Ultra-short-term power load prediction method, terminal and medium

Abstract

The invention discloses an ultra-short-term power load prediction method, a terminal and a medium in the technical field of power load prediction. The method comprises the following steps: acquiring power load data and corresponding temperature data of all time nodes in a plurality of days; according to the collected data, a scatter diagram between the power load and the average temperature of the current day is drawn for each time node; respectively calculating power load change rates of different time nodes for non-holidays and holidays, and storing the power load change rates into a corresponding power load change rate table; solving an optimal optimization change rate corresponding to each power load change rate on holidays and festivals by using a particle swarm optimization algorithm; and matching a corresponding power load change rate table according to the average temperature of the prediction day, and calculating a power load prediction value of a future time node in combination with the optimized change rate. According to the ultra-short-term power load prediction method provided by the invention, the accuracy, reliability and adaptability of power load prediction are remarkably improved through comprehensive data acquisition, scientific data analysis and an optimization algorithm.

Description

Technical Field

[0001] The present invention relates to a very short - term electric load forecasting method, a terminal and a medium, belonging to the technical field of electric load forecasting. Background Art

[0002] Very short - term electric load forecasting plays a crucial role in power system dispatching and optimal operation, and its accuracy directly affects the economy and stability of the power grid. Traditional very short - term load forecasting methods rely solely on historical load data and use time - series models for forecasting, ignoring the influence of many external factors such as meteorology and social activities on electric load, resulting in a large deviation between the forecasting results and the actual values.

[0003] Meteorological conditions have a significant impact on electric load. Changes in meteorological elements such as temperature, humidity, and illumination will change the electricity - using behaviors of users and the electricity - using demands of various industries. However, traditional forecasting models fail to deeply explore the internal relationship between meteorology and electric load and cannot optimize the forecasting model based on real - time meteorological data, seriously reducing the forecasting accuracy.

[0004] During holidays and large - scale social activities, the electricity - using patterns of residents' lives and enterprises' production will change, and the distribution and change law of electric load will also change accordingly. However, traditional forecasting methods have insufficient understanding of the load change characteristics during special periods, and the forecasting errors during special periods increase significantly.

[0005] Therefore, how to construct a more robust very short - term load forecasting method to improve the adaptability and accuracy of forecasting remains an important challenge in current research and applications. Summary of the Invention

[0006] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a very short - term electric load forecasting method, a terminal and a medium. Through comprehensive data collection, scientific data analysis, and optimized forecasting algorithms, the accuracy, reliability, and adaptability of electric load forecasting are significantly improved.

[0007] To achieve the above - mentioned purpose, the present invention is implemented by the following technical solutions:

[0008] In the first aspect, the present invention provides a very short - term electric load forecasting method, including:

[0009] Collecting electric load data and corresponding temperature data at all time nodes for several days;

[0010] According to the collected electric load data and corresponding temperature data, respectively drawing a scatter plot of electric load versus the average temperature of the day for each time node to analyze the trend of electric load change with temperature;

[0011] Identify and correct the abnormal power load data in the scatter plot. Based on the corrected power load data, calculate the power load change rates at different time nodes for non-holiday and holiday respectively, and store them in the power load change rate table corresponding to the temperature range according to the average temperature of the day.

[0012] Use the particle swarm optimization algorithm to solve the optimal optimized change rate corresponding to each power load change rate during holidays, so as to minimize the error between the power load prediction value and the actual power load value.

[0013] Match the corresponding power load change rate table according to the average temperature of the prediction day, and calculate the power load prediction value of the future time node in combination with the optimized change rate.

[0014] Furthermore, the abscissa of each scatter point in each scatter plot is the average temperature of the corresponding date at each time node, and the ordinate is the power load value of that date at that time node; the number of scatter plots is the same as the number of time nodes divided in a day.

[0015] Furthermore, identifying and correcting the abnormal power load data in the scatter plot includes:

[0016] Divide the power load data in each scatter plot into intervals according to the average temperature of each day; conduct statistical analysis on the power load data in different temperature intervals, calculate the abnormal range of the power load data of each group of time nodes in each temperature interval by using the quartile method, and mark the power load data that exceeds the normal quartile range as abnormal power load data.

[0017] For the marked abnormal power load data, based on the power load data of adjacent time nodes, use the cubic spline interpolation method to calculate the power load correction value, and replace the abnormal power load data with the calculation result.

[0018] Furthermore, based on the corrected power load data, calculate the power load change rates at different time nodes for non-holiday and holiday respectively, and store them in the corresponding power load change rate table, including:

[0019] Divide the power load value of the current time node by the power load value of the previous time node to obtain the

[0020] corresponding power load change rate of the current time node; Among them, represents the number of previous time nodes used to calculate the power load prediction value. When the day is a non-holiday, = 1; when the day is a holiday,

[0021] Calculate the power load change rate for all time nodes in each day. According to the average temperature of the day, store the obtained power load change rate in the non-holiday power load change rate table or holiday power load change rate table corresponding to the temperature range.

[0022] When storing the load change rate, if there is existing stock data for the corresponding time node, take the average of the new and old data and update the stock data with this average.

[0023] Among them, the power load change rate on non-holidays is directly used to calculate the power load prediction value on non-holidays, while the power load change rate on holidays is used to calculate the optimal optimization change rate. The optimal optimization change rate, the power load change rate, and the power load value at the previous time node are jointly used to calculate the power load prediction value on holidays.

[0024] Furthermore, use the particle swarm optimization algorithm to solve the optimal optimization change rate corresponding to each power load change rate on holidays to minimize the error between the power load prediction value and the actual power load value, including:

[0025] For the power load data on holidays, calculate the power load prediction value and the actual power load value The deviation rate between them is as shown in the following formula:

[0026] ;

[0027] ;

[0028] In the formula: Only on holidays, = 3; is the power load prediction value at the next time node , is the power load change rate between the current time node and the previous th time node, is the power load value at the previous th time node, is the initial optimization change rate corresponding to each power load change rate;

[0029] Calculate the sum of the deviation rates for all time nodes in a day to obtain the optimal score as shown in the following formula:

[0030] ;

[0031] In the formula: is the total number of time nodes divided in a day;

[0032] The optimal score is used to measure the prediction error under the combination of the current power load change rate and the optimized change rate. With the goal of minimizing the prediction error, the particle swarm optimization algorithm is used to optimize the particle positions;

[0033] Each particle position represents a potential solution, and the particle position includes the power load change rate and the optimized change rate , and the particle position is expressed as ;

[0034] The particles adjust their velocities according to the current position and the global optimal position. The formula for updating the particle position is shown as follows:

[0035] ;

[0036] In the formula: represents the velocity of particle at the -th iteration; represents the inertia weight, which controls the global search ability of the particle swarm; and represent the learning factors, which control the learning of the particle from its own and the global optimal positions; and represent random numbers, which ensure the randomness of the search; represents the best position of particle ; represents the global best position of the particle swarm;

[0037] After each update of the particle position, the optimal score is used to evaluate the fitness of the particle position; if the position of a certain particle brings a lower error, the best position of the particle is updated ;

[0038] Meanwhile, all the particles in the particle swarm will update their positions according to the current global optimal solution until the termination condition of the iterative optimization is met, at which point the iteration terminates, and the optimal optimized change rate corresponding to each power load change rate is output .

[0039] Furthermore, according to the average temperature of the prediction day, the corresponding power load change rate table is matched, and the power load prediction value at the future time node is calculated in combination with the optimized change rate, including:

[0040] When the day is a non-holiday, the corresponding non-holiday power load change rate table is matched according to the average temperature of the prediction day to obtain the power load value and the power load change rate , calculate the predicted value of the power load at the next time node as shown in the following formula:

[0041] ;

[0042] When the day is a holiday, match the corresponding holiday power load change rate table according to the average temperature of the prediction day to obtain the power load value at the previous time node , the optimal optimization change rate and the power load change rate , calculate the predicted value of the power load at the next time node as shown in the following formula:

[0043] .

[0044] Furthermore, the ultra-short-term power load prediction method further includes:

[0045] After obtaining the actual value of the power load at a future time node, use the quartile method to verify its rationality. If it is determined to be an outlier, replace the outlier with the predicted value of the power load and output an error message to remind manual inspection.

[0046] In a second aspect, the present invention provides an electronic terminal, including a processor and a memory connected to the processor. A computer program is stored in the memory. When the computer program is executed by the processor, the steps of the above-mentioned ultra-short-term power load prediction method are executed.

[0047] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. The program is characterized in that when it is executed by a processor, the steps of the above-mentioned ultra-short-term power load prediction method are realized.

[0048] Compared with the prior art, the beneficial effects achieved by the present invention:

[0049] The ultra-short-term power load prediction method provided by the present invention can accurately grasp the change trend of the power load with temperature by collecting the power load and temperature data at each time node for several days and drawing a scatter plot, providing a reliable basis for subsequent prediction. The identification and correction of abnormal data in the scatter plot ensure the data quality and avoid interfering with the prediction results. Calculating the power load change rate for holidays and non-holidays and storing it according to temperature intervals fully considers the differences in electricity consumption characteristics in different periods, making the prediction more in line with the actual situation. Using the particle swarm optimization algorithm to solve the optimal optimization change rate for holidays effectively reduces the prediction error and improves the prediction accuracy for holidays. Calculating the predicted value by matching the change rate table according to the average temperature of the prediction day combined with the optimization change rate can flexibly adapt to different meteorological conditions, providing timely and accurate decision support for the power system dispatching and management, and enhancing the stability and reliability of the power system operation. Brief Description of the Drawings

[0050] Figure 1 It is a flowchart of the ultra - short - term power load forecasting method provided in the first embodiment of the present invention. Detailed Embodiments

[0051] The technical solution of the present invention will be described in detail below through the drawings and specific embodiments. It should be understood that the specific features in the embodiments of the present application are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. Without conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other.

[0052] Embodiment 1

[0053] Figure 1 It is a flowchart of an ultra - short - term power load forecasting method in the first embodiment of the present invention. This flowchart only shows the logical sequence of the method described in this embodiment. On the premise of non - conflict, in other possible embodiments of the present invention, the steps shown or described can be completed in a different order from Figure 1 the order shown. Referring to Figure 1 , the method of this embodiment specifically includes the following steps:

[0054] Collect the power load data and corresponding temperature data of all time nodes in several days;

[0055] According to the collected power load data and corresponding temperature data, draw a scatter plot of the power load and the average temperature of the day for each time node respectively, to analyze the trend of the power load changing with the temperature;

[0056] Identify and correct the abnormal power load data in the scatter plot. Based on the corrected power load data, calculate the power load change rates of different time nodes for non - holiday and holiday respectively, and store them in the power load change rate table of the corresponding temperature interval according to the average temperature of the day;

[0057] Use the particle swarm optimization algorithm to solve the optimal optimized change rate corresponding to each power load change rate during holidays, so that the error between the power load prediction value and the actual power load value is minimized;

[0058] Match the corresponding power load change rate table according to the average temperature of the prediction day, and calculate the power load prediction value of the future time node in combination with the optimized change rate.

[0059] Specifically, first, carry out data collection work. Within the time span of several days, collect the power load data corresponding to all time nodes and the corresponding temperature data. These data will be used as the basis for subsequent analysis and prediction. In this embodiment, a day is divided into 96 time nodes, and the interval duration between every two consecutive time nodes is 15 minutes.

[0060] Next, based on the collected power load data and the corresponding temperature data, for each time node, a scatter plot between the power load and the average temperature of the day is drawn respectively. Through these scatter plots, the trend of the power load changing with temperature can be visually analyzed. In each scatter plot, the abscissa of each scatter point is the average temperature of the corresponding date of each time node, and the ordinate is the power load value of that date at that time node. Moreover, the number of scatter plots is the same as the number of time nodes divided in a day, so as to comprehensively and meticulously display the relationship between the power load and temperature at different time nodes.

[0061] In this embodiment, a day is divided into 96 time nodes, corresponding to 96 scatter plots, namely the scatter plot of the 0:00 time node, the scatter plot of the 0:15 time node, the scatter plot of the 0:30 time node,..., the scatter plot of the 23:45 time node. Taking the scatter plot of the 0:00 time node as an example for illustration, in actual data collection, to ensure the accuracy of calculating parameters such as the power load change rate in the subsequent process, this embodiment collects relevant data within a relatively long time span, rather than being limited to only a few days. Here, for the sake of simple illustration, it is assumed that the collected data is for the 3 days from January 1st to January 3rd; when drawing the scatter plot of the 0:00 time node, each scatter point in the figure corresponds to the data of one day. Its abscissa is the average temperature of each date, and this average temperature comprehensively considers the temperature fluctuations of the 96 time nodes on that day and can more accurately represent the overall temperature condition of that day; the ordinate is the power load value of the corresponding date at the 0:00 time node. Illustrated with specific numerical values, for the three points in the scatter plot, the abscissa is the average temperature, corresponding to 5.0°C on January 1st, 6.0°C on January 2nd, and 5.5°C on January 3rd respectively; the ordinate is the load value at 0:00, corresponding to the three points being 150MW, 155MW, and 153MW respectively.

[0062] That is to say, the specific meaning of each point is as follows:

[0063] (5.0°C, 150MW) means that the average temperature on January 1st is 5.0°C, and the load value at the 0:00 time node on January 1st is 150MW;

[0064] (6.0°C, 155MW) means that the average temperature on January 2nd is 5.0°C, and the load value at the 0:00 time node on January 2nd is 155MW;

[0065] (5.5°C, 153MW) means that the average temperature on January 3rd is 5.5°C, and the load value at the 0:00 time node on January 3rd is 153MW;

[0066] Subsequently, it is necessary to identify and correct the abnormal power load data in the scatter plot. The specific operation is to divide the power load data in each scatter plot into intervals according to the average temperature of each day. Statistical analysis is carried out on the power load data in different temperature intervals, and the quartile method is used to calculate the abnormal range of the power load data of each group of time nodes in each temperature interval. Once the power load data exceeds the normal quartile range, it is marked as abnormal power load data. For the marked abnormal power load data, based on the power load data of adjacent time nodes, the cubic spline interpolation method is used to calculate the power load correction value, and the calculated result is used to replace the abnormal power load data to ensure the accuracy of the data.

[0067] In this embodiment, the power load data is grouped according to an interval of every 5°C, such as 0°C - 5°C, 5°C - 10°C, etc. For each temperature group, the data of all dates at the same time node within the group is collected. For example, in the 5°C - 10°C group, the power load values at the 0:00 time node of all dates are collected to form a set of data. This set of data is derived from the power load data corresponding to each time node when the scatter plot was drawn previously.

[0068] For each group of time node data within each temperature group, the quartiles are calculated. Taking the data of the 0:00 time node in the 5°C - 10°C group as an example, the 25% quantile (Q1) and 75% quantile (Q3) are calculated, and then the interquartile range (IQR = Q3 - Q1) is obtained. According to the quartile method, the abnormal value range is determined, with the lower bound being Q1 - 1.5×IQR and the upper bound being Q3 + 1.5×IQR. If the load value of a certain date at the 0:00 time node exceeds this range, it is marked as an abnormal power load value.

[0069] Traverse the data of 96 time nodes under each temperature group to find all power load values that exceed the abnormal value range. When an abnormal power load value is determined, based on the power load data of adjacent time nodes, the cubic spline interpolation method is used to calculate the power load correction value, and the calculated result is used to replace the abnormal power load data. In this embodiment, the load values of the 5 adjacent time nodes before and after the date where the abnormal power load value is located are obtained. Suppose the load value at 0:00 on January 3 is determined to be an abnormal value, then the load value data of these 10 time nodes, namely 23:00, 23:15, 23:30, 23:45, 0:00 on January 2 and 0:15, 0:30, 0:45, 1:00, 1:15 on January 3, are collected.

[0070] Using the collected load value data at 5 time nodes before and after, the cubic spline interpolation method is used to calculate the reasonable load value at the position of the outlier. The cubic spline interpolation method fits the data points by constructing a cubic polynomial function, making the curve have continuous first and second derivatives at each data point, thus ensuring the smoothness of the curve. The calculated interpolation result is used to replace the original outlier. For example, if the interpolation of the outlier at 0:00 on January 3rd is calculated to be 160MW, then the original abnormal load value is replaced with 160MW to complete the data correction.

[0071] Based on the corrected power load data, calculate the power load change rates at different time nodes for non-holiday and holiday respectively, and store them in the corresponding power load change rate tables. The specific calculation method is to divide the power load value at the current time node by the power load value at the previous time node to obtain the corresponding power load change rate at the current time node. Here, when the day is a non-holiday, = 1; when the day is a holiday, = 3. After calculating the power load change rates at all time nodes in each day, according to the average temperature of the day, store the obtained power load change rates in the non-holiday power load change rate table or holiday power load change rate table in the corresponding temperature range. When storing the load change rate, if there is already existing data at the corresponding time node, take the average of the new and old data and update the existing data with this average. The calculation steps of the power load change rate are illustrated as follows with specific data:

[0072] Suppose the power load data from January 1st to January 5th are collected, and some of the data are as follows (the actual data are 96 time nodes per day, simplified here for display):

[0073]

[0074] Calculate the power load change rates at different time nodes for non-holiday and holiday:

[0075] 1. Non-holiday power load change rate (in the form of a 1×96 list, simplified to an example of 3 time nodes here)

[0076] At 0:15 on January 1st, the load value is 102MW, and the load value at the previous time node (0:00) is 100MW. The power load change rate = 102÷100 = 1.02.

[0077] At 0:30 on January 1st, the load value is 105MW, and the load value at the previous time node (0:15) is 102MW. The power load change rate = 105÷102≈1.03.

[0078] At 0:15 on January 3rd, the load value was 104 MW, and the load value at the previous time node (0:00) was 103 MW. The power load change rate = 104÷103≈1.01.

[0079] At 0:30 on January 3rd, the load value was 106 MW, and the load value at the previous time node (0:15) was 104 MW. The power load change rate = 106÷104≈1.02.

[0080] At 0:15 on January 4th, the load value was 112 MW, and the load value at the previous time node (0:00) was 110 MW. The power load change rate = 112÷110≈1.02.

[0081] At 0:30 on January 4th, the load value was 115 MW, and the load value at the previous time node (0:15) was 112 MW. The power load change rate = 115÷112≈1.03.

[0082] Taking the temperature grouping of 5°C - 10°C as an example, the power load change rate corresponding to each time node is stored in the non-holiday power load change rate table. Taking the power load change rate at 0:15 as an example:

[0083] The power load change rate at 0:15 on January 1st was 1.02 and was stored in the table;

[0084] The power load change rate at 0:15 on January 3rd was 1.01. At this time, there was already data 1.02 in the table. Then, the average of 1.02 and 1.01 was taken, (1.02 + 1.01) / 2≈1.02, and this average value of 1.02 was stored in the table, overwriting the original 1.02 (here, 1.02 refers to the power load change rate of 1.02 at 0:15 on January 1st)

[0085] The power load change rate at 0:15 on January 4th was 1.02,... The same steps were followed as above. At this time, the average was taken again, and the final value stored in the table was 1.02.

[0086] Through the above steps, under non-holiday conditions, when the temperature range is 5°C - 10°C, the power load change rate corresponding to the time node 0:15 is obtained. By analogy, the power load change rates corresponding to other time nodes in this temperature range are calculated. Finally, the power load change rates corresponding to 96 time nodes in the temperature range of 5°C - 10°C under non-holiday conditions can be obtained, and these change rates are integrated into a 1×96 list.

[0087] 2. Holiday power load change rate (in the form of a 3×96 list, simplified to 3 time point examples here)

[0088] At 0:15 on January 2nd, with a load value of 107 MW, divide it by the load values of the previous three nodes (assumed load value of 105 MW at 0:00, assumed load value of 98 MW at 23:45 on January 1st, and assumed load value of 96 MW at 23:30 on January 1st) respectively to obtain 3 power load change rates:

[0089] Power load change rate = 107 ÷ 105 ≈ 1.02

[0090] Power load change rate = 107 ÷ 98 ≈ 1.09

[0091] Power load change rate = 107 ÷ 96 ≈ 1.11

[0092] At 0:30 on January 2nd, with a load value of 110 MW, divide it by the load values of the previous three nodes (assumed load value of 107 MW at 0:15, assumed load value of 100 MW at 23:59 on January 1st, and assumed load value of 98 MW at 23:45 on January 1st) respectively to obtain 3 power load change rates:

[0093] Power load change rate = 110 ÷ 107 ≈ 1.03

[0094] Power load change rate = 110 ÷ 100 = 1.10

[0095] Power load change rate = 110 ÷ 98 ≈ 1.12

[0096] At 0:15 on January 5th, with a load value of 114 MW, divide it by the load values of the previous three nodes (assumed load value of 112 MW at 0:00, assumed load value of 108 MW at 23:45 on January 4th, and assumed load value of 106 MW at 23:30 on January 4th) respectively to obtain 3 power load change rates:

[0097] Power load change rate = 114 ÷ 112 ≈ 1.02

[0098] Power load change rate = 114 ÷ 108 ≈ 1.06

[0099] Power load change rate = 114 ÷ 106 ≈ 1.08

[0100] Taking the temperature grouping of 5°C - 10°C as an example, store the 3 power load change rates corresponding to each time node in the holiday power load change rate table. When storing the load change rate, if there is existing stock data for the corresponding time node, take the average of the new and old data and update the stock data with this average value.

[0101] Through the above steps, the power load change rates corresponding to 96 time nodes in the temperature range of 5°C - 10°C during holidays are obtained, and these power load change rates are integrated into a 3×96 list. Considering the obvious differences in people's electricity consumption behaviors between holidays and non-holidays, this method of handling cases separately can more accurately reflect the power load characteristics in different time periods, making the prediction results more in line with the actual electricity consumption situation.

[0102] The power load change rates during non-holidays are directly used to calculate the power load prediction values for non-holidays, while the power load change rates during holidays are used to calculate the optimal optimization change rates. The optimal optimization change rates, power load change rates, and the power load values of the previous time nodes are jointly used to calculate the power load prediction values for holidays.

[0103] After that, the particle swarm optimization algorithm is used to solve the optimal optimization change rate corresponding to each power load change rate during holidays, with the goal of minimizing the error between the power load prediction value and the actual power load value. Through continuous iterative optimization, this algorithm can find the prediction parameters most suitable for the holiday electricity consumption pattern, effectively improving the accuracy of holiday power load prediction. The specific process is as follows: for the holiday power load data, calculate the deviation rate between the power load prediction value and the actual power load value The deviation rate is shown in the following formula:

[0104] ;

[0105] ;

[0106] In the formula: only during holidays, = 3; is the power load prediction value of the next time node , is the power load change rate between the current time node and the th previous time node, is the power load value of the th previous time node, is the initial optimization change rate corresponding to each power load change rate.

[0107] Next, calculate the sum of the deviation rates of all time nodes in a day to obtain the optimal score , as shown in the following formula:

[0108] ;

[0109] Wherein: is the total number of time nodes divided within a day;

[0110] In this embodiment, a day is divided into 96 time nodes, and the sum of the deviation rates corresponding to the 96 time nodes is calculated to obtain the optimal score. This optimal score is used to measure the prediction error under the combination of the current power load change rate and the optimized change rate. With the goal of minimizing the prediction error, the particle swarm optimization algorithm is used to optimize the particle positions. Each particle position represents a potential solution, and the particle position includes the power load change rate and the optimized change rate , and the particle position is expressed as .

[0111] The particle adjusts its velocity according to the current position and the global optimal position. The formula for updating the particle position is shown as follows:

[0112] ;

[0113] Wherein: represents the velocity of particle at the th iteration; represents the inertia weight, which controls the global search ability of the particle swarm; and represent the learning factors, which control the learning of the particle from its own and the global optimal positions; and represent random numbers, which ensure the randomness of the search; represents the best position of particle ; represents the global best position of the particle swarm;

[0114] After each particle position update, the optimal score is used to evaluate the fitness of the particle position; if the position of a certain particle brings a lower error, the best position of the particle is updated;

[0115] Meanwhile, all the particles in the particle swarm will update their positions according to the current global optimal solution until the termination condition of the iterative optimization is satisfied, at which point the iteration terminates and the optimal optimized change rate corresponding to each power load change rate is output .

[0116] Finally, match the corresponding power load change rate table according to the average temperature on the prediction day, and combine the optimized change rate to calculate the power load prediction value at the future time node. When the day is a non-holiday, match the corresponding non-holiday power load change rate table according to the average temperature on the prediction day to obtain the power load value at the current time node and the power load change rate , calculate the power load prediction value at the next time node, as shown in the following formula:

[0117] ;

[0118] When the day is a holiday, match the corresponding holiday power load change rate table according to the average temperature on the prediction day to obtain the power load value at the previous time node and the optimal optimized change rate and the power load change rate , calculate the power load prediction value at the next time node, as shown in the following formula:

[0119] .

[0120] In addition, this method also includes a verification work. After obtaining the actual value of the power load at the future time node, the quartile method is used to verify its rationality. If it is determined as an outlier, the outlier is replaced with the power load prediction value, and an error message is output to remind manual inspection. This mechanism further ensures the reliability of the prediction result, can timely detect and handle possible abnormal situations, and improves the operation stability of the power system

[0121] In summary, this ultra-short-term power load prediction method significantly improves the accuracy, reliability and adaptability of power load prediction through comprehensive data collection, scientific data analysis and optimized prediction algorithms, and can provide strong support for the production, scheduling and planning of power enterprises, with important practical application value

[0122] Embodiment 2:

[0123] The embodiment of the present invention also provides an electronic terminal, which is characterized in that it includes a processor and a memory connected to the processor, and a computer program is stored in the memory. When the computer program is executed by the processor, the steps of the ultra-short-term power load prediction method described in Embodiment 1 above are executed

[0124] Embodiment 3:

[0125] The embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the steps of the ultra-short-term power load prediction method described in Embodiment 1 above are first implemented

[0126] The computer-readable storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs, etc.

[0127] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) that contain computer-usable program codes.

[0128] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in Figure 1 one or more flows Figure 1 or a combination of multiple flows and / or blocks

[0129] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device realizes the functions specified in Figure 1 one or more flows Figure 1 or a combination of multiple flows and / or blocks

[0130] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for realizing the functions specified in Figure 1 one or more flows Figure 1 or a combination of multiple flows and / or blocks

[0131] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A very short-term electric load forecasting method, characterized in that, Including: Collecting the power load data and corresponding temperature data at all time nodes for several days; According to the collected power load data and corresponding temperature data, drawing a scatter plot between the power load and the average temperature of the day for each time node to analyze the trend of the power load changing with temperature; Identifying and correcting the abnormal power load data in the scatter plot, and based on the corrected power load data, calculating the power load change rates of different time nodes for non-holiday and holiday respectively, and storing them in the power load change rate table corresponding to the temperature range according to the average temperature of the day; Using the particle swarm optimization algorithm to solve the optimal optimized change rate corresponding to each power load change rate during holidays, so as to minimize the error between the power load prediction value and the actual power load value; Matching the corresponding power load change rate table according to the average temperature of the prediction day, and calculating the power load prediction value of the future time node by combining the optimized change rate.

2. The ultra-short-term electric load forecasting method according to claim 1, wherein The abscissa of each scatter point in each scatter plot is the average temperature of the corresponding date of each time node, and the ordinate is the power load value of the date at this time node; the number of scatter plots is the same as the number of time nodes divided in a day.

3. The ultra-short-term electric load forecasting method according to claim 1, characterized in that, Identifying and correcting the abnormal power load data in the scatter plot, including: Dividing the power load data in each scatter plot into intervals according to the average temperature of each day; Conducting statistical analysis on the power load data in different temperature ranges, using the quartile method to calculate the abnormal range of the power load data of each group of time nodes in each temperature range, and marking the power load data exceeding the normal quartile range as abnormal power load data; For the marked abnormal power load data, based on the power load data of adjacent time nodes, using the cubic spline interpolation method to calculate the power load correction value, and replacing the abnormal power load data with the calculation result.

4. The ultra-short-term electric load forecasting method according to claim 1, wherein Based on the corrected power load data, calculating the power load change rates of different time nodes for non-holiday and holiday respectively, and storing them in the corresponding power load change rate table, including: Divide the power load value at the current time node by the power load value at the previous time node to obtain the power load change rate corresponding to the current time node; Among them, represents the number of previous time nodes used to calculate the predicted value of the power load. When the day is a non-holiday, = 1; when the day is a holiday, = 3; Calculating the power load change rates of all time nodes in each day, and storing the obtained power load change rates in the non-holiday power load change rate table or holiday power load change rate table corresponding to the temperature range according to the average temperature of the day; When storing the load change rate, if there is existing stock data for the corresponding time node, take the average value of the new and old data, and update the stock data with this average value; Among them, the power load change rate of non-holiday is directly used to calculate the power load prediction value of non-holiday, while the power load change rate of holiday is used to calculate the optimal optimized change rate, and the optimal optimized change rate, power load change rate and the power load value of the previous time node are jointly used to calculate the power load prediction value of holiday.

5. The ultra-short-term electric load forecasting method according to claim 4, wherein Using the particle swarm optimization algorithm to solve the optimal optimized change rate corresponding to each power load change rate during holidays, so as to minimize the error between the power load prediction value and the actual power load value, including: Calculate the power load prediction value for the power load data during holidays and the actual power load value to obtain the deviation rate , as shown in the following formula: ； ； Wherein: only on holidays, = 3; is the predicted value of the power load at the next time node , is the current time node and the th power load change rate from the previous time node, is the power load value at the th time node before, is the initial optimization change rate corresponding to each power load change rate; Calculate the sum of deviation rates at all time nodes within a day Obtain the optimal score , as shown in the following formula: ； Wherein: is the total number of time nodes divided within one day; The optimal score is used to measure the prediction error under the combination of the current power load change rate and the optimized change rate. With the goal of minimizing the prediction error, the particle swarm optimization algorithm is used to optimize the particle positions; Each particle position represents a potential solution, and the particle position contains the power load change rate and the optimized change rate , and the particle position is expressed as ; The particle adjusts its speed according to the current position and the global optimal position, and the formula for updating the particle position is shown as follows: ； Where: represents the velocity of the particle at the th iteration; represents the inertia weight, which controls the global search ability of the particle swarm; and represent the learning factors, which control the learning of the particle from its own and the global optimal positions; and represent random numbers, which ensure the randomness of the search; represents the best position of the particle ; represents the global best position of the particle swarm; After each update of the particle position, the optimal score is used to evaluate the fitness of the particle position; if the position of a certain particle brings a lower error, update the best position of the particle ; Meanwhile, all particles in the particle swarm will update their positions according to the current global optimal solution until the termination condition of iterative optimization is met. When this occurs, the iteration terminates, and the optimal optimization rate corresponding to each power load change rate is output. .

6. The ultra-short-term electric load forecasting method according to claim 5, characterized in that, Match the corresponding power load change rate table according to the average temperature on the prediction date, and combine the optimized change rate to calculate the power load prediction value at future time nodes, including: When the day is not a holiday, match the corresponding non-holiday power load change rate table according to the average temperature of the prediction day to obtain the power load value at the current time node and the power load change rate , calculate the predicted power load value at the next time node as shown in the following formula: ； When the day is a holiday, match the corresponding holiday power load change rate table according to the average temperature of the prediction day to obtain the power load value of the previous time node , the optimal optimization change rate and the power load change rate , calculate the power load prediction value of the next time node as shown in the following formula: 。 7. The ultra-short-term electric load forecasting method according to claim 3, wherein Also included are: After obtaining the actual value of the power load at the future time node, the quartile method is used to verify its rationality. If it is determined to be an outlier, the outlier is replaced with the power load prediction value, and an error message is output to remind manual inspection.

8. An electronic terminal, characterized in that, It includes a processor and a memory connected to the processor. A computer program is stored in the memory. When the computer program is executed by the processor, the steps of the ultra-short-term power load prediction method according to any one of claims 1 to 7 are executed.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, the steps of the ultra-short-term power load prediction method according to any one of claims 1 to 7 are implemented.

Images