A power prediction method under influence of multi-source data of new energy access

Abstract

The application discloses a power prediction method under the influence of multi-source data under new energy access, which aims to solve the problem that the traditional prediction model is strongly dependent on historical data and is difficult to cope with external sudden factors. The core is to build a monthly and annual collaborative prediction framework that integrates multi-source data. The monthly prediction is based on the actual data of the previous period of the month, the historical data is corrected by temperature and rainfall, and the decadal coefficient is calculated for fine sample prediction. At the same time, multi-scenario analysis is introduced and the correction model for special events such as holidays and extreme weather is integrated, and finally the monthly issued power with confidence interval is output. The annual prediction adopts the proportion method based on historical power proportion and the three exponential smoothing method based on growth trend for cross-validation and rolling update. The application significantly improves the power prediction accuracy and robustness in the face of unexpected situations by quantifying the influence of external factors, and has strong practical application value.

Description

Technical Field

[0001] This invention specifically relates to a power generation prediction method for the influence of multi-source data under large-scale renewable energy access, belonging to the field of power generation prediction technology. Background Technology

[0002] To effectively improve the efficiency of power grid operation and the scientific nature of decision-making, the key to conducting research on high-precision power generation forecasting lies in improving the accuracy of the forecast results. Developing power generation forecasting models with high accuracy and strong adaptability has become a key research focus in the industry. Currently, several mature forecasting methods exist for power generation, such as ARIMA, grey prediction, and neural networks, and they play important roles in various application scenarios.

[0003] However, with the large-scale integration of new energy sources and the increasing complexity of electricity consumption structures, existing forecasting solutions face significant challenges. First, at the data utilization level, most models rely excessively on the inherent patterns of historical electricity data, failing to effectively integrate external multi-source information such as real-time weather, holiday shifts, and industrial restructuring. This leads to a sharp decline in forecast accuracy when facing scenarios such as the Spring Festival effect, extreme weather, and sudden changes in economic policies. Second, at the model adaptability level, traditional statistical models struggle to characterize the high volatility of new power systems, while complex intelligent algorithms often face difficulties in parameter tuning, high dependence on historical data, weak interpretability, and high costs for cross-regional deployment. Particularly noteworthy is the lack of a systematic, quantifiable, and scalable fine-grained correction mechanism for special periods in monthly forecasts, such as the Spring Festival effect in January-February and high-temperature loads in July-August, limiting the guiding value of forecast results in actual business operations. To address these issues, a collaborative forecasting method that can integrate multi-source data, possess fine-grained correction capabilities, and is easily scalable is urgently needed.

[0004] To address the aforementioned issues affecting the accuracy of electricity forecasting, this study focuses on constructing an accurate forecasting model that integrates multi-source data. By analyzing historical data patterns and establishing correction mechanisms for special factors such as the shift of the Spring Festival, temperature changes, and changes in economic policies, the aim is to provide an automated forecasting tool with practical application value.

[0005] A search revealed that Chinese invention application CN118673463A discloses a method, system, device, and storage medium for power supply prediction based on multi-source data, which can improve the accuracy and reliability of power supply prediction. The method includes: generating a historical time-series representation matrix of power supply in the area to be predicted based on historical power supply data records of each power supply unit in the area to be predicted; the historical time-series representation matrix of power supply includes the electricity consumption of all electricity consumption categories in each power supply unit; obtaining the power supply environment impact characteristics corresponding to the area where each power supply unit is located within the time period to be predicted; and mapping and extracting the environmental correlation coefficient set corresponding to each power supply unit from the pre-analyzed power supply environment correlation coefficient matrix for the power supply environment impact characteristics of each power supply unit; the environmental correlation coefficient set includes the correlation coefficients between the power supply environment impact characteristics and the electricity consumption of each electricity consumption category in the power supply unit.

[0006] 1. This application focuses on constructing a monthly and annual collaborative forecasting framework with ten-day forecasting as the core. The core is to use the ten-day coefficient of the same period in history and combine it with real-time weather forecasts to map and extrapolate the known samples in the first ten days to the middle and last ten days. It also systematically integrates quantitative correction models of multi-source data such as temperature, continuous rainfall, holidays, special weather (such as typhoons and plum rains) and economic policies.

[0007] The core technology of the invention application "Power Supply Prediction Method, System, Device, and Storage Medium Based on Multi-Source Data" lies in constructing a historical time-series representation matrix of power supply and quantifying the correlation between environmental characteristics and various types of electricity consumption by mapping a pre-analyzed power supply environment correlation coefficient matrix. Essentially, it is a feature weighting and influence quantification mechanism based on predefined correlation coefficient mapping. The two differ fundamentally in their core prediction logic and the way they integrate external factors.

[0008] 2. This application adopts a strategy that combines a variety of classic statistical and time series methods, such as the ten-day coefficient method, the triple exponential smoothing method, and the proportion method. It also introduces three scenario analyses, namely optimistic, stable, and pessimistic, to improve the robustness of the prediction. The final output is the predicted value of the point and its fluctuation range.

[0009] In contrast, the approach of "power supply prediction method, system, equipment and storage medium based on multi-source data" focuses more on data representation and feature correlation mapping. It structurally processes multi-source data by constructing time-series matrices and correlation coefficient sets, and the overall framework relies on the pre-analyzed environmental correlation coefficient matrix. The former is a contextualized prediction framework that integrates multiple methods, while the latter is a quantitative evaluation framework based on the mapping of feature matrices and correlation coefficients.

[0010] 3. This application emphasizes dynamic correction and rolling updates. For example, in monthly forecasts, forecasts are initiated in real time based on the data of the previous 9 days, and in annual forecasts, the annual forecasts are updated on a rolling basis based on the latest monthly data. Its output includes the specific forecast values ​​of monthly / annual electricity consumption, adjusted electricity consumption, and confidence intervals.

[0011] The "Power Supply Prediction Method, System, Equipment, and Storage Medium Based on Multi-Source Data" focuses on using historical data records to construct time-series representations at the power supply unit level and extracting environmental impact characteristics within the prediction period for correlation analysis. Its output is a power supply prediction result adjusted based on environmental correlation coefficients. The former utilizes data with the characteristics of time-series rolling and external impact response, while the latter focuses more on the correlation between power supply unit division and static environmental characteristics.

[0012] A search revealed that Chinese invention application CN118572702A discloses a centralized control method and system for a multi-source energy storage system. The method includes the following steps: S1: collecting historical data and cleaning the data; the historical data includes historical electricity price data, historical power generation data, historical load data, historical energy storage data, and historical discharge data; S2: performing power data prediction based on the historical data, including electricity price prediction, power generation prediction, and load prediction; S3: determining the energy storage capacity based on the historical energy storage data and the power data prediction results; S4: statistically analyzing the historical discharge data to generate a discharge sequence; S5: planning the energy storage strategy for the multi-source energy storage system based on the energy storage capacity and the discharge sequence. This invention can promote multi-source collaborative optimization of multi-source energy storage systems, enhance the stability of multi-source energy storage systems, and promote the consumption of renewable energy.

[0013] 1. This application belongs to the field of power generation forecasting technology. Its core objective is to improve the accuracy and robustness of power generation forecasting for power grids, thereby serving power grid operation and planning.

[0014] The patent "A Centralized Control Method and System for a Multi-Source Energy Storage System" belongs to the field of multi-source energy storage system control technology. Its core objective is to achieve centralized control and collaborative optimization of the energy storage system, and to plan energy storage strategies through prediction results in order to improve system stability and promote the consumption of renewable energy. The two patents address completely different technical problems and have entirely different application scenarios.

[0015] 2. The multi-source data integrated in this application mainly revolves around external factors that affect electricity demand, such as weather, holidays, and economic policies.

[0016] The patent "A Centralized Control Method and System for a Multi-Source Energy Storage System" processes multi-source data that directly serves the operation and market behavior of the energy storage system, including historical electricity prices, power generation data, load data, and energy storage / discharge data. Regarding models, this application focuses on demand-side power forecasting models; while the latter requires the comprehensive use of multiple models such as electricity price forecasting, power generation forecasting, and load forecasting, ultimately integrating them into the energy storage control strategy. The two differ significantly in their data processing sources and model service objectives. Summary of the Invention

[0017] The purpose of this invention is to propose a power generation forecasting method under the influence of multiple data sources under the access of new energy sources, which is based on ten-day weather forecasts and multi-source data fusion correction. For monthly forecasts, a ten-day weather forecasting model based on historical patterns and real-time weather forecasts is developed, integrating holiday correction and special weather handling mechanisms. For annual forecasts, a monthly proportion method and a triple exponential smoothing method are adopted, and optimistic, stable, and pessimistic multi-scenario analysis is introduced to cope with different emergencies and improve the stability and accuracy of forecasts.

[0018] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0019] A method for predicting electricity generation based on the impact of multi-source data under renewable energy access includes the following steps:

[0020] S1, Monthly Electricity Consumption Forecast:

[0021] S11. Collect actual electricity sales, weather forecast data and holiday information for the 1st to 9th of the current month, as well as corresponding data for the same period in history, and perform temperature correction and restoration and continuous rainfall correction and restoration on the historical electricity sales data.

[0022] S12. Classify dates based on weather type and holiday type, calculate the ten-day period coefficient under different categories, and establish electricity samples for the first, middle and last ten days of the month based on the data of the first 10 days of the month and the ten-day period coefficient;

[0023] S13. Map the electricity samples from the first, middle, and last ten days of the month to the electricity sales forecast for unknown dates of the month, accumulate the monthly forecast electricity sales, and make predictions based on three scenarios: optimistic, stable, and pessimistic.

[0024] S14. Adjust the predicted electricity sales volume in response to special events that are predicted to occur in the current month, and calculate the predicted adjusted electricity volume for the current month based on the historical adjusted electricity volume ratio. Finally, sum them up to obtain the predicted monthly electricity issuance volume and confidence interval.

[0025] S2, Annual Electricity Forecast:

[0026] S21. Based on the historical monthly electricity sales ratio and the distribution of holidays in the current year, calculate the monthly electricity sales ratio for the current year, and combine it with the annual total electricity sales forecast to use the ratio method for annual electricity sales forecasting and rolling updates.

[0027] S22. Use the triple exponential smoothing method and historical data to predict the annual electricity sales trend, and compare and verify the prediction results with the percentage method, and output the annual electricity sales prediction results.

[0028] S3. Based on the monthly electricity generation forecast and confidence interval obtained in steps S14 and S22, and the annual electricity sales forecast, the final output is the record forecast set.

[0029] As a preferred technical solution of the present invention: in step S11, the temperature correction and restoration specifically includes:

[0030] For July and August, to eliminate the difference between historical electricity sales and current month's electricity sales caused by temperature, the correction formula is as follows:

[0031] (1);

[0032] in, This indicates historical electricity sales after temperature correction; This represents the original historical electricity sales volume; It is the temperature difference coefficient; It is calculated as the difference between the historical daily high temperature of the current month and the average high temperature of the current month this year. When the temperature is higher than 37℃, it is calculated as 37℃.

[0033] As a preferred technical solution of the present invention: in step S11, the continuous rainfall correction and restoration specifically includes:

[0034] The historical electricity sales, after temperature correction, are further adjusted based on the number of consecutive days of rainfall, using the following formula:

[0035] (2);

[0036] in, d represents the historical electricity sales volume after correction for continuous rainfall; d represents the number of consecutive rainfall days.

[0037] As a preferred technical solution of the present invention: in step S12, the ten-day period coefficient includes the ten-day period coefficient for the first and middle ten days and the ten-day period coefficient for the first and last ten days, and is calculated separately for working days and non-working days.

[0038] As a preferred technical solution of the present invention: in step S13, the calculation formula for the monthly predicted electricity sales is:

[0039] (3);

[0040] in, This indicates the monthly forecast for electricity sales. This indicates the volume of electricity sold on a known date. To predict electricity sales volume in the first ten days of the month; To predict electricity sales volume in the middle of the month; To predict electricity sales volume for the second half of the month.

[0041] As a preferred technical solution of the present invention: in step S14, the special events include national statutory holidays, continuous high temperatures, plum rain season and typhoon weather, and corresponding correction models are used to correct the predicted electricity sales for different special events.

[0042] As a preferred technical solution of the present invention: in step S14, the calculation formula for the monthly electricity generation is:

[0043] (4);

[0044] in, Forecast monthly electricity generation; This is the final monthly forecast of electricity sales after adjustments for special events; The formula for calculating the monthly electricity forecast adjustment is as follows:

[0045] (5);

[0046] Adjusting the electricity consumption to match historical data for the same month. This represents the original historical electricity sales volume.

[0047] As a preferred technical solution of the present invention: In step S21, the calculation formula for the monthly electricity sales ratio in the ratio prediction is as follows:

[0048] (6);

[0049] Where p represents the percentage of monthly electricity sales; This indicates the electricity sales volume in a certain month last year; This refers to the electricity sales volume in a certain month of the year before last. This represents the total electricity sales volume for the entire previous year. This represents the total electricity sales volume for the entire year before last.

[0050] As a preferred technical solution of the present invention: In step S21, the annual total electricity sales forecast is based on the GDP growth rate forecast. The annual total electricity sales can be predicted in January of this year. Based on the predicted monthly electricity sales and combined with the monthly electricity sales ratio, the annual total electricity sales forecast data is updated on a rolling basis.

[0051] As a preferred technical solution of the present invention: In step S22, the prediction formula of the triple exponential smoothing method is:

[0052] (7);

[0053] in, For time t, the future... The predicted value for each cycle; The intercept term represents the horizontal component of the current time t; The linear trend component represents the linear growth per unit period; It is a secondary trend component, reflecting the acceleration or deceleration of the trend; This represents the number of leading cycles predicted.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] This invention, based on ten-day electricity consumption forecasting, systematically quantifies the impact of weather, holidays, economic policies, and various special events on electricity consumption, constructing a complete monthly and annual electricity consumption forecasting scheme. This scheme effectively improves forecast accuracy through refined adjustments to electricity consumption data and demonstrates strong robustness in the face of unexpected events. Furthermore, this invention has low dependence on historical data and regional information; when applied to electricity consumption forecasting in different regions, it can be quickly adapted and forecasted simply by adjusting the corresponding ten-day coefficients and various influence coefficients, possessing strong cross-regional scalability and practical value. Attached Figure Description

[0056] Figure 1 This is a flowchart illustrating a power prediction method influenced by multi-source data. Detailed Implementation

[0057] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0058] This invention relates to, for example Figure 1 This paper presents a method for predicting electricity sales based on the influence of multi-source data under renewable energy access. It quantifies the impact of weather, holidays, temperature, and special events on electricity sales and provides rolling updates for annual electricity sales forecasts, achieving accurate predictions of monthly and annual electricity generation. The specific scheme is as follows:

[0059] This includes monthly and annual electricity forecasting steps:

[0060] S1. The monthly electricity consumption forecasting steps include:

[0061] S11. Collect actual electricity sales, weather forecast data, and holiday information for the period from the 1st to the 9th of the current month, as well as corresponding historical data for the same period. Then, perform temperature correction and continuous rainfall correction on the historical electricity sales data. The specific process is as follows:

[0062] S111. Collect electricity sales data and weather forecast data for the period from the 1st to the 9th of the current month, including weather conditions, daily maximum and minimum temperatures, and the distribution of holidays for the month. Electricity sales data, weather information, and holiday details for the previous year should also be collected.

[0063] S112. Perform temperature correction and restoration on historical electricity sales data, as well as continuous rainfall correction and restoration. Temperature correction and restoration, for July and August, eliminates the temperature-related differences between historical electricity sales and current month's sales. The historical daily maximum temperature for the current month is defined as the difference between the average daily maximum temperature for the current month this year and the historical daily maximum temperature for the current month. (If the temperature is higher than 37 degrees, it will be calculated as 37 degrees), as shown in the following formula:

[0064] ;

[0065] in, This indicates historical electricity sales after temperature correction; This represents the original historical electricity sales volume; This is the temperature difference coefficient.

[0066] S113. Perform continuous rainfall correction and restoration on the historical electricity sales data processed in S112 to eliminate the impact of continuous rainfall on electricity sales. The number of consecutive rainfall days is defined as... The specific formula is as follows:

[0067] ;

[0068] in, Historical electricity sales figures are adjusted for continuous rainfall.

[0069] S12. Based on weather type and holiday type, classify dates into categories, calculate the ten-day period coefficients for different categories, and establish electricity consumption samples for the first, middle, and last ten days of the month based on the data of the first 10 days of the month and the ten-day period coefficients, as follows:

[0070] S121. Divide the dates of the month into 8 categories based on weather and holiday types: sunny, cloudy, overcast, rainy / snowy, weekdays, and non-working days.

[0071] S122. For each case in S121, calculate the coefficients for the first and middle ten-day periods and the coefficients for the upper and lower ten-day periods. The specific formulas are as follows:

[0072] (1) Working day coefficients for the first and middle ten days of the month:

[0073] ;

[0074] in This represents the working day coefficient for the first and middle ten days of the month; The effective electricity volume for the first ten days of the same period in history is the average of the electricity sales volume for the first 10 days of the same period in history. Here, "the same period in history" refers to the average of the data from the previous two years. The same applies to the effective electricity volume for the middle and late ten days of the same period in history. This represents the effective electricity consumption for the same period in history during the middle of the working day.

[0075] (2) Working day coefficients for the first and second ten-day periods:

[0076] ;

[0077] in This indicates the coefficient for dividing the working days in the first and second ten-day periods of the month; This represents the effective electricity consumption for the last working day of the same period in history.

[0078] (3) Ten-day coefficient for non-working days in the first and middle ten days of the month:

[0079] ;

[0080] in This represents the coefficient for non-working days in the first and middle ten days of the month; This represents the highest effective electricity consumption on non-working days during the first ten days of the month in history. This represents the highest effective electricity consumption during non-working days in the same period in history.

[0081] (4) Ten-day division coefficient for non-working days in the first and second ten-day periods:

[0082] ;

[0083] in This indicates the coefficient for non-working days in the first and second ten-day periods of the month; This represents the highest effective electricity consumption during the same period in history for non-working days in the latter half of the month.

[0084] S123. Take the data from the previous 10 days (the electricity sales data on the 10th day is usually a forecast) as the sample for the first ten days of the month. Calculate the samples for the middle and latter ten days using the ten-day division coefficient in S122. Unknown samples for the current month are calculated using historical data. There are 6 scenarios in the sample calculation, and the specific formulas are as follows. First, the calculation of the sample for the first ten days:

[0085] (1) The electricity sales data corresponding to the weather and holidays can be found in the first ten days of the month, and the average value is taken as the sample for the first ten days.

[0086] (2) Data for the first ten days of the month is missing, but data for the first ten days of the month can be found in the past. The sample for the first ten days of the month is calculated according to the growth pattern of electricity sales. The specific formula is as follows:

[0087] ;

[0088] in This refers to samples from the first ten days of the month. This represents the average electricity sales volume for the corresponding weather and holiday periods in the past ten days.

[0089] (3) Data for the current month is missing, historical data is missing in the first ten days of the month, found in the middle ten days, and missing in the last ten days of the month. The formula for calculating the sample for the first ten days of the current month is as follows:

[0090] ;

[0091] in This represents the coefficient for the first and middle ten-day periods of the month. This represents the average electricity sales volume during the corresponding weather and holiday periods in the middle of the historical month.

[0092] (4) Data for the current month is missing, and historical data for the first and middle ten days of the month is missing, but data for the last ten days of the month is found. The formula for calculating the sample for the first ten days of the current month is as follows:

[0093] ;

[0094] in This represents the coefficient for dividing the first and second ten-day periods of the month. This represents the average electricity sales volume during the corresponding weather and holiday periods in the latter part of the historical period.

[0095] (5) Data for the current month is missing, and historical data for the first ten days of the month is missing, but data for the middle and last ten days of the month are found. The sample for the first ten days of the current month is calculated as follows:

[0096] ;

[0097] (6) Data not found.

[0098] The second step is to calculate the sample size in mid-month:

[0099] (1) The calculation method for the sample in the first ten days of the month (1) is as follows:

[0100] ;

[0101] in This indicates a sample from mid-month.

[0102] (2) Data for the first ten days of the month is missing, but historical data for the middle ten days can be found. The sample for the middle ten days of the month is calculated based on the growth pattern of electricity sales. The specific formula is as follows:

[0103] ;

[0104] (3) Data for the current month is missing, and historical data for the middle and late parts of the month is missing, but data for the first part of the month is found. The formula for calculating the sample for the middle part of the month is as follows:

[0105] ;

[0106] (4) Data for the current month is missing. Historical data for the first and middle ten days of the month is missing, but data for the last ten days of the month is found. The formula for calculating the sample for the middle ten days of the month is as follows:

[0107] ;

[0108] (5) Data for the current month is missing, historical data for the middle of the month is missing, but data for the first and last ten days of the month are available. The formula for calculating the sample for the middle of the month is as follows:

[0109] ;

[0110] (6) Data not found.

[0111] Finally, the sample calculation for the latter part of the month:

[0112] (1) The calculation method for the sample in the first ten days of the month (1) is as follows:

[0113] ;

[0114] in This indicates samples taken in the latter part of the month.

[0115] (2) Data for the first ten days of the month is missing, but historical data for the last ten days of the month can be found. The sample for the last ten days of the month is calculated based on the growth pattern of electricity sales. The specific formula is as follows:

[0116] ;

[0117] (3) Data for the current month is missing, and historical data for the middle and late parts of the month is missing, but data for the first ten days of the month is found. The formula for calculating the sample for the second ten days of the current month is as follows:

[0118] ;

[0119] (4) Data for the current month is missing, and historical data for the first and last ten days of the month is missing, but data for the middle ten days is found. The formula for calculating the sample for the last ten days of the current month is as follows:

[0120] ;

[0121] (5) Data for the current month is missing, historical data for the latter part of the month is missing, but data for the first and middle parts of the month are available. The formula for calculating the sample for the latter part of the month is as follows:

[0122] ;

[0123] (6) Data not found.

[0124] S13. Map the electricity consumption samples from the first, middle, and last ten days of the month to the electricity sales forecast for unknown dates in the current month, accumulate them to obtain the monthly forecast electricity sales, and make predictions based on three scenarios: optimistic, stable, and pessimistic, as follows:

[0125] S131. Map the samples from the first, middle, and last ten days of the month to the dates with unknown electricity sales for the current month, and sum them up to obtain the predicted electricity sales for the current month. The specific formula is as follows:

[0126] ;

[0127] in, This indicates the monthly forecast for electricity sales. This indicates the sales volume on a known date, typically from the 1st to the 9th of the first ten days of the month. To predict electricity sales in the first ten days of the month, the electricity sales volume on the 10th day is usually used; To predict electricity sales volume in the middle of the month; To predict electricity sales volume for the second half of the month.

[0128] S132. Predict electricity sales under three scenarios: optimistic, pessimistic, and stable development. The prediction result obtained in S131 is the electricity sales under the stable development state. For the optimistic and pessimistic scenarios, the average electricity sales used in step S123 needs to be replaced with the maximum and minimum electricity sales values.

[0129] S14. Adjust the predicted electricity sales volume for any special events predicted for the current month, and calculate the adjusted electricity sales volume for the current month based on the historical proportion of adjusted electricity sales. Finally, sum the results to obtain the predicted monthly electricity sales volume and confidence interval, as detailed below:

[0130] S141. Adjustments will be made to the predicted electricity sales volume to account for special circumstances such as short holidays, typhoons, policy changes, and prolonged periods of high temperatures during the predicted month. The first adjustment is for electricity sales volume forecasts related to national statutory holidays:

[0131] (1) For holidays with fixed dates, such as New Year's Day, Labor Day, and National Day:

[0132] ;

[0133] in Historical benchmark electricity sales for fixed-date holidays Before the start of the holiday Historical electricity sales are calculated over a 7-day period, and historical electricity sales are usually taken from the same day last year.

[0134] ;

[0135] in The annual benchmark electricity sales volume for fixed-date holidays. Before the start of the holiday Daily electricity sales.

[0136] The predicted electricity consumption during the holiday period is calculated as follows:

[0137] ;

[0138] in This indicates the daily electricity sales volume during historical holidays.

[0139] (2) For holidays with non-fixed dates, such as Qingming Festival, Dragon Boat Festival, and Mid-Autumn Festival:

[0140] ;

[0141] in This is the historical benchmark electricity sales volume for non-fixed date holidays. Before the start of the holiday The historical electricity sales volume of the day.

[0142] ;

[0143] in The benchmark electricity sales volume for the current year is based on non-fixed date holidays. Before the start of the holiday Daily electricity sales.

[0144] The predicted electricity consumption during the holiday period is calculated as follows:

[0145] ;

[0146] The method for determining the electricity sales sample during the Spring Festival, which has a long time span, is the same as that for fixed-date holidays.

[0147] Secondly, there are revisions to electricity sales forecasts for special weather conditions such as continuous high temperatures, continuous rainfall, the plum rain season, and typhoons.

[0148] (1) The continuous rainfall correction is involved in step S113 and will not be repeated here.

[0149] (2) Continuous high temperature correction:

[0150] A continuous high temperature is defined as three consecutive days with a maximum daily temperature of ≥35℃. Based on electricity consumption data under continuous high temperatures in a certain region during summer, the curve of the growth rate of electricity consumption under continuous high temperatures was obtained. It was concluded that the impact of continuous high temperatures with fewer days (within 13 days) on electricity consumption shows a local linear relationship. The growth rate of electricity consumption reaches its peak at around 13 days, which is about 7%. Long-term high temperatures after 22 days can also maintain a relatively high growth rate of about 6.5%, after which the growth rate of electricity consumption remains unchanged.

[0151] (3) Rainy season correction:

[0152] The impact of the plum rain season on electricity sales in a certain region was quantified using a group comparison method. The ratio of average electricity sales during the plum rain season to that during the non-plum rain season, under the same conditions, was selected, and the plum rain coefficient was calculated to be approximately 1.08. Therefore, the electricity sales during the plum rain season should be multiplied by the plum rain coefficient based on the original forecast data (the previous year's electricity sales during the plum rain season need to exclude the plum rain coefficient before calculating the ten-day coefficient).

[0153] (4) Typhoon Correction:

[0154] In a certain region, electricity sales typically see a slight increase the day before a typhoon arrives, followed by a continuous decline after the typhoon's arrival, reaching its lowest point 2-3 days later, before slowly recovering to normal levels. During the high temperatures of July and August, the continuous rainfall brought by typhoons causes a drop in temperature, resulting in lower electricity sales compared to hot weather. This effect usually begins to take hold 5-7 days before the typhoon's arrival, and the impact of the temperature drop increases with the escalation of the warning level. Under other non-high-temperature weather conditions, the temperature drop brought by typhoons is limited, and the impact of temperature on electricity sales is relatively small; therefore, electricity sales do not typically experience a significant decrease before the typhoon's arrival.

[0155] S142. The forecast for adjusted electricity volume is calculated based on the completed monthly electricity sales forecast, according to the historical proportion of adjusted electricity volume for the current month. The specific formula is as follows:

[0156] ;

[0157] in, To adjust the electricity consumption for the current month, Adjusting the electricity consumption to match historical data for the same month. This is the final revised electricity sales forecast data.

[0158] S143. Calculate the monthly electricity generation and set upper and lower limits based on this, expanding the confidence interval. The upper and lower limits are adjusted by 25 million kWh above and below the predicted results, as shown in the following formula:

[0159] ;

[0160] in This refers to the monthly electricity generation.

[0161] S2, the annual electricity forecasting steps include:

[0162] S21. Based on the historical monthly electricity sales ratio and the distribution of holidays in the current year, calculate the monthly electricity sales ratio for the current year, and combine it with the annual total electricity sales forecast. Use the ratio method to forecast and update the annual electricity sales, as follows:

[0163] S211. Calculate the average monthly electricity sales ratio over the years, excluding the impact of holidays. Define the monthly electricity sales ratio as... The specific formula is as follows:

[0164] ;

[0165] in This indicates the electricity sales volume in a certain month last year; This refers to the electricity sales volume in a certain month of the year before last. This represents the total electricity sales volume for the entire previous year. This represents the total electricity sales for the entire year before last. The average of the electricity sales percentage for a specific month from the previous year and the year before last is used here to increase the validity of this percentage.

[0166] S212. The total electricity sales forecast for this year is based on the GDP growth rate, and the total electricity sales for the whole year can be predicted in January of this year. Based on the electricity sales forecast for the current month obtained from the above steps, and combined with the monthly electricity sales ratio, the total electricity sales forecast data is updated on a rolling basis.

[0167] S22. Using the triple exponential smoothing method combined with historical data, predict the annual electricity sales trend, and compare and verify the results with the percentage method prediction. Output the annual electricity sales prediction results as follows:

[0168] S221. Using the triple exponential smoothing method, and combining known data from the current year and two years ago (a total of two years), predict the unknown monthly electricity sales for the current year. The specific formula for this method is as follows:

[0169] ;

[0170] in, For time t, the future... The predicted value for each cycle; The intercept term represents the horizontal component of the current time t; The linear trend component represents the linear growth per unit period; It is a secondary trend component, reflecting the acceleration or deceleration of the trend; This represents the number of leading cycles predicted.

[0171] S222. Combine the electricity sales volume of unknown months in the current year predicted by the triple exponential smoothing method with the electricity sales volume of known months in the current year to calculate the total electricity sales volume for the current year, and compare it with the prediction results of the percentage method to enhance the reliability of the annual forecast.

[0172] In a specific implementation case, an experiment was conducted using the electricity generation forecast for a certain region from March to October 2025 as an example:

[0173] The monthly electricity generation forecast results and errors are shown in Table 1: Table 1

[0174]

[0175] The monthly electricity generation forecast results show that the minimum forecast error is mostly within 1%, and the maximum does not exceed 2%, indicating that the forecasting scheme is highly feasible. Even when facing special weather conditions, such as the high temperatures in July and August 2025 which significantly impact electricity sales, the scheme maintains high forecast accuracy, demonstrating its strong resilience.

[0176] Taking November as an example, the real-time annual electricity generation forecast data is shown in Table 2: Table 2

[0177]

[0178] S3. Based on the monthly electricity generation forecast and confidence interval obtained in steps S14 and S22, and the annual electricity sales forecast, the final output is the record forecast set.

[0179] In summary, this invention proposes a power generation forecasting method based on the influence of multi-source data under large-scale renewable energy integration. This method constructs a monthly and annual collaborative forecasting framework that integrates multi-source data such as real-time weather, holidays, and special events. By quantifying the impact of external factors such as temperature, rainfall, continuous high temperatures, plum rains, and typhoons on power generation, and establishing a systematic correction model, the forecasting accuracy and robustness to unforeseen circumstances are significantly improved. Practical examples show that the monthly forecasting error of this scheme is mostly controlled within 1%, verifying its feasibility and strong practical value.

[0180] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any modifications or equivalent changes made based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A method for predicting electricity generation under the influence of multi-source data in the context of renewable energy access, characterized in that, Includes the following steps: S1, Monthly Electricity Consumption Forecast: S11. Collect actual electricity sales, weather forecast data and holiday information for the 1st to 9th of the current month, as well as corresponding data for the same period in history, and perform temperature correction and restoration and continuous rainfall correction and restoration on the historical electricity sales data. S12. Classify dates based on weather type and holiday type, calculate the ten-day period coefficient under different categories, and establish electricity samples for the first, middle and last ten days of the month based on the data of the first 10 days of the month and the ten-day period coefficient; S13. Map the electricity samples from the first, middle, and last ten days of the month to the electricity sales forecast for unknown dates of the month, accumulate the monthly forecast electricity sales, and make predictions based on three scenarios: optimistic, stable, and pessimistic. S14. Adjust the predicted electricity sales volume in response to special events that are predicted to occur in the current month, and calculate the predicted adjusted electricity volume for the current month based on the historical adjusted electricity volume ratio. Finally, sum them up to obtain the predicted monthly electricity issuance volume and confidence interval. S2, Annual Electricity Forecast: S21. Based on the historical monthly electricity sales ratio and the distribution of holidays in the current year, calculate the monthly electricity sales ratio for the current year, and combine it with the annual total electricity sales forecast to use the ratio method for annual electricity sales forecasting and rolling updates. S22. Use the triple exponential smoothing method and historical data to predict the annual electricity sales trend, and compare and verify the prediction results with the percentage method, and output the annual electricity sales prediction results. S3. Based on the monthly electricity generation forecast and confidence interval obtained in steps S14 and S22, and the annual electricity sales forecast, the final output is the record forecast set.

2. The method for predicting electricity generation under the influence of multi-source data in the context of new energy access, as described in claim 1, is characterized in that... In step S11, the temperature correction and restoration specifically involves: For July and August, to eliminate the difference between historical electricity sales and current month's electricity sales caused by temperature, the correction formula is as follows: （1）； in, This indicates historical electricity sales after temperature correction; This represents the original historical electricity sales volume; It is the temperature difference coefficient; It is calculated as the difference between the historical daily high temperature of the current month and the average high temperature of the current month this year. When the temperature is higher than 37℃, it is calculated as 37℃.

3. The method for predicting electricity generation under the influence of multi-source data in the context of new energy access, as described in claim 2, is characterized in that... In step S11, the continuous rainfall correction and restoration specifically involves: The historical electricity sales, after temperature correction, are further adjusted based on the number of consecutive days of rainfall, using the following formula: （2）； in, d represents the historical electricity sales volume after correction for continuous rainfall; d represents the number of consecutive rainfall days.

4. The method for predicting electricity generation under the influence of multi-source data in the context of new energy access, as described in claim 1, is characterized in that... In step S12, the ten-day period coefficient includes the ten-day period coefficient for the first and middle ten days and the ten-day period coefficient for the first and last ten days, and is calculated separately for working days and non-working days.

5. The method for predicting electricity generation under the influence of multi-source data in the context of new energy access, as described in claim 1, is characterized in that... In step S13, the formula for calculating the monthly predicted electricity sales is: （3）； in, This indicates the monthly forecast for electricity sales. This indicates the volume of electricity sold on a known date. To predict electricity sales volume in the first ten days of the month; To predict electricity sales volume in the middle of the month; To predict electricity sales volume for the second half of the month.

6. The method for predicting electricity generation under the influence of multi-source data in the context of new energy access, as described in claim 1, is characterized in that... In step S14, the special events include national statutory holidays, continuous high temperatures, plum rain season and typhoon weather, and corresponding correction models are used to correct the predicted electricity sales for different special events.

7. The method for predicting electricity generation under the influence of multi-source data in the context of new energy access, as described in claim 1, is characterized in that... In step S14, the formula for calculating the monthly electricity generation is: （4）； in, Forecast electricity generation for the month; This is the final monthly forecast of electricity sales after adjustments for special events; The formula for calculating the monthly electricity forecast adjustment is as follows: （5）； Adjusting the electricity consumption to match historical data for the same month. This represents the original historical electricity sales volume.

8. The method for predicting electricity generation under the influence of multi-source data in the context of new energy access, as described in claim 1, is characterized in that... In step S21, the formula for calculating the monthly electricity sales ratio in the ratio-based prediction is as follows: （6）； Where p represents the percentage of monthly electricity sales; This indicates the electricity sales volume in a certain month last year; This refers to the electricity sales volume in a certain month of the year before last. This represents the total electricity sales volume for the entire previous year. This represents the total electricity sales volume for the entire year before last.

9. The method for predicting electricity generation under the influence of multi-source data in the context of new energy access, as described in claim 1, is characterized in that... In step S21, the annual total electricity sales forecast is based on the GDP growth rate. The total annual electricity sales can be predicted in January of this year. Based on the predicted monthly electricity sales and the monthly electricity sales ratio, the annual total electricity sales forecast data is updated on a rolling basis.

10. The method for predicting electricity generation under the influence of multi-source data in the context of new energy access, as described in claim 1, is characterized in that... In step S22, the prediction formula for the triple exponential smoothing method is: （7）； in, For time t, the future... The predicted value for each cycle; The intercept term represents the horizontal component of the current time t; The linear trend component represents the linear growth per unit period; It is a secondary trend component, reflecting the acceleration or deceleration of the trend; This represents the number of leading cycles predicted.

Images