Cold-wave weather occurrence period identification method based on Prophet algorithm

Abstract

The invention discloses a cold-wave weather occurrence time period identification method based on a Prophet algorithm, and belongs to the technical field of cold-wave weather time period identification, and the identification method specifically comprises the following steps: I, selecting a time sequence of historical wind power data and meteorological data of a wind power plant power station, and cleaning the data; iI, establishing a Prophet prediction model, and setting model parameters; iII, constructing a trend model, detecting potential temperature mutation points, and performing sparse prior on the amplitude of trend change; iV, trend flexibility is adjusted, and the degree of sparse prior is adjusted by using input parameters; v, modeling is carried out on the periodic change, and meteorological early warning information published by a meteorological bureau is used as a special node accidental event influence part to carry out modeling; compared with a previous identification scheme, the method has the advantages that the accuracy and integrity of data are improved, reliable input is provided for subsequent data analysis and modeling work, and the occurrence time period of the cold wave event can be extracted more completely.

Description

Technical Field

[0001] The present invention relates to the technical field of cold wave weather period identification, and in particular to a cold wave weather occurrence period identification method based on Prophet algorithm. Background Art

[0002] In the context of developing a new power system with new energy as the main supplier, expanding the proportion of clean energy consumption represented by wind power and building a low-carbon, safe and efficient energy system are one of the important measures to achieve my country's "dual carbon" strategic goals. In recent years, affected by climate change with global warming as the main feature, the occurrence of extreme and abnormal weather has shown a significant increasing trend. Among them, cold wave weather represented by strong winds and freezing rain has the characteristics of low probability, high risk and great harm, and has the most significant impact on wind turbines. According to the current national standard "Cold Air Level": After the cold air passes through a certain area, the daily minimum temperature drops by 8℃ or more within 24 hours, or drops by 10℃ or more within 48 hours, or drops by 12℃ or more within 72 hours, and the daily minimum temperature drops to 4℃ or below, it can be considered that a cold wave has occurred. Cold wave is a complex extreme weather scene with multiple meteorological characteristics coupled, usually manifested as a large-scale drastic drop in temperature, strong winds and rain and snow in a certain area, often occurring in October, November and March of the following year. In the cold wave weather scenario, wind power output shows fluctuations of different characteristics on a short-term time scale, which may cause instantaneous power mutations, ramps and "cliff jumps", seriously affecting the accuracy of wind power prediction. Therefore, accurate extraction of cold wave periods is of vital importance to wind power prediction.

[0003] The key meteorological factor affecting the impact of cold waves on wind power is temperature. During the period of cold waves, the fluctuation characteristics of wind power are complex and changeable, and wind turbines may even shut down because the ambient temperature is lower than the minimum allowable operating temperature. The main impacts of cold waves on wind turbines are as follows: (1) The impact of low temperature on wind power generation efficiency. In a low temperature environment, the output characteristics of wind turbines will change. Low temperature will cause the components of wind turbines to become stiff, increase mechanical operation resistance, and thus affect the operating efficiency of wind turbines. In extremely low temperature conditions, mechanical components are easily affected by temperature changes and become brittle, increasing the risk of equipment damage. At the same time, extreme low temperatures may also cause failures of components such as cables and sensors, affecting the normal operation of the entire wind power system. (2) Uncertainty caused by strong winds. In cold wave weather, wind speed changes are large and unstable. The increased uncertainty may cause large fluctuations in the output power of wind turbines, further affecting the stability and reliability of the power grid. (3) Icing risk. Cold waves are often accompanied by meteorological characteristics such as low temperature and freezing rain. Under these conditions, the components of wind turbines are prone to ice, especially ice on the blade surface, which will lead to a decrease in aerodynamic performance and even cause the unit to shut down. (4) Rime affects fan performance. In cold and humid climates, rime may occur, causing the surface roughness of the fan blades to increase, affecting the aerodynamic performance of the blades and further affecting the fan output power. In severe cases, there is a risk of breakage.

[0004] At present, the means of observing cold wave weather mainly rely on the numerical weather forecast service provided by the meteorological department, and the real-time monitoring data provided by the meteorological measurement equipment near the wind farm station is insufficiently utilized. In addition, during the period of cold wave weather, the observation environment is poor, making it difficult to guarantee the quality of the measured data of the wind farm. Therefore, before extracting the cold wave period, the historical data of the wind farm needs to be cleaned. At present, data cleaning mainly includes two parts: outlier removal and missing value processing. Especially when the proportion of outlier data is high, the increase in the amount of outlier removal will destroy the integrity of the original data. Therefore, in order to ensure the adequacy and effectiveness of the data, the removed data needs to be reconstructed. Summary of the invention

[0005] The purpose of the present invention is to solve the defects in the prior art and to propose a method for identifying the period of cold wave weather based on the Prophet algorithm.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for identifying the period of cold wave weather based on the Prophet algorithm, the specific steps of the identification method are as follows:

[0008] Ⅰ. Select the historical wind power data and meteorological data time series of the wind farm power station and clean the data;

[0009] Ⅱ. Establish the Prophet prediction model and set the model parameters;

[0010] III. Construct a trend model to detect potential temperature mutation points and make a sparse prior for the magnitude of trend changes;

[0011] IV. Adjust the trend flexibility and use the input parameters to adjust the degree of sparse priors;

[0012] V. Model the periodic changes and take the meteorological warning information published by the Meteorological Bureau as the impact of accidental events at special nodes;

[0013] VI. Establish an additive regression model to obtain the overall prediction model of the Prophet algorithm and predict the temperature change trend;

[0014] Ⅶ. Based on the trend change amplitude and the time of historical cold wave events, define the rate change amplitude, monitor the temperature mutation points, and identify the time period when cold wave events occur.

[0015] As a further solution of the present invention, the specific steps of cleaning data in step I are as follows:

[0016] S1.1: The change point grouping algorithm is used to identify the shutdown data and some scattered abnormal data in the wind power curve data set. Then, the quartile algorithm is used to divide a set of data into 4 parts in ascending order, each of which accounts for 25% of the entire data set. The lower quartile is recorded as Q1, the median is recorded as Q2, and the upper quartile is recorded as Q3;

[0017] S1.2: Calculate the difference between the upper quartile and the lower quartile and record it as the interquartile range I QR , according to Q1-1.5I QR and Q3+1.5I QR Get the inner limit of outliers [F L , F U ] and remove the data beyond this range;

[0018] S1.3: Use the three-moment interpolation technique to construct a cubic spline interpolation function, and use the cubic spline interpolation method to fill in missing values ​​in the data. When the missing values ​​are less than 20%, the mean or median can be used to fill in the continuous variables, and the categorical variables are not filled in or are treated as a single category, or the mode is used to fill in. When the missing values ​​are between 20% and 80%, the mean or median can be used to fill in the continuous variables, and an indicator dummy variable is created for each variable with missing values. When the missing values ​​are greater than 80%, an indicator dummy variable is created for each variable with missing values ​​and is used in subsequent modeling. At the same time, the original variable is no longer used.

[0019] As a further solution of the present invention, the specific mathematical model of the Prophet prediction model in step II is as follows:

[0020] y(t)=g(t)+s(t)+h(t)+ε t

[0021] In the formula, g(t) represents the trend function, which is used to analyze the periodic changes in time series; s(t) represents the periodic changes, such as one year or one week; h(t) represents the influence of accidental factors at special time nodes; ε t represents random fluctuations, indicating the impact of errors not considered by the model;

[0022] The model parameters described in step II specifically include trend parameters, priority scale parameters and seasonal parameters;

[0023] The trend parameters specifically include growth parameters, changepoints parameters, n_changepoints parameters, changepoint_range and holidays;

[0024] The priority scale parameters specifically include a changepoint_prior_scale parameter, a seasonality_prior_scale parameter, and a holiday_prior_scale parameter;

[0025] The seasonal parameters specifically include a seasonality_yearly parameter, a weekly_seasonality parameter, and a daily_seasonality parameter.

[0026] As a further solution of the present invention, the growth parameter: specifies a growth model to determine the growth trend of the time series, and the default is "linear" linear trend;

[0027] The changepoints parameter includes a list of dates for potential mutation points. If no name is specified, it will be automatically identified by default.

[0028] The n_changepoints parameter: the number of potential change points automatically selected, set to 25;

[0029] The changepoint_range: selects the historical data ratio of the change point, which is set to the first 80% of the time series;

[0030] Said holidays: a data frame containing holiday information (special node accidental event information, i.e., weather warning information published by the Meteorological Bureau);

[0031] The changepoint_prior_scale parameter controls the sensitivity of the trend change point and needs to be optimized through cross-validation or grid search;

[0032] The seasonality_prior_scale parameter controls the sensitivity of the seasonal pattern and needs to be optimized through cross-validation or grid search;

[0033] The holiday_prior_scale parameter controls the sensitivity of the holiday effect and needs to be optimized through cross-validation or grid search;

[0034] The seasonality_yearly parameter is the annual seasonal component. If it is True, it will automatically calculate according to one year. Otherwise, you can enter an integer such as 2 or 3 years.

[0035] The weekly_seasonality parameter: weekly seasonal component, if True, it will automatically calculate based on one year, otherwise you can enter an integer 2, 3 years, etc.;

[0036] The daily_seasonality parameter: daily seasonal component, if True, it will automatically calculate according to one year, otherwise you can enter an integer 2, 3 years, etc.

[0037] As a further solution of the present invention, the trend model in step III is specifically generated by modeling the trend function g(t). The trend model consists of two parts: a piecewise linear model and a saturated growth model. The specific mathematical expression of the trend model is as follows:

[0038]

[0039] In the formula, g(t) represents the trend function; k represents the rate parameter of saturated growth; m(t) represents the linear trend;

[0040] The sparse prior described in step III can be regarded as L1 regularization, and the specific mathematical expression is as follows:

[0041]

[0042] Where w represents the weight vector of the model; n represents the length of the weight vector, that is, the number of weights; in the L1 regularization, the penalty term can be written as the L1 norm of the weight:

[0043] penalty = λ||w|| l

[0044] In the formula, λ represents the regularization parameter, which is used to control the strength of regularization;

[0045] The specific calculation formula of the L1 regularized loss function is as follows:

[0046]

[0047] Where, L data Represents the data loss of the model, usually the error between the model's predicted value and the true label; |w i | represents the absolute value of the model weight.

[0048] As a further solution of the present invention, the periodic variation modeling in step V is used to automatically detect and model seasonality in data, and to process various types of seasonal patterns, including daily, weekly and annual seasonality, which is modeled using Fourier series, and its specific mathematical expression is as follows:

[0049]

[0050] Where N represents the seasonal order and T represents the seasonal period.

[0051] As a further solution of the present invention, the special node accidental events in step V will cause predictable fluctuations in the time series. The dates before and after the accidental events of each special node are considered separately, and the effects of the accidental events are simulated by fitting additional parameters. The specific expression of the data model is as follows:

[0052] h(t)=∑(k j I)

[0053] In the formula, k j represents the intensity parameter; I represents the indicator function.

[0054] As a further solution of the present invention, the overall prediction model of the Prophet algorithm in step VI is specifically obtained by adding a trend function, a periodic change function, an accidental factor influence function and a random fluctuation term.

[0055] As a further solution of the present invention, when the rate change amplitude in step VII satisfies |α|≥10, it indicates that a short-term strong temperature change occurs, wherein α represents the rate change amplitude.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] This method for identifying the period of cold wave weather based on the Prophet algorithm first preprocesses the historical wind power data and meteorological data of the wind farm, and takes a series of measures to clean and optimize the data, thereby improving the accuracy and completeness of the data, and providing reliable input for subsequent data analysis and modeling work. Based on the definition standard of cold wave weather and comprehensive consideration of the meteorological warning information provided by the Meteorological Bureau, the Prophet algorithm is used to predict the short-term and approaching change trend of the real-time measured temperature of the wind farm station in the future. According to the rate change amplitude of the potential change point, the temperature mutation point is accurately identified, thereby more completely extracting the period of cold wave events. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0059] Figure 1 A flowchart of a method for identifying the period of occurrence of cold wave weather based on the Prophet algorithm proposed by the present invention;

[0060] Figure 2 A schematic diagram of the logical framework of a method for identifying the period of occurrence of cold wave weather based on the Prophet algorithm proposed in the present invention;

[0061] Figure 3 This is a wind farm abnormal data distribution diagram of a cold wave weather occurrence period identification method based on the Prophet algorithm proposed by the present invention;

[0062] Figure 4 This is a temperature change trend prediction effect diagram of a cold wave weather occurrence period identification method based on the Prophet algorithm proposed by the present invention;

[0063] Figure 5 This is a temperature change trend decomposition diagram of a method for identifying the period of cold wave weather based on the Prophet algorithm proposed by the present invention;

[0064] Figure 6 This is a schematic diagram of the rate change amplitude of potential change points in a method for identifying the period of cold wave weather based on the Prophet algorithm proposed in the present invention. DETAILED DESCRIPTION

[0065] Reference Figure 1-6 , a method for identifying the period of cold wave weather based on Prophet algorithm, the specific steps of the identification method are as follows:

[0066] Select the historical wind power data and meteorological data time series of the wind farm power station, and clean the data.

[0067] Specifically, refer to Figure 2 and Figure 3 It can be seen that the change point grouping algorithm is used to identify the shutdown data and some scattered abnormal data in the wind power curve data set. Then the quartile algorithm is used to divide a set of data into 4 parts in ascending order, each accounting for 25% of the entire data set. The lower quartile is recorded as Q1, the median is recorded as Q2, and the upper quartile is recorded as Q3. The difference between the upper quartile and the lower quartile is recorded as the interquartile range I QR , according to Q1-1.5I QR and Q3+1.5I QR Get the inner limit of outliers [F L , F U ], and the data beyond this range are eliminated, the cubic spline interpolation function is constructed using the three-moment interpolation technique, and the missing values ​​in the data are filled by the cubic spline interpolation method. When the missing values ​​are less than 20%, the mean or median can be used to fill in the continuous variables, and the categorical variables are not filled or are treated as a single category, or the mode is used to fill in. When the missing values ​​are between 20% and 80%, the mean or median can be used to fill in the continuous variables, and an indicator dummy variable is created for each variable with missing values. When the missing values ​​are greater than 80%, an indicator dummy variable is created for each variable with missing values ​​and is used in subsequent modeling, and the original variables are no longer used.

[0068] Establish the Prophet forecasting model and set the model parameters.

[0069] Specifically, the specific mathematical model of the Prophet prediction model is as follows:

[0070] y(t)=g(t)+s(t)+h(t)+ε t

[0071] In the formula, g(t) represents the trend function, which is used to analyze the periodic changes in time series; s(t) represents the periodic changes, such as one year or one week; h(t) represents the influence of accidental factors at special time nodes; ε t represents random fluctuations, indicating the impact of errors not considered by the model;

[0072] The Prophet forecasting model parameters specifically include trend parameters, priority scale parameters, and seasonal parameters.

[0073] The trend parameters specifically include growth parameters, changepoints parameters, n_changepoints parameters, changepoint_range and holidays;

[0074] The priority scale parameters specifically include the changepoint_prior_scale parameter, the seasonality_prior_scale parameter, and the holiday_prior_scale parameter;

[0075] Seasonality parameters include seasonality_yearly, weekly_seasonality, and daily_seasonality.

[0076] As a further solution of the present invention, the growth parameter: specifies the growth model to determine the growth trend of the time series, and the default is "linear" linear trend;

[0077] changepoints parameter: a list of dates including potential mutation points. If not specified, it will be automatically identified by default.

[0078] n_changepoints parameter: the number of potential change points automatically selected, set to 25;

[0079] changepoint_range: select the historical data ratio of the change point, set to the first 80% of the time series;

[0080] Holidays: A data frame containing holiday information (special node accidental event information, i.e., weather warning information published by the Meteorological Bureau);

[0081] Changepoint_prior_scale parameter: controls the sensitivity of trend change points, which needs to be optimized through cross-validation or grid search;

[0082] It should be further explained that if the trend change is overfitted (that is, too flexible) or underfitted (that is, not flexible enough), the input parameter changepoint_prior_scale is used to adjust the degree of sparse prior. The default value of this parameter is 0.05. Increasing the parameter value will make the trend fitting more flexible; reducing the parameter value will make the trend fitting less flexible.

[0083] seasonality_prior_scale parameter: controls the sensitivity of seasonal patterns and needs to be optimized through cross-validation or grid search;

[0084] holiday_prior_scale parameter: controls the sensitivity of the holiday effect, which needs to be optimized through cross-validation or grid search;

[0085] The seasonality_yearly parameter is the annual seasonal component. If it is True, it will automatically calculate based on one year. Otherwise, you can enter an integer such as 2 or 3 years.

[0086] weekly_seasonality parameter: weekly seasonal component, if True is enabled, it will automatically calculate based on one year, otherwise you can enter an integer of 2, 3 years, etc.;

[0087] daily_seasonality parameter: daily seasonal component. If True, it will be automatically calculated based on one year. Otherwise, you can enter an integer such as 2 or 3 years.

[0088] Build a trend model, detect potential temperature mutation points, and make sparse priors for the magnitude of trend changes.

[0089] Specifically, the trend model is generated by modeling the trend function g(t). The trend model consists of two parts: a piecewise linear model and a saturated growth model. The specific mathematical expression of the trend model is as follows:

[0090]

[0091] In the formula, g(t) represents the trend function; k represents the rate parameter of saturated growth; m(t) represents the linear trend;

[0092] The sparse prior can be regarded as L1 regularization, and the specific mathematical expression is as follows:

[0093]

[0094] Where w represents the weight vector of the model; n represents the length of the weight vector, that is, the number of weights; in L1 regularization, the penalty term can be written as the L1 norm of the weight:

[0095] penalty = λ||w|| l

[0096] In the formula, λ represents the regularization parameter, which is used to control the strength of regularization;

[0097] The specific calculation formula of the loss function of L1 regularization is as follows:

[0098]

[0099] Where, L data Represents the data loss of the model, usually the error between the model's predicted value and the true label; |w i | represents the absolute value of the model weight.

[0100] Adjust the trend flexibility and the degree of sparse prior using input parameters.

[0101] The periodic changes are modeled, and the meteorological warning information published by the Meteorological Bureau is modeled as the impact of accidental events at special nodes.

[0102] Specifically, refer to Figure 2 , Figure 4 as well as Figure 5 It can be seen that cyclical change modeling is used to automatically detect and model seasonality in data, and handle multiple types of seasonal patterns, including daily, weekly, and annual seasonality. It uses Fourier series for modeling, and its specific mathematical expression is as follows:

[0103]

[0104] Where N represents the seasonal order and T represents the seasonal period.

[0105] Specifically, accidental events at special nodes will lead to predictable fluctuations in the time series. The dates before and after the accidental events at each special node are considered separately, and the effects of the accidental events are simulated by fitting additional parameters. The specific expression of the data model is as follows:

[0106] h(t)=∑(k j I)

[0107] In the formula, k j represents the intensity parameter; I represents the indicator function.

[0108] An additive regression model was established to obtain the overall prediction model of the Prophet algorithm and predict the temperature change trend.

[0109] Specifically, the overall prediction model of the Prophet algorithm is obtained by adding the trend function, the periodic change function, the accidental factor influence function and the random fluctuation term.

[0110] Based on the trend change amplitude and the time of historical cold wave events, the rate change amplitude is defined, the temperature mutation points are monitored, and the time period when cold wave events occur is identified.

[0111] Specifically, refer to Figure 6 It can be seen that when the rate change amplitude satisfies |α|≥10, it indicates that the temperature has undergone a short-term strong change, where α represents the rate change amplitude.

Claims

1. A method for identifying cold wave weather period based on Prophet algorithm, characterized in that: The specific steps of the identification method are as follows: Ⅰ. Select the historical wind power data and meteorological data time series of the wind farm power station and clean the data; Ⅱ. Establish the Prophet prediction model and set the model parameters; III. Construct a trend model to detect potential temperature mutation points and make a sparse prior for the magnitude of trend changes; IV. Adjust the trend flexibility and use the input parameters to adjust the degree of sparse priors; V. Model the periodic changes and take the meteorological warning information published by the Meteorological Bureau as the impact of accidental events at special nodes; VI. Establish an additive regression model to obtain the overall prediction model of the Prophet algorithm and predict the temperature change trend; Ⅶ. Based on the trend change amplitude and the time of historical cold wave events, define the rate change amplitude, monitor the temperature mutation points, and identify the time period when cold wave events occur.

2. The method for identifying the period of occurrence of cold wave weather based on the Prophet algorithm according to claim 1 is characterized in that: The specific steps for cleaning data in step Ⅰ are as follows: S1.1: The change point grouping algorithm is used to identify the shutdown data and some scattered abnormal data in the wind power curve data set. Then, the quartile algorithm is used to divide a set of data into 4 parts in ascending order, each of which accounts for 25% of the entire data set. The lower quartile is recorded as Q1, the median is recorded as Q2, and the upper quartile is recorded as Q3; S1.2: Calculate the difference between the upper quartile and the lower quartile and record it as the interquartile range I QR , according to Q1-1.5I QR and Q3+1.5I QR Get the inner limit of outliers [F L , F U ] and remove the data beyond this range; S1.3: Use the three-moment interpolation technique to construct a cubic spline interpolation function, and use the cubic spline interpolation method to fill in missing values ​​in the data. When the missing values ​​are less than 20%, the mean or median can be used to fill in the continuous variables, and the categorical variables are not filled in or are treated as a single category, or the mode is used to fill in. When the missing values ​​are between 20% and 80%, the mean or median can be used to fill in the continuous variables, and an indicator dummy variable is created for each variable with missing values. When the missing values ​​are greater than 80%, an indicator dummy variable is created for each variable with missing values ​​and is used in subsequent modeling. At the same time, the original variable is no longer used.

3. The method for identifying the period of occurrence of cold wave weather based on Prophet algorithm according to claim 1 is characterized in that: The specific mathematical model of the Prophet prediction model described in step II is as follows: y(t)=g(t)+s(t)+h(t)+ε t In the formula, g(t) represents the trend function, which is used to analyze the periodic changes in time series; s(t) represents the periodic changes, such as one year or one week; h(t) represents the influence of accidental factors at special time nodes; ε t represents random fluctuations, indicating the impact of errors not considered by the model; The model parameters described in step II specifically include trend parameters, priority scale parameters and seasonal parameters; The trend parameters specifically include growth parameters, changepoints parameters, n_changepoints parameters, changepoint_range and holidays; The priority scale parameters specifically include a changepoint_prior_scale parameter, a seasonality_prior_scale parameter, and a holiday_prior_scale parameter; The seasonal parameters specifically include a seasonality_yearly parameter, a weekly_seasonality parameter, and a daily_seasonality parameter.

4. The method for identifying the period of occurrence of cold wave weather based on Prophet algorithm according to claim 3 is characterized in that: The trend model described in step III is specifically generated by modeling the trend function g(t). The trend model consists of two parts: a piecewise linear model and a saturated growth model. The specific mathematical expression of the trend model is as follows: In the formula, g(t) represents the trend function; k represents the rate parameter of saturated growth; m(t) represents the linear trend; The sparse prior described in step III can be regarded as L1 regularization, and the specific mathematical expression is as follows: Where w represents the weight vector of the model; n represents the length of the weight vector, that is, the number of weights; in the L1 regularization, the penalty term can be written as the L1 norm of the weight: penalty=λ||w|| l In the formula, λ represents the regularization parameter, which is used to control the strength of regularization; The specific calculation formula of the L1 regularized loss function is as follows: Where, L data Represents the data loss of the model, usually the error between the model's predicted value and the true label; |w i | represents the absolute value of the model weight.

5. The method for identifying the period of occurrence of cold wave weather based on Prophet algorithm according to claim 3 is characterized in that: The cyclical change modeling described in step V is used to automatically detect and model seasonality in the data, and handles multiple types of seasonal patterns, including daily, weekly, and annual seasonality. It uses Fourier series for modeling, and its specific mathematical expression is as follows: Where N represents the seasonal order and T represents the seasonal period.

6. The method for identifying the period of occurrence of cold wave weather based on Prophet algorithm according to claim 3 is characterized in that: The accidental events at the special nodes described in step V will lead to predictable fluctuations in the time series. The dates before and after the accidental events at each special node are considered separately, and the effects of the accidental events are simulated by fitting additional parameters. The specific expression of the data model is as follows: h(t)=∑(k j ·I) In the formula, k j represents the intensity parameter; I represents the indicator function.

7. The method for identifying the period of occurrence of cold wave weather based on Prophet algorithm according to claim 3 is characterized in that: The overall prediction model of the Prophet algorithm described in step VI is specifically obtained by adding a trend function, a periodic change function, an accidental factor influence function and a random fluctuation term.

8. The method for identifying the period of occurrence of cold wave weather based on Prophet algorithm according to claim 1 is characterized in that: When the rate change amplitude in step VII satisfies |α|≥10, it indicates that a short-term and strong temperature change occurs, wherein α represents the rate change amplitude.

Images