-7 They represent the electricity consumption of market users from t to t-7 respectively.

4. The method for predicting market-based transaction electricity based on sliding average according to claim 1, characterized in that: The time series prediction model is as follows: The first aspect of the time series prediction model includes three variables: independent variables, covariates and dependent variables, where the independent variables are time d1, d2, ..., d t , the time interval is 1 day, that is, d t -d t-1 = 1 day; dependent variables are electricity Q1, Q2, …, Q t ; Covariates are also called characteristic quantities. The characteristic quantities considered in this model include daily maximum temperature, daily minimum temperature, daily rainfall, daily humidity, daily weather conditions, sunrise time, sunset time, time-of-use electricity price and GDP, denoted as z1, z2,…, z n The second aspect includes the functional relationship between the independent variable and the covariate and the dependent variable, namely Q t =f(t)=g(d1,d2,…,…,d t ,Q1,Q2,…,Q t-1 ,z1,z2,…,z n ); the parameters in the function f(t) are obtained by training with historical data, where the historical data involved in the training are all cleaned data rather than original data; the time series model contains a periodic term and considers annual, monthly, and weekly periodicity at the same time, that is, f(t) contains a periodic term s(t): Among them, a year and b year is the annual cycle coefficient; a month and b month is the monthly cycle coefficient; a week and b week are the weekly cycle coefficients, all obtained by least squares fitting.

5. A market-based transaction electricity forecasting system based on sliding average, used to implement the method according to any one of claims 1 to 4, characterized in that: include: A data query module is used to derive daily electricity consumption data from the power marketing database according to settlement units, electricity consumption categories, and market attributes; Data cleaning module, used to judge, clean and correct abnormal data; Rolling forecast module, used to predict the market electricity consumption in the future; The result display module is used to summarize data according to preset rules and output visual reports.

6. The system for predicting market-based electricity volume based on sliding average according to claim 5, characterized in that: The data query module decomposes the provincial power grid electricity according to power supply units and electricity consumption categories, where the electricity consumption categories include large industrial electricity consumption, non-general industrial electricity consumption, commercial electricity consumption, and non-residential lighting electricity consumption. Power supply units are divided according to settlement units and output daily electricity consumption data.

7. The market-based transaction electricity forecasting system based on sliding average according to claim 5 is characterized in that: The rolling forecast module predicts the proportion of market-based user electricity consumption for each power supply unit and electricity consumption category in the future, and predicts the daily electricity consumption of all users of each power supply unit and electricity consumption category in the future, thereby calculating the daily electricity consumption of market-based users in the future, and then summarizing it into monthly electricity consumption.

Images