A method for optimizing the application decision of controlled-release nitrogen fertilizer combined with the characteristics of soil water content change

Abstract

The present application relates to the field of data analysis, and more particularly to a method for optimizing controlled-release nitrogen fertilizer application decision-making combined with soil moisture change characteristics, which comprises: obtaining an effective infiltration potential optimization factor by thresholding the field water-holding capacity of the soil volume moisture content sequence and evaluating the cumulative difference of the effective infiltration potential time window; obtaining a water potential lag direction optimization factor by coupling the gradient direction consistency of the first-order difference of the soil volume moisture content sequence and the trend amplitude gating; correcting the DTW local distance by the effective infiltration potential optimization factor and the water potential lag direction optimization factor; and obtaining precise controlled-release nitrogen fertilizer application decision-making suggestions by weighted prediction of the optimal matching historical sequence corresponding to the optimized DTW distance, so as to solve the problem of historical scene matching distortion and controlled-release nitrogen fertilizer application decision-making deviation caused by the neglect of effective cumulative effect and water potential lag effect in the existing Euclidean distance measurement.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a method for optimizing controlled-release nitrogen fertilizer application decisions by incorporating soil moisture content variation characteristics. Background Technology

[0002] Controlled-release nitrogen fertilizer, as an important input for improving nitrogen use efficiency in modern agriculture, releases nutrients into the soil environment at a relatively slow and adjustable rate by coating the outer surface of urea, nitrate nitrogen fertilizer, or other nitrogen-containing fertilizer granules with a coating material. This better aligns with the nutrient requirements of plants at different growth stages. In actual production, the release process of controlled-release nitrogen fertilizer is not a simple linear diffusion but is influenced by multiple physical mechanisms, including water absorption and swelling of the coating material, osmotic pressure driving, and changes in pore structure. Among these, soil moisture conditions are a key external environmental factor determining the permeability of the coating and the ability of nutrients to migrate. Especially under conditions of alternating droughts and floods, sudden rainfall, frequent irrigation, and significant differences in soil texture, the temporal changes in soil volumetric water content can significantly alter the water potential difference and solute diffusion channels inside and outside the coating, leading to significant differences in the actual release curves of the same fertilizer application rate in different plots or years. Since changes in the release curve are directly related to the timeliness and excess of crop nitrogen supply, as well as environmental risks such as nitrogen leaching and volatilization, optimizing controlled-release nitrogen fertilizer application decisions based on soil moisture content changes at the field scale has become an important research direction in precision agriculture and intelligent fertilization management. With the development of the Internet of Things and agricultural big data platforms, more and more intelligent agricultural decision-making systems are deploying soil moisture sensors in fields to continuously acquire time-series data on soil volumetric moisture content. This data is then compared and analyzed with historical monitoring records from the same or similar fields in historical databases to support precise recommendations for fertilization timing and amount. In existing systems, a common approach is to use the current growing season's soil volumetric moisture content time-series data as the query sequence and multiple historical soil volumetric moisture content records from the historical database as a reference sequence set. A time-series matching algorithm is used to retrieve the most similar historical scenario from the reference sequence set to the query sequence. Then, fertilizer release monitoring data and yield data from similar historical scenarios are reused to infer the release trend under current conditions and form fertilization decisions. Dynamic Time Warping (DTW), a classic algorithm capable of aligning two sequences through time-axis stretching and compression, is widely used for similarity retrieval and pattern matching of agricultural environmental time-series data due to its ability to handle water change processes at different time scales or stages. However, in traditional DTW algorithms, local distance is typically measured using Euclidean distance, which accumulates the overall matching distance by measuring the numerical differences between the query and reference sequences at corresponding time points. This metric primarily reflects the geometrical degree of numerical closeness. Since controlled-release nitrogen fertilizer release exhibits significant threshold triggering and cumulative lag characteristics, and even if the soil moisture content is the same during hygroscopic and desiccant processes, different matrix suction states may correspond to different conditions, time-series matching relying solely on Euclidean distance can easily misjudge scenarios with significantly different physical release characteristics as highly similar. This leads to inconsistencies between historical scene retrieval results and the actual release potential, ultimately causing biases in controlled-release nitrogen fertilizer application decisions.Therefore, it is necessary to propose a similarity measurement and decision optimization method that can address the release mechanism of controlled-release fertilizers and reflect the effective infiltration accumulation effect and water potential lag effect during the time-series matching process, so as to improve the reliability and applicability of intelligent fertilization decisions. Summary of the Invention

[0003] In view of this, the present invention aims to propose an optimization method for controlled-release nitrogen fertilizer application decisions that incorporates soil moisture variation characteristics, in order to solve the problem that existing Euclidean distance metrics ignore the effective cumulative effect and water potential lag effect, resulting in historical scenario matching distortion and controlled-release nitrogen fertilizer application decision bias.

[0004] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0005] A method for optimizing controlled-release nitrogen fertilizer application decisions by incorporating soil moisture content variation characteristics, the method comprising:

[0006] Step S1: Obtain and preprocess the soil volumetric moisture content query sequence and reference sequence set through the Internet of Things sensor network and historical database;

[0007] Step S2: Obtain the effective infiltration potential optimization factor by evaluating the cumulative difference between the field holding capacity threshold cutoff and the effective infiltration potential time window of the soil volumetric water content sequence.

[0008] Step S3: Obtain the water potential lag direction optimization factor by evaluating the coupling of the first-order difference gradient direction consistency and trend amplitude of the soil volumetric water content sequence;

[0009] Step S4: By multiplicatively fusing the effective wetting potential energy optimization factor and the water potential hysteresis direction optimization factor and correcting the DTW local distance, the physical consistency optimized DTW distance is obtained and the time series matching is completed;

[0010] Step S5: Obtain accurate controlled-release nitrogen fertilizer application decision recommendations by performing release curve weighted prediction on the optimal matching historical sequence corresponding to the DTW distance optimized by physical consistency.

[0011] Furthermore, the acquisition and preprocessing of soil volumetric moisture content query sequences and reference sequence sets through IoT sensor networks and historical databases includes:

[0012] Soil moisture sensors were deployed in the crop root activity layer of the plot to be decided, and sampling time intervals were set. The soil moisture sensors were continuously collected according to the sampling time intervals to obtain the time series data of soil volumetric water content in the current growing season, and the soil volumetric water content time series data was constructed into a soil volumetric water content query sequence.

[0013] Based on the historical database of the agricultural big data platform, historical records of soil volumetric moisture content for the same period in several past years for the plot to be decided or similar plots are retrieved to construct a soil volumetric moisture content reference sequence set, where each historical record of soil volumetric moisture content corresponds to a soil volumetric moisture content reference sequence; the field capacity of the plot to be decided is obtained by consulting the soil physical property test report or by using soil moisture characteristic curve measurement experiments, and the field capacity is used as a known soil physical property constant; the soil volumetric moisture content query sequence and the soil volumetric moisture content reference sequence set are preprocessed, and the data preprocessing includes at least outlier data extraction, missing data linear interpolation to complete the time step consistency of one season.

[0014] Furthermore, the method of obtaining the effective infiltration potential optimization factor by evaluating the cumulative difference between the field holding capacity threshold cutoff and the effective infiltration potential time window of the soil volumetric water content sequence includes:

[0015] By performing nonlinear squared enhancement processing on the field water holding capacity threshold truncated data of soil volume water content query sequence and reference sequence set, the effective infiltration potential energy accumulation characteristic data is obtained.

[0016] By performing time window integral accumulation and difference normalization evaluation on the accumulated characteristic data of effective wetting potential energy, the optimization factor of effective wetting potential energy is obtained.

[0017] Furthermore, the method of obtaining effective infiltration potential energy accumulation characteristic data by performing nonlinear squared enhancement processing on the field water holding capacity threshold truncated data of the soil volumetric water content query sequence and the reference sequence set includes:

[0018] For any target time point in the soil volumetric moisture content query sequence, extract the soil volumetric moisture content data for each time step within the target time point and the preset historical backtracking time window before the target time point. Use the difference between the soil volumetric moisture content data and field capacity at each time step as the moisture content exceeding the threshold difference data. When the moisture content exceeding the threshold difference data is less than or equal to a constant 0, set the field capacity threshold truncation data for the corresponding time step to a constant 0. When the moisture content exceeding the threshold difference data is greater than a constant 0, use the ratio of the moisture content exceeding the threshold difference data to the field capacity as the field capacity threshold truncation data for the corresponding time step. Perform a square operation on the field capacity threshold truncation data to obtain the squared enhanced threshold truncation data corresponding to the soil volumetric moisture content query sequence.

[0019] For any soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set, for any target time point in the soil volumetric moisture content reference sequence, extract the soil volumetric moisture content data of the target time point and each time step within a preset historical backtracking time window before the target time point. The difference between the soil volumetric moisture content data and field capacity at each time step is used as the moisture content exceeding threshold difference data. When the moisture content exceeding threshold difference data is less than or equal to a constant 0, the field capacity threshold truncation data of the corresponding time step is set to a constant 0. When the moisture content exceeding threshold difference data is greater than a constant 0, the ratio of the moisture content exceeding threshold difference data to field capacity is used as the field capacity threshold truncation data of the corresponding time step. The field capacity threshold truncation data is squared to obtain the square-enhanced threshold truncation data corresponding to the soil volumetric moisture content reference sequence. The square-enhanced threshold truncation data corresponding to the soil volumetric moisture content query sequence and the soil volumetric moisture content reference sequence are used as the effective infiltration potential energy accumulation feature data.

[0020] Furthermore, the method of obtaining the effective wetting potential energy optimization factor by performing time window integral accumulation and difference normalization evaluation on the effective wetting potential energy accumulation characteristic data includes:

[0021] For any first target time point in the soil volumetric moisture content query sequence and any second target time point in the soil volumetric moisture content reference sequence, extract the square-strengthened threshold truncated data corresponding to each time step within a preset historical backtracking time window before the first target time point from the effective infiltration potential energy cumulative feature data. Then, accumulate the square-strengthened threshold truncated data corresponding to each time step within the preset historical backtracking time window to obtain the effective infiltration potential energy integral cumulative value within the query window corresponding to the first target time point. Extract the square-strengthened threshold truncated data corresponding to each time step within a preset historical backtracking time window before the second target time point from the effective infiltration potential energy cumulative feature data. Then, accumulate the square-strengthened threshold truncated data corresponding to each time step within the preset historical backtracking time window to obtain the effective infiltration potential energy integral cumulative value within the reference window corresponding to the second target time point.

[0022] The absolute value of the difference between the cumulative effective immersion potential energy integral value in the query window and the cumulative effective immersion potential energy integral value in the reference window is used as the effective immersion potential energy integral difference assessment; the minimum value between the cumulative effective immersion potential energy integral value in the query window and the cumulative effective immersion potential energy integral value in the reference window is used as the effective immersion potential energy integral normalization benchmark; a minimum positive number stability protection term is set, the result of adding the stability protection term and the effective immersion potential energy integral normalization benchmark is used as the normalization denominator, the effective immersion potential energy integral difference assessment is used as the normalization numerator, and the ratio of the normalization numerator to the normalization denominator is used as the effective immersion potential energy difference normalization assessment value.

[0023] The calculation result of adding the normalized evaluation value of the effective wetting potential energy difference to the natural constant is used as the input value of the logarithmic mapping. The natural logarithmic operation is performed on the logarithmic mapping input value to obtain the effective wetting potential energy optimization factor corresponding to the first target time point and the second target time point.

[0024] Furthermore, the method of obtaining the water potential lag direction optimization factor through the coupled evaluation of the consistency of the first-order difference gradient direction and the trend amplitude of the soil volumetric water content sequence includes:

[0025] By performing first-order difference processing on the moisture content data of adjacent time steps in the soil volume moisture content query sequence and the reference sequence set, the gradient data of the query sequence and the gradient data of the reference sequence are obtained.

[0026] Gradient direction consistency evaluation data is obtained by performing normalized inner product calculation on the gradient data of the query sequence and the gradient data of the reference sequence.

[0027] By performing trend amplitude gating coupling processing on gradient direction consistency assessment data, query sequence gradient data, and reference sequence gradient data, the water potential lag direction optimization factor is obtained.

[0028] Furthermore, the step of obtaining gradient data of the query sequence and gradient data of the reference sequence by performing first-order difference processing on the moisture content data of adjacent time steps in the soil volumetric moisture content query sequence and the reference sequence set includes:

[0029] For any first target time point in the soil volumetric moisture content query sequence, extract the soil volumetric moisture content data corresponding to the first target time point and the soil volumetric moisture content data corresponding to the adjacent time steps before the first target time point from the soil volumetric moisture content query sequence. Use the difference between the soil volumetric moisture content data corresponding to the first target time point and the soil volumetric moisture content data corresponding to the adjacent time steps as the query sequence gradient data corresponding to the first target time point. Construct the query sequence gradient dataset by using the query sequence gradient data corresponding to each first target time point in the soil volumetric moisture content query sequence.

[0030] For any soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set, for any second target time point in the soil volumetric moisture content reference sequence, extract the soil volumetric moisture content data corresponding to the second target time point and the soil volumetric moisture content data corresponding to the adjacent time steps before the second target time point from the soil volumetric moisture content reference sequence. Use the difference between the soil volumetric moisture content data corresponding to the second target time point and the soil volumetric moisture content data corresponding to the adjacent time steps as the reference sequence gradient data corresponding to the second target time point. Construct a reference sequence gradient dataset from the reference sequence gradient data corresponding to each second target time point in the soil volumetric moisture content reference sequence. Use the query sequence gradient dataset as the query sequence gradient data and the reference sequence gradient dataset as the reference sequence gradient data.

[0031] Furthermore, the step of obtaining gradient direction consistency evaluation data by performing normalized inner product calculation on the query sequence gradient data and the reference sequence gradient data includes:

[0032] For any first target time point in the soil volumetric moisture content query sequence and any second target time point in the soil volumetric moisture content reference sequence, extract the query sequence gradient data corresponding to the first target time point from the query sequence gradient data, and extract the reference sequence gradient data corresponding to the second target time point from the reference sequence gradient data; use the product of the query sequence gradient data and the reference sequence gradient data as the gradient product evaluation corresponding to the first target time point and the second target time point;

[0033] A minimum positive stability protection term is set. The sum of the square of the gradient data of the query sequence and the stability protection term is used as the first square stability term, and the sum of the square of the gradient data of the reference sequence and the stability protection term is used as the second square stability term. Square root operations are performed on the first square stability term and the second square stability term to obtain the first gradient magnitude evaluation and the second gradient magnitude evaluation respectively. The product of the first gradient magnitude evaluation and the second gradient magnitude evaluation is used as the normalized denominator, the gradient product evaluation is used as the normalized numerator, and the ratio of the normalized numerator to the normalized denominator is used as the gradient direction consistency evaluation data.

[0034] Furthermore, the step of obtaining the water potential hysteresis direction optimization factor by performing trend amplitude gating coupling processing on the gradient direction consistency evaluation data, the query sequence gradient data, and the reference sequence gradient data includes:

[0035] For any first target time point in the soil volumetric moisture content query sequence and any second target time point in the soil volumetric moisture content reference sequence, extract the gradient direction consistency assessment data corresponding to the first target time point and the second target time point from the gradient direction consistency assessment data, extract the query sequence gradient data corresponding to the first target time point from the query sequence gradient data, and extract the reference sequence gradient data corresponding to the second target time point from the reference sequence gradient data; use the difference between the constant 1 and the gradient direction consistency assessment data as the gradient direction conflict assessment.

[0036] The absolute value of the difference between the gradient data of the query sequence and the gradient data of the reference sequence is used as the gradient difference magnitude assessment; the gradient difference magnitude assessment is processed by hyperbolic tangent function mapping to obtain the corresponding trend magnitude gating assessment data.

[0037] The result of multiplying the gradient direction conflict assessment data and the trend amplitude gating assessment data is used as the gating coupling modulation quantity. The gating coupling modulation quantity is then subjected to exponential mapping with the natural constant as the base to obtain the water potential lag direction optimization factor corresponding to the first target time point and the second target time point.

[0038] Furthermore, the step of multiplicatively fusing the effective wetting potential energy optimization factor and the water potential hysteresis direction optimization factor, correcting the DTW local distance to obtain the physically consistent optimized DTW distance, and completing the time series matching includes:

[0039] For any first target time point in the soil volumetric moisture content query sequence and any second target time point in any soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set, obtain the effective infiltration potential energy optimization factor and the water potential lag direction optimization factor corresponding to the first target time point and the second target time point; and take the product of the effective infiltration potential energy optimization factor and the water potential lag direction optimization factor as the multiplicative fusion optimization factor.

[0040] Soil volumetric moisture content data corresponding to the first target time point is extracted from the soil volumetric moisture content query sequence, and soil volumetric moisture content data corresponding to the second target time point is extracted from the soil volumetric moisture content reference sequence. The difference between the two is used as the moisture content difference assessment. The square of the moisture content difference assessment is used as the basic DTW local distance squared term. The result of multiplying the multiplicative fusion optimization factor with the basic DTW local distance squared term is used as the corrected DTW local distance between the first target time point and the second target time point.

[0041] A cumulative distance matrix is ​​constructed with the time steps of the soil volumetric moisture content query sequence as rows and the time steps of the soil volumetric moisture content reference sequence as columns. The modified DTW local distance is used as the local distance term of the corresponding matrix element in the cumulative distance matrix. According to the dynamic programming recursion rule, for any matrix element in the cumulative distance matrix, the calculated result of adding the modified DTW local distance corresponding to the matrix element to the minimum cumulative distance of its adjacent preceding matrix elements is used as the cumulative distance value of the matrix element. The adjacent preceding matrix elements include at least the left matrix element, the lower matrix element, and the lower left matrix element. The cumulative distance value corresponding to the terminal matrix element of the cumulative distance matrix is ​​used as the physical consistency optimized DTW distance between the soil volumetric moisture content query sequence and the soil volumetric moisture content reference sequence.

[0042] The physical consistency optimization DTW distance is calculated for each soil volumetric moisture content query sequence and each soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set. The soil volumetric moisture content reference sequence with the smallest physical consistency optimization DTW distance is determined as the optimal matching historical sequence, thereby completing the time series matching.

[0043] Compared with the prior art, the present invention has the following advantages:

[0044] This invention presents a controlled-release nitrogen fertilizer application decision optimization method that incorporates soil moisture content variation characteristics. It introduces the actual mechanism of soil moisture content's influence on membrane penetration and nutrient release into time-series similarity calculations. By using a field capacity threshold to suppress background fluctuations in low-moisture ranges, and by nonlinearly enhancing and accumulating time windows for effective moisture content exceeding the threshold, the similarity judgment shifts from "simultaneous numerical similarity" to "simultaneous effective infiltration contribution." Therefore, even if a moisture content spike occurs due to short-term heavy rainfall, this method can distinguish it from a sustained high-humidity state caused by continuous rain, avoiding misjudging pulsed high humidity with minimal contribution to nutrient release as a high-release-potential scenario. This significantly reduces historical case retrieval bias and improves the reliability of release trend prediction and the effectiveness of fertilizer parameter reuse.

[0045] In the context of dynamic soil moisture changes, this solution further utilizes the first-order differential gradient to characterize the direction and intensity of moisture changes. By assessing directional consistency and using trend amplitude gating to suppress pseudo-directional reversals caused by sensor perturbations during the steady-state phase, it effectively identifies the differences in "matrix suction at the same moisture content" caused by water potential lag during hygroscopic and desiccant processes. Therefore, when retrieving historical scenarios, the system not only matches moisture content levels but also the stage of moisture change and the microenvironmental water potential state, reducing the probability of confusing active and inhibited release periods. This results in stronger temporal consistency in the predicted controlled-release fertilizer release curve, thereby improving the stability and accuracy of fertilization time and dosage recommendations. This ensures yield while reducing resource waste and environmental risks caused by nitrogen leaching and volatilization. Attached Figure Description

[0046] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0047] Figure 1 This is a flowchart illustrating a method for optimizing controlled-release nitrogen fertilizer application decisions based on soil moisture variation characteristics, as described in an embodiment of the present invention. Detailed Implementation

[0048] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0049] See Figure 1 This is a flowchart of a method for optimizing controlled-release nitrogen fertilizer application decisions based on soil moisture variation characteristics, as provided in Embodiment 1 of the present invention. Figure 1 As shown, a method for optimizing controlled-release nitrogen fertilizer application decisions by incorporating soil moisture variation characteristics may include:

[0050] Step S1: Obtain and preprocess the soil volumetric moisture content query sequence and reference sequence set through the Internet of Things sensor network and historical database.

[0051] First, soil moisture sensors are deployed in the crop root activity layer of the plot to be decided, and a sampling time interval is set. In this embodiment of the invention, the sampling time interval is set to 1 hour. The soil moisture sensors are continuously collected according to the sampling time interval to obtain the time series data of soil volumetric water content in the current growing season, and the soil volumetric water content time series data is constructed into a soil volumetric water content query sequence.

[0052] Based on the historical database of the agricultural big data platform, historical records of soil volumetric moisture content for the same period in several past years for the plot to be decided or similar plots are retrieved to construct a soil volumetric moisture content reference sequence set, where each historical record of soil volumetric moisture content corresponds to a soil volumetric moisture content reference sequence; the field capacity of the plot to be decided is obtained by consulting the soil physical property test report or by using soil moisture characteristic curve measurement experiments, and the field capacity is used as a known soil physical property constant; the soil volumetric moisture content query sequence and the soil volumetric moisture content reference sequence set are preprocessed, and the data preprocessing includes at least outlier data extraction, missing data linear interpolation to complete the time step consistency of one season.

[0053] This completes the acquisition and preprocessing of soil volumetric moisture content query sequences and reference sequence sets through IoT sensor networks and historical databases.

[0054] Step S2: The effective infiltration potential optimization factor is obtained by evaluating the cumulative difference between the field water holding capacity threshold cutoff and the effective infiltration potential time window of the soil volume water content sequence.

[0055] This step aims to address the physical contradiction that existing algorithms cannot distinguish the differences in the impact of instantaneous high humidity fluctuations and continuous effective infiltration on fertilizer release. In actual controlled-release nitrogen fertilizer application scenarios, the release of coated nutrients is not linearly determined by the current instantaneous moisture content, but exhibits significant threshold characteristics and a cumulative effect over time. Soil moisture can only become free water capable of moving freely and permeating the coating after exceeding the field capacity; furthermore, continuous infiltration for a certain duration is required for the osmotic pressure inside and outside the coating to reach the level that triggers nutrient release. However, existing distance metrics only focus on the geometric proximity of values. This leads to a typical misjudgment scenario: when a query sequence shows a rapid spike in moisture content due to short-term heavy rainfall (high in value but extremely short-lived, insufficient to fully activate the coating), the algorithm easily misjudges it as highly similar to a continuous high-humidity state caused by prolonged overcast and rainy weather in historical sequences (moderate in value but long-lasting, resulting in massive release). This misjudgment ignores the crucial physical variable of effective infiltration duration, which determines the release rate, leading to an erroneous overestimation of fertilizer release rate under short-duration rainfall. Therefore, this application corrects the aforementioned physical bias by introducing an effective infiltration potential energy optimization factor based on physical threshold truncation and nonlinear accumulation, which forces the inclusion of verification of historical effective infiltration energy differences when calculating local distances.

[0056] In summary, this invention first performs nonlinear squared enhancement processing on the field capacity threshold truncated data of the soil volumetric moisture content query sequence and the reference sequence set to obtain effective infiltration potential energy accumulation feature data. Specifically, for any target time point in the soil volumetric moisture content query sequence, soil volumetric moisture content data of each time step within a preset historical backtracking time window before the target time point are extracted. The difference between the soil volumetric moisture content data and the field capacity at each time step is used as the moisture content exceeding threshold difference data. When the moisture content exceeding threshold difference data is less than or equal to a constant 0, the field capacity threshold truncated data of the corresponding time step is set to a constant 0. When the moisture content exceeding threshold difference data is greater than a constant 0, the ratio of the moisture content exceeding threshold difference data to the field capacity is used as the field capacity threshold truncated data of the corresponding time step. The field capacity threshold truncated data is then squared to obtain the squared enhanced threshold truncated data corresponding to the soil volumetric moisture content query sequence.

[0057] For any soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set, for any target time point in the soil volumetric moisture content reference sequence, extract the soil volumetric moisture content data of the target time point and each time step within a preset historical backtracking time window before the target time point. The difference between the soil volumetric moisture content data and field capacity at each time step is used as the moisture content exceeding threshold difference data. When the moisture content exceeding threshold difference data is less than or equal to a constant 0, the field capacity threshold truncation data of the corresponding time step is set to a constant 0. When the moisture content exceeding threshold difference data is greater than a constant 0, the ratio of the moisture content exceeding threshold difference data to field capacity is used as the field capacity threshold truncation data of the corresponding time step. The field capacity threshold truncation data is squared to obtain the square-enhanced threshold truncation data corresponding to the soil volumetric moisture content reference sequence. The square-enhanced threshold truncation data corresponding to the soil volumetric moisture content query sequence and the soil volumetric moisture content reference sequence are used as the effective infiltration potential energy accumulation feature data.

[0058] After obtaining the accumulated feature data of effective infiltration potential energy, the effective infiltration potential energy optimization factor is obtained by performing time window integral accumulation and difference normalization evaluation on the accumulated feature data of effective infiltration potential energy. Specifically, for any first target time point in the soil volumetric moisture content query sequence and any second target time point in the soil volumetric moisture content reference sequence, the square-enhanced threshold truncation data corresponding to each time step within the preset historical backtracking time window before the first target time point is extracted from the accumulated feature data of effective infiltration potential energy. The square-enhanced threshold truncation data corresponding to each time step within the preset historical backtracking time window is accumulated and calculated to obtain the accumulated value of effective infiltration potential energy within the query window corresponding to the first target time point. In this embodiment of the invention, the length of the historical backtracking time window is set to 72 hours, because it usually takes 1 to 3 days for the controlled-release fertilizer coating to absorb water, swell, and reach osmotic equilibrium. This time window length can be adjusted according to the actual scenario and is not required. The second target time point and the time step within the preset historical backtracking time window before the second target time point are extracted from the accumulated feature data of effective infiltration potential energy. The data is truncated using the squared enhancement threshold corresponding to each time step, and the truncated data of the squared enhancement threshold corresponding to each time step within the preset historical backtracking time window are accumulated to obtain the cumulative value of the effective immersion potential energy integral within the reference window corresponding to the second target time point; the absolute value of the difference between the cumulative value of the effective immersion potential energy integral within the query window and the cumulative value of the effective immersion potential energy integral within the reference window is used as the effective immersion potential energy integral difference assessment; the minimum value between the cumulative value of the effective immersion potential energy integral within the query window and the cumulative value of the effective immersion potential energy integral within the reference window is used as the effective immersion potential energy integral normalization benchmark; let... A minimum positive stability protection term is defined. The result of adding the stability protection term to the normalized baseline of the effective wetting potential energy integral is used as the normalized denominator, and the effective wetting potential energy integral difference assessment is used as the normalized numerator. The ratio of the normalized numerator to the normalized denominator is used as the normalized assessment value of the effective wetting potential energy difference. The result of adding the normalized assessment value of the effective wetting potential energy difference to the natural constant is used as the logarithmic mapping input value. The logarithmic mapping input value is then processed by natural logarithmic operation to obtain the effective wetting potential energy optimization factor corresponding to the first target time point and the second target time point.

[0059] In one implementation, assume the first Before the [time]th moment Soil volumetric moisture content query sequence data at each time step is ;No. Before the [time]th moment The soil volumetric water content reference sequence data at each time step are as follows: The field water holding capacity of the soil in the plot to be decided is: The length of the historical backtracking time window is Then in the query sequence, the first... The time point and the reference sequence The calculation expression for the effective wetting potential energy optimization factor corresponding to each time point is:

[0060]

[0061] in, Indicates the first position in the query sequence The time point and the reference sequence The effective wetting potential energy optimization factor corresponding to each time point; This represents the logarithmic function with the natural constant e as the base. Denotes the natural constant e; This indicates a very small positive number used to prevent division by zero in the denominator, and is set in the embodiments of the present invention. ; Indicates the length of the historical backtracking time window; Indicates the first Before the [time]th moment Soil volumetric moisture content query sequence data at each time step; Indicates the first Before the [time]th moment Reference sequence data of soil volumetric moisture content at each time step; This indicates the field water holding capacity of the soil in the plot to be decided. Represents the maximum value function; Describes the minimum value function; This indicates absolute value calculation.

[0062] It should be noted that the above formula is designed to address the problem that existing algorithms struggle to distinguish between short-term numerical fluctuations and sustained high numerical states by extracting and comparing data features. Firstly, the formula utilizes... Combination The function preprocesses the original sequence. In real-world scenarios, the water content is below field capacity. The moisture content has a relatively small impact on fertilizer coating; this step sets this portion of the data to zero, eliminating background fluctuations in the low moisture content range and focusing the calculation on the high moisture content data segment that actually contributes to release. Secondly, the data exceeding a threshold is processed using a squared term. This nonlinear mapping increases the weight of high moisture content data points in feature calculation to simulate the nonlinear effect of high humidity on fertilizer release rate. Finally, the past time windows are summed... The data within the time frame is accumulated to obtain integral feature values ​​that reflect the degree to which the sequence maintains a high numerical state over a period of time. This step converts instantaneous values ​​into time-period cumulative features, used to distinguish between short-duration rainfall and continuous rain. When calculating the difference, the formula uses... The smaller integral eigenvalue from the two sequences is chosen as the denominator. This is to address the issue of orders-of-magnitude differences: when one sequence contains only short-duration pulses, while the other is a continuous high-humidity sequence, the smaller denominator significantly increases the value of the ratio term, thus keenly capturing the essential difference in their cumulative characteristics. Finally, using... The function maps the aforementioned ratio to the final optimization factor. This logarithmic function acts as a numerical smoother, converting the resulting large ratio into a numerically stable penalty coefficient. This effectively increases the calculated local distance when the cumulative features of the two sequences differ significantly, thus avoiding algorithmic misjudgment.

[0063] Thus, the optimization factor for effective infiltration potential energy was obtained by evaluating the cumulative difference between the field water holding capacity threshold cutoff and the effective infiltration potential energy time window of the soil volumetric water content sequence.

[0064] Step S3: The water potential lag direction optimization factor is obtained by evaluating the consistency of the first-order difference gradient direction and the trend amplitude gating of the soil volumetric water content sequence.

[0065] After addressing the energy accumulation difference between instantaneous fluctuations and effective infiltration in step S2 using the effective wetting potential energy optimization factor, the algorithm still suffers from a deficiency in measuring local distance, namely, neglecting the hysteresis effect of soil water potential. In the actual soil-fertilizer microenvironment, soil matrix suction is not monotonically determined by the moisture content value, but is closely related to the direction of moisture change. Specifically, at the same moisture content, soil undergoing hygroscopic processes has lower matrix suction than soil undergoing desiccation processes, making it easier for the fertilizer coating to absorb and utilize the water; while the latter has a stronger water-holding capacity, and fertilizer release is significantly inhibited.

[0066] After the correction in step S2, although the algorithm can identify the differences in the accumulation of effective moisture, it will still classify two scenarios with similar moisture content and effective accumulation, but with diametrically opposed trends (e.g., the query sequence is rapidly becoming wet, while the reference sequence is rapidly drying) as a high match. This directional misjudgment can cause the system to confuse scenarios in the active release phase with those in the inhibition phase, thus leading to temporal bias in dynamic fertilization decisions. Therefore, based on step S2, it is necessary to further introduce an optimization factor that can sense the direction of moisture change gradient and the intensity of dynamic trends, thereby capturing the differences in the direction of sequence movement in phase space and performing a secondary correction on the local distance.

[0067] In summary, this invention first obtains query sequence gradient data and reference sequence gradient data by performing first-order difference processing on the moisture content data of adjacent time steps in the soil volumetric moisture content query sequence and the reference sequence set. Specifically, for any first target time point in the soil volumetric moisture content query sequence, the soil volumetric moisture content data corresponding to the first target time point and the soil volumetric moisture content data corresponding to the adjacent time steps before the first target time point are extracted from the soil volumetric moisture content query sequence. The difference between the soil volumetric moisture content data corresponding to the first target time point and the soil volumetric moisture content data corresponding to the adjacent time steps is used as the query sequence gradient data corresponding to the first target time point. The query sequence gradient data corresponding to each first target time point in the soil volumetric moisture content query sequence are then used to construct the query sequence gradient data. For any soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set, for any second target time point in the soil volumetric moisture content reference sequence, extract the soil volumetric moisture content data corresponding to the second target time point and the soil volumetric moisture content data corresponding to the adjacent time steps before the second target time point from the soil volumetric moisture content reference sequence. Use the difference between the soil volumetric moisture content data corresponding to the second target time point and the soil volumetric moisture content data corresponding to the adjacent time steps as the reference sequence gradient data corresponding to the second target time point. Construct the reference sequence gradient dataset for each second target time point in the soil volumetric moisture content reference sequence. Use the query sequence gradient dataset as the query sequence gradient data and the reference sequence gradient dataset as the reference sequence gradient data.

[0068] After obtaining the query sequence gradient data and the reference sequence gradient data, the gradient direction consistency assessment data is obtained by performing normalized inner product calculation on the query sequence gradient data and the reference sequence gradient data. Specifically, for any first target time point in the soil volumetric moisture content query sequence and any second target time point in the soil volumetric moisture content reference sequence, the query sequence gradient data corresponding to the first target time point is extracted from the query sequence gradient data, and the reference sequence gradient data corresponding to the second target time point is extracted from the reference sequence gradient data. The product of the query sequence gradient data and the reference sequence gradient data is taken as the first target time point and the second target time point. The gradient product is evaluated at the corresponding time points; a minimum positive stability protection term is set, and the result of adding the square of the gradient data of the query sequence to the stability protection term is used as the first square stability term, and the result of adding the square of the gradient data of the reference sequence to the stability protection term is used as the second square stability term; the square root operation is performed on the first square stability term and the second square stability term respectively to obtain the first gradient magnitude evaluation and the second gradient magnitude evaluation; the product of the first gradient magnitude evaluation and the second gradient magnitude evaluation is used as the normalized denominator, the gradient product evaluation is used as the normalized numerator, and the ratio of the normalized numerator to the normalized denominator is used as the gradient direction consistency evaluation data.

[0069] After obtaining the gradient direction consistency assessment data, the final step is to perform trend amplitude gating coupling processing on the gradient direction consistency assessment data, query sequence gradient data, and reference sequence gradient data to obtain the water potential lag direction optimization factor. Specifically, for any first target time point in the soil volumetric water content query sequence and any second target time point in the soil volumetric water content reference sequence, the gradient direction consistency assessment data corresponding to the first and second target time points is extracted from the gradient direction consistency assessment data, the query sequence gradient data corresponding to the first target time point is extracted from the query sequence gradient data, and the gradient direction consistency assessment data corresponding to the first target time point is extracted from the reference sequence gradient data. Take the gradient data of the reference sequence corresponding to the second target time point; take the difference between the constant 1 and the gradient direction consistency evaluation data as the gradient direction conflict evaluation; take the absolute value of the difference between the query sequence gradient data and the reference sequence gradient data as the gradient difference magnitude evaluation; perform hyperbolic tangent function mapping on the gradient difference magnitude evaluation to obtain the corresponding trend magnitude gating evaluation data; take the calculation result of multiplying the gradient direction conflict evaluation and the trend magnitude gating evaluation data as the gating coupling modulation quantity, and perform exponential mapping processing on the gating coupling modulation quantity with the natural constant as the base to obtain the water potential lag direction optimization factor corresponding to the first target time point and the second target time point.

[0070] In one implementation, assume the query sequence is in the _th ... The gradient data at each time step are The reference sequence is in the first... The gradient data at each time step are Then the query sequence's first... The time interval and the reference sequence at time n The calculation expression for the optimization factor of the water potential lag direction at each time point is:

[0071]

[0072] in, Represents the first query sequence The time interval and the reference sequence at time n The optimization factor for the water potential lag direction at each time point; This represents an exponential function with the natural constant e as the base. Indicates the query sequence at the 1st position. Gradient data at each time step; Indicates the reference sequence at the 1st Gradient data at each time step; This represents a stability protection term to prevent division by zero in the denominator; This represents the hyperbolic tangent function.

[0073] It should be noted that this formula constructs optimization factors based on the direction and magnitude characteristics of the sequence gradient, aiming to address the state matching bias problem caused by different trends in soil moisture variation. First, the terms within the square brackets... This formula is specifically designed to capture the degree of conflict between two sequences in their changing trends. In real-world soil physics scenarios, moisture changes have a clear directionality; that is, rainfall-induced hygroscopicity and evaporative dehumidification correspond to distinctly different soil matrix suction states. The formula determines this physical state by calculating the normalized inner product of two gradient vectors: when the changing trends of the two sequences are in the same direction (both increasing or decreasing moisture), the gradient product is positive, causing the fractional terms to approach a certain value. Thus, with the preceding They cancel each other out, causing the entire direction term to approach the mean. This indicates that the physical processes of the two are consistent and require no correction; however, when their trends are opposite (one increasing and the other decreasing), the gradient product is negative, causing the fractional term to approach the mean. At this point, the sum of the directional terms approaches the value. This quantifies the fundamental conflict between moisture absorption and desiccation. Secondly, This approach addresses the signal-to-noise ratio (SNR) issue in practical data acquisition. When soil moisture is relatively static or changes slowly, inherent minor fluctuations in the sensor can cause random reversals in the calculated gradient direction. Without control, the aforementioned direction determination logic might misinterpret these background noises. Therefore, the hyperbolic tangent function is introduced as an amplitude gate: when the absolute value of the gradient difference is small (i.e., the change is not significant), the function value approaches... The output of the direction term is automatically masked; the function value only approaches the direction term when there is a significant difference in dynamic trends between the two. This ensures that the algorithm only handles real, dynamic physical conflicts. Finally, through... The function outputs the final optimization factor. When two sequences are detected to not only change drastically but also in completely opposite directions (i.e., the most severe physical state conflict occurs), the input value of the exponential function approaches [the desired value]. , making The value increases to approximately (about This computational logic assigns significant weight to sequences whose values ​​are geometrically close but whose physical stages of change are contradictory in the distance metric, thereby forcibly widening their matching distance and effectively correcting decision-making errors caused by the lag effect of water potential.

[0074] Thus, the optimization factor for the water potential lag direction was obtained by coupling the first-order difference gradient direction consistency and trend amplitude of the soil volumetric water content sequence.

[0075] Step S4: The physical consistency optimized DTW distance is obtained by multiplicatively fusing the effective wetting potential energy optimization factor and the water potential hysteresis direction optimization factor and correcting the DTW local distance, and the time sequence matching is completed.

[0076] Optimization factors based on effective wetting potential energy were constructed respectively. Optimization factor with water potential lag direction Subsequently, this step integrates these two factors into the core iteration process of the Dynamic Time Warping (DTW) algorithm to replace the traditional Euclidean distance metric. Specifically, for any first target time point in the soil volumetric moisture content query sequence and any second target time point in any soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set, the effective infiltration potential energy optimization factor and the water potential lag direction optimization factor corresponding to the first target time point and the second target time point are obtained; the product of the effective infiltration potential energy optimization factor and the water potential lag direction optimization factor is used as the multiplicative fusion optimization factor.

[0077] Soil volumetric moisture content data corresponding to the first target time point is extracted from the soil volumetric moisture content query sequence, and soil volumetric moisture content data corresponding to the second target time point is extracted from the soil volumetric moisture content reference sequence. The difference between the two is used as the moisture content difference assessment. The square of the moisture content difference assessment is used as the basic DTW local distance squared term. The calculation result of multiplying the multiplicative fusion optimization factor with the basic DTW local distance squared term is used as the corrected DTW local distance between the first target time point and the second target time point.

[0078] A cumulative distance matrix is ​​constructed with the time steps of the soil volumetric moisture content query sequence as rows and the time steps of the soil volumetric moisture content reference sequence as columns. The modified DTW local distance is used as the local distance term of the corresponding matrix element in the cumulative distance matrix. According to the dynamic programming recursion rule, for any matrix element in the cumulative distance matrix, the calculated result of adding the modified DTW local distance corresponding to the matrix element to the minimum cumulative distance of its adjacent preceding matrix elements is used as the cumulative distance value of the matrix element. The adjacent preceding matrix elements include at least the left matrix element, the lower matrix element, and the lower left matrix element. The cumulative distance value corresponding to the terminal matrix element of the cumulative distance matrix is ​​used as the physical consistency optimized DTW distance between the soil volumetric moisture content query sequence and the soil volumetric moisture content reference sequence.

[0079] The physical consistency optimization DTW distance is calculated for each soil volumetric moisture content query sequence and each soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set. The soil volumetric moisture content reference sequence with the smallest physical consistency optimization DTW distance is determined as the optimal matching historical sequence, thereby completing the time series matching.

[0080] Thus, the process of obtaining physically consistent DTW distance by multiplicatively fusing the effective wetting potential energy optimization factor and the water potential hysteresis direction optimization factor and correcting the DTW local distance has been completed, and time-series matching has been finished.

[0081] Step S5: Obtain accurate controlled-release nitrogen fertilizer application decision recommendations by performing release curve weighted prediction on the optimal matching historical sequence corresponding to the DTW distance optimized by physical consistency.

[0082] Based on the best-matching historical sequences selected in step S4, the system retrieves the actual crop yield data and fertilizer release monitoring data corresponding to these historical years. Using a weighted average method, the optimal fertilization parameters under these historical scenarios are fused to predict the expected release curve of controlled-release fertilizer under current weather and soil conditions.

[0083] If the prediction results show that the release rate under current conditions is lower than the crop's needs (e.g., due to insufficient release identified by the optimization algorithm because of excessive ineffective rainfall in the early stages), the system generates a decision suggestion to apply additional accelerated nitrogen fertilizer. If the predicted release rate closely matches the crop's nutrient requirements, the current controlled-release fertilizer application plan is maintained. Ultimately, the system recommends the optimized fertilizer dosage, application time, and fertilizer formula to the user, thereby achieving precise controlled-release nitrogen fertilizer application based on soil moisture changes, ensuring crop yield while reducing nitrogen fertilizer loss and environmental pollution.

[0084] Thus, the goal of obtaining accurate controlled-release nitrogen fertilizer application decision recommendations by weighted prediction of release curves based on the optimal matching historical sequence corresponding to the DTW distance optimized by physical consistency is achieved.

[0085] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing controlled-release nitrogen fertilizer application decisions based on soil moisture content variation characteristics, characterized in that, The method includes: Step S1: Obtain and preprocess the soil volumetric moisture content query sequence and reference sequence set through the Internet of Things sensor network and historical database; Step S2: Obtain the effective infiltration potential optimization factor by evaluating the cumulative difference between the field holding capacity threshold cutoff and the effective infiltration potential time window of the soil volumetric water content sequence. Step S3: Obtain the water potential lag direction optimization factor by evaluating the coupling of the first-order difference gradient direction consistency and trend amplitude of the soil volumetric water content sequence; Step S4: By multiplicatively fusing the effective wetting potential energy optimization factor and the water potential hysteresis direction optimization factor and correcting the DTW local distance, the physical consistency optimized DTW distance is obtained and the time series matching is completed; Step S5: Obtain accurate controlled-release nitrogen fertilizer application decision recommendations by performing release curve weighted prediction on the optimal matching historical sequence corresponding to the DTW distance optimized by physical consistency. The method of obtaining the effective infiltration potential energy optimization factor by evaluating the difference between the field holding capacity threshold truncation of the soil volumetric water content sequence and the cumulative difference of the effective infiltration potential energy over a time window includes: obtaining the cumulative characteristic data of effective infiltration potential energy by performing nonlinear squared enhancement processing on the field holding capacity threshold truncation data of the soil volumetric water content query sequence and the reference sequence set; and obtaining the effective infiltration potential energy optimization factor by performing time window integral accumulation and difference normalization evaluation processing on the cumulative characteristic data of effective infiltration potential energy. The method of obtaining the water potential lag direction optimization factor by coupling the first-order difference gradient direction consistency and trend amplitude of the soil volumetric water content sequence includes: performing first-order difference processing on water content data of adjacent time steps in the soil volumetric water content query sequence and reference sequence sets to obtain query sequence gradient data and reference sequence gradient data; performing normalized inner product calculation processing on the query sequence gradient data and reference sequence gradient data to obtain gradient direction consistency evaluation data; and performing trend amplitude gating coupling processing on the gradient direction consistency evaluation data, the query sequence gradient data, and the reference sequence gradient data to obtain the water potential lag direction optimization factor.

2. The method for optimizing controlled-release nitrogen fertilizer application decisions based on soil moisture variation characteristics as described in claim 1, characterized in that, The process of acquiring and preprocessing soil volumetric moisture content query sequences and reference sequence sets through IoT sensor networks and historical databases includes: Soil moisture sensors were deployed in the crop root activity layer of the plot to be decided, and sampling time intervals were set. The soil moisture sensors were continuously collected according to the sampling time intervals to obtain the time series data of soil volumetric water content in the current growing season, and the soil volumetric water content time series data was constructed into a soil volumetric water content query sequence. Based on the historical database of the agricultural big data platform, historical records of soil volumetric moisture content for the same period in several past years for the plot to be decided or similar plots are retrieved to construct a soil volumetric moisture content reference sequence set, where each historical record of soil volumetric moisture content corresponds to a soil volumetric moisture content reference sequence; the field capacity of the plot to be decided is obtained by consulting the soil physical property test report or by using soil moisture characteristic curve measurement experiments, and the field capacity is used as a known soil physical property constant; the soil volumetric moisture content query sequence and the soil volumetric moisture content reference sequence set are preprocessed, and the data preprocessing includes at least outlier data extraction, missing data linear interpolation to complete the time step consistency of one season.

3. The controlled-release nitrogen fertilizer application decision optimization method based on soil moisture content variation characteristics as described in claim 1, characterized in that, The method involves performing nonlinear squared enhancement processing on the field water holding capacity threshold truncated data of the soil volumetric water content query sequence and the reference sequence set to obtain effective infiltration potential energy accumulation characteristic data, including: For any target time point in the soil volumetric moisture content query sequence, extract the soil volumetric moisture content data for each time step within the target time point and the preset historical backtracking time window before the target time point. Use the difference between the soil volumetric moisture content data and field capacity at each time step as the moisture content exceeding the threshold difference data. When the moisture content exceeding the threshold difference data is less than or equal to a constant 0, set the field capacity threshold truncation data for the corresponding time step to a constant 0. When the moisture content exceeding the threshold difference data is greater than a constant 0, use the ratio of the moisture content exceeding the threshold difference data to the field capacity as the field capacity threshold truncation data for the corresponding time step. Perform a square operation on the field capacity threshold truncation data to obtain the squared enhanced threshold truncation data corresponding to the soil volumetric moisture content query sequence. For any soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set, for any target time point in the soil volumetric moisture content reference sequence, extract the soil volumetric moisture content data of the target time point and each time step within a preset historical backtracking time window before the target time point. The difference between the soil volumetric moisture content data and field capacity at each time step is used as the moisture content exceeding threshold difference data. When the moisture content exceeding threshold difference data is less than or equal to a constant 0, the field capacity threshold truncation data of the corresponding time step is set to a constant 0. When the moisture content exceeding threshold difference data is greater than a constant 0, the ratio of the moisture content exceeding threshold difference data to field capacity is used as the field capacity threshold truncation data of the corresponding time step. The field capacity threshold truncation data is squared to obtain the square-enhanced threshold truncation data corresponding to the soil volumetric moisture content reference sequence. The square-enhanced threshold truncation data corresponding to the soil volumetric moisture content query sequence and the soil volumetric moisture content reference sequence are used as the effective infiltration potential energy accumulation feature data.

4. The method for optimizing controlled-release nitrogen fertilizer application decisions based on soil moisture content variation characteristics as described in claim 1, characterized in that, The process involves evaluating the accumulated feature data of effective wetting potential energy through time window integration and difference normalization to obtain the effective wetting potential energy optimization factor, including: For any first target time point in the soil volumetric moisture content query sequence and any second target time point in the soil volumetric moisture content reference sequence, extract the square-strengthened threshold truncated data corresponding to each time step within a preset historical backtracking time window before the first target time point from the effective infiltration potential energy cumulative feature data. Then, accumulate the square-strengthened threshold truncated data corresponding to each time step within the preset historical backtracking time window to obtain the effective infiltration potential energy integral cumulative value within the query window corresponding to the first target time point. Extract the square-strengthened threshold truncated data corresponding to each time step within a preset historical backtracking time window before the second target time point from the effective infiltration potential energy cumulative feature data. Then, accumulate the square-strengthened threshold truncated data corresponding to each time step within the preset historical backtracking time window to obtain the effective infiltration potential energy integral cumulative value within the reference window corresponding to the second target time point. The absolute value of the difference between the cumulative effective immersion potential energy integral value in the query window and the cumulative effective immersion potential energy integral value in the reference window is used as the effective immersion potential energy integral difference assessment; the minimum value between the cumulative effective immersion potential energy integral value in the query window and the cumulative effective immersion potential energy integral value in the reference window is used as the effective immersion potential energy integral normalization benchmark; a minimum positive number stability protection term is set, the result of adding the stability protection term and the effective immersion potential energy integral normalization benchmark is used as the normalization denominator, the effective immersion potential energy integral difference assessment is used as the normalization numerator, and the ratio of the normalization numerator to the normalization denominator is used as the effective immersion potential energy difference normalization assessment value. The calculation result of adding the normalized evaluation value of the effective wetting potential energy difference to the natural constant is used as the logarithmic mapping input value. The logarithmic mapping input value is then processed by natural logarithmic operation to obtain the effective wetting potential energy optimization factor corresponding to the first target time point and the second target time point.

5. The method for optimizing controlled-release nitrogen fertilizer application decisions based on soil moisture content variation characteristics according to claim 1, characterized in that, The process involves performing first-order difference processing on the soil volumetric moisture content query sequence and the moisture content data of adjacent time steps in the reference sequence set to obtain the gradient data of the query sequence and the gradient data of the reference sequence, including: For any first target time point in the soil volumetric moisture content query sequence, extract the soil volumetric moisture content data corresponding to the first target time point and the soil volumetric moisture content data corresponding to the adjacent time steps before the first target time point from the soil volumetric moisture content query sequence. Use the difference between the soil volumetric moisture content data corresponding to the first target time point and the soil volumetric moisture content data corresponding to the adjacent time steps as the query sequence gradient data corresponding to the first target time point. Construct the query sequence gradient dataset by using the query sequence gradient data corresponding to each first target time point in the soil volumetric moisture content query sequence. For any soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set, for any second target time point in the soil volumetric moisture content reference sequence, extract the soil volumetric moisture content data corresponding to the second target time point and the soil volumetric moisture content data corresponding to the adjacent time steps before the second target time point from the soil volumetric moisture content reference sequence. Use the difference between the soil volumetric moisture content data corresponding to the second target time point and the soil volumetric moisture content data corresponding to the adjacent time steps as the reference sequence gradient data corresponding to the second target time point. Construct a reference sequence gradient dataset from the reference sequence gradient data corresponding to each second target time point in the soil volumetric moisture content reference sequence. Use the query sequence gradient dataset as the query sequence gradient data and the reference sequence gradient dataset as the reference sequence gradient data.

6. The controlled-release nitrogen fertilizer application decision optimization method based on soil moisture content variation characteristics according to claim 1, characterized in that, The step of obtaining gradient direction consistency evaluation data by performing normalized inner product calculation on the gradient data of the query sequence and the gradient data of the reference sequence includes: For any first target time point in the soil volumetric moisture content query sequence and any second target time point in the soil volumetric moisture content reference sequence, extract the query sequence gradient data corresponding to the first target time point from the query sequence gradient data, and extract the reference sequence gradient data corresponding to the second target time point from the reference sequence gradient data; use the product of the query sequence gradient data and the reference sequence gradient data as the gradient product evaluation corresponding to the first target time point and the second target time point; A minimum positive stability protection term is set. The sum of the square of the gradient data of the query sequence and the stability protection term is used as the first square stability term, and the sum of the square of the gradient data of the reference sequence and the stability protection term is used as the second square stability term. Square root operations are performed on the first square stability term and the second square stability term to obtain the first gradient magnitude evaluation and the second gradient magnitude evaluation respectively. The product of the first gradient magnitude evaluation and the second gradient magnitude evaluation is used as the normalized denominator, the gradient product evaluation is used as the normalized numerator, and the ratio of the normalized numerator to the normalized denominator is used as the gradient direction consistency evaluation data.

7. The method for optimizing controlled-release nitrogen fertilizer application decisions based on soil moisture content variation characteristics as described in claim 1, characterized in that, The process of obtaining the water potential hysteresis direction optimization factor by performing trend amplitude gating coupling processing on the gradient direction consistency evaluation data, the query sequence gradient data, and the reference sequence gradient data includes: For any first target time point in the soil volumetric moisture content query sequence and any second target time point in the soil volumetric moisture content reference sequence, extract the gradient direction consistency assessment data corresponding to the first target time point and the second target time point from the gradient direction consistency assessment data, extract the query sequence gradient data corresponding to the first target time point from the query sequence gradient data, and extract the reference sequence gradient data corresponding to the second target time point from the reference sequence gradient data; use the difference between the constant 1 and the gradient direction consistency assessment data as the gradient direction conflict assessment. The absolute value of the difference between the gradient data of the query sequence and the gradient data of the reference sequence is used as the gradient difference magnitude assessment; the gradient difference magnitude assessment is processed by hyperbolic tangent function mapping to obtain the corresponding trend magnitude gating assessment data. The result of multiplying the gradient direction conflict assessment data and the trend amplitude gating assessment data is used as the gating coupling modulation quantity. The gating coupling modulation quantity is then subjected to exponential mapping processing with the natural constant as the base to obtain the water potential lag direction optimization factor corresponding to the first target time point and the second target time point.

8. The method for optimizing controlled-release nitrogen fertilizer application decisions based on soil moisture content variation characteristics according to claim 1, characterized in that, The process of obtaining physically consistent optimized DTW distance and completing temporal matching by multiplicatively fusing the effective wetting potential energy optimization factor and the water potential hysteresis direction optimization factor, correcting the DTW local distance, includes: For any first target time point in the soil volumetric moisture content query sequence and any second target time point in any soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set, obtain the effective infiltration potential energy optimization factor and the water potential lag direction optimization factor corresponding to the first target time point and the second target time point; and take the product of the effective infiltration potential energy optimization factor and the water potential lag direction optimization factor as the multiplicative fusion optimization factor. Soil volumetric moisture content data corresponding to the first target time point is extracted from the soil volumetric moisture content query sequence, and soil volumetric moisture content data corresponding to the second target time point is extracted from the soil volumetric moisture content reference sequence. The difference between the two is used as the moisture content difference assessment. The square of the moisture content difference assessment is used as the basic DTW local distance squared term. The result of multiplying the multiplicative fusion optimization factor with the basic DTW local distance squared term is used as the corrected DTW local distance between the first target time point and the second target time point. A cumulative distance matrix is ​​constructed with the time steps of the soil volumetric moisture content query sequence as rows and the time steps of the soil volumetric moisture content reference sequence as columns. The modified DTW local distance is used as the local distance term of the corresponding matrix element in the cumulative distance matrix. According to the dynamic programming recursion rule, for any matrix element in the cumulative distance matrix, the calculated result of adding the modified DTW local distance corresponding to the matrix element to the minimum cumulative distance of its adjacent preceding matrix elements is used as the cumulative distance value of the matrix element. The adjacent preceding matrix elements include at least the left matrix element, the lower matrix element, and the lower left matrix element. The cumulative distance value corresponding to the terminal matrix element of the cumulative distance matrix is ​​used as the physical consistency optimized DTW distance between the soil volumetric moisture content query sequence and the soil volumetric moisture content reference sequence. The physical consistency optimization DTW distance is calculated for each soil volumetric moisture content query sequence and each soil volumetric moisture content reference sequence in the soil volumetric moisture content reference sequence set. The soil volumetric moisture content reference sequence with the smallest physical consistency optimization DTW distance is determined as the optimal matching historical sequence, thereby completing the time series matching.

Images