An input data anomaly identification evaluation method and system of a power prediction model

Abstract

The application provides an input data anomaly identification evaluation method and system of a power prediction model, comprising: obtaining historical measured data of meteorological elements of a new energy station, and classifying each meteorological element; obtaining a probability density function and a cumulative probability density function corresponding to the historical measured data of each meteorological element at any moment under each type; setting a confidence interval of the cumulative probability density function of each meteorological element at any moment under each type, obtaining the upper and lower bounds of the normal data range of each meteorological element at any moment under each type; determining the envelope line of the normal data corresponding to each meteorological element at different moments under each type, and performing real-time input data online anomaly identification; based on the result of the anomaly identification, designing an evaluation index of the proportion and degree of input data anomaly, and completing the quantitative evaluation of the real-time input data quality of the power prediction model. The application solves the problems of low prediction result accuracy and poor stability of the existing photovoltaic power prediction model.

Description

Technical Field

[0001] This invention relates to the field of new energy power prediction technology, and in particular to a method and system for identifying and evaluating anomalies in the input data of a power prediction model. Background Technology

[0002] Guided by the goal of energy conservation and emission reduction, a high proportion of renewable energy integration will become the future development trend of the power system. Solar energy, with its ubiquitous availability, cleanliness, vast quantity, and inexhaustible nature, stands out as an ideal primary energy source among many energy sources. Solar photovoltaic (PV) power generation, as the most efficient way to utilize solar energy, boasts a clean and green power generation process, convenient and flexible installation, safe and reliable operation, and relatively simple maintenance, making it one of the main research and development directions in the current renewable energy power generation field. However, the volatility, randomness, and intermittency of PV output pose serious challenges to the safe, stable, and economical operation of the power system. Accurate and reliable PV power forecasting can provide information support for the planning, operation, stability analysis and control, and trading of new power systems, which is of great significance for improving the overall energy efficiency of the system and promoting the consumption of renewable energy. It is an important guarantee for the safe, economical, and efficient operation of new power systems.

[0003] Photovoltaic power prediction is a data mining-based research area, and its accuracy depends on the quality of the training data for the prediction model. For data-driven machine learning models, outliers in low-quality data act as noise, affecting model training and leading to biased learning results. Due to the presence of human, natural, and random factors, a large number of outliers inevitably exist in measured photovoltaic meteorological data, which can have a significant impact on the parameter training of photovoltaic power probabilistic prediction models. Therefore, before using measured photovoltaic meteorological data to train the parameters of photovoltaic power prediction models, it is necessary to process outliers using data anomaly identification and repair techniques.

[0004] Existing anomaly identification algorithms mostly assume that historical data follows a normal distribution and identify anomalies by setting confidence intervals. However, historical measured data of photovoltaic meteorological elements may be affected by various nonlinear factors, resulting in a skewed distribution or a heavy tail, which is inconsistent with the symmetry of the normal distribution. Therefore, for historical measured data of photovoltaic meteorological elements, the method of using a normal distribution as an assumption and setting confidence intervals for anomaly identification is not suitable and can easily lead to inaccurate photovoltaic power prediction. Summary of the Invention

[0005] This invention provides a method and system for identifying and evaluating anomalies in the input data of a power prediction model, in order to solve the problems of low accuracy and poor stability of existing photovoltaic power prediction models.

[0006] This invention provides a method for identifying and evaluating anomalies in input data of a power prediction model, comprising:

[0007] Acquire historical measured data of meteorological elements at new energy power stations and classify each meteorological element;

[0008] Find the probability density function and cumulative probability density function of each meteorological element at any time under each type of historical measured data;

[0009] Set the confidence interval of the cumulative probability density function of each meteorological element at any time under each type, and obtain the upper and lower bounds of the normal data range of each meteorological element at any time under each type.

[0010] Based on the upper and lower bounds of the normal data range of each meteorological element at any time under each type, the envelope of the normal data corresponding to different times under each type of meteorological element is determined, and online anomaly identification of real-time input data is carried out.

[0011] Based on the results of anomaly identification, evaluation indicators for the proportion and degree of anomalies in the input data are designed to quantitatively evaluate the quality of real-time input data for the power prediction model.

[0012] According to the present invention, a method for identifying and evaluating anomalies in input data of a power prediction model includes, in part, acquiring historical measured data of meteorological elements at new energy power plants and classifying each meteorological element, specifically comprising:

[0013] Collect historical measured meteorological data for power prediction of new energy power plants, including historical measured data of irradiance and historical measured data of wind speed;

[0014] Historical measured data of various meteorological elements are classified using classification standards that are appropriate to their distribution characteristics.

[0015] According to the power prediction model input data anomaly identification and evaluation method provided by the present invention, the step of obtaining the probability density function and cumulative probability density function corresponding to the historical measured data of each meteorological element at any time under each type specifically includes:

[0016] The probability density function is obtained using the nonparametric kernel density estimation method, and the calculation formula is as follows:

[0017]

[0018] In the formula, Let f(x) be the kernel density estimation function of the probability density function f(x), where n is the number of sample points, h is the bandwidth, and K(x) is the kernel function.

[0019] The cumulative probability density function gives the probability that a random variable is less than or equal to a certain value. As the value increases, the cumulative probability density function gradually increases from 0 to 1. The formula for calculating the cumulative probability density function is:

[0020]

[0021] In the formula, X is a random variable and x is any real number.

[0022] According to the power prediction model input data anomaly identification and evaluation method provided by the present invention, the step of setting the confidence interval of the cumulative probability density function of each meteorological element at any time under each type to obtain the upper and lower bounds of the normal data range of each meteorological element at any time under each type specifically includes:

[0023] Based on the shape of the probability density curves of each meteorological element at any time under each type, different confidence intervals are set for its cumulative probability density function;

[0024] The cumulative probability density function curve takes the upper and lower bounds of the meteorological element values ​​corresponding to the confidence interval, and the upper and lower bounds can be represented by matrices respectively.

[0025] According to the power prediction model provided by the present invention, the method for identifying and evaluating anomalies in input data, wherein the envelope of normal data corresponding to different times under each meteorological element under each type is determined based on the upper and lower bounds of the normal data range of each meteorological element at any time under each type, and real-time online anomaly identification of input data is performed, specifically including:

[0026] Based on the classification criteria, the type label of the real-time input data is determined. By referring to the envelope of the normal data at different times under the corresponding meteorological element type, the input data anomaly identification is performed and the identification result is generated.

[0027] The identification results include three categories: abnormally large, normal, and abnormally small.

[0028] According to the present invention, a method for identifying and evaluating anomalies in input data of a power prediction model is provided. Based on the results of anomaly identification, evaluation indicators for the proportion and degree of anomalies in the input data are designed to quantitatively evaluate the quality of real-time input data for the power prediction model. Specifically, this includes:

[0029] Based on the number of abnormal results in the identification results, calculate the proportion of abnormal matrix elements in the input data matrix to the total number of elements, and determine the proportion of abnormal data.

[0030] Based on the abnormal elements in the actual input matrix, the degree of anomaly at a single point and the overall degree of anomaly of the row vector corresponding to the input physical quantity are calculated. The distance between the abnormal value and the upper and lower bounds of the normal value of the meteorological element is determined, thereby determining the degree of data anomaly.

[0031] The present invention also provides an input data anomaly identification and evaluation system for a power prediction model, the system comprising:

[0032] The data acquisition module is used to acquire historical measured data of meteorological elements at new energy power stations and to classify each meteorological element.

[0033] The probability and statistics module is used to obtain the probability density function and cumulative probability density function of each meteorological element at any time under each type of historical measured data.

[0034] The upper and lower bound determination module is used to set the confidence interval of the cumulative probability density function of each meteorological element at any time under each type, and obtain the upper and lower bounds of the normal data range of each meteorological element at any time under each type.

[0035] The anomaly identification module is used to determine the envelope of normal data corresponding to different times under each meteorological element and type based on the upper and lower bounds of the normal data range of each meteorological element at any time under each type, and to perform real-time online anomaly identification of input data.

[0036] The quality evaluation module is used to design evaluation indicators for the proportion and degree of anomalies in the input data based on the results of anomaly identification, and to complete the quantitative evaluation of the quality of real-time input data of the power prediction model.

[0037] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement an input data anomaly identification and evaluation method for any of the power prediction models described above.

[0038] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements an input data anomaly identification and evaluation method for any of the power prediction models described above.

[0039] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements an input data anomaly identification and evaluation method for any of the power prediction models described above.

[0040] This invention provides a method and system for identifying and evaluating anomalies in input data of a power prediction model. Based on the upper and lower bounds of the normal data range for each meteorological element at any given time under each type, the system determines the envelope of normal data corresponding to different times under each meteorological element type. Real-time online anomaly identification of the input data is then performed. Based on the anomaly identification results, evaluation indicators for the proportion and degree of anomalies in the input data are designed to quantitatively evaluate the quality of real-time input data for the power prediction model. Statistical modeling and confidence interval setting of historical data for non-normally distributed meteorological elements improve the flexibility of real-time input data anomaly identification for the power prediction model. The design of comprehensive evaluation indicators for quantitative analysis of real-time input data quality provides strong support for anomaly data processing. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0042] Figure 1 This is one of the flowcharts illustrating an anomaly identification and evaluation method for input data of a power prediction model provided by the present invention;

[0043] Figure 2 This is the second flowchart of a method for identifying and evaluating anomalies in input data of a power prediction model provided by the present invention.

[0044] Figure 3 This is the third flowchart of a method for identifying and evaluating anomalies in input data of a power prediction model provided by the present invention.

[0045] Figure 4 This is the fourth flowchart of a method for identifying and evaluating anomalies in input data of a power prediction model provided by the present invention.

[0046] Figure 5 This is the fifth flowchart of a method for identifying and evaluating anomalies in input data of a power prediction model provided by the present invention.

[0047] Figure 6 This is a schematic diagram of the module connections of an input data anomaly identification and evaluation system for a power prediction model provided by the present invention;

[0048] Figure 7 This is a schematic diagram of the measured irradiance probability density function at 11:00 AM on a sunny day in spring, provided by the present invention.

[0049] Figure 8This is a schematic diagram of the cumulative probability density function of measured irradiance at 11:00 AM on a sunny day in spring, provided by the present invention.

[0050] Figure 9 This is a schematic diagram of the measured irradiance probability density function at 8:45 AM on a sunny autumn day, provided by the present invention.

[0051] Figure 10 This is a schematic diagram of the cumulative probability density function of measured irradiance at 8:45 AM on a sunny autumn day, provided by the present invention.

[0052] Figure 11 This is a schematic diagram of the measured irradiance probability density function at 8:45 AM on a sunny day in spring, provided by the present invention.

[0053] Figure 12 This is a schematic diagram of the cumulative probability density function of measured irradiance at 8:45 AM on a sunny day in spring, provided by the present invention.

[0054] Figure 13 This is a schematic diagram of the measured irradiance probability density function at 8:45 AM on a cloudy day in autumn, provided by the present invention.

[0055] Figure 14 This is a schematic diagram of the cumulative probability density function of measured irradiance at 8:45 AM on a cloudy day in autumn, provided by the present invention.

[0056] Figure 15 This is a schematic diagram of the measured irradiance probability density function at 7:45 am on a rainy day in spring, provided by the present invention.

[0057] Figure 16 This is a schematic diagram of the cumulative probability density function of measured irradiance at 7:45 am on a rainy day in spring, provided by the present invention.

[0058] Figure 17 This is a schematic diagram of the measured irradiance probability density function at 10:30 am on a sunny winter day, provided by the present invention.

[0059] Figure 18 This is a schematic diagram of the cumulative probability density function of measured irradiance at 10:30 am on a sunny winter day, provided by the present invention.

[0060] Figure 19 This is a schematic diagram of anomaly degree calculation provided by the present invention;

[0061] Figure 20 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0062] Figure label:

[0063] 110: Data Acquisition Module; 120: Probability and Statistics Module; 130: Upper and Lower Bound Determination Module; 140: Anomaly Identification Module; 150: Quality Evaluation Module;

[0064] 2010: Processor; 2020: Communication interface; 2030: Memory; 2040: Communication bus. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0066] The following is combined with Figures 1-5 The present invention describes a method for identifying and evaluating anomalies in input data of a power prediction model, comprising:

[0067] S100. Obtain historical measured data of meteorological elements at new energy power stations and classify each meteorological element.

[0068] S200: Calculate the probability density function and cumulative probability density function of each meteorological element at any time under each type of historical measured data.

[0069] S300. Set the confidence interval of the cumulative probability density function of each meteorological element at any time under each type, and obtain the upper and lower bounds of the normal data range of each meteorological element at any time under each type.

[0070] S400: Based on the upper and lower bounds of the normal data range of each meteorological element at any time under each type, determine the envelope of the normal data corresponding to different times under each type of meteorological element, and perform online anomaly identification of real-time input data.

[0071] Based on the results of anomaly identification, S500 designs evaluation indicators for the proportion and degree of anomalies in the input data, and completes the quantitative evaluation of the quality of real-time input data for the power prediction model.

[0072] In this invention, based on probability and statistics theory, the normal range of each input meteorological element at any time under each type is obtained by analyzing the probability distribution of historical input data. This enables the identification of anomalies and the evaluation of the degree of anomalies in real-time meteorological input data, helping decision-makers understand the quality of input data and guide them to take corresponding corrective measures for bad data, thereby improving the stability and accuracy of power prediction models.

[0073] Historical measured data of meteorological elements at new energy power stations were obtained, and each meteorological element was classified, including:

[0074] S101. Collect historical measured data of meteorological elements for power prediction of new energy power plants, including historical measured data of irradiance and historical measured data of wind speed;

[0075] S102. Classify the historical measured data of each meteorological element using a classification standard that is appropriate to its distribution characteristics.

[0076] In this invention, historical measured data of meteorological elements for power prediction of new energy power plants are collected, including historical measured data of irradiance, historical measured data of wind speed, etc., with a time resolution of 15 minutes.

[0077] Historical measured data of various meteorological elements were classified using classification criteria adapted to their distribution characteristics. Taking irradiance as an example, seasonal type and weather type affect the probability distribution of historical irradiance data by influencing sunrise and sunset times, irradiance magnitude and fluctuation characteristics. To consider the impact of weather type and seasonal type on the anomaly identification effect of input data, the historical irradiance dataset was divided into 12 categories according to four seasonal types (spring, summer, autumn, and winter) and three weather types (sunny, cloudy, and rainy).

[0078] In this invention, the probability density function is represented by PDF, and the cumulative probability density function is represented by CDF. The process of obtaining the probability density function and cumulative probability density function corresponding to historical measured data of each meteorological element at any time under each type specifically includes:

[0079] The probability density function is calculated by integrating it over a given interval. The shape of the probability density function reflects the distribution characteristics of the random variable, such as symmetry, skewness, and kurtosis. Different shapes correspond to different distribution types, such as normal, uniform, and exponential distributions. Methods for calculating the probability density function can be categorized into parametric and non-parametric methods. Parametric methods assume a known distribution; common methods include box plots, t-tests, Grubbs' tests, and chi-square tests. Non-parametric methods make no assumptions about the statistical distribution of the data, better adapting to the complexities of the actual data distribution; common non-parametric methods include histogram methods, finite state machine methods, and kernel function methods. This invention utilizes the non-parametric kernel density estimation method to calculate the probability density function, as shown in the following formula:

[0080]

[0081] In the formula, Let f(x) be the kernel density estimation function of the probability density function f(x), where n is the number of sample points, h is the bandwidth, and K(x) is the kernel function.

[0082] The cumulative probability density function is obtained by calculating the probability that a random variable is less than or equal to a certain value. As the value increases, the cumulative probability density function gradually increases from 0 to 1.

[0083] The formula for calculating the cumulative probability density function is shown below:

[0084]

[0085] In the formula, X is a random variable and x is any real number.

[0086] We set confidence intervals for the cumulative probability density functions of each meteorological element at any time under each type, thus obtaining the upper and lower bounds of the normal data range for each meteorological element at any time under each type. Specifically, this includes:

[0087] S201. Based on the shape of the probability density curve of each meteorological element at any time under each type, set different confidence intervals for its cumulative probability density function.

[0088] S202. The cumulative probability density function curve takes the upper and lower bounds of the meteorological element values ​​corresponding to the confidence interval. The upper and lower bounds can be represented by matrices respectively.

[0089] Taking measured irradiance as an example, the probability density curve shape presents three cases: "approximately normal distribution", "part of normal distribution", and "combination of two normal distributions".

[0090] Define the real-time input matrix of the power prediction model as C = (x mn )∈R M×V Where M is the number of input meteorological elements, N is the input time span, and x is the input time range. mn This represents the measurement value of the m-th input meteorological element at the n-th time.

[0091] When the probability density curve is "approximately a normal distribution" or "part of a normal distribution", the cumulative probability density function curve takes the upper and lower bounds of the meteorological element values ​​corresponding to the confidence interval. The upper and lower bounds can be represented by two matrices respectively, namely, matrix and in, This is the upper bound value corresponding to the measurement value of the m-th input meteorological element at the n-th time. This is the lower bound value corresponding to the measurement value of the m-th input meteorological element at the n-th time.

[0092] When the probability density curve is a combination of two normal distributions, the cumulative probability density function curve takes two sets of upper and lower bounds of the meteorological element values ​​corresponding to the confidence interval. The upper and lower bounds can be represented by two sets of matrices, namely the first set of matrices. and Second set of matrices and in, This is the upper bound of the first normal range corresponding to the measurement value of the m-th input meteorological element at the n-th time. This is the lower bound of the first normal range corresponding to the measurement value of the m-th input meteorological element at the n-th time. This is the upper bound of the second normal range corresponding to the measurement value of the m-th input meteorological element at the n-th time. This is the lower bound of the second normal range corresponding to the measurement value of the m-th input meteorological element at the n-th time.

[0093] Specifically, based on the shape of the probability density curves of each meteorological element at any time under each type, different confidence intervals are set for its cumulative probability density function to obtain the normal data envelopment curves corresponding to each meteorological element at any time under each type. Taking the measured irradiance of the target station as an example, the probability density curve shapes are as follows:

[0094] Case 1: As Figure 7 The probability density function curve shown can be approximated as a normal distribution. Therefore, the confidence interval for the cumulative probability density function is taken as [0.05, 0.95]. Figure 8 As shown.

[0095] Case 2: For example Figure 9 The probability density function curve shown can be considered as part of a normal distribution. Therefore, the confidence interval for the cumulative probability density function is taken as [0.05, 1], as follows. Figure 10 As shown.

[0096] Case 3: For example Figure 11 The probability density function curve shown can be considered as part of a normal distribution. Therefore, the confidence interval for the cumulative probability density function is [0, 0.95], as shown below. Figure 12 As shown.

[0097] Case 4: For example Figure 13 The probability density function curve shown can be considered as a combination of two normal distributions. Therefore, its cumulative probability density function confidence interval is divided into two segments: the first segment is [0, the cumulative probability density function value corresponding to the minimum point of the probability density function -0.05], and the second segment is [the cumulative probability density function value corresponding to the minimum point of the probability density function +0.05, 1]. Figure 14 As shown.

[0098] Case 5: For example Figure 15The probability density function curve shown can be considered as a combination of two normal distributions. Therefore, its cumulative probability density function confidence interval is divided into two segments: the first segment is [0, cumulative probability density function value corresponding to the minimum point of the probability density function -0.05], and the second segment is [cumulative probability density function value corresponding to the minimum point of the probability density function +0.05, 0.95], as shown below. Figure 16 As shown.

[0099] Case 6: For example Figure 17 The probability density function curve shown can be considered as a combination of two normal distributions. Therefore, its cumulative probability density function confidence interval is divided into two segments: the first segment is [0.05, the cumulative probability density function value corresponding to the minimum point of the probability density function -0.05], and the second segment is [the cumulative probability density function value corresponding to the minimum point of the probability density function +0.05, 0.95], as shown below. Figure 18 As shown.

[0100] For Cases 1-3, each cumulative probability density function curve represents the upper and lower bounds of the meteorological element values ​​corresponding to the confidence interval. These upper and lower bounds can be represented by two matrices, i.e., matrices. and By representing the upper and lower bounds in the figure, the envelope of the normal data corresponding to each meteorological element at 96 time points can be obtained.

[0101] For Cases 4-6, each cumulative probability density function curve takes two sets of upper and lower bounds for the meteorological element values ​​corresponding to the confidence interval. These upper and lower bounds can be represented by two sets of matrices, namely the first set of matrices. and Second set of matrices and

[0102] Based on the upper and lower bounds of the normal data range for each meteorological element at any time under each type, the envelope of the normal data corresponding to different times under each meteorological element is determined, and real-time online anomaly identification of input data is performed, specifically including:

[0103] S301. Based on the classification criteria, determine the type label of the real-time input data, refer to the envelope of the normal data at different times under the corresponding meteorological element type, identify the anomalies in the input data, and generate the identification results.

[0104] S302, The identification results include three cases: abnormally large, normal, and abnormally small.

[0105] In this invention, real-time online anomaly detection of input data is performed based on the envelope of normal data at 96 time points corresponding to each meteorological element and type. Specifically, the type label of the real-time input data is determined based on classification criteria, and anomaly detection is performed by referring to the envelope of normal data at 96 time points corresponding to the meteorological element and type. The detection results include three categories: excessively large anomaly, normal, and excessively small anomaly.

[0106] For Case 1-Case 3, the anomaly matrix B = (b mn )∈{0,1} M×N For x mn ∈C, ( like Then determine x mn For normal values, the corresponding b mn =1; if huo Then determine x mb For outliers, the corresponding b mb =0.

[0107] For Case 4-Case 6, the anomaly matrix B = (b mb )∈{0,1} M×N For x mn ∈C,

[0108] like or Then determine x mn For normal values, the corresponding b mn =1; if or or Then determine x mn For outliers, the corresponding b mn =0.

[0109] Based on the results of anomaly identification, evaluation indicators for the proportion and severity of anomalies in the input data are designed to quantitatively evaluate the quality of real-time input data for the power prediction model. Specifically, this includes:

[0110] S401. Based on the number of abnormal results in the identification results, calculate the proportion of abnormal matrix elements in the input data matrix to the total number of elements, and determine the proportion of abnormal data.

[0111] S402. Based on the abnormal elements in the actual input matrix, calculate the degree of single-point anomaly and the overall degree of anomaly of the row vector corresponding to the input physical quantity, determine the distance between the abnormal value and the upper and lower bounds of the normal value of the meteorological element, and thus determine the degree of data anomaly.

[0112] In this invention, the percentage of data anomalies is defined as the percentage of anomalies in the input data matrix, B = (b...). mn )∈{0,1} M×N The proportion of zero elements out of all elements. This indicator ranges from [0,1]. For each input matrix, there is a unique anomaly proportion indicator, without calculating the anomaly proportion of each row vector corresponding to each meteorological element separately. The calculation formula is:

[0113]

[0114] In the formula, α is the anomaly percentage index, M represents the number of input meteorological elements, N is the input time span, and b mn It is an element in the anomaly matrix corresponding to the measurement value of the m-th input meteorological element at the n-th time.

[0115] Data anomaly level. For outlier elements in the actual input matrix, the anomaly level index is the distance between the outlier value and the upper or lower bound of the normal meteorological element value; for non-outlier elements in the actual input matrix, the value of this anomaly level index is 0. The calculation process is as follows:

[0116] 1) Calculation of the degree of single-point anomaly.

[0117] For Cases 1-3, the following formula applies. This step is performed once for each element of the input data matrix, and the result is the anomaly severity matrix β = (β... mn )∈{0,R} M×N .

[0118]

[0119] In the formula, β mn x represents the degree of anomaly corresponding to the measured value of the m-th input meteorological element at time n. mn It is the measurement value of the m-th input meteorological element at the n-th time. It is the upper bound of the m-th input meteorological element at time n. It is the lower bound of the m-th input meteorological element at the n-th time.

[0120] In the molecule, For data point x mn Distance to the upper boundary, For data point x mn Distance to the lower bound, For data point x mn The distance between the upper and lower bounds. If but like When the outlier is above the upper bound (twice the distance between the outlier and the upper bound), when the outlier is below the lower bound. (twice the distance between the outlier and the lower bound), therefore multiplying the denominator by 2 gives the distance from the outlier to the lower bound or the lower bound. In the denominator, For data point x mn The distance to the midpoint of the upper and lower bounds, where It is the midpoint between the upper and lower bounds.

[0121] Therefore, when β mn =0, when At that time, β mn This represents the distance from the outlier to the lower bound or the lower bound relative to x. mn The ratio of the distances to the midpoints of the upper and lower bounds, 0 < β mn <1.

[0122] For Case 4-Case 6, the formula is as follows.

[0123]

[0124]

[0125]

[0126] In the formula, β mn_1 β represents the degree of abnormality of the first segment of the normal range of the measured value of the m-th input meteorological element at time n. mn_2 β represents the degree of abnormality in the second segment of the normal range of the measured value of the m-th input meteorological element at time n. nn The degree of anomaly in the measured value of the m-th input meteorological element at time n is represented by x. mn For the m-th input meteorological element, the measurement value at time n is given. This is the upper bound of the first normal range corresponding to the measurement value of the m-th input meteorological element at the n-th time. This is the lower bound of the first normal range corresponding to the measurement value of the m-th input meteorological element at the n-th time. This is the upper bound of the second normal range corresponding to the measurement value of the m-th input meteorological element at the n-th time. This is the lower bound of the second normal range corresponding to the measurement value of the m-th input meteorological element at the n-th time.

[0127] Anomaly degree calculation diagram as shown below Figure 19 As shown.

[0128] 2) Calculate the overall anomaly degree of the row vectors corresponding to the input physical quantities.

[0129] For the anomaly degree matrix β=(β mn )∈{0,R} M×N The arithmetic mean of each row of elements is calculated to obtain the overall anomaly degree column vector β. ′ =(β) ′ m )∈{0,R} M×1 The resulting overall anomaly degree column vector β is shown in the following formula. ′ =(β) ′ m )∈{0,R} M×1 This is the final indicator of the degree of abnormality.

[0130] β ′ =[β ′ 1,β ′ 2,…,β ′ M ] T ,

[0131] In the formula, β mn β represents the degree of anomaly corresponding to the measured value of the m-th input meteorological element at time n. ′ m It is the average value of the m-th input meteorological element measured at N time points.

[0132] This invention provides a method for identifying and evaluating anomalies in input data for a power prediction model. Based on the upper and lower bounds of the normal data range for each meteorological element at any given time under each type, the method determines the envelope of normal data corresponding to different times under each meteorological element type. Real-time online anomaly identification is then performed. Based on the anomaly identification results, evaluation indicators for the proportion and severity of anomalies in the input data are designed to quantitatively evaluate the quality of real-time input data for the power prediction model. Statistical modeling and confidence interval setting of historical data for non-normally distributed meteorological elements improve the flexibility of real-time input data anomaly identification for the power prediction model. The design of comprehensive evaluation indicators for quantitative analysis of real-time input data quality provides strong support for anomaly data processing.

[0133] refer to Figure 6 The present invention also discloses an input data anomaly identification and evaluation system for a power prediction model, the system comprising:

[0134] The data acquisition module 110 is used to acquire historical measured data of meteorological elements at new energy power stations and classify each meteorological element.

[0135] The probability and statistics module 120 is used to obtain the probability density function and cumulative probability density function of each meteorological element at any time under each type of historical measured data.

[0136] The upper and lower bound determination module 130 is used to set the confidence interval of the cumulative probability density function of each meteorological element at any time under each type, so as to obtain the upper and lower bounds of the normal data range of each meteorological element at any time under each type.

[0137] The anomaly identification module 140 is used to determine the envelope of the normal data corresponding to different times under each type of meteorological element based on the upper and lower bounds of the normal data range of each meteorological element at any time under each type, and to perform real-time online anomaly identification of input data.

[0138] The quality evaluation module 150 is used to design evaluation indicators for the proportion and degree of anomalies in the input data based on the results of anomaly identification, and to complete the quantitative evaluation of the quality of real-time input data of the power prediction model.

[0139] The data acquisition module collects historical measured data of meteorological elements for power prediction of new energy power plants, including historical measured data of irradiance and historical measured data of wind speed.

[0140] Historical measured data of various meteorological elements are classified using classification standards that are appropriate to their distribution characteristics.

[0141] The calculation module uses a nonparametric kernel density estimation method to obtain the probability density function. The calculation formula is as follows:

[0142]

[0143] In the formula, Let f(x) be the kernel density estimation function of the probability density function f(x), where n is the number of sample points, h is the bandwidth, and K(x) is the kernel function.

[0144] The cumulative probability density function gives the probability that a random variable is less than or equal to a certain value. As the value increases, the cumulative probability density function gradually increases from 0 to 1. The formula for calculating the cumulative probability density function is:

[0145]

[0146] In the formula, X is a random variable and x is any real number.

[0147] The upper and lower bound determination module sets different confidence intervals for the cumulative probability density function of each meteorological element at any time under each type, based on the shape of the probability density curve of each meteorological element at any time.

[0148] The cumulative probability density function curve takes the upper and lower bounds of the meteorological element values ​​corresponding to the confidence interval, and the upper and lower bounds can be represented by matrices respectively.

[0149] The anomaly identification module determines the type label of real-time input data based on classification criteria, and identifies anomalies in the input data by referring to the envelope of normal data at different times under the corresponding meteorological element type, and generates identification results.

[0150] The identification results include three categories: abnormally large, normal, and abnormally small.

[0151] The evaluation module calculates the proportion of abnormal matrix elements in the input data matrix to the total number of elements based on the number of abnormal results in the identification results, and determines the proportion of abnormal data.

[0152] Based on the abnormal elements in the actual input matrix, the degree of anomaly at a single point and the overall degree of anomaly of the row vector corresponding to the input physical quantity are calculated. The distance between the abnormal value and the upper and lower bounds of the normal value of the meteorological element is determined, thereby determining the degree of data anomaly.

[0153] This invention provides an input data anomaly identification and evaluation system for a power prediction model. Based on the upper and lower bounds of the normal data range for each meteorological element at any given time under each type, the system determines the envelope of normal data corresponding to different times under each meteorological element type. Real-time online anomaly identification of the input data is then performed. Based on the anomaly identification results, evaluation indicators for the proportion and degree of anomalies in the input data are designed to quantitatively evaluate the quality of real-time input data for the power prediction model. Statistical modeling and confidence interval setting of historical data for non-normally distributed meteorological elements improve the flexibility of real-time input data anomaly identification for the power prediction model. The design of comprehensive evaluation indicators for quantitative analysis of real-time input data quality provides strong support for anomaly data processing.

[0154] Figure 20 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 20 As shown, the electronic device may include a processor 2010, a communications interface 2020, a memory 2030, and a communication bus 2040. The processor 2010, communications interface 2020, and memory 2030 communicate with each other via the communication bus 2040. The processor 2010 can call logic instructions from the memory 2030 to execute a method for identifying and evaluating anomalies in the input data of a power prediction model.

[0155] Furthermore, the logical instructions in the aforementioned memory 2030 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0156] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute a power prediction model input data anomaly identification and evaluation method provided by the above methods.

[0157] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements an input data anomaly identification and evaluation method for a power prediction model provided by the methods described above.

[0158] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0159] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying and evaluating anomalies in input data of a power prediction model, characterized in that, include: Acquire historical measured data of meteorological elements at new energy power stations and classify each meteorological element; Find the probability density function and cumulative probability density function of each meteorological element at any time under each type of historical measured data; Set the confidence interval of the cumulative probability density function of each meteorological element at any time under each type, and obtain the upper and lower bounds of the normal data range of each meteorological element at any time under each type. Based on the upper and lower bounds of the normal data range of each meteorological element at any time under each type, the envelope of the normal data corresponding to different times under each type of meteorological element is determined, and online anomaly identification of real-time input data is carried out. Based on the results of anomaly identification, evaluation indicators for the proportion and degree of anomalies in the input data are designed to complete the quantitative evaluation of the quality of real-time input data of the power prediction model. This includes: calculating the proportion of the number of anomaly elements in the input data matrix to the total number of elements based on the number of anomalies in the identification results, and determining the proportion of data anomalies. Based on the abnormal elements in the actual input matrix, the degree of anomaly at a single point and the overall degree of anomaly of the row vector corresponding to the input physical quantity are calculated. The distance between the abnormal value and the upper and lower bounds of the normal value of the meteorological element is determined, thereby determining the degree of data anomaly.

2. The method for identifying and evaluating anomalies in input data of the power prediction model according to claim 1, characterized in that, The acquisition of historical measured data of meteorological elements at new energy power stations, and the classification of each meteorological element, specifically includes: Collect historical measured meteorological data for power prediction of new energy power plants, including historical measured data of irradiance and historical measured data of wind speed; Historical measured data of various meteorological elements are classified using classification standards that are appropriate to their distribution characteristics.

3. The method for identifying and evaluating anomalies in input data of the power prediction model according to claim 1, characterized in that, The process of obtaining the probability density function and cumulative probability density function of each meteorological element at any time under each type specifically includes: The probability density function is obtained using the nonparametric kernel density estimation method, and the calculation formula is as follows: In the formula, The probability density function The kernel density estimation function, The number of sample points For bandwidth, For kernel functions; The cumulative probability density function gives the probability that a random variable is less than or equal to a certain value; as the value increases, the cumulative probability density function gradually increases from 0 to 1. The formula for calculating the cumulative probability density function is: In the formula, For random variables, Let be any real number.

4. The method for identifying and evaluating anomalies in input data of the power prediction model according to claim 1, characterized in that, The step of setting the confidence interval of the cumulative probability density function of each meteorological element at any time under each type, and obtaining the upper and lower bounds of the normal data range of each meteorological element at any time under each type, specifically includes: Based on the shape of the probability density curves of each meteorological element at any time under each type, different confidence intervals are set for its cumulative probability density function; The cumulative probability density function curve takes the upper and lower bounds of the meteorological element values ​​corresponding to the confidence interval, and the upper and lower bounds can be represented by matrices respectively.

5. The method for identifying and evaluating anomalies in input data of the power prediction model according to claim 1, characterized in that, The process of determining the envelope of normal data corresponding to different times under each meteorological element type based on the upper and lower bounds of the normal data range of each meteorological element at any time under each type, and performing real-time online anomaly identification of input data, specifically includes: Based on the classification criteria, the type label of the real-time input data is determined. By referring to the envelope of the normal data at different times under the corresponding meteorological element type, the input data anomaly identification is performed and the identification result is generated. The identification results include three categories: abnormally large, normal, and abnormally small.

6. A system for identifying and evaluating anomalies in input data of a power prediction model, characterized in that, The system includes: The data acquisition module is used to acquire historical measured data of meteorological elements at new energy power stations and to classify each meteorological element. The probability and statistics module is used to obtain the probability density function and cumulative probability density function of each meteorological element at any time under each type of historical measured data. The upper and lower bound determination module is used to set the confidence interval of the cumulative probability density function of each meteorological element at any time under each type, and obtain the upper and lower bounds of the normal data range of each meteorological element at any time under each type. The anomaly identification module is used to determine the envelope of normal data corresponding to different times under each meteorological element and type based on the upper and lower bounds of the normal data range of each meteorological element at any time under each type, and to perform real-time online anomaly identification of input data. The quality evaluation module is used to design evaluation indicators for the proportion and degree of anomalies in the input data based on the results of anomaly identification, and to complete the quantitative evaluation of the quality of real-time input data of the power prediction model. This includes: calculating the proportion of the number of abnormal matrix elements in the input data matrix to the total number of elements based on the number of abnormal results in the identification results, and determining the proportion of data anomalies; and calculating the degree of anomalies at individual points and the overall degree of anomalies of the row vectors corresponding to the input physical quantities based on the abnormal elements in the actual input matrix, and determining the distance between the abnormal values ​​and the upper and lower bounds of the normal values ​​of meteorological elements, thereby determining the degree of data anomalies.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the input data anomaly identification and evaluation method of the rate prediction model as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the input data anomaly identification and evaluation method of the rate prediction model as described in any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the input data anomaly identification and evaluation method of the rate prediction model as described in any one of claims 1 to 5.

Images