Virtual power plant gateway data processing method and system based on edge computing

Abstract

The present invention discloses a virtual power plant gateway data processing method and system based on edge computing, which relates to the field of new energy power generation prediction, and comprises the following steps: obtaining an initial sample set from a database, the initial sample set comprising historical meteorological data and historical power generation; eliminating abnormal data in the initial sample set; calculating meteorological correlation coefficients between historical meteorological data and historical power generation, and dividing historical meteorological data into historical relevant meteorological data and historical irrelevant meteorological data based on the meteorological correlation coefficients; fitting each type of historical relevant meteorological data with historical power generation to obtain m power generation prediction functions, and deploying the m power generation prediction functions on an edge gateway. The present invention is used to solve the problems in the existing new energy power generation prediction technology that the meteorological data selection is accurate and the power generation abnormality judgment link is missing after the prediction, resulting in inaccurate power generation prediction and untimely discovery of power generation abnormalities.

Description

Technical Field

[0001] The present invention relates to the field of renewable energy power generation prediction, and specifically to a virtual power plant gateway data processing method and system based on edge computing. Background Art

[0002] A virtual power plant is a system that uses advanced communication, metering, and control technologies to aggregate distributed energy resources such as renewable energy generation with traditional power generation, energy storage systems, and controllable loads to form a unified, dispatchable energy supply system. This system can optimize resource allocation, improve the stability and economy of grid operation, and participate in the operation of the power market. With the continuous development of the power grid system, there are more and more terminal devices in the power system, and the cloud computing model has large latency and high cloud computing and storage pressure. Edge computing is proposed, which transfers data processing and analysis tasks from the central server to the edge of the network, that is, closer to the data source. This architecture can reduce data transmission delays, improve response speed, and reduce network bandwidth usage.

[0003] Renewable energy power generation will become the main development trend, among which renewable energy power generation includes wind power generation, photovoltaic power generation and hydropower generation. Renewable energy power generation is related to meteorological data, so the power generation can be predicted based on meteorological data. At the same time, the instability of meteorological data also leads to the instability of renewable energy power generation. When there is a small abnormality in renewable energy power generation, it is not easy to find the problem. For example, in the patent application with the authorization announcement number CN114281846B, a new energy power generation prediction method based on machine learning is disclosed. The scheme only considers the prediction of renewable energy power generation under normal power generation conditions. When power generation is abnormal, the prediction method cannot find the power generation abnormality, resulting in a deviation between the predicted power generation and the actual power generation. At the same time, different renewable energy power generation is affected by different meteorological data. If there is no correlation analysis between meteorological data and power generation, the meteorological data selection is inaccurate, resulting in reduced prediction accuracy. The existing new energy power generation prediction technology still has the problem of accurate meteorological data selection and lack of power generation abnormality judgment link after prediction, resulting in inaccurate power generation prediction and untimely discovery of power generation abnormality. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems in the prior art to a certain extent, that is, to solve the problem that the existing renewable energy power generation prediction technology does not perform a fault detection link after the power generation prediction, resulting in inaccurate power generation prediction and untimely detection of power generation anomalies.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a virtual power plant gateway data processing method based on edge computing, comprising the following steps:

[0006] Acquire an initial sample set from a database, the initial sample set including historical meteorological data and historical power generation;

[0007] The abnormal data in the initial sample set is eliminated; the meteorological correlation coefficient between the historical meteorological data and the historical power generation is calculated, and the historical meteorological data is divided into historically relevant meteorological data and historically irrelevant meteorological data based on the meteorological correlation coefficient;

[0008] Each type of historical meteorological data is fitted with the historical power generation to obtain m power generation prediction functions, and the m power generation prediction functions are deployed on the edge gateway, where m is the total number of types of historical meteorological data;

[0009] Based on the historical relevant meteorological data, the corresponding real-time relevant meteorological data is obtained to obtain the real-time power generation, the real-time relevant meteorological data is input into the corresponding power generation prediction function to obtain m function power generation, the m function power generation is weighted and summed to obtain the real-time predicted power generation, and the power generation is judged to be abnormal according to the difference between the real-time predicted power generation and the real-time power generation. If the power generation is abnormal, an abnormal signal is sent to the processing end;

[0010] The processing end determines whether the abnormal signal is accurate, and if not, updates the power generation prediction function.

[0011] Furthermore, removing abnormal data from the initial sample set includes the following sub-steps:

[0012] Obtain historical power generation and historical meteorological data once every first time interval, and mark the total amount of historical power generation obtained as A;

[0013] Sort the historical power generation from small to large, and set a sequence number corresponding to each historical power generation, which is an integer starting from 1;

[0014] Determine whether a1*A is an integer. If it is an integer, set the historical power generation with the serial number a1*A as the first quantile Q1. If it is not an integer, calculate the average value of the historical power generation corresponding to the serial numbers on both sides of a1*A and mark it as the first quantile Q1; where a1 is the first quantile ratio;

[0015] Determine whether a2*A is an integer. If it is an integer, set the historical power generation with the serial number a2*A as the second quantile Q2. If it is not an integer, calculate the average value of the historical power generation corresponding to the serial numbers on both sides of a2*A and mark it as the second quantile Q2; where a2 is the second quantile ratio, a2=(1 / a1-1)*a1;

[0016] Set q1=Q1-b*(Q2-Q1), q2=Q2+b*(Q2-Q1); where q1 is the lower threshold, q2 is the upper threshold, and b is the range constant;

[0017] The historical power generation power less than q1 and greater than q2 and the corresponding historical meteorological data are deleted.

[0018] Furthermore, the calculation of the meteorological correlation coefficient between the historical meteorological data and the historical power generation includes the following sub-steps:

[0019] The total number of historical power generation after statistical deletion is n;

[0020] Historical meteorological data include: Solar radiation Xty i 、Temperature Xwd i 、Air humidity Xyc i 、Air density Xmd i 、Precipitation Xjs i 、Snowfall Xjx i and wind speed Xfs i , the historical power generation is Y i ;

[0021] The average value of each historical meteorological data is , , , , , as well as The historical average power generation is ;

[0022] The meteorological correlation coefficient formula is used to obtain the meteorological correlation coefficient of each historical meteorological data. The meteorological correlation coefficient formula is as follows:

[0023] ;

[0024] Where Xn i Xty i 、Xwd i ,Xyc i 、Xmd i 、Xjs i 、Xjx i and Xfs iCn represents one of Cty, Cwd, Cyc, Cmd, Cjs, Cjx and Cfs, wherein Cty is the meteorological correlation coefficient of solar radiation, Cwd is the meteorological correlation coefficient of temperature, Cyc is the meteorological correlation coefficient of air humidity, Cmd is the meteorological correlation coefficient of air density, Cjs is the meteorological correlation coefficient of precipitation, Cjx is the meteorological correlation coefficient of snowfall, and Cfs is the meteorological correlation coefficient of wind speed; express , , , , , as well as One of;

[0025] After sorting the absolute values ​​of each meteorological correlation coefficient from large to small, the historical meteorological data corresponding to the meteorological correlation coefficient of the first proportion before sorting are marked as historically relevant meteorological data, and the historical meteorological data excluding the historically relevant meteorological data are marked as historically irrelevant meteorological data.

[0026] Furthermore, fitting and calculating each type of historical meteorological data with the historical power generation to obtain m power generation prediction functions includes the following sub-steps:

[0027] Get the total number of types m of historical meteorological data, set each historical meteorological data to correspond to a type number, and the type number is an integer from 1 to m;

[0028] All historical relevant meteorological data are normalized using the max-min normalization method and marked as historical normalized meteorological data;

[0029] The size of each historical normalized meteorological data is used as the X-axis data, and the power generation power is used as the Y-axis data to establish a plane rectangular coordinate system, which is marked as the prediction coordinate system;

[0030] Plot the historical normalized meteorological data and the power generation into the corresponding prediction coordinate system respectively to obtain m related scatter plots;

[0031] According to the distribution of each relevant scatter plot, appropriate preset power generation prediction functions are selected respectively, and the coefficients in the preset power generation prediction function are calculated by fitting the relevant scatter plots respectively, so as to obtain m power generation prediction functions.

[0032] Furthermore, based on the historical relevant meteorological data, corresponding real-time relevant meteorological data are obtained to obtain real-time power generation, the real-time relevant meteorological data are input into the corresponding power generation prediction function to obtain m function power generation, and the m function power generation is weighted and summed to obtain the real-time predicted power generation, including the following sub-steps:

[0033] Obtain real-time relevant meteorological data and real-time power generation Ws;

[0034] The real-time related meteorological data are normalized using the max-min normalization method and marked as real-time normalized meteorological data;

[0035] Each type of real-time normalized meteorological data is input into the corresponding power generation prediction function to obtain the function power generation F i ; Where i is the type number of real-time relevant meteorological data from 1 to m;

[0036] Find F i The weight ratio k i ;

[0037] The real-time predicted power generation is obtained as:

[0038] ; Where F is the real-time predicted power generation.

[0039] Furthermore, we can obtain F i The weight ratio k i The following sub-steps are included:

[0040] Substitute all relevant historical meteorological data into the corresponding power generation prediction function to obtain m groups of historical predicted power generation Yx i ; Where i is the historical meteorological data from 1 to n;

[0041] Find each group of Yx i With Y i The root mean square error between them is:

[0042] ; where e i represents one of the root mean square errors of m types of power generation prediction functions;

[0043] F i According to e i Arrange the values ​​from large to small, and then i Arrange from small to large and mark it as u i , F i The corresponding e i Replace with u in order i , find F i The weight ratio is:

[0044] .

[0045] Further, judging whether power generation is abnormal based on the difference between the real-time predicted power generation and the real-time power generation, and if the power generation is abnormal, sending an abnormal signal to the processing end includes the following sub-steps:

[0046] The difference between F and Ws is: z = F-Ws; where z is the difference between F and Ws;

[0047] Obtain the negative threshold t1 and positive threshold t2 of the difference z between F and Ws;

[0048] Determine whether z satisfies the inequality: t1 <z<t2;

[0049] If not satisfied, an exception signal is sent to the processing end.

[0050] Furthermore, obtaining the negative threshold value t1 and the positive threshold value t2 of the difference z between F and Ws includes the following sub-steps:

[0051] A first amount of historical relevant meteorological data and historical power generation is obtained, and simulation training is performed on the historical relevant meteorological data and historical power generation, which is marked as threshold simulation training. T1 and t2 are obtained through the threshold simulation training. The threshold simulation training includes:

[0052] Substitute historical relevant meteorological data into m power generation prediction functions to obtain m training function power generation, and then perform weighted summation of the m training function power generation to obtain the training prediction power generation b;

[0053] When the historical power generation corresponding to the historical meteorological data is in a normal state, the historical power generation in the normal state is marked as the training normal power generation o;

[0054] The normal power generation difference is calculated as: Zz=bo; where Zz is the normal power generation difference;

[0055] When the historical power generation corresponding to the historical meteorological data is in an abnormal state, the abnormal historical power generation is marked as training abnormal power generation p;

[0056] The abnormal power generation difference is obtained as follows: Zy=bp; where Zy is the abnormal power generation difference;

[0057] Establish a normal histogram and an abnormal histogram, and the horizontal axes of the normal histogram and the abnormal histogram are divided into u intervals;

[0058] Count the frequency of normal power generation difference in each interval respectively;

[0059] The normal histogram is drawn with the normal power generation difference as the x-axis and the frequency as the y-axis;

[0060] Count the frequency of abnormal power generation difference in each interval respectively;

[0061] The abnormal histogram is drawn with the abnormal power generation difference as the x-axis and the frequency as the y-axis;

[0062] Obtain the interval with the largest total number of normal power generation differences and the interval with the largest total number of abnormal power generation differences on the negative half axis of the x-axis, and mark them as the negative axis normal interval and the negative axis abnormal interval respectively;

[0063] Calculate the mean of the normal interval of the negative axis and the abnormal interval of the negative axis, and mark them as the normal value of the negative axis tz1 and the abnormal value of the negative axis ty1 respectively;

[0064] The negative threshold is: t1=(tz1+ty1) / 2;

[0065] Obtain the interval with the largest total number of normal power generation differences and the interval with the largest total number of abnormal power generation differences on the positive half axis of the x-axis, and mark them as positive axis normal value tz2 and positive axis abnormal value ty2 respectively;

[0066] Calculate the mean of the negative axis normal interval and the negative axis abnormal interval, and mark them as the positive axis normal value tz2 and the positive axis abnormal value ty2 respectively;

[0067] The positive threshold is: t2=(tz2+ty2) / 2;

[0068] The newly added first amount of historical relevant meteorological data and the historical power generation input threshold are simulated and trained to update t1 and t2.

[0069] Furthermore, the processing end determines whether the abnormal signal is accurate. If it is inaccurate, updating the power generation prediction function includes the following sub-steps:

[0070] The processing end determines whether the abnormal signal is correct. If the abnormal signal is incorrect, a second amount of historical relevant meteorological data and historical power generation are added to re-fit the calculation to obtain an updated power generation prediction function, and the power generation prediction function of the edge gateway is replaced with the updated power generation prediction function.

[0071] In a second aspect, the present invention provides a virtual power plant gateway data processing system based on edge computing, including: a data acquisition module, a data processing module, a function fitting module, an anomaly detection module and a function update module;

[0072] The data acquisition module is used to acquire an initial sample set from a database, the initial sample set including historical meteorological data and historical power generation;

[0073] The data processing module is used to remove abnormal data in the initial sample set, calculate the meteorological correlation coefficient between the historical meteorological data and the historical power generation, and divide the historical meteorological data into historically relevant meteorological data and historically irrelevant meteorological data based on the meteorological correlation coefficient;

[0074] The function fitting module is used to fit each type of historical relevant meteorological data with the historical power generation, obtain m power generation prediction functions, and deploy the m power generation prediction functions on the edge gateway, where m is the total number of types of historical relevant meteorological data;

[0075] The anomaly detection module is used to obtain corresponding real-time related meteorological data based on historical related meteorological data, obtain real-time power generation, input the real-time related meteorological data into the corresponding power generation prediction function, obtain m function power generation, perform weighted summation on the m function power generation to obtain the real-time predicted power generation, and judge whether power generation is abnormal according to the difference between the real-time predicted power generation and the real-time power generation. If power generation is abnormal, an abnormal signal is sent to the processing end;

[0076] The function updating module is used by the processing end to determine whether the abnormal signal is accurate, and if it is inaccurate, to update the power generation prediction function.

[0077] Beneficial effects of the present invention: The present invention obtains m power generation prediction functions by fitting each type of historical relevant meteorological data with the historical power generation. The advantage is that, because different new energy power generation may be affected by different meteorological data, the historical meteorological data and the power generation are correlated before the fitting calculation. The power generation prediction function fitted with historical meteorological data with strong correlation is more accurate. The m power generation prediction functions predict m function power generation, and the m function power generation powers are weighted and summed to obtain the real-time predicted power generation, making the predicted data more accurate. After the prediction, power generation abnormality judgment is added, which can promptly detect the deviation between the predicted power generation and the actual power generation.

[0078] The present invention obtains t1 and t2 through threshold simulation training. The advantage is that simulation training is performed by training relevant meteorological data, training normal power generation and training abnormal power generation, and t1 and t2 are calculated through a large amount of historical data, so that t1 and t2 are more universal and accurate, and the threshold simulation training data is updateable, so that t1 and t2 obtained by training are more accurate, thereby improving the calculation accuracy of the threshold simulation training. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 is a functional block diagram of the system of the present invention;

[0080] Figure 2 It is a wind speed scatter diagram of the present invention;

[0081] Figure 3 is an air density scatter plot of the present invention;

[0082] Figure 4 is a normal histogram of the present invention;

[0083] Figure 5 is an abnormal histogram of the present invention;

[0084] Figure 6 A flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0085] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0086] Example 1, please refer to Figure 1 As shown, in the first aspect, the present application provides a virtual power plant gateway data processing system based on edge computing, including a data acquisition module, a data processing module, a function fitting module, an anomaly detection module and a function update module;

[0087] The data acquisition module is used to obtain an initial sample set from a database, where the initial sample set includes historical meteorological data and historical power generation;

[0088] The data processing module is used to remove abnormal data from the initial sample set, calculate the meteorological correlation coefficient between the historical meteorological data and the historical power generation, and divide the historical meteorological data into historically relevant meteorological data and historically irrelevant meteorological data based on the meteorological correlation coefficient. The data processing module includes an abnormal data removal unit and a correlation analysis unit;

[0089] The abnormal data elimination unit is configured with an abnormal data elimination strategy, which includes: obtaining historical power generation and historical meteorological data once every first time interval, marking the total amount of historical power generation obtained as A; the first time is set to one hour;

[0090] Sort the historical power generation from small to large, and set a serial number corresponding to each historical power generation, where the serial number is an integer starting from 1; the setting of the serial number facilitates the acquisition and calculation of the historical power generation;

[0091] Determine whether a1*A is an integer. If it is an integer, set the historical power generation with the serial number a1*A as the first quantile Q1. If it is not an integer, find the average value of the historical power generation corresponding to the serial numbers on both sides of a1*A and mark it as the first quantile Q1; where a1 is the first quantile ratio; the first quantile ratio setting can be set according to the distribution range of the historical power generation. If the distribution of the historical power generation is relatively concentrated, a1 can be appropriately increased; if the distribution of the historical power generation is relatively dispersed, a1 can be appropriately reduced. The setting range of the first quantile ratio is between 0 and 0.5, and 0.25 is preferred;

[0092] Determine whether a2*A is an integer. If it is an integer, set the historical power generation with the serial number a2*A as the second quantile Q2. If it is not an integer, calculate the average value of the historical power generation corresponding to the serial numbers on both sides of a2*A and mark it as the second quantile Q2; where a2 is the second quantile ratio, a2=(1 / a1-1)*a1;

[0093] Set q1=Q1-b*(Q2-Q1), q2=Q2+b*(Q2-Q1); q1 is the lower threshold, q2 is the upper threshold, and b is the range constant; the setting of b can further adjust the upper and lower thresholds according to the discrete degree of historical power generation distribution. If the discrete degree is large, the range between the upper and lower thresholds can be appropriately increased to filter out excessive and small power generation; the overall method is to determine the power generation value in the middle between Q1 and Q2, and then subtract the range between Q1 and Q2 from Q1 to get the smaller power generation, and then add the middle distribution value range multiplied by the value of b to get the larger power generation. Deleting the smaller power generation and the larger power generation facilitates subsequent calculations;

[0094] Delete the historical power generation less than q1 and greater than q2 and the corresponding historical meteorological data, and delete the abnormal data that is too large or too small to make subsequent calculations more accurate;

[0095] In practical applications, if the total number of historical power generation A obtained is 1000, the historical power generation is sorted from small to large, and each historical power generation is set to correspond to a serial number, which is an integer starting from 1; where a1 is 0.25, a2=(1 / a1-1)*a1=0.75, then a1*A=1000*0.25=250, and the historical power generation of 1 million kWh with serial number 250 is obtained as the first quantile Q 1, a2*A=1000*0.75=750, then the historical power generation of 1.6 million kWh with serial number 750 is obtained as the second quantile Q2, when the b range constant is set to 1.5, q1=Q1-b*(Q2-Q1)=300,000 kWh, q2=Q2+b*(Q2-Q1)=2.5 million kWh, then the historical power generation less than 300,000 kWh and greater than 2.5 million kWh and the corresponding historical meteorological data are deleted.

[0096] The correlation analysis unit is configured with a correlation analysis strategy, which includes: the total number of historical power generation after statistical deletion is n;

[0097] Historical meteorological data include: Solar radiation Xty i 、Temperature Xwd i 、Air humidity Xyc i 、Air density Xmd i 、Precipitation Xjs i 、Snowfall Xjx i and wind speed Xfs i , the historical power generation is Y i ;

[0098] In the specific application, historical meteorological data include more than the above seven types of historical meteorological data. For the convenience of calculation, the historical meteorological data are unified in units, such as solar radiation Xty i The unit is kWh / m², temperature Xwd i The unit is °C, air humidity Xyc i The unit is g / m³, air density Xmd i Unit is kg / m³, precipitation Xjs i Unit: mm, snowfall amount Xjx i The unit is mm, wind speed Xfs i The unit is km / h, the historical power generation is Y i The unit is 10,000 kWh, where the temperature, air humidity and air density are the average values ​​within one hour, and the precipitation and snowfall are the total amounts within one hour;

[0099] The average value of each historical meteorological data is , , , , , as well as The historical average power generation is ;

[0100] The meteorological correlation coefficient formula is used to obtain the meteorological correlation coefficient of each historical meteorological data. The meteorological correlation coefficient formula is as follows:

[0101] ;

[0102] Where Xn i Xty i 、Xwd i ,Xyc i 、Xmd i 、Xjs i 、Xjx i and Xfs i Cn represents one of Cty, Cwd, Cyc, Cmd, Cjs, Cjx and Cfs, wherein Cty is the meteorological correlation coefficient of solar radiation, Cwd is the meteorological correlation coefficient of temperature, Cyc is the meteorological correlation coefficient of air humidity, Cmd is the meteorological correlation coefficient of air density, Cjs is the meteorological correlation coefficient of precipitation, Cjx is the meteorological correlation coefficient of snowfall, and Cfs is the meteorological correlation coefficient of wind speed; express , , , , , as well as One of;

[0103] After sorting the absolute values ​​of each meteorological correlation coefficient from large to small, the historical meteorological data corresponding to the meteorological correlation coefficient of the first proportion before sorting is marked as historically relevant meteorological data, and the historical meteorological data excluding the historically relevant meteorological data is marked as historically irrelevant meteorological data. The setting of the first proportion can be combined with the number of historical meteorological data types and the number of historical meteorological data types with higher correlation coefficients;

[0104] In practical applications, taking a wind power plant as an example, each historical meteorological data and historical power generation are calculated, and the correlation coefficient between wind speed and historical power generation is calculated as an example to calculate the average wind speed. The speed is 20km / h, and the average historical power generation is calculated as 1.3 million kilowatts, and then Cfs=0.852 is obtained through the meteorological correlation coefficient formula, where the correlation coefficient calculation result retains three decimal places, and then all Cn are calculated by calculating Cfs as an example, and the calculation results are as follows: Cty=-0.125, Cwd=0.198, Cyc=0.162, Cmd=0.251, Cjs=0.021, Cjx=-0.052, and the absolute values ​​of all Cn are sorted as 0.852, 0.251, 0.198, 0.162, 0.125, 0.052, 0.021; the first ratio is set to 2 / 7, and the historical meteorological data corresponding to the meteorological correlation coefficient of the first ratio before sorting is marked as historical relevant meteorological data. The historical relevant meteorological data are wind speed and air density, and the historical irrelevant meteorological data are solar radiation, air humidity, temperature, precipitation and snowfall.

[0105] The function fitting module is used to fit each type of historical meteorological data with the historical power generation, obtain m power generation prediction functions, and deploy the m power generation prediction functions on the edge gateway, where m is the total number of types of historical meteorological data;

[0106] The function fitting module includes obtaining m power generation prediction function strategies, and obtaining m power generation prediction function strategies includes:

[0107] Get the total number of types m of historical meteorological data, set each historical meteorological data to correspond to a type number, and the type number is an integer from 1 to m;

[0108] All historical relevant meteorological data are normalized using the max-min normalization method and marked as historical normalized meteorological data;

[0109] The size of each historical normalized meteorological data is used as the X-axis data, and the power generation power is used as the Y-axis data to establish a plane rectangular coordinate system, which is marked as the prediction coordinate system;

[0110] Plot the historical normalized meteorological data and the power generation into the corresponding prediction coordinate system respectively to obtain m related scatter plots;

[0111] According to the distribution of each relevant scatter plot, appropriate preset power generation prediction functions are selected respectively, and the coefficients in the preset power generation prediction function are calculated by fitting the relevant scatter plots respectively, so as to obtain m power generation prediction functions.

[0112] In practical applications, when obtaining a kind of historical meteorological data, it is necessary to ensure that the rest of the historical meteorological data is around the mean value to prevent the influence of the rest of the historical meteorological data on the power generation. The relevant meteorological data obtained according to the first ratio are wind speed and air density. Therefore, m is 2, the wind speed is set to the corresponding type number 1, and the air density is set to the corresponding type number 2. The wind speed and air density are normalized by the max-min normalization method respectively. The calculation formula of the wind speed max-min normalization method is taken as an example: ; Among them Xfsg i is the normalized wind speed, Xfs max is the historical maximum wind speed, Xfs min is the minimum historical wind speed; if Xfs2 is 20km / h, Xfs max 50km / h, Xfs min is 2km / h, and by calculation we get Xfsg2=18 / 48=0.357, and all Xfsg i , similarly calculate all normalized air densities Xmdg i , respectively taking the normalized wind speed and normalized air density as the X-axis data, and the generated power as the Y-axis data, to establish a plane rectangular coordinate system, marked as the prediction coordinate system;

[0113] See also Figure 2 and Figure 3 As shown in the figure, the historical normalized meteorological data and the power generation are plotted into the corresponding prediction coordinate system to obtain the wind speed scatter plot and the air density scatter plot; according to the wind speed scatter plot and the air density scatter plot, the preset power generation prediction function for each wind speed is determined to be F1=h1*X1 2 +h2, where h1 and h2 are coefficients. The preset power generation prediction function of air density is F2=h3*X2+h4, where h3 and h4 are coefficients. The preset power generation prediction function is fitted with relevant scatter plots to obtain the wind speed fitting function and the air density fitting function. According to the wind speed fitting function and the air density fitting function, the corresponding coefficients h1=182.35, h2=3.52, h3=198.25, h4=-18.82 are obtained. Substituting all coefficients into the corresponding formula, the power generation prediction function is F1=182.35*X1 2 +3.52, F2=198.25*X2-18.82.

[0114] The anomaly detection module is used to obtain the corresponding real-time relevant meteorological data based on the historical relevant meteorological data, obtain the real-time power generation, input the real-time relevant meteorological data into the corresponding power generation prediction function, obtain m function power generation, perform weighted summation on the m function power generation to obtain the real-time predicted power generation, and judge whether the power generation is abnormal according to the difference between the real-time predicted power generation and the real-time power generation. If the power generation is abnormal, an abnormal signal is sent to the processing end;

[0115] The anomaly detection module includes a strategy for obtaining real-time predicted power generation, which includes: obtaining real-time relevant meteorological data and real-time power generation Ws;

[0116] The real-time relevant meteorological data are normalized using the max-min normalization method and marked as real-time normalized meteorological data; the max-min normalization method uses the maximum and minimum values ​​of the historical relevant meteorological data for calculation;

[0117] Each type of real-time normalized meteorological data is input into the corresponding power generation prediction function to obtain the function power generation F i ; Where i is the type number of real-time relevant meteorological data from 1 to m;

[0118] Find F i The weight ratio k i ;

[0119] The real-time predicted power generation is obtained as:

[0120] ; Where F is the real-time predicted power generation. Weighted summation of each power generation prediction function makes the data more accurate;

[0121] In practical applications, the real-time wind speed and real-time air density are obtained, and the real-time wind speed and real-time air density are normalized to 0.85 and 0.75 respectively, and substituted into the formula F1=182.35*X1 2 +3.52, F2=198.25*X2-18.82, then F1=135.27, F2=129.87, according to the obtained k1 and k2, F=k1*F1+k2*F2=135.27*0.75+129.87*0.25=133.92.

[0122] The anomaly detection module includes obtaining F i The weight ratio k i Strategy, find F i The weight ratio k i The strategy includes: Substituting all relevant historical meteorological data into the corresponding power generation prediction function to obtain m groups of historical predicted power generation Yx i ; Where i is the historical meteorological data from 1 to n;

[0123] Find each group of Yx i With Y i The root mean square error between them is:

[0124] ; where e i represents one of the root mean square errors of m types of power generation prediction functions;

[0125] F i According to e i Arrange the values ​​from large to small, and then i Arrange from small to large and mark it as u i , F i The corresponding e i Replace with u in order i , find F i The weight ratio is:

[0126] Among them, F i The weight ratio k i , while F i The smaller the error, the better, so e i The smaller the power generation prediction function, the more accurate the prediction, and the weight proportion should be increased. Therefore, the errors are sorted and replaced to make the weight proportion of the calculated small error the largest.

[0127] In practical applications, all historical wind speeds and historical air densities are substituted into the corresponding formula F1=182.35*X1 2 +3.52 and F2=198.25*X2-18.82, and get 2 sets of historical predicted power generation Yx i , find each group of Yx i With the historical power generation Y i The root mean square error between them is e1=2, e2=6, u1=6, u2=2, so we get k1=u1 / (e1+e2)=6 / 8=0.75 and k2=u2 / (e1+e2)=2 / 8=0.25.

[0128] The abnormality detection module includes a strategy for determining whether power generation is abnormal. The strategy for determining whether power generation is abnormal includes: obtaining the difference between F and Ws as: z=F-Ws; wherein z is the difference between F and Ws;

[0129] Obtain the negative threshold t1 and positive threshold t2 of the difference z between F and Ws;

[0130] Determine whether z satisfies the inequality: t1 <z<t2;

[0131] If not, an abnormal signal is sent to the processing end; when the real-time power generation differs too much from the predicted power generation, it can be considered that there is damage to the machine or abnormality in the power plant, resulting in a large difference between the real-time power generation and the predicted power generation, so an abnormal signal is sent to the processing end;

[0132] In practical applications, the real-time power generation Ws=131.25, z=F-Ws=133.92-131.25=2.67, the negative threshold t1 of the difference z between F and Ws is -4.5 and the positive threshold t2 is 5; determine whether z satisfies the inequality: -4.5<2.67<5; if so, no signal needs to be sent to the processing end.

[0133] The anomaly detection module includes a strategy for obtaining a negative threshold value t1 and a positive threshold value t2 of the difference z between F and Ws. The strategy for obtaining a negative threshold value t1 and a positive threshold value t2 of the difference z between F and Ws includes: obtaining a first amount of historical relevant meteorological data and historical power generation, performing simulation training on the historical relevant meteorological data and historical power generation, marking them as threshold simulation training, obtaining t1 and t2 through threshold simulation training, and the threshold simulation training includes:

[0134] Substitute historical relevant meteorological data into m power generation prediction functions to obtain m training function power generation, and then perform weighted summation of the m training function power generation to obtain the training prediction power generation b;

[0135] When the historical power generation corresponding to the historical meteorological data is in a normal state, the historical power generation in the normal state is marked as the training normal power generation o;

[0136] The normal power generation difference is calculated as: Zz=bo; where Zz is the normal power generation difference;

[0137] When the historical power generation corresponding to the historical meteorological data is in an abnormal state, the abnormal historical power generation is marked as training abnormal power generation p;

[0138] The abnormal power generation difference is obtained as follows: Zy=bp; where Zy is the abnormal power generation difference;

[0139] See also Figure 4 and Figure 5 , establish a normal histogram and an abnormal histogram, and the horizontal axes of the normal histogram and the abnormal histogram are divided into u intervals;

[0140] Count the frequency of normal power generation difference in each interval respectively;

[0141] The normal histogram is drawn with the normal power generation difference as the x-axis and the frequency as the y-axis;

[0142] Count the frequency of abnormal power generation difference in each interval respectively;

[0143] The abnormal histogram is drawn with the abnormal power generation difference as the x-axis and the frequency as the y-axis;

[0144] The normal histogram and the abnormal histogram are drawn to facilitate observation of the distribution of the normal power generation difference and the abnormal power generation difference, and to facilitate finding the negative threshold t1 and the positive threshold t2;

[0145] Obtain the interval with the largest total number of normal power generation differences and the interval with the largest total number of abnormal power generation differences on the negative half axis of the x-axis, and mark them as the negative axis normal interval and the negative axis abnormal interval respectively;

[0146] Calculate the mean of the normal interval of the negative axis and the abnormal interval of the negative axis, and mark them as the normal value of the negative axis tz1 and the abnormal value of the negative axis ty1 respectively;

[0147] The negative threshold is: t1=(tz1+ty1) / 2;

[0148] Obtain the interval with the largest total number of normal power generation differences and the interval with the largest total number of abnormal power generation differences on the positive half axis of the x-axis, and mark them as positive axis normal value tz2 and positive axis abnormal value ty2 respectively;

[0149] Calculate the mean of the negative axis normal interval and the negative axis abnormal interval, and mark them as the positive axis normal value tz2 and the positive axis abnormal value ty2 respectively;

[0150] The positive threshold is: t2=(tz2+ty2) / 2;

[0151] Simulate the training of the newly added first quantity of historical meteorological data and historical power generation input thresholds to update t1 and t2. By continuously updating t1 and t2, t1 and t2 are made more accurate;

[0152] In practical applications, the interval u between the normal power generation difference and the normal power generation difference is 20. If t1 and t2 are more precise, the interval u can be appropriately increased. Please refer to Figure 4 As shown, a normal histogram is drawn, see Figure 5 As shown, draw an abnormal histogram, observe the two histograms and find that the normal value tz1 of the negative axis is -1.5 and the abnormal value ty1 of the negative axis is -7.5, and the negative threshold is: t1=(tz1+ty1) / 2=-4.5, observe the two histograms and find that the normal value tz2 of the positive axis is 1.5 and the abnormal value ty2 of the positive axis is 8.5, and the positive threshold is: t2=(tz2+ty2) / 2=5.

[0153] The function update module is used by the processing end to determine whether the abnormal signal is accurate. If it is not accurate, the power generation prediction function is updated;

[0154] The function update module includes a strategy for updating the power generation prediction function, and the strategy for updating the power generation prediction function includes:

[0155] The processing end determines whether the abnormal signal is correct. If the abnormal signal is incorrect, a second amount of historical relevant meteorological data and historical power generation are added to re-fit the calculation to obtain an updated power generation prediction function, and the power generation prediction function of the edge gateway is replaced with the updated power generation prediction function.

[0156] In actual applications, the processing end can be the power plant maintenance personnel. When an abnormal signal comes, the power plant maintenance personnel determine whether there is an abnormality in the power plant. If no abnormality is found, it means that the abnormal signal is an error signal and the electric power prediction function is inaccurate. Therefore, the electric power prediction function needs to be updated.

[0157] Example 2, please refer to Figure 6 As shown, the virtual power plant gateway data processing method based on edge computing includes the following steps: Step S1, obtaining an initial sample set from a database, the initial sample set including historical meteorological data and historical power generation power.

[0158] Step S2, remove abnormal data in the initial sample set; calculate the meteorological correlation coefficient between the historical meteorological data and the historical power generation, and divide the historical meteorological data into historically relevant meteorological data and historically irrelevant meteorological data based on the meteorological correlation coefficient. Step S2 includes the following sub-steps:

[0159] Step S21, removing abnormal data from the initial sample set, step S21 includes the following sub-steps: acquiring historical power generation and historical meteorological data once every first time interval, marking the total amount of historical power generation acquired as A;

[0160] Step S2101, sort the historical power generation from small to large, and set a sequence number corresponding to each historical power generation, where the sequence number is an integer starting from 1;

[0161] Step S2102, determine whether a1*A is an integer. If it is an integer, set the historical power generation with the serial number a1*A as the first quantile Q1. If it is not an integer, calculate the average value of the historical power generation corresponding to the serial numbers on both sides of a1*A and mark it as the first quantile Q1; where a1 is the first quantile ratio;

[0162] Step S2103, determine whether a2*A is an integer. If it is an integer, set the historical power generation with the serial number a2*A as the second quantile Q2. If it is not an integer, calculate the average value of the historical power generation corresponding to the serial numbers on both sides of a2*A and mark it as the second quantile Q2; where a2 is the second quantile ratio, a2=(1 / a1-1)*a1;

[0163] Step S2104, set q1=Q1-b*(Q2-Q1), q2=Q2+b*(Q2-Q1); where q1 is the lower threshold, q2 is the upper threshold, and b is the range constant;

[0164] Step S2105, deleting the historical power generation power less than q1 and greater than q2 and the corresponding historical meteorological data.

[0165] Step S22, calculating the meteorological correlation coefficient between the historical meteorological data and the historical power generation, and dividing the historical meteorological data into historically relevant meteorological data and historically irrelevant meteorological data based on the meteorological correlation coefficient, step S22 includes the following sub-steps: the total number of historical power generation after statistical deletion is n;

[0166] Step S2201, historical meteorological data includes: solar radiation Xty i 、Temperature Xwd i 、Air humidity Xyc i 、Air density Xmd i 、Precipitation Xjs i 、Snowfall Xjx i and wind speed Xfs i , the historical power generation is Y i ;

[0167] Step S2202, calculate the average value of each historical meteorological data , , , , , as well as The historical average power generation is ;

[0168] Step S2203, the meteorological correlation coefficient of each historical meteorological data is obtained by using the meteorological correlation coefficient formula, and the meteorological correlation coefficient formula is as follows:

[0169] ;

[0170] Step S2204, after sorting from large to small according to the absolute value of each meteorological correlation coefficient, mark the historical meteorological data corresponding to the meteorological correlation coefficient of the first proportion before sorting as historically relevant meteorological data, and mark the historical meteorological data excluding the historically relevant meteorological data as historically irrelevant meteorological data.

[0171] Step S3, each type of historical meteorological data is fitted with the historical power generation to obtain m power generation prediction functions, and the m power generation prediction functions are deployed on the edge gateway, where m is the total number of types of historical meteorological data. Step S3 includes the following sub-steps:

[0172] Step S301, obtaining the total number m of types of historical relevant meteorological data, setting each historical relevant meteorological data to correspond to a type number, where the type number is an integer from 1 to m;

[0173] Step S302, all historical relevant meteorological data are normalized using the max-min normalization method and marked as historical normalized meteorological data;

[0174] Step S303, using the size of each historical normalized meteorological data as X-axis data and the power generation power as Y-axis data, respectively, to establish a plane rectangular coordinate system, marked as a prediction coordinate system;

[0175] Step S303, respectively plotting the historical normalized meteorological data and the power generation into the corresponding prediction coordinate system to obtain m related scatter plots;

[0176] Step S304, according to the distribution of each relevant scatter plot, respectively select appropriate preset power generation prediction functions, respectively use the relevant scatter plot fitting to calculate the coefficients in the preset power generation prediction function, and obtain m power generation prediction functions.

[0177] Step S4, based on the historical relevant meteorological data, corresponding real-time relevant meteorological data is obtained to obtain real-time power generation, the real-time relevant meteorological data is input into the corresponding power generation prediction function to obtain m function power generation, the m function power generation is weighted and summed to obtain the real-time predicted power generation, and the power generation is judged to be abnormal according to the difference between the real-time predicted power generation and the real-time power generation. If the power generation is abnormal, an abnormal signal is sent to the processing end. Step S4 includes the following sub-steps:

[0178] Step S401, obtaining real-time relevant meteorological data and real-time power generation Ws;

[0179] Step S402, normalizing the real-time related meteorological data using the max-min normalization method and marking it as real-time normalized meteorological data;

[0180] Step S403: Input each type of real-time normalized meteorological data into the corresponding power generation prediction function to obtain the function power generation F i ; Where i is the type number of real-time relevant meteorological data from 1 to m;

[0181] Step S404, obtain F i The weight ratio k i ;

[0182] Step S404 includes the following sub-steps:

[0183] Step S4041: Substitute all relevant historical meteorological data into the corresponding power generation prediction function to obtain m groups of historical predicted power generation Yx i ; Where i is the historical meteorological data from 1 to n;

[0184] Step S4042, find each group of Yx i With Y i The root mean square error between is:

[0185] ;

[0186] Step S4043: F i According to e i Arrange the values ​​from large to small, and then i Arrange from small to large and mark it as u i , F i The corresponding e i Replace with u in order i , find F i The weight ratio is:

[0187] .

[0188] Step S405, obtaining the real-time predicted power generation:

[0189] ; Where F is the real-time predicted power generation.

[0190] Step S406, the difference between F and Ws is calculated as: z=F-Ws; wherein z is the difference between F and Ws;

[0191] Step S407, obtaining a negative threshold value t1 and a positive threshold value t2 of the difference z between F and Ws;

[0192] Step S407 includes the following sub-steps:

[0193] Step S4071, respectively substitute the historical relevant meteorological data into m power generation prediction functions to obtain m training function power generation, and then perform weighted summation of the m training function power generation to obtain the training predicted power generation b;

[0194] Step S4072, when the historical power generation corresponding to the historical meteorological data is in a normal state, the historical power generation in the normal state is marked as the training normal power generation o; the normal power generation difference is obtained as: Zz=bo; wherein Zz is the normal power generation difference;

[0195] Step S4073: When the historical power generation corresponding to the historical relevant meteorological data is in an abnormal state, mark the historical power generation in the abnormal state as the training abnormal power generation p; calculate the abnormal power generation difference as: Zy = b - p; where Zy is the abnormal power generation difference.

[0196] Step S4074: Establish a normal histogram and an abnormal histogram. The horizontal axes of both the normal histogram and the abnormal histogram are divided into u intervals.

[0197] Step S4075: Count the frequencies of the normal power generation differences in each interval respectively; use the normal power generation difference as the x-axis and the frequency as the y-axis for the normal histogram, and draw the normal histogram.

[0198] Step S4075: Count the frequencies of the abnormal power generation differences in each interval respectively; use the abnormal power generation difference as the x-axis and the frequency as the y-axis for the abnormal histogram, and draw the abnormal histogram.

[0199] Step S4076: Obtain the intervals with the largest total number of normal power generation differences and the largest total number of abnormal power generation differences on the negative x-axis respectively, and mark them as the negative-axis normal interval and the negative-axis abnormal interval; calculate the means of the negative-axis normal interval and the negative-axis abnormal interval respectively, and mark them as the negative-axis normal value tz1 and the negative-axis abnormal value ty1; calculate the negative threshold as: t1 = (tz1 + ty1) / 2.

[0200] Step S4077: Obtain the intervals with the largest total number of normal power generation differences and the largest total number of abnormal power generation differences on the positive x-axis respectively, and mark them as the positive-axis normal value tz2 and the positive-axis abnormal value ty2; calculate the means of the negative-axis normal interval and the negative-axis abnormal interval respectively, and mark them as the positive-axis normal value tz2 and the positive-axis abnormal value ty2; calculate the positive threshold as: t2 = (tz2 + ty2) / 2.

[0201] Step S4078: Input the newly added first quantity of historical relevant meteorological data and historical power generation into the threshold simulation training to update t1 and t2.

[0202] Step S408: Determine whether z satisfies the inequality: t1 < z < t2; if not, send an abnormal signal to the processing end.

[0203] Step S5: The processing end determines whether the abnormal signal is accurate. If not, update the power generation prediction function. Step S5 includes the following sub-steps:

[0204] Step S501: The processing end determines whether the abnormal signal is correct. If the abnormal signal is incorrect, add the second quantity of historical relevant meteorological data and historical power generation to re-fit and calculate, obtain the updated power generation prediction function, and replace the power generation prediction function of the edge gateway with the updated power generation prediction function.

[0205] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. Among them, the storage medium may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, magnetic disk or optical disk. These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0206] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

Claims

1. A virtual power plant gateway data processing method based on edge computing, characterized in that: The steps include: Acquire an initial sample set from a database, the initial sample set including historical meteorological data and historical power generation; The abnormal data in the initial sample set is eliminated; the meteorological correlation coefficient between the historical meteorological data and the historical power generation is calculated, and the historical meteorological data is divided into historically relevant meteorological data and historically irrelevant meteorological data based on the meteorological correlation coefficient; Each type of historical meteorological data is fitted with the historical power generation to obtain m power generation prediction functions, and the m power generation prediction functions are deployed on the edge gateway, where m is the total number of types of historical meteorological data; Based on the historical relevant meteorological data, the corresponding real-time relevant meteorological data is obtained to obtain the real-time power generation, the real-time relevant meteorological data is input into the corresponding power generation prediction function to obtain m function power generation, the m function power generation is weighted and summed to obtain the real-time predicted power generation, and the power generation is judged to be abnormal according to the difference between the real-time predicted power generation and the real-time power generation. If the power generation is abnormal, an abnormal signal is sent to the processing end; The processing end determines whether the abnormal signal is accurate, and if it is not accurate, updates the power generation prediction function; Judging whether power generation is abnormal based on the difference between the real-time predicted power generation and the real-time power generation, and if power generation is abnormal, sending an abnormal signal to the processing end includes the following sub-steps: The difference between F and Ws is obtained as: z=F-Ws; where z is the difference between F and Ws, Ws is the real-time generated power, and F is the real-time predicted generated power; Obtain the negative threshold t1 and positive threshold t2 of the difference z between F and Ws; Determine whether z satisfies the inequality: t1 <z<t2; If not satisfied, an exception signal is sent to the processing end; Obtaining the negative threshold value t1 and the positive threshold value t2 of the difference z between F and Ws includes the following sub-steps: A first amount of historical relevant meteorological data and historical power generation is obtained, and simulation training is performed on the historical relevant meteorological data and historical power generation, which is marked as threshold simulation training. T1 and t2 are obtained through the threshold simulation training. The threshold simulation training includes: Substitute historical relevant meteorological data into m power generation prediction functions to obtain m training function power generation, and then perform weighted summation of the m training function power generation to obtain the training prediction power generation b; When the historical power generation corresponding to the historical meteorological data is in a normal state, the historical power generation in the normal state is marked as the training normal power generation o; The normal power generation difference is calculated as: Zz=bo; where Zz is the normal power generation difference; When the historical power generation corresponding to the historical meteorological data is in an abnormal state, the abnormal historical power generation is marked as training abnormal power generation p; The abnormal power generation difference is obtained as follows: Zy=bp; where Zy is the abnormal power generation difference; Establish a normal histogram and an abnormal histogram, and the horizontal axes of the normal histogram and the abnormal histogram are divided into u intervals; Count the frequency of normal power generation difference in each interval respectively; The normal histogram is drawn with the normal power generation difference as the x-axis and the frequency as the y-axis; Count the frequency of abnormal power generation difference in each interval respectively; The abnormal histogram is drawn with the abnormal power generation difference as the x-axis and the frequency as the y-axis; Obtain the interval with the largest total number of normal power generation differences and the interval with the largest total number of abnormal power generation differences on the negative half axis of the x-axis, and mark them as the negative axis normal interval and the negative axis abnormal interval respectively; Calculate the mean of the normal interval of the negative axis and the abnormal interval of the negative axis, and mark them as the normal value of the negative axis tz1 and the abnormal value of the negative axis ty1 respectively; The negative threshold is: t1=(tz1+ty1) / 2; Obtain the interval with the largest total number of normal power generation differences and the interval with the largest total number of abnormal power generation differences on the positive half axis of the x-axis, and mark them as positive axis normal value tz2 and positive axis abnormal value ty2 respectively; Calculate the mean of the negative axis normal interval and the negative axis abnormal interval, and mark them as the positive axis normal value tz2 and the positive axis abnormal value ty2 respectively; The positive threshold is: t2=(tz2+ty2) / 2; The newly added first amount of historical relevant meteorological data and the historical power generation input threshold are simulated and trained to update t1 and t2.

2. The virtual power plant gateway data processing method based on edge computing according to claim 1 is characterized in that: Eliminating abnormal data from the initial sample set includes the following sub-steps: Obtain historical power generation and historical meteorological data once every first time interval, and mark the total amount of historical power generation obtained as A; Sort the historical power generation from small to large, and set each historical power generation to correspond to a sequence number, which is an integer starting from 1; Determine whether a1*A is an integer. If it is an integer, set the historical power generation with the serial number a1*A as the first quantile Q1. If it is not an integer, calculate the average value of the historical power generation corresponding to the serial numbers on both sides of a1*A and mark it as the first quantile Q1; where a1 is the first quantile ratio; Determine whether a2*A is an integer. If it is an integer, set the historical power generation with the serial number a2*A as the second quantile Q2. If it is not an integer, calculate the average value of the historical power generation corresponding to the serial numbers on both sides of a2*A and mark it as the second quantile Q2; where a2 is the second quantile ratio, a2=(1 / a1-1)*a1; Set q1=Q1-b*(Q2-Q1), q2=Q2+b*(Q2-Q1); where q1 is the lower threshold, q2 is the upper threshold, and b is the range constant; The historical power generation power less than q1 and greater than q2 and the corresponding historical meteorological data are deleted.

3. The virtual power plant gateway data processing method based on edge computing according to claim 2 is characterized in that: The calculation of meteorological correlation coefficient between historical meteorological data and historical power generation includes the following sub-steps: The total number of historical power generation after statistical deletion is n; Historical meteorological data include: Solar radiation Xty i 、Temperature Xwd i 、Air humidity Xyc i 、Air density Xmd i 、Precipitation Xjs i 、Snowfall Xjx i and wind speed Xfs i , the historical power generation is Y i ; The average value of each historical meteorological data is , , , , , as well as The historical average power generation is ; The meteorological correlation coefficient of each historical meteorological data is obtained through the meteorological correlation coefficient formula. The meteorological correlation coefficient formula is as follows: ； Where Xn i Xty i 、Xwd i ,Xyc i 、Xmd i 、Xjs i 、Xjx i and Xfs i Cn represents one of Cty, Cwd, Cyc, Cmd, Cjs, Cjx and Cfs, wherein Cty is the meteorological correlation coefficient of solar radiation, Cwd is the meteorological correlation coefficient of temperature, Cyc is the meteorological correlation coefficient of air humidity, Cmd is the meteorological correlation coefficient of air density, Cjs is the meteorological correlation coefficient of precipitation, Cjx is the meteorological correlation coefficient of snowfall, and Cfs is the meteorological correlation coefficient of wind speed; express , , , , , as well as One of; After sorting the absolute values ​​of each meteorological correlation coefficient from large to small, the historical meteorological data corresponding to the meteorological correlation coefficient of the first proportion before sorting are marked as historically relevant meteorological data, and the historical meteorological data excluding the historically relevant meteorological data are marked as historically irrelevant meteorological data.

4. The virtual power plant gateway data processing method based on edge computing according to claim 3 is characterized in that: Fitting each historical meteorological data with the historical power generation to obtain m power generation prediction functions includes the following sub-steps: Get the total number of types m of historical meteorological data, set each historical meteorological data to correspond to a type number, and the type number is an integer from 1 to m; All historical relevant meteorological data are normalized using the max-min normalization method and marked as historical normalized meteorological data; The size of each historical normalized meteorological data is used as the X-axis data, and the power generation power is used as the Y-axis data to establish a plane rectangular coordinate system, which is marked as the prediction coordinate system; Plot the historical normalized meteorological data and the power generation into the corresponding prediction coordinate system respectively to obtain m related scatter plots; According to the distribution of each relevant scatter plot, appropriate preset power generation prediction functions are selected respectively, and the coefficients in the preset power generation prediction function are calculated by fitting the relevant scatter plots respectively, so as to obtain m power generation prediction functions.

5. The virtual power plant gateway data processing method based on edge computing according to claim 4 is characterized in that: Based on the historical relevant meteorological data, the corresponding real-time relevant meteorological data is obtained to obtain the real-time power generation, the real-time relevant meteorological data is input into the corresponding power generation prediction function to obtain m function power generation, and the m function power generation is weighted and summed to obtain the real-time predicted power generation, including the following sub-steps: Obtain real-time relevant meteorological data and real-time power generation Ws; The real-time related meteorological data are normalized using the max-min normalization method and marked as real-time normalized meteorological data; Each type of real-time normalized meteorological data is input into the corresponding power generation prediction function to obtain the function power generation F i ; Where i is the type number of real-time relevant meteorological data from 1 to m; Find F i The weight ratio k i ; The real-time predicted power generation is obtained as: ; Where F is the real-time predicted power generation.

6. The virtual power plant gateway data processing method based on edge computing according to claim 5 is characterized in that: Find F i The weight ratio k i The following sub-steps are included: Substitute all relevant historical meteorological data into the corresponding power generation prediction function to obtain m groups of historical predicted power generation Yx i ; Where i is the historical meteorological data from 1 to n; Find each group of Yx i With Y i The root mean square error between them is: ; where e i represents one of the root mean square errors of m types of power generation prediction functions; F i According to e i Arrange the values ​​from large to small, and then i Arrange from small to large and mark it as u i , F i The corresponding e i Replace with u in order i , find F i The weight ratio is: 。 7. The virtual power plant gateway data processing method based on edge computing according to claim 6 is characterized in that: The processing end determines whether the abnormal signal is accurate. If it is inaccurate, the power generation prediction function is updated, including the following sub-steps: The processing end determines whether the abnormal signal is correct. If the abnormal signal is incorrect, a second amount of historical relevant meteorological data and historical power generation are added to re-fit the calculation to obtain an updated power generation prediction function, and the power generation prediction function of the edge gateway is replaced with the updated power generation prediction function.

8. A virtual power plant gateway data processing system based on edge computing, applicable to a virtual power plant gateway data processing method based on edge computing according to any one of claims 1 to 7, characterized in that: include: Data acquisition module, data processing module, function fitting module, anomaly detection module and function update module; The data acquisition module is used to acquire an initial sample set from a database, the initial sample set including historical meteorological data and historical power generation; The data processing module is used to remove abnormal data in the initial sample set, calculate the meteorological correlation coefficient between the historical meteorological data and the historical power generation, and divide the historical meteorological data into historically relevant meteorological data and historically irrelevant meteorological data based on the meteorological correlation coefficient; The function fitting module is used to fit each type of historical relevant meteorological data with the historical power generation, obtain m power generation prediction functions, and deploy the m power generation prediction functions on the edge gateway, where m is the total number of types of historical relevant meteorological data; The anomaly detection module is used to obtain corresponding real-time related meteorological data based on historical related meteorological data, obtain real-time power generation, input the real-time related meteorological data into the corresponding power generation prediction function, obtain m function power generation, perform weighted summation on the m function power generation to obtain the real-time predicted power generation, and judge whether power generation is abnormal according to the difference between the real-time predicted power generation and the real-time power generation. If power generation is abnormal, an abnormal signal is sent to the processing end; The function updating module is used by the processing end to determine whether the abnormal signal is accurate, and if it is inaccurate, to update the power generation prediction function.

Images