Power sales prediction method based on big data analysis

Abstract

The invention relates to the technical field of data processing, in particular to a power sales prediction method based on big data analysis, and the method comprises the steps: obtaining historical power sales data; obtaining an electricity sale quantity fitting curve based on historical electricity sale data, and obtaining an electricity sale quantity fluctuation degree of each month; obtaining an initial sequence of each month and a plurality of corresponding differential sequences; the stability degree of each differential sequence of each month is obtained, and then the stability degree of the electricity sale quantity of each month in each year is obtained; obtaining a stability degree sequence of each month in each year; obtaining the importance of each month to a prediction month, and further obtaining an update importance degree sequence of the prediction month; and obtaining an electricity sale quantity adjustment weight of each month, adjusting the ARIMA model, and obtaining an electricity sale quantity prediction value of a prediction month in combination with the electricity sale quantity of each month. According to the method, historical power sales data are analyzed, ARIMA model parameters are adjusted, and the predicted power sales quantity with higher reference value is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for predicting electricity sales based on big data analysis. Background Art

[0002] The power industry is a capital-intensive enterprise. By predicting the sales of electricity, we can better plan resources such as power generation, transmission and distribution, improve resource utilization, and formulate more targeted marketing and pricing strategies. Therefore, it is necessary to predict the monthly electricity sales. The ARIMA model is usually used to predict data. However, the ARIMA model using the original data weight has defects when processing non-stationary data, ignoring the cyclical changes and long-term growth trends of electricity sales data, which leads to the inability of the electricity sales forecast model to combine the fluctuation trend of historical data for forecasting, thus affecting the accuracy of the predicted value of electricity sales. Summary of the invention

[0003] The present invention provides a power sales forecasting method based on big data analysis to solve the existing problems.

[0004] The power sales forecasting method based on big data analysis of the present invention adopts the following technical solutions:

[0005] The present invention proposes a method for predicting electricity sales based on big data analysis, which includes the following steps:

[0006] Acquire historical electricity sales data; the historical electricity sales data includes electricity sales in several months in several years;

[0007] Based on the historical electricity sales data, a sales volume fitting curve is obtained, and then several extreme value points on the sales volume fitting curve are obtained, and several sales volume segment curves are obtained; according to the sales volume segment curve and the sales volume of the same month in different years, the sales volume fluctuation degree of each month is obtained;

[0008] According to the fluctuation degree of electricity sales in adjacent months, the initial sequence of each month and the corresponding several difference sequences are obtained; according to the numerical difference and change of data points in the difference sequence, the stability of each difference sequence of each month is obtained, and then the stability of electricity sales in each month of each year is obtained;

[0009] Obtain the adjacent reference months of each month, and combine the stability of electricity sales in each month of each year to obtain the stability sequence of each month of each year; according to the stability sequence of each month and the stability sequence of the predicted month, obtain the importance of each month to the predicted month, and then obtain the updated importance sequence of the predicted month;

[0010] According to the update importance sequence, the adjustment weight of electricity sales for each month is obtained, and the ARIMA model is adjusted. Combined with the electricity sales for each month, the predicted value of electricity sales for the predicted month is obtained.

[0011] Furthermore, the method of obtaining a power sales amount fitting curve based on historical power sales data, and then obtaining a number of extreme value points on the power sales amount fitting curve, and dividing the curve to obtain a number of power sales amount segmented curves, includes the following specific steps:

[0012] A coordinate system is constructed with the month as the horizontal axis and the monthly electricity sales as the vertical axis. The electricity sales of each month in the historical electricity sales data are converted into coordinate points in the coordinate system. The coordinate points of adjacent months are connected to obtain the electricity sales fitting curve of the power company.

[0013] All extreme value points of the electricity sales amount fitting curve are obtained, and the electricity sales amount fitting curve is segmented according to the extreme value points to obtain a plurality of electricity sales amount segmented curves.

[0014] Furthermore, the fluctuation degree of the electricity sales volume for each month is obtained according to the electricity sales volume segmentation curve and the electricity sales volume of the same month in different years. The corresponding specific calculation formula is:

[0015]

[0016] Among them, B i represents the fluctuation degree of electricity sales in the i-th month; N represents the number of years of electricity sales; Z n,i represents the electricity sales in the i-th month of the n-th year; Z n,i,min It represents the minimum value on the electricity sales segment curve of the electricity sales in the i-th month of the n-th year; Z n,i,max It represents the minimum value on the electricity sales segment curve of the electricity sales in the i-th month of the n-th year; Z n+1,i represents the electricity sales in the i-th month of the n+1th year, and || represents the absolute value function.

[0017] Furthermore, the initial sequence of each month and the corresponding differential sequences are obtained according to the fluctuation degree of electricity sales in adjacent months, and the specific steps include the following:

[0018] On the electricity sales fitting curve, take one month in any year as the center, record several months within the first radius of the month as the reference month of the month, and the degree of fluctuation of electricity sales between the month and all the reference months of the month constitute the initial sequence of the month in the order of months;

[0019] Perform iterative differentiation on the initial sequence of the month to obtain all the difference sequences of the month.

[0020] Furthermore, the stability of each difference sequence of each month is obtained according to the numerical difference and change of the data points in the difference sequence. The corresponding specific calculation formula is:

[0021]

[0022] Among them, P j,x Indicates the stability of the xth difference sequence in the jth month; R j,x represents the number of data points in the xth difference sequence in the jth month; B ′ j,x,r Represents the value of the rth data point in the xth difference sequence in the jth month; It represents the mean of all data points in the xth difference sequence in the jth month; exp() is an exponential function with a natural constant as the base.

[0023] Furthermore, the step of obtaining the stability of electricity sales volume in each month of each year includes the following specific steps:

[0024] The stationarity of all difference sequences in any month is linearly normalized, and the stationarity of the difference sequence with the smallest difference order among several difference sequences whose normalized stationarity is greater than the stationarity threshold in that month is taken as the stationarity of electricity sales in that month.

[0025] Furthermore, the step of obtaining the adjacent reference months of each month and combining the stability of electricity sales in each month of each year to obtain a stability sequence of each month of each year includes the following specific steps:

[0026] Taking any month except the first number of months in the electricity sales fitting curve as the starting point, the electricity sales stability of all adjacent reference months of this month are combined in the order of months to form a stability sequence for this month.

[0027] Furthermore, the adjacent reference months of each month are obtained by the following specific steps:

[0028] On the electricity sales fitting curve, take any month other than the first number of months in the electricity sales fitting curve as the starting point, and record several months within the first number before this month as adjacent reference months of this month.

[0029] Furthermore, the importance of each month to the predicted month is obtained according to the stability sequence of each month and the stability sequence of the predicted month, and then the updated importance sequence of the predicted month is obtained, including the following specific steps:

[0030]

[0031] Among them, G jrepresents the importance of the jth month to the predicted month; R represents the second quantity; P j,r P represents the stability of the rth month in the adjacent reference month of the jth month; r Indicates the stability of the rth month in the adjacent reference month of the forecast month; P j represents the stability sequence of the jth month; P represents the stability sequence of the predicted month; Cov() represents the correlation function; exp() is an exponential function with a natural constant as the base;

[0032] The importance of all months except the first number of months to the predicted month is calculated in order of the months to form an updated importance sequence for the predicted month.

[0033] Furthermore, the specific calculation formula for obtaining the monthly electricity sales adjustment weight is as follows:

[0034] W k =sigmoid(w k ×(1+G k ))

[0035] Among them, W k represents the adjustment weight of electricity sales in the kth month; w k represents the initial weight of electricity sales in the kth month; G k Indicates the importance of the kth month to the predicted month; sigmoid() is the sigmoid function.

[0036] The beneficial effects of the technical solution of the present invention are:

[0037] By statistically analyzing the historical electricity sales data, the accuracy and reference value of the electricity sales forecast can be effectively improved. Specifically, based on the historical electricity sales data, the electricity sales fitting curve is obtained, and the extreme points of the electricity sales change are identified, which can effectively divide the electricity sales segment curve, so as to more finely analyze the fluctuation degree of electricity sales, and more accurately reflect the cyclical changes of electricity sales; then, by calculating the fluctuation degree of electricity sales each month, the initial sequence and the corresponding difference sequence of each month are obtained, and the stability of electricity sales in each month of each year is obtained by analyzing the stability of the difference sequence, which provides a more stable data basis for the electricity sales forecast, thereby improving the reliability of the forecast; using the stability sequence, combined with the adjacent reference months of each month, it is possible to accurately evaluate the importance of each month to the forecast month, and construct an updated importance sequence for the forecast month, so as to ensure that the contribution of each month's electricity sales data to the forecast is dynamically adjusted according to the actual fluctuation situation, thereby improving the adaptability and flexibility of the model. Finally, by adjusting the parameters in the ARIMA model, it can better adapt to the cyclical changes and long-term growth trends of electricity sales data. The updated importance sequence is used to obtain the adjustment weights of electricity sales for each month. Finally, a more accurate prediction value is obtained by combining the electricity sales data. This makes the ARIMA model more flexible in dealing with non-stationary data and solves the defects of the traditional ARIMA model in dealing with cyclical changes and long-term trends. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 This is a flowchart of the steps of the electricity sales forecasting method based on big data analysis of the present invention. DETAILED DESCRIPTION

[0040] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, features and effects of the power sales forecasting method based on big data analysis proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0041] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0042] The specific scheme of the power sales forecasting method based on big data analysis provided by the present invention is described in detail below with reference to the accompanying drawings.

[0043] See also Figure 1 , which shows a flowchart of a method for predicting power sales based on big data analysis provided by an embodiment of the present invention, the method comprising the following steps:

[0044] Step S001: Obtain historical electricity sales data.

[0045] The main purpose of this embodiment is to predict the power sales volume of the power company, and the change trend of the data is analyzed through the historical power sales data of the power company.

[0046] Specifically, this embodiment collects the electricity sales data of the power company every month in the past ten years, and performs two preprocessing operations of data cleaning and data integration, which are unified as historical electricity sales data. All the collected electricity sales data after preprocessing are recorded as historical electricity sales data.

[0047] Step S002: Obtain a power sales fitting curve based on historical power sales data, and then obtain several extreme points on the power sales fitting curve, and divide it to obtain several power sales segment curves; according to the power sales segment curve and the power sales of the same month in different years, obtain the degree of fluctuation of power sales for each month.

[0048] It should be noted that the ARIMA model does not work well when using the original weights of historical electricity sales when dealing with non-stationary data. The main reason for the non-stationary data is the seasonal changes in electricity consumption and the increase in overall electricity consumption caused by long-term economic growth. Therefore, it is necessary to first analyze the seasonal changes and overall trend characteristics of historical electricity sales to understand the degree of fluctuation in electricity sales in different months. Then, according to the degree of fluctuation in different months, determine the importance of historical electricity sales to the current predicted monthly electricity sales. Finally, the model performance is optimized by adjusting the weights of historical electricity sales in the ARIMA model.

[0049] It should be further explained that electricity sales are related to the current season and weather conditions. Electricity consumption in summer and winter will be higher than in spring and autumn, and each year's electricity sales will produce corresponding seasonal or cyclical changes. At the same time, with the growth of the economy, users' electricity consumption will also increase accordingly. The weight of historical electricity sales in the original model ignores these fluctuations. If you want to adjust the weight of historical electricity sales in the model, you need to first consider the degree of fluctuation in electricity sales.

[0050] Specifically, a coordinate system is constructed with the month as the horizontal axis and the monthly electricity sales as the vertical axis. The monthly electricity sales in the historical electricity sales data are converted into coordinate points in the coordinate system, and the coordinate points of adjacent months are connected to obtain the power company's electricity sales fitting curve; all extreme points of the electricity sales fitting curve are obtained, and the electricity sales fitting curve is segmented according to the extreme points to obtain several electricity sales segmented curves. The specific calculation formula for the fluctuation degree of electricity sales in each month is:

[0051]

[0052] Among them, B i represents the fluctuation degree of electricity sales in the i-th month; N represents the number of years of electricity sales; Z n,i represents the electricity sales in the i-th month of the n-th year; Z n,i,min It represents the minimum value on the electricity sales segment curve of the electricity sales in the i-th month of the n-th year; Z n,i,max It represents the minimum value on the electricity sales segment curve of the electricity sales in the i-th month of the n-th year; Z n+1,i represents the electricity sales in the i-th month of the n+1th year, and || represents the absolute value function.

[0053] It should be noted that |Z n,i -Z n,i,min | represents the difference between the electricity sales volume in the ith month of the nth year and the minimum value on the electricity sales volume segmentation curve of that month; |Z n,i,max -Z n,i,min | represents the difference between the maximum and minimum values ​​on the electricity sales segment curve in the i-th month of the n-th year; Indicates the position of the electricity sales volume in the i-th month of the n-th year in the curve segment, and measures the degree of fluctuation of the electricity sales volume in the seasonal changes of that month. The larger its value is, the greater the degree of fluctuation of the electricity sales volume is; |Z n+1,i -Z n,i | represents the difference between the electricity sales in the i-th month of the n+1th year and the electricity sales in the i-th month of the nth year; It represents the average growth of electricity sales in the i-th month over the years, reflecting the change of electricity sales in that month in the overall long-term trend. The larger the value is, the greater the change of electricity sales in the long-term overall trend and the greater the degree of fluctuation.

[0054] Step S003: According to the fluctuation degree of electricity sales in adjacent months, the initial sequence of each month and the corresponding differential sequences are obtained; according to the numerical difference and change of the data points in the differential sequence, the stability of each differential sequence of each month is obtained, and then the stability of electricity sales in each month of each year is obtained.

[0055] It should be noted that the ARIMA model needs to perform differential processing to stabilize the data and make predictions based on the stable data series. By performing differential processing on the initial series, as the order increases, a differential series of the number of months in the initial series minus one is obtained. The stationarity of multiple differential series is determined, and the stability of the electricity sales volume is obtained based on the degree of fluctuation of the electricity sales volume series after differential processing.

[0056] Specifically, the first radius is preset to 3, and this is used as an example for description. Other implementations may be set to other values. On the power sales fitting curve, take one month in any year as the center, and record several months within the first radius of the month as the reference month of the month. The degree of fluctuation of power sales of the month and all reference months of the month constitute the initial sequence of the month in the order of the month; perform iterative differentiation on the initial sequence of the month to obtain all differential sequences of the month. It is a well-known technology to perform iterative differentiation on the sequence to obtain several differential sequences, in which the order of each differential is different, and the specific process will not be repeated in this embodiment; according to the values ​​of all data points and the number of data points in each differential sequence of each month, the degree of stability in each differential sequence of each month is obtained. The specific calculation formula is:

[0057]

[0058] Among them, P j,x Indicates the stability of the xth difference sequence in the jth month; R j,x represents the number of data points in the xth difference sequence in the jth month; B ′ j,x,r Represents the value of the rth data point in the xth difference sequence in the jth month; It represents the mean of all data points in the x-th difference sequence in the j-th month; exp() is an exponential function with a natural constant as the base. This embodiment adopts the exp(-x) model to present the inverse proportional relationship. x is the input of the model. The implementer can set the inverse proportional function according to the actual situation.

[0059] It should be noted that It indicates the consistency of the fluctuation degree of the xth difference sequence in the jth month. The smaller the consistency, the more consistent the fluctuation degree of the elements in the difference sequence. At the same time, the smaller the mean of all data points in the xth difference sequence in the jth month, the smaller the fluctuation degree of the elements in the difference sequence. The fluctuation degree refers to the change of electricity sales in the difference sequence. The more consistent the change and the smaller the mean, the higher the stability of the elements in the difference sequence.

[0060] Furthermore, the preset stability threshold is 0.9, which is used as an example for description, and other values ​​may be set in other implementations. The stability of all difference sequences in any month is linearly normalized, and the stability of the difference sequence with the smallest difference order among several difference sequences in the month whose normalized stability is greater than the stability threshold is taken as the stability of electricity sales in that month.

[0061] Step S004: Obtain the adjacent reference months of each month, and combine the stability of electricity sales in each month of each year to obtain the stability sequence of each month of each year; based on the stability sequence of each month and the stability sequence of the predicted month, obtain the importance of each month to the predicted month, and then obtain the updated importance sequence of the predicted month.

[0062] It should be noted that the importance of each month's electricity sales is obtained according to the stability of each month, so as to adjust the weight. The smaller the difference between the stability of electricity sales in the previous month of a certain historical month and the stability of electricity sales in the previous month of the forecast month, the higher the reference value of electricity sales in this month for the electricity sales in the forecast month.

[0063] Specifically, the first number is preset to 5, which is used as an example for description. Other implementations may be set to other values. On the power sales fitting curve, take any month other than the months before the first number in the power sales fitting curve as the starting point, and record several months within the first number before this month as the adjacent reference months of this month. The power sales stability of all adjacent reference months of this month are combined in the order of months to form the stability sequence of this month; it should be noted that in the subsequent calculation of historical power sales data, the months before the first number are not involved in the subsequent calculation; the next month of the current month is taken as the predicted month, and similarly, according to the power sales stability of the most recent first number of months, the stability sequence of the predicted month is obtained; then the specific calculation formula for the importance of each month to the predicted month is:

[0064]

[0065] Among them, G j represents the importance of the jth month to the predicted month; R represents the second quantity; P j,r P represents the stability of the rth month in the adjacent reference month of the jth month; r Indicates the stability of the rth month in the adjacent reference month of the forecast month; P j represents the stability sequence of the jth month; P represents the stability sequence of the predicted month; Cov() represents the correlation function; exp() is an exponential function with a natural constant as the base. This embodiment adopts the exp(-x) model to present the inverse proportional relationship and normalization processing. x is the input of the model. The implementer can set the inverse proportional function and normalization function according to the actual situation.

[0066] It should be noted that It represents the square difference between the stationarity of the adjacent reference month of the jth month and the adjacent reference month of the forecast month, which shows the difference between the two series. The smaller the difference, the smaller the difference between the stationarity, the more reference significance it has for the forecast month's electricity sales, and the higher its importance for the forecast month's electricity sales. Indicates that the difference is normalized to maintain the same magnitude; Cov(P j ,P) represents the correlation between the stability sequence of monthly electricity sales in the neighborhood of the j-th month and the stability sequence of monthly electricity sales in the neighborhood of the predicted month, reflecting the influence of the stability sequence of historical electricity sales on the stability sequence of predicted month electricity sales. The larger the value, the greater its importance to the predicted month electricity sales.

[0067] Furthermore, the importance of all the months except the first number of months to the predicted month is combined in the order of the months to form an updated importance sequence for the predicted month.

[0068] Step S005: According to the update importance sequence, the electricity sales adjustment weight of each month is obtained, and the ARIMA model is adjusted. Combined with the electricity sales of each month, the electricity sales forecast value of the forecast month is obtained.

[0069] It should be noted that, according to the update importance sequence, the initial weight of the electricity sales in the corresponding historical months in the ARIMA model is adjusted to obtain the adjusted weight of electricity sales for each month, wherein the initial weight of electricity sales is an existing parameter in the ARIMA model, which will not be described in detail in this embodiment; the corresponding specific formula is:

[0070] W k =sigmoid(w k ×(1+G k ))

[0071] Among them, W k represents the adjustment weight of electricity sales in the kth month; w k represents the initial weight of electricity sales in the kth month; G k Indicates the importance of the kth month to the predicted month; sigmoid() is the sigmoid function, and this embodiment uses the sigmoid function for normalization processing.

[0072] It should be noted that w k ×(1+G k ) indicates that according to the importance of the electricity sales in the kth month to the electricity sales in the predicted month, the weight of the electricity sales in the kth month in the ARIMA model is adjusted, and then the electricity sales adjustment weight is obtained.

[0073] Furthermore, the weights are adjusted according to the electricity sales volume of each month to obtain the adjusted ARIMA model, and the electricity sales volume of each month in each year in the historical sales data is input to predict the electricity sales volume of the forecast month, and the electricity sales volume forecast value of the forecast month is obtained, thus completing the forecast of the electricity sales volume of the power company in that month.

[0074] At this point, this embodiment is completed.

[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A power sales forecasting method based on big data analysis, characterized in that: The method comprises the following steps: Acquire historical electricity sales data; the historical electricity sales data includes electricity sales in several months in several years; Based on the historical electricity sales data, a sales volume fitting curve is obtained, and then several extreme value points on the sales volume fitting curve are obtained, and several sales volume segment curves are obtained; according to the sales volume segment curve and the sales volume of the same month in different years, the sales volume fluctuation degree of each month is obtained; According to the fluctuation degree of electricity sales in adjacent months, the initial sequence of each month and the corresponding several difference sequences are obtained; according to the numerical difference and change of data points in the difference sequence, the stability of each difference sequence of each month is obtained, and then the stability of electricity sales in each month of each year is obtained; Obtain the adjacent reference months of each month, and combine the stability of electricity sales in each month of each year to obtain the stability sequence of each month of each year; according to the stability sequence of each month and the stability sequence of the predicted month, obtain the importance of each month to the predicted month, and then obtain the updated importance sequence of the predicted month; According to the update importance sequence, the adjustment weight of electricity sales for each month is obtained, and the ARIMA model is adjusted. Combined with the electricity sales for each month, the predicted value of electricity sales for the predicted month is obtained.

2. The power sales forecasting method based on big data analysis according to claim 1 is characterized in that: The method of obtaining a power sales fitting curve based on historical power sales data, and then obtaining a number of extreme points on the power sales fitting curve, and dividing the curve to obtain a number of power sales segment curves, includes the following specific steps: A coordinate system is constructed with the month as the horizontal axis and the monthly electricity sales as the vertical axis. The electricity sales of each month in the historical electricity sales data are converted into coordinate points in the coordinate system. The coordinate points of adjacent months are connected to obtain the electricity sales fitting curve of the power company. All extreme value points of the electricity sales amount fitting curve are obtained, and the electricity sales amount fitting curve is segmented according to the extreme value points to obtain a plurality of electricity sales amount segmented curves.

3. The power sales forecasting method based on big data analysis according to claim 1 is characterized in that: According to the electricity sales segmentation curve and the electricity sales of the same month in different years, the fluctuation degree of electricity sales of each month is obtained, and the corresponding specific calculation formula is: Among them, B i represents the fluctuation degree of electricity sales in the i-th month; N represents the number of years of electricity sales; Z n,i represents the electricity sales in the i-th month of the n-th year; Z n,i,min It represents the minimum value on the electricity sales segment curve of the electricity sales in the i-th month of the n-th year; Z n,i,max It represents the minimum value on the electricity sales segment curve of the electricity sales in the i-th month of the n-th year; Z n+1,i represents the electricity sales in the i-th month of the n+1th year, and || represents the absolute value function.

4. The power sales forecasting method based on big data analysis according to claim 1 is characterized in that: The specific steps of obtaining the initial sequence of each month and the corresponding differential sequences according to the fluctuation degree of electricity sales in adjacent months are as follows: On the electricity sales fitting curve, take one month in any year as the center, record several months within the first radius of the month as the reference month of the month, and the degree of fluctuation of electricity sales between the month and all the reference months of the month constitute the initial sequence of the month in the order of months; Perform iterative differentiation on the initial sequence of the month to obtain all the difference sequences of the month.

5. The power sales forecasting method based on big data analysis according to claim 1 is characterized in that: According to the numerical differences and changes of the data points in the difference sequence, the stability of each difference sequence of each month is obtained, and the corresponding specific calculation formula is: Among them, P j,x Indicates the stability of the xth difference sequence in the jth month; R j,x represents the number of data points in the xth difference sequence in the jth month; B ′ j,x,r Represents the value of the rth data point in the xth difference sequence in the jth month; It represents the mean of all data points in the xth difference sequence in the jth month; exp() is an exponential function with a natural constant as the base.

6. The power sales forecasting method based on big data analysis according to claim 1 is characterized in that: The specific steps of obtaining the stability of electricity sales in each month of each year are as follows: The stationarity of all difference sequences in any month is linearly normalized, and the stationarity of the difference sequence with the smallest difference order among several difference sequences whose normalized stationarity is greater than the stationarity threshold in that month is taken as the stationarity of electricity sales in that month.

7. The power sales forecasting method based on big data analysis according to claim 1 is characterized in that: The step of obtaining the adjacent reference months of each month and combining the stability of electricity sales in each month of each year to obtain the stability sequence of each month of each year includes the following specific steps: Taking any month except the first number of months in the electricity sales fitting curve as the starting point, the electricity sales stability of all adjacent reference months of this month are combined in the order of months to form a stability sequence for this month.

8. The power sales forecasting method based on big data analysis according to claim 7 is characterized in that: The specific steps for obtaining the adjacent reference months of each month are as follows: On the electricity sales fitting curve, take any month other than the first number of months in the electricity sales fitting curve as the starting point, and record several months within the first number before this month as adjacent reference months of this month.

9. The power sales forecasting method based on big data analysis according to claim 1 is characterized in that: The importance of each month to the predicted month is obtained according to the stability sequence of each month and the stability sequence of the predicted month, and then the updated importance sequence of the predicted month is obtained, and the specific steps include the following: Among them, G j represents the importance of the jth month to the predicted month; R represents the second quantity; P j，r P represents the stability of the rth month in the adjacent reference month of the jth month; r Indicates the stability of the rth month in the adjacent reference month of the forecast month; P j represents the stability sequence of the jth month; P represents the stability sequence of the predicted month; Cov() represents the correlation function; exp() is an exponential function with a natural constant as the base; The importance of all months except the first number of months to the predicted month is calculated in order of the months to form an updated importance sequence for the predicted month.

10. The power sales forecasting method based on big data analysis according to claim 1, characterized in that: According to the update importance sequence, the adjustment weight of the electricity sales volume for each month is obtained, and the corresponding specific calculation formula is: Wk=sigmoid(w k ×(1+G k )) Among them, W k represents the adjustment weight of electricity sales in the kth month; w k represents the initial weight of electricity sales in the kth month; G k Indicates the importance of the kth month to the predicted month; sigmoid() is the sigmoid function.

Images