Flowering period prediction method based on GOD model

Abstract

The invention relates to the technical field of florescence prediction methods, in particular to a florescence prediction method based on a GOD model, which comprises the following steps: acquiring historical daily average temperature data of a target area, and preprocessing the historical daily average temperature data; based on the preprocessed historical daily average temperature data, constructing and training a long and short-term memory network model for predicting day-by-day temperature in a future time period, and outputting a day-by-day temperature sequence; on the basis of the predicted day-by-day temperature sequence, a growth day model is adopted to calculate a daily growth day and an accumulated growth day, and threshold values of an initial flowering stage and an end flowering stage are determined; the florescence length is obtained by subtracting the last florescence from the early florescence, and temperature data predicted by combining the GOD model with the LSTM model is adopted, so that high-precision phenological prediction is realized, and reliable support is provided for flower appreciation tourism and agricultural planning; the method can improve the prediction precision of the flowering phase.

Description

Technical Field

[0001] This invention relates to the field of flowering period prediction methods, and in particular to a flowering period prediction method based on the GOD model. Background Technology

[0002] As an important traditional festival, Qingming Festival not only carries profound humanistic emotions but also marks a significant stage in the transition of the natural seasons. At this time, spring is in full bloom in the south, while the north is beginning to warm up, making it a suitable time for spring outings and ancestor worship.

[0003] The period around Qingming Festival is also a time when many flowers bloom in concentrated bloom, such as apricot blossoms, rapeseed flowers, azaleas, cherry blossoms, and peonies, which bloom in succession from south to north, creating a magnificent spring flower viewing scene. However, the flowering period is significantly affected by meteorological factors such as temperature, sunlight, and precipitation, and is highly time-sensitive and difficult to predict, resulting in insufficient accuracy of existing flowering period prediction methods. Summary of the Invention

[0004] The purpose of this invention is to provide a flowering period prediction method based on the GOD model, which can improve the accuracy of flowering period prediction.

[0005] To achieve the above objectives, this invention provides a flowering period prediction method based on the GOD model, comprising:

[0006] Obtain historical daily average temperature data for the target area and preprocess the historical daily average temperature data;

[0007] Based on preprocessed historical daily average temperature data, a long short-term memory network model is constructed and trained to predict the daily temperature in future time periods and output the daily temperature sequence.

[0008] Based on the predicted daily temperature series, the daily growth days and cumulative growth days are calculated using the growth day model to determine the thresholds for the initial flowering period and the final flowering period.

[0009] The length of the flowering period is obtained by subtracting the end of the flowering period from the beginning of the flowering period.

[0010] The specific steps for preprocessing historical daily average temperature data include:

[0011] Missing values ​​in historical daily average temperature data are filled in using linear interpolation, and all daily average temperature data are standardized.

[0012] In the step of constructing and training a long short-term memory network model based on preprocessed historical daily average temperature data to predict daily temperatures for future periods and output daily temperature sequences,

[0013] The Long Short-Term Memory (LSTM) network model adopts a network architecture with two LSTM units superimposed on a fully connected output layer. The LSM network model uses MSE as the loss function and is trained using the Adam optimizer.

[0014] Among the steps, based on the predicted daily temperature series, a growth degree-day model is used to calculate the daily growth degree-days and cumulative growth degree-days, and to determine the thresholds for the initial flowering period and the final flowering period.

[0015] In the aforementioned growth-day model, the formula for calculating the cumulative growth-days is:

[0016]

[0017] in, This indicates the average temperature of the day. This represents the baseline temperature, which is the minimum temperature threshold for plant development. This represents the cumulative growth days from the start date to the current date. Indicates the current date.

[0018] Among the steps, based on the predicted daily temperature series, a growth degree-day model is used to calculate the daily growth degree-days and cumulative growth degree-days, and to determine the thresholds for the initial flowering period and the final flowering period.

[0019] The formula for calculating the daily threshold of growth degree at the initial flowering stage is:

[0020]

[0021] in, This indicates the required degree-day threshold for the initial flowering period. The baseline parameter representing the daily threshold of growth at the initial flowering stage. The magnitude parameter of the exponential term representing the daily threshold of growth at the initial flowering stage. The temperature scaling factor represents the daily threshold of growth degree at the initial flowering stage. This represents the average temperature within the optimal time window. , and Determined using a double annealing optimization algorithm.

[0022] Among the steps, based on the predicted daily temperature series, a growth degree-day model is used to calculate the daily growth degree-days and cumulative growth degree-days, and to determine the thresholds for the initial flowering period and the final flowering period.

[0023] The formula for calculating the daily threshold of growth degree at the end of flowering period is:

[0024]

[0025] in, This represents the life-day threshold required for the final flowering stage. The baseline parameter representing the daily threshold of growth degree at the end of flowering period. The magnitude parameter of the exponential term representing the daily threshold of growth at the end of the flowering period. The temperature scaling factor represents the daily threshold of growth degree at the end of the flowering period. , and Determined using a double annealing optimization algorithm.

[0026] This invention discloses a flowering period prediction method based on the GOD model. By combining the temperature data predicted by the GOD model with the LSTM model, it achieves high-precision phenological prediction and provides reliable support for flower viewing tourism and agricultural planning. This invention can improve the accuracy of flowering period prediction. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0028] Figure 1 This is a flowchart of a flowering period prediction method based on the GOD model according to the present invention.

[0029] Figure 2 This is a schematic diagram of the validation set effect of cherry blossoms in Wuhan.

[0030] Figure 3 This is a table of error indicators for predicting flowering period.

[0031] Figure 4 This is a table showing the predicted flowering period of rapeseed flowers in Wuyuan.

[0032] Figure 5 This is a table showing the predicted blooming period of cherry blossoms in Wuhan. Detailed Implementation

[0033] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0034] Please see Figures 1-5 This invention provides a flowering period prediction method based on the GOD model, comprising:

[0035] S1 acquires historical daily average temperature data for the target area and preprocesses the historical daily average temperature data;

[0036] The specific steps for preprocessing historical daily average temperature data include:

[0037] Missing values ​​in historical daily average temperature data are filled in using linear interpolation, and all daily average temperature data are standardized.

[0038] S2 constructs and trains a long short-term memory network model based on preprocessed historical daily average temperature data to predict daily temperatures in future time periods and output daily temperature sequences.

[0039] In this step, the Long Short-Term Memory (LSTM) network model adopts a network architecture with two LSTM units superimposed on a fully connected output layer. The LSM network model uses MSE as the loss function and is trained using the Adam optimizer.

[0040] S3 uses a growth degree-day model based on predicted daily temperature sequences to calculate daily growth degree-days and cumulative growth degree-days, and determines the thresholds for the initial flowering period and the final flowering period.

[0041] In this step, the formula for calculating the cumulative growth days in the growth-day model is as follows:

[0042]

[0043] in, This indicates the average temperature of the day. This represents the baseline temperature, which is the minimum temperature threshold for plant development. This represents the cumulative growth days from the start date to the current date. Indicates the current date.

[0044] The formula for calculating the daily threshold of growth degree at the initial flowering stage is:

[0045]

[0046] in, This indicates the required degree-day threshold for the initial flowering period. The baseline parameter representing the daily threshold of growth at the initial flowering stage. The magnitude parameter of the exponential term representing the daily threshold of growth at the initial flowering stage. The temperature scaling factor represents the daily threshold of growth degree at the initial flowering stage. This represents the average temperature within the optimal time window. , and Determined using a double annealing optimization algorithm.

[0047] The formula for calculating the daily threshold of growth degree at the end of flowering period is:

[0048]

[0049] in, This represents the life-day threshold required for the final flowering stage. The baseline parameter representing the daily threshold of growth degree at the end of flowering period. The magnitude parameter of the exponential term representing the daily threshold of growth at the end of the flowering period. The temperature scaling factor represents the daily threshold of growth degree at the end of the flowering period. , and Determined using a double annealing optimization algorithm.

[0050] S4 obtains the length of the flowering period by subtracting the end of the flowering period from the beginning of the flowering period;

[0051] To better understand the present invention, an embodiment is described below.

[0052] In this embodiment, the goal is to accurately predict the blooming times of rapeseed flowers and cherry blossoms, typical flowers in Wuyuan and Wuhan, in the spring of 2026. Flowering period data from 2006 to 2018 and temperature data from 2006 to 2025 were collected online. A Long Short-Term Memory (LSTM) network model was then trained to predict the temperature data for 2026. Based on common knowledge, cherry blossoms and peonies typically bloom between March and April, so the predicted temperature range in this embodiment is also between January and April 2026. Next, a Growing Days (GDD) model was constructed. This embodiment established a single-flower-species GDD model, calculating its daily and cumulative GDD to determine the thresholds for the beginning and end of the flowering period. The length of the flowering period was then obtained by subtracting the end of the flowering period from the beginning of the flowering period.

[0053] The first step is to calculate the cumulative GDD formula. First, calculate the daily GDD, and then sum them to obtain the cumulative GDD. The parameter t used is the prediction result of the LSTM model.

[0054]

[0055] in, This indicates the average temperature of the day. This represents the baseline temperature, which is the minimum temperature threshold for plant development. This represents the cumulative growth days from the start date to the current date. Indicates the current date.

[0056] Below is the formula for the GDD threshold at the initial flowering stage. Plants need to accumulate a specific amount of GDD to enter the initial flowering stage, and this GDD is the GDD threshold. The three parameters d, e, and f are determined through double annealing optimization.

[0057]

[0058] in, This indicates the required degree-day threshold for the initial flowering period. The baseline parameter representing the daily threshold of growth at the initial flowering stage. The magnitude parameter of the exponential term representing the daily threshold of growth at the initial flowering stage. The temperature scaling factor represents the daily threshold of growth degree at the initial flowering stage. This represents the average temperature within the optimal time window.

[0059] The GDD threshold formula for the end-of-flowering period is similar to that for the beginning-of-flowering period. The end-of-flowering period (EFD) also requires the accumulation of a certain amount of GDD, but the threshold is usually higher than that for the beginning-of-flowering period, reflecting the extra heat required to continue the flowering period.

[0060]

[0061] in, This represents the life-day threshold required for the final flowering stage. The baseline parameter representing the daily threshold of growth degree at the end of flowering period. The magnitude parameter of the exponential term representing the daily threshold of growth at the end of the flowering period. The temperature scaling factor represents the daily threshold of growth degree at the end of the flowering period.

[0062] Taking cherry blossoms in Wuhan as an example, preprocessed flowering period and temperature data are input into the aforementioned GDD and LSTM models. Python code is used to obtain predicted temperature data for 2026, thereby yielding predictions for the flowering period and initial flowering date in 2026. This embodiment uses data from 2015 to 2018 as the validation set. Figure 2 This study evaluates the effectiveness of the prediction of the initial flowering date and flowering period of cherry blossoms in Wuhan on the validation set. The final validation set evaluation results are as follows: Figure 3 As shown, the predicted temperature for 2026 was finally input into the model to predict the onset and flowering periods for that time. The corresponding prediction results are as follows. Figures 4-5 As shown, taking Wuhan cherry blossoms as an example, the root mean square error (RMSE) for predicting the initial flowering period is approximately 5.6 days, while the RMSE for the length of the flowering period is only 1.32 days.

[0063] In another embodiment, the present invention can introduce an online learning mechanism to dynamically update the model parameters after acquiring new flowering period and temperature data each year. An incremental double-annealing algorithm is used to fine-tune only the newly added data; simultaneously, since the number of influencing factors referenced is relatively small, corresponding influencing factors can be added to enable the model to predict flowering periods over a wider range.

[0064] This invention discloses a flowering period prediction method based on the GOD model. By combining the temperature data predicted by the GOD model with the LSTM model, it achieves high-precision phenological prediction and provides reliable support for flower viewing tourism and agricultural planning. This invention can improve the accuracy of flowering period prediction.

[0065] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.

Claims

1. A method for predicting flowering time based on a GOD model, characterized in that, The method comprises the following steps: obtaining historical daily average temperature data of a target area, and preprocessing the historical daily average temperature data; based on the preprocessed historical daily average temperature data, constructing and training a long short-term memory network model for predicting daily temperature in a future period, and outputting a daily temperature sequence; based on the predicted daily temperature sequence, calculating daily length-of-day and cumulative length-of-day by using a length-of-day model, and determining threshold values of the beginning and end of the flowering period; obtaining the length of the flowering period by subtracting the end of the flowering period from the beginning of the flowering period.

2. The flower stage prediction method based on the GOD model according to claim 1, wherein, The specific steps of preprocessing the historical daily average temperature data comprise: linear interpolation is performed on missing values of the historical daily average temperature data, and all the daily average temperature data are standardized.

3. The flower stage prediction method based on the GOD model according to claim 2, characterized in that, In the step of constructing and training the long short-term memory network model based on the preprocessed historical daily average temperature data for predicting daily temperature in a future period and outputting a daily temperature sequence, the long short-term memory network model adopts a network architecture of two layers of LSTM units stacked with one layer of fully connected output layer, and the long short-term memory network model uses MSE as a loss function and uses an Adam optimizer for training.

4. The flower stage prediction method based on the GOD model according to claim 3, characterized in that, In the step of calculating daily length-of-day and cumulative length-of-day by using a length-of-day model based on the predicted daily temperature sequence, and determining threshold values of the beginning and end of the flowering period, in the length-of-day model, the calculation formula of the cumulative length-of-day is: ； wherein, represents the average temperature of the day, represents the base temperature, i.e. the minimum temperature threshold for plant development, represents the accumulated growing degree days from the start day to the current day, represents the current date.

5. The GOD model-based flowering date prediction method of claim 4, wherein, In the step of calculating daily length-of-day and cumulative length-of-day by using a length-of-day model based on the predicted daily temperature sequence, and determining threshold values of the beginning and end of the flowering period, the calculation formula of the beginning-of-flowering-period length-of-day threshold value is: ； wherein, denotes the threshold of the length of day required for the onset of flowering, denotes a baseline parameter for the threshold of the length of day for the onset of flowering, denotes an amplitude parameter for the exponential term of the threshold of the length of day for the onset of flowering, denotes a temperature scaling factor for the threshold of the length of day for the onset of flowering, denotes the average temperature of the optimal time window, , and is determined by a double annealing optimization algorithm.

6. The flower stage prediction method based on the GOD model according to claim 5, wherein, In the step of calculating daily length-of-day and cumulative length-of-day by using a length-of-day model based on the predicted daily temperature sequence, and determining threshold values of the beginning and end of the flowering period, the calculation formula of the end-of-flowering-period length-of-day threshold value is: ； wherein, represents the threshold of the growing degree days required at the end of the flowering period, represents a baseline parameter of the threshold of the growing degree days at the end of the flowering period, represents an amplitude parameter of the exponential term of the threshold of the growing degree days at the end of the flowering period, represents a temperature scaling factor of the threshold of the growing degree days at the end of the flowering period, , and are determined by a double annealing optimization algorithm.

Images