Polypeptide fertilizer proportioning optimization method and system for promoting color turning and sweetening of fruits

Abstract

The invention discloses a polypeptide fertilizer proportion optimization method and system for promoting fruit coloring and sweetening, and relates to the related field of organic fertilizer formula optimizing.The method comprises the steps that application big data mining is conducted according to a current scheme, and the color score, the sweetness score and other performance scores of current fertilizer application are determined; introducing a target plant growth regulator to perform multi-parameter adjustment on the current scheme to obtain an adjustment space; performing color and luster and sweetness promotion optimization and performance loss optimization on the adjustment space according to a current fertilizer application color score, a sweetness score and other performance scores, and establishing a color and luster promotion adjustment domain, a sweetness promotion adjustment domain and a loss optimization adjustment domain; and performing multi-domain guide joint optimization on the adjustment space according to the three adjustment domains to generate an optimization result. According to the method, the technical problem that the overall effect is poor due to the fact that the optimization accuracy of the fertilizer ratio is insufficient in an existing method for promoting fruit coloring and sweetening is solved, and the technical effect that the optimization accuracy is improved, so that the overall quality of fruits is improved is achieved.

Description

Technical Field

[0001] This application relates to the field of organic fertilizer formulation optimization, and in particular to a method and system for optimizing the ratio of polypeptide fertilizers to promote fruit color change and sweetness. Background Technology

[0002] Fruit color and sweetness are core quality indicators that determine the commercial value and market competitiveness of fruits, and their optimization and improvement are crucial for enhancing the quality and efficiency of the fruit industry. Currently, production mainly relies on empirically based fertilization ratios or the application of single-function regulators, such as spraying ethephon during the fruit color-changing period or increasing potassium fertilizer during the sweetening period. These existing methods often focus on improving a single performance aspect or use fixed formulas, lacking precise adaptation to the needs of fruit growth. This often leads to problems such as asynchronous color-changing and sweetening, uneven color-changing, and decreased fruit resistance, making it difficult to simultaneously improve quality and stabilize overall performance.

[0003] At present, the methods for promoting fruit color change and sweetness have technical problems such as insufficient precision in optimizing fertilizer ratios, resulting in poor overall effects. Summary of the Invention

[0004] This application provides a method and system for optimizing the ratio of polypeptide fertilizers to promote fruit color change and sweetness. It utilizes big data mining to determine the scores of the current polypeptide fertilizer ratio in terms of fruit color, sweetness, and other properties. Then, it introduces a target plant growth regulator to adjust the current ratio using multiple parameters to construct a fertilizer ratio adjustment space. Based on the current color, sweetness, and other performance scores, optimization is performed within this adjustment space to establish corresponding color-promoting, sweetness-promoting, and loss-optimizing fertilizer adjustment domains. Finally, these three adjustment domains are combined to guide joint optimization within the adjustment space, thereby generating an optimized fertilizer ratio scheme. These technical methods solve the technical problem of insufficient precision in fertilizer ratio optimization in existing methods for promoting fruit color change and sweetness, leading to poor overall results. This approach improves the precision of optimization, thereby enhancing the overall quality of the fruit.

[0005] This application provides a method for optimizing the ratio of polypeptide fertilizers to promote fruit color change and sweetness, comprising: conducting big data mining based on the current polypeptide fertilizer ratio scheme of the target fruit to determine the current fertilizer application color score, current fertilizer application sweetness score, and current fertilizer application other performance scores; introducing a target plant growth regulator to adjust the current polypeptide fertilizer ratio scheme through multiple parameters to obtain a fertilizer ratio adjustment space; optimizing the fertilizer ratio adjustment space for color promotion based on the current fertilizer application color score to establish a color-promoting fertilizer adjustment domain; optimizing the fertilizer ratio adjustment space for sweetness promotion based on the current fertilizer application sweetness score to establish a sweetness-promoting fertilizer adjustment domain; optimizing the fertilizer ratio adjustment space for performance loss based on the current fertilizer application other performance scores to obtain a loss-optimized fertilizer adjustment domain; and performing multi-domain guided joint optimization of the fertilizer ratio adjustment space based on the color-promoting fertilizer adjustment domain, the sweetness-promoting fertilizer adjustment domain, and the loss-optimized fertilizer adjustment domain to generate a fertilizer ratio optimization result.

[0006] In a possible implementation, big data mining is performed based on the current peptide fertilizer formulation scheme for the target fruit, and the following processing is executed: Big data retrieval is performed on the target fruit according to the current peptide fertilizer formulation scheme to obtain historical evaluation sets of fertilizer application color, fertilizer application sweetness, and other fertilizer application performance; attention allocation aggregation under support analysis is performed on the historical evaluation set of fertilizer application color to obtain the current fertilizer application color score; attention allocation aggregation under support analysis is performed on the historical evaluation set of fertilizer application sweetness to obtain the current fertilizer application sweetness score; attention allocation aggregation under support analysis is performed on the historical evaluation set of fertilizer application performance to obtain the current fertilizer application performance scores.

[0007] In a possible implementation, attention allocation aggregation under support analysis is performed based on the historical evaluation set of fertilizer application color to obtain the current fertilizer application color score, and the following processing is performed: support evaluation is performed on each historical evaluation score of fertilizer application color in the historical evaluation set to obtain a color evaluation support sequence; credible attention allocation is performed on the historical evaluation set of fertilizer application color based on the color evaluation support sequence to obtain a color evaluation attention configuration sequence; weighted aggregation is performed on the historical evaluation set of fertilizer application color based on the color evaluation attention configuration sequence to generate the current fertilizer application color score.

[0008] In a possible implementation, the fertilizer ratio adjustment space is optimized for color promotion based on the current fertilizer application color score to establish a color-promoting fertilizer adjustment domain. The following processes are then performed: Extracting the Dth fertilizer ratio adjustment scheme from the fertilizer ratio adjustment space, where D is a positive integer; retrieving fertilizer application color evaluation records based on the target fruit to obtain a fertilizer ratio sample set and a fruit color evaluation record set; performing multi-dimensional perturbation aggregation learning based on the fertilizer ratio sample set and the fruit color evaluation record set to establish a fertilizer application color score prediction channel; inputting the Dth fertilizer ratio adjustment scheme into the fertilizer application color score prediction channel to output the Dth fertilizer application color prediction score; evaluating the color promotion of the Dth fertilizer application color prediction score based on the current fertilizer application color score to obtain the Dth fertilizer color promotion coefficient; if the Dth fertilizer color promotion coefficient is greater than or equal to the fertilizer color promotion threshold, adding the Dth fertilizer ratio adjustment scheme to the color-promoting fertilizer adjustment domain.

[0009] In a possible implementation, multidimensional perturbation aggregation learning is performed based on the fertilizer ratio sample set and the fruit color evaluation record set to establish a fertilizer application color score prediction channel. The following processes are then performed: aligning the fertilizer ratio sample set and the fruit color evaluation record set to obtain a first color evaluation training set; supervising learning of the SVR model based on the first color evaluation training set to obtain a first color score prediction model and a first model loss characteristic corresponding to the first color score prediction model; injecting multidimensional adversarial perturbation into the first color evaluation training set based on the first model loss characteristic to obtain a second color evaluation training set; supervising learning of the GRU model based on the second color evaluation training set to obtain a second color score prediction model; and federated aggregation learning is performed based on the first color score prediction model and the second color score prediction model to generate the fertilizer application color score prediction channel.

[0010] In a possible implementation, the fertilizer ratio adjustment space is subjected to multi-domain guided joint optimization based on the color-promoting fertilizer adjustment domain, the sweetness-promoting fertilizer adjustment domain, and the loss-optimizing fertilizer adjustment domain to generate fertilizer ratio optimization results. The following processes are then performed: Differential guided propagation optimization is performed on the fertilizer ratio adjustment space based on the color-promoting fertilizer adjustment domain to obtain a first fertilizer adjustment optimization space; differential guided propagation optimization is performed on the fertilizer ratio adjustment space based on the sweetness-promoting fertilizer adjustment domain to obtain a second fertilizer adjustment optimization space; differential guided propagation optimization is performed on the fertilizer ratio adjustment space based on the loss-optimizing fertilizer adjustment domain to obtain a third fertilizer adjustment optimization space; a fruit quality promotion evaluation model is constructed based on fruit quality promotion evaluation conditions; and joint optimization is performed on the first, second, and third fertilizer adjustment optimization spaces based on the fruit quality promotion evaluation model to obtain the fertilizer ratio optimization results.

[0011] In a possible implementation, the fertilizer ratio adjustment space is subjected to differential-guided propagation optimization based on the color-promoting fertilizer regulation domain to obtain a first fertilizer regulation optimization space. The following processes are then performed: Differential identification is performed on the fertilizer ratio adjustment space based on the color-promoting fertilizer regulation domain to obtain a first fertilizer regulation differential distribution; the fertilizer ratio adjustment space is guided to undergo mutation propagation based on the first fertilizer regulation differential distribution to obtain a first fertilizer regulation propagation group; the first fertilizer regulation propagation group is optimized for color promotion using the current fertilizer application color score to establish a second fertilizer regulation propagation group; the second fertilizer regulation propagation group is optimized for sweetness promotion using the current fertilizer application sweetness score to establish a third fertilizer regulation propagation group; and the third fertilizer regulation propagation group is optimized for performance loss using other performance scores applied to the current fertilizer application to obtain the first fertilizer regulation optimization space.

[0012] In a possible implementation, the first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, and the third fertilizer adjustment optimization space are jointly optimized according to the fruit quality promotion evaluation model to obtain the fertilizer ratio optimization result. The following processing is then performed: the intersection analysis of the color-promoting fertilizer adjustment domain, the sweetness-promoting fertilizer adjustment domain, and the loss-optimizing fertilizer adjustment domain is performed to obtain a fourth fertilizer adjustment optimization space; the first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, the third fertilizer adjustment optimization space, and the fourth fertilizer adjustment optimization space are merged to obtain a fifth fertilizer adjustment optimization space; the fifth fertilizer adjustment optimization space is iteratively optimized according to the fruit quality promotion evaluation model to generate the fertilizer ratio optimization result.

[0013] In a possible implementation, the following processing is performed: the fruit quality improvement evaluation conditions include color improvement weight, sweetness improvement weight, and other performance loss weights.

[0014] This application also provides a peptide fertilizer ratio optimization system for promoting fruit color change and sweetness, comprising: a big data mining module, used to perform big data mining based on the current peptide fertilizer ratio scheme of the target fruit to determine the current fertilizer application color score, current fertilizer application sweetness score, and other performance scores of the current fertilizer application; a multi-parameter adjustment module, used to introduce target plant growth regulators to adjust the current peptide fertilizer ratio scheme using multiple parameters to obtain a fertilizer ratio adjustment space; and a color promotion optimization module, used to perform color promotion optimization on the fertilizer ratio adjustment space based on the current fertilizer application color score to establish a color... The system includes: a color-enhancing fertilizer regulation domain; a sweetness-enhancing optimization module, used to optimize the fertilizer ratio adjustment space based on the sweetness score of the current fertilizer application, and establish a sweetness-enhancing fertilizer regulation domain; a performance loss optimization module, used to optimize the fertilizer ratio adjustment space based on other performance scores of the current fertilizer application, and obtain a loss-optimized fertilizer regulation domain; and a multi-domain guided joint optimization module, used to perform multi-domain guided joint optimization of the fertilizer ratio adjustment space based on the color-enhancing fertilizer regulation domain, the sweetness-enhancing fertilizer regulation domain, and the loss-optimized fertilizer regulation domain, and generate fertilizer ratio optimization results.

[0015] This application proposes a method and system for optimizing the ratio of polypeptide fertilizers to promote fruit color change and sweetness. First, based on the current polypeptide fertilizer ratio scheme for the target fruit, big data mining is performed to determine the current fertilizer application's color score, sweetness score, and other performance scores. Next, a target plant growth regulator is introduced to adjust multiple parameters of the current polypeptide fertilizer ratio scheme, obtaining a fertilizer ratio adjustment space. Then, the color score of the current fertilizer application is used to optimize the fertilizer ratio adjustment space for color promotion, establishing a color-promoting fertilizer adjustment domain. The sweetness score of the current fertilizer application is used to optimize the fertilizer ratio adjustment space for sweetness promotion, establishing a sweetness-promoting fertilizer adjustment domain. Other performance scores of the current fertilizer application are used to optimize the fertilizer ratio adjustment space for performance loss, obtaining a loss-optimized fertilizer adjustment domain. Finally, based on the color-promoting fertilizer adjustment domain, the sweetness-promoting fertilizer adjustment domain, and the loss-optimized fertilizer adjustment domain, multi-domain guided joint optimization is performed on the fertilizer ratio adjustment space to generate the optimized fertilizer ratio result. Through the above process, the method and system proposed in this application achieve the technical effect of improving the accuracy of fertilizer ratio optimization, thereby improving the overall quality of fruit. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a flowchart illustrating a method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness, as provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of a polypeptide fertilizer ratio optimization system for promoting fruit color change and sweetness, provided in an embodiment of this application.

[0019] Figure labeling: Big Data Mining Module 10, Multi-Parameter Adjustment Module 20, Color-Enhanced Optimization Module 30, Sweetness-Enhanced Optimization Module 40, Performance Loss Optimization Module 50, Multi-Domain Guided Joint Optimization Module 60. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] This application provides a method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness, such as... Figure 1 As shown, the method includes:

[0022] Step S100: Based on the current peptide fertilizer formulation scheme for the target fruit, big data mining is performed to determine the color score, sweetness score, and other performance scores of the current fertilizer application.

[0023] Specifically, the target fruit is tomato. Distributed web crawling is used to retrieve historical application data related to the current polypeptide fertilizer formulation from agricultural big data platforms, farmer planting record systems, and agricultural input company product feedback databases. The retrieved unstructured data, such as planting logs and product evaluation texts, is then structured to extract a comprehensive score including color, sweetness, and other performance scores such as disease resistance, yield, and maturity. After data cleaning, a weighted average method is used to calculate the current comprehensive score.

[0024] In one possible implementation, big data mining is performed based on the current peptide fertilizer formulation scheme for the target fruit. Step S100 further includes step S110, which involves performing a big data retrieval on the target fruit based on the current peptide fertilizer formulation scheme to obtain historical evaluation sets for fertilizer application color, sweetness, and other performance characteristics. Specifically, key parameters of the current peptide fertilizer formulation scheme, such as the nitrogen-phosphorus-potassium ratio, peptide type and concentration, and application time, are obtained. Based on these key parameters, search keywords are constructed, such as: tomato, peptide fertilizer, nitrogen-phosphorus-potassium 2:1:3, and color score. The Elasticsearch search engine is used to perform a full-text search on the historical data in the distributed database, filtering out matching records. For example, from 100,000 tomato planting records, 500 records using the same formulation scheme are retrieved, and the color evaluation data, sweetness evaluation data, and other performance evaluation data are extracted to form three evaluation sets.

[0025] Step S120: Based on the historical evaluation set of fertilizer application color, attention allocation aggregation under support analysis is performed to obtain the current fertilizer application color score. Specifically, the number of times each color score appears in the historical evaluation set is counted, i.e., the support. For example, a score of 8.5 appears 100 times, with a support of 100; a score of 7.8 appears 80 times, with a support of 80. The support is normalized, such as by dividing by the total number of evaluations, to obtain the attention weight. The higher the support, the greater the weight. For example, with a total of 500 evaluations, the weight of a score of 8.5 is 100 / 500 = 0.2, and the weight of a score of 7.8 is 80 / 500 = 0.16. Each score is multiplied by its corresponding attention weight, and the sum is obtained to obtain the current fertilizer application color score.

[0026] Step S130: Based on the historical evaluation set of fertilizer application sweetness, attention allocation aggregation under support analysis is performed to obtain the current fertilizer application sweetness score. Specifically, similar to step S120, the frequency of each sweetness score in the historical evaluation set is counted, normalized to obtain the attention weight, and then weighted and aggregated to obtain the current sweetness score.

[0027] Step S140: Based on the historical evaluation set of other performance of the fertilizer application, attention allocation aggregation under support analysis is performed to obtain the current other performance score of the fertilizer application. Specifically, similar to step S120, other performance includes disease resistance, yield, maturity period, etc. First, each indicator is standardized, such as mapped to 1-10 points, then support is calculated, attention weights are assigned, and weighted aggregation is performed to obtain the current other performance score.

[0028] In one possible implementation, attention allocation aggregation under support analysis is performed based on the historical evaluation set of fertilizer application color to obtain the current fertilizer application color score. Step S120 further includes step S121, which evaluates the support of each historical evaluation score of fertilizer application color in the historical evaluation set to obtain a color evaluation support sequence. Specifically, the historical evaluation set of fertilizer application color is traversed, and the Python collections.Counter function is used to count the occurrence frequency of each score to generate a support sequence. For example, Counter({8.5:100, 7.8:80, 9.0:60, ...}) has a support sequence of [100, 80, 60, ...].

[0029] Step S122: Based on the color evaluation support sequence, perform reliable attention allocation on the fertilizer application color history evaluation set to obtain a color evaluation attention configuration sequence. Specifically, divide each value in the support sequence by the total number of evaluations to obtain a normalized attention weight.

[0030] Step S123: The fertilizer application color history evaluation set is weighted and aggregated according to the color evaluation attention configuration sequence to generate the current fertilizer application color score. Specifically, the fertilizer application color history evaluation set and the attention configuration sequence are multiplied element-wise, and the sum is obtained to obtain a weighted average score.

[0031] Step S200: Introduce the target plant growth regulator to adjust multiple parameters of the current polypeptide fertilizer ratio scheme to obtain the fertilizer ratio adjustment space.

[0032] Specifically, the target plant growth regulator is heme chloride, with an effective concentration range of 0.1-1.0 mmol / L. Key parameters of the current polypeptide fertilizer formulation are adjusted. For example, the nitrogen, phosphorus, and potassium ratio is adjusted from the original 2:1:3 to (1.5:1:3.5), (2:1:3), and (2.5:1:2.5), etc.; the polypeptide concentration is adjusted from the original 0.5% to 0.3%, 0.5%, and 0.7%, etc.; the heme chloride concentration is set to 0.1 mmol / L, 0.3 mmol / L, 0.5 mmol / L, 0.7 mmol / L, and 0.9 mmol / L, etc.; and the application time is expanded from the original fruit-setting period to include application during the fruit expansion period, or application during both the fruit-setting and fruit expansion periods. Through a full-factor experimental design, a fertilizer formulation adjustment space is generated, for example, the above example contains 3×3×5×3=135 adjustment schemes.

[0033] Step S300: Optimize the fertilizer ratio adjustment space by using the current fertilizer application color score to promote color enhancement, and establish a color-enhancing fertilizer adjustment domain.

[0034] Specifically, all possible schemes in the fertilizer ratio adjustment space are traversed, and the color score of each scheme is predicted by a machine learning model. Schemes with predicted scores higher than the current fertilizer application color score are selected to form a color-promoting fertilizer adjustment domain.

[0035] In one possible implementation, the fertilizer ratio adjustment space is optimized for color promotion using the current fertilizer application color score to establish a color-promoting fertilizer adjustment domain. Step S300 further includes step S310, extracting the Dth fertilizer ratio adjustment scheme according to the fertilizer ratio adjustment space, where D is a positive integer. Specifically, all schemes in the fertilizer ratio adjustment space are arranged in random order, and the 1st, 2nd...th schemes are extracted sequentially. For example, the 1st scheme is: NPK 1.5:1:3.5 + polypeptide 0.3% + heme chloride 0.1 mmol / L + applied during fruit setting.

[0036] Step S320: Retrieve fertilizer application color evaluation records based on the target fruit to obtain a fertilizer ratio sample set and a fruit color evaluation record set. Specifically, use a distributed crawler to retrieve fertilizer application records related to tomatoes, extract fertilizer ratio parameters and corresponding color scores, and form a fertilizer ratio sample set and a fruit color evaluation record set.

[0037] Step S330: Perform multi-dimensional perturbation aggregation learning based on the fertilizer ratio sample set and the fruit color evaluation record set to establish a fertilizer application color score prediction channel. Specifically, the fertilizer ratio sample set and the fruit color evaluation record set are mapped by index to ensure that each ratio scheme corresponds to a color score. First, a Support Vector Regression (SVR) model is used for supervised learning of the data, with the ratio parameters as input and the color score as output, resulting in a first color score prediction model after training. The loss function of the first color score prediction model is calculated, and multi-dimensional adversarial perturbation is added to the training data according to the loss distribution to generate an augmented dataset. A gated recurrent unit (GRU) model is used to perform secondary training on the augmented dataset to obtain a second color score prediction model. The prediction results of the two models are weighted and averaged to construct the final fertilizer application color score prediction channel.

[0038] Step S340: Input the Dth fertilizer ratio adjustment scheme into the fertilizer application color score prediction channel, and output the Dth fertilizer application color prediction score. Specifically, convert the parameters of the Dth adjustment scheme into a model input vector, input it into the prediction channel, and obtain the prediction score.

[0039] Step S350: Evaluate the color promotion of the predicted color of fertilizer application D based on the current fertilizer application color score, and obtain the color promotion coefficient of fertilizer D. Specifically, For example, the current rating is 8.2, and the predicted rating is 8.6. .

[0040] Step S360: If the color-promoting coefficient of fertilizer D is greater than or equal to the color-promoting threshold, the fertilizer ratio adjustment scheme of fertilizer D is added to the color-promoting fertilizer adjustment domain. Specifically, the fertilizer color-promoting threshold is set according to agricultural production needs. If the promotion coefficient is greater than or equal to the fertilizer color-promoting threshold, the scheme is added to the color-promoting fertilizer adjustment domain.

[0041] In one possible implementation, multidimensional perturbation aggregation learning is performed based on the fertilizer ratio sample set and the fruit color evaluation record set to establish a fertilizer application color score prediction channel. Step S330 further includes step S331, aligning the fertilizer ratio sample set and the fruit color evaluation record set to obtain a first color evaluation training set. Specifically, the fertilizer ratio sample set and the fruit color evaluation record set are combined into a DataFrame using Python's pandas library, ensuring that each row of data contains complete ratio parameters and corresponding color scores.

[0042] Step S332: Supervised learning is performed on the SVR model based on the first color evaluation training set to obtain a first color score prediction model and a first model loss characteristic corresponding to the first color score prediction model. Specifically, using... The SVR model in the library divides the first color evaluation training set into a training set and a validation set in a 7:3 ratio. The training set is used for model fitting, and the validation set is used to evaluate model performance. After training, loss metrics such as mean squared error (MSE) and mean absolute error (MAE) on the validation set are calculated as the first model loss characteristics.

[0043] Step S333: Inject multidimensional adversarial perturbations into the first color evaluation training set based on the loss characteristics of the first model to obtain a second color evaluation training set. Specifically, by analyzing the loss characteristics of the first color scoring prediction model, such as the distribution of MSE and the types of incorrectly predicted samples, the model's weaknesses are identified. For example, the model may have a large prediction error for samples with a "low nitrogen, high potassium ratio + high heme chloride concentration," or be sensitive to samples with "peptide concentrations close to the threshold." For the training data corresponding to these weaknesses, multidimensional adversarial perturbations are designed and injected. For example, small Gaussian noise is added to numerical parameters, such as... For proportional parameters, random shifts are performed, such as adjusting the nitrogen-phosphorus-potassium ratio from 2:1:3 to 2.1:1:2.9. For classification parameters, probabilistic replacements are performed, such as replacing the application time (10% probability during fruit setting) with the fruit setting period plus the fruit expansion period. Finally, the perturbed samples are merged with the original training set to generate a second color evaluation training set. This augmented data forces the subsequent GRU model to learn more robust features, avoiding overfitting to specific samples and improving its adaptability to complex ratio scenarios.

[0044] Step S334: Supervised learning is performed on the GRU model based on the second color evaluation training set to obtain a second color score prediction model. Specifically, a GRU model is constructed using the TensorFlow library. The input layer is a vector representation of the matching parameters, the hidden layer consists of 64 GRU units, and the output layer is a single neuron used to predict the color score. The model is trained using the Adam optimizer and the MSE loss function. After 100 iterations, the MSE on the validation set drops below 0.1, thus obtaining the second color score prediction model.

[0045] Step S335: Federated aggregation learning is performed based on the first color score prediction model and the second color score prediction model to generate the fertilizer application color score prediction channel. Specifically, the prediction results of the two models are weighted and averaged, with the weights determined based on the accuracy of the models on the validation set. For example, if the validation accuracy of the SVR model is 85% and the validation accuracy of the GRU model is 90%, then the weights are 0.49 and 0.51, respectively. The final output of the prediction channel is: .

[0046] Step S400: Optimize the fertilizer ratio adjustment space by using the current fertilizer application sweetness score to promote sweetness, and establish a sweetness-promoting fertilizer adjustment domain.

[0047] Specifically, similar to step S300, the sweetness score of each regulation scheme is predicted using the SVR+GRU aggregation model, and schemes with predicted scores higher than the current fertilizer application sweetness score are selected to form a sweetness-promoting fertilizer regulation domain.

[0048] Step S500: Optimize the performance loss of the fertilizer ratio adjustment space by applying other performance scores to the current fertilizer application, and obtain the loss-optimized fertilizer adjustment domain.

[0049] Specifically, other performance loss evaluation indicators include disease resistance decline rate, yield reduction rate, and maturity extension rate. A performance loss threshold is set; for example, a performance loss threshold of 5% means that the other performance scores of the adjustment scheme must not be lower than 95% of the current fertilizer application's other performance scores. If the current fertilizer application's other performance score is 8.1, then the lowest acceptable score is 8.1 × (1-5%) = 7.7. If the score is lower than 7.7, it indicates that the scheme leads to excessive decline in disease resistance, yield reduction, or maturity extension, which is an unacceptable loss. Using the same model as in step S300, the parameters of the adjustment scheme are input to predict the other performance scores for each scheme. The predicted scores of each scheme are compared with the lowest acceptable score, and schemes with predicted scores greater than or equal to the lowest acceptable score are selected to form the loss optimization fertilizer adjustment domain.

[0050] Step S600: Based on the color-promoting fertilizer regulation domain, the sweetness-promoting fertilizer regulation domain, and the loss-optimizing fertilizer regulation domain, the fertilizer ratio regulation space is subjected to multi-domain guided joint optimization to generate fertilizer ratio optimization results.

[0051] Specifically, the intersection of the three fertilizer regulation domains is calculated to obtain a set of schemes that simultaneously satisfy the requirements of color promotion, sweetness promotion, and acceptable performance loss. A fruit quality promotion evaluation model is constructed, and for each scheme in the intersection set, the comprehensive quality score is calculated using the fruit quality promotion evaluation model. The scheme with the highest score is selected as the optimization result.

[0052] In one possible implementation, multi-domain guided joint optimization is performed on the fertilizer ratio adjustment space based on the color-promoting fertilizer regulation domain, the sweetness-promoting fertilizer regulation domain, and the loss-optimizing fertilizer regulation domain to generate fertilizer ratio optimization results. Step S600 further includes step S610, which involves differential-guided propagation optimization on the fertilizer ratio adjustment space based on the color-promoting fertilizer regulation domain to obtain a first fertilizer regulation optimization space. Specifically, the parameter differences between each scheme in the color-promoting fertilizer regulation domain and the current polypeptide fertilizer ratio scheme are calculated to generate a difference distribution heatmap. Based on the difference distribution, the schemes in the fertilizer ratio adjustment space are mutated, such as by simulating crossover and mutation operations in a genetic algorithm, to generate a new scheme group. Color-promoting optimization (similar to step S300), sweetness-promoting optimization (similar to step S400), and performance loss optimization (similar to step S500) are performed on the new scheme group to screen out schemes that meet the conditions and form the first fertilizer regulation optimization space.

[0053] Step S620: Based on the sweetness-promoting fertilizer regulation domain, the fertilizer ratio regulation space is subjected to differential-guided propagation optimization to obtain a second fertilizer regulation optimization space. Specifically, similar to step S610, based on the parameter difference distribution of the sweetness-promoting fertilizer regulation domain, the fertilizer ratio regulation space is subjected to mutation propagation, and after multiple rounds of screening, a second fertilizer regulation optimization space is obtained.

[0054] Step S630: Based on the loss-optimized fertilizer regulation domain, the fertilizer ratio regulation space is subjected to differential-guided propagation optimization to obtain a third fertilizer regulation optimization space. Specifically, similar to step S610, based on the parameter difference distribution of the loss-optimized fertilizer regulation domain, the fertilizer ratio regulation space is subjected to mutation propagation, and after multiple rounds of screening, a third fertilizer regulation optimization space is obtained.

[0055] Step S640: Based on the fruit quality promotion evaluation conditions, a fruit quality promotion evaluation model is constructed. These conditions include weights for color promotion, sweetness promotion, and other performance loss. Specifically, the function expression of the fruit quality promotion evaluation model is: Q = w1 × C + w2 × S + w3 × O, where Q is the overall quality score, C is the color score, S is the sweetness score, O is the other performance score, w1 is the color promotion weight, w2 is the sweetness promotion weight, and w3 is the other performance loss weight.

[0056] Step S650: Based on the fruit quality promotion evaluation model, jointly optimize the first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, and the third fertilizer adjustment optimization space to obtain the fertilizer ratio optimization result. Specifically, the first, second, and third fertilizer adjustment optimization spaces are merged, and duplicate schemes are removed to obtain a joint optimization space. For each scheme in the joint optimization space, a comprehensive quality score is calculated using the fruit quality promotion evaluation model. The scheme with the highest comprehensive score is selected as the fertilizer ratio optimization result. For example, if a scheme has a comprehensive score of 8.8, which is the highest value, it is the optimization result.

[0057] In one possible implementation, the fertilizer ratio adjustment space is differentially guided for optimization based on the color-promoting fertilizer regulation domain to obtain a first fertilizer regulation optimization space. Step S610 further includes step 611, which involves differentially identifying the fertilizer ratio adjustment space based on the color-promoting fertilizer regulation domain to obtain a first fertilizer regulation differential distribution. Specifically, the parameter differences between each scheme within the color-promoting fertilizer regulation domain and the current polypeptide fertilizer ratio scheme are calculated using Python's NumPy library. For example, if the current scheme has a nitrogen-phosphorus-potassium ratio of 2:1:3, and a scheme within the color-promoting fertilizer regulation domain has a ratio of 1.5:1:3.5, then the difference is [-0.5, 0, +0.5]. All differences are statistically analyzed to generate a differential distribution histogram or heatmap, identifying parameters with large differences, such as the heme chloride concentration, which has the largest difference.

[0058] Step 612: Guide the fertilizer ratio adjustment space for mutation propagation based on the first fertilizer regulation difference distribution to obtain the first fertilizer regulation propagation group. Specifically, analyze the first fertilizer regulation difference distribution to identify the key parameters that contribute most to improving color and the optimal adjustment range and direction of these parameters. Then, randomly select a portion of basic schemes from the fertilizer ratio adjustment space as parent schemes. Next, simulate the mutation process of biological inheritance and modify the parameters of these parent schemes. However, this modification is not completely random but directional, that is, it is more likely to be adjusted towards the parameter direction and range indicated by the first fertilizer regulation difference distribution, which is beneficial to improving color. For example, if the difference distribution shows that the effect is best when the heme chloride concentration is around 0.3 mmol / L, then during mutation, the heme chloride concentration of the parent scheme will tend to be adjusted to be near this range, rather than arbitrarily setting a value. Through this difference distribution-guided mutation propagation strategy, a batch of new offspring schemes that inherit excellent genes are generated. These new schemes together constitute the first fertilizer regulation propagation group.

[0059] Step 613: The color-enhancing optimization of the first fertilizer regulation breeding group is performed using the current fertilizer application color score to establish a second fertilizer regulation breeding group. Specifically, the fertilizer application color score prediction channel trained in step S330 is invoked, and each fertilizer ratio scheme in the first fertilizer regulation breeding group is used as input to obtain the corresponding color prediction score. An optimization objective is set, namely, selecting schemes whose color prediction scores are greater than or equal to the current fertilizer application color score. All schemes that pass this selection condition are retained and collectively form the second fertilizer regulation breeding group. This process ensures that subsequent optimization iterations continuously improve the color score, a key indicator.

[0060] Step 614: The second fertilizer regulation breeding group is optimized for sweetness enhancement using the current fertilizer application sweetness score to establish a third fertilizer regulation breeding group. Specifically, while ensuring that color performance is not lagging behind, the optimization target shifts to sweetness. A fertilizer application sweetness score prediction channel, similar to step S330 but specifically designed for predicting sweetness, is invoked. Each scheme in the second fertilizer regulation breeding group is input into this prediction channel to obtain its predicted sweetness score. The screening condition is set as a sweetness prediction score greater than or equal to the current fertilizer application sweetness score. Only schemes that meet this condition can enter the next round; these schemes collectively constitute the third fertilizer regulation breeding group. This step ensures that while pursuing color, sweetness performance is simultaneously improved or at least maintained at the current level.

[0061] Step 615: The performance loss of the third fertilizer regulation propagation group is optimized using the current fertilizer application's other performance scores to obtain the first fertilizer regulation optimization space. Specifically, to avoid the deterioration of other important agronomic traits, such as disease resistance and yield, due to excessive pursuit of color and sweetness, a final performance loss verification is performed. A fertilizer application other performance score prediction channel, similar to that in step S330, is invoked to comprehensively evaluate each scheme in the third fertilizer regulation propagation group and predict its other performance scores. A performance loss threshold is set, and a minimum acceptable score is calculated. Schemes with other performance prediction scores greater than or equal to this minimum acceptable score are selected. The schemes retained after these three rounds of rigorous screening constitute the first fertilizer regulation optimization space that considers color, sweetness, and other comprehensive performance.

[0062] In one possible implementation, the first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, and the third fertilizer adjustment optimization space are jointly optimized according to the fruit quality promotion evaluation model to obtain the fertilizer ratio optimization result. Step S650 further includes step S651, performing intersection analysis on the color-promoting fertilizer adjustment domain, the sweetness-promoting fertilizer adjustment domain, and the loss-optimizing fertilizer adjustment domain to obtain a fourth fertilizer adjustment optimization space. Specifically, the intersection operation in set theory is used to calculate the common schemes of the three fertilizer adjustment domains. For example, the color-promoting fertilizer adjustment domain has 30 schemes, the sweetness-promoting fertilizer adjustment domain has 25 schemes, and the loss-optimizing fertilizer adjustment domain has 40 schemes. The intersection of the three is 10 schemes, forming the fourth fertilizer adjustment optimization space.

[0063] Step S652: Merge the first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, the third fertilizer adjustment optimization space, and the fourth fertilizer adjustment optimization space to obtain a fifth fertilizer adjustment optimization space. Specifically, merge the schemes in the four optimization spaces, and use the drop_duplicates function of the pandas library to remove duplicate schemes to obtain the fifth fertilizer adjustment optimization space.

[0064] Step S653: Based on the fruit quality promotion evaluation model, iterative optimization of the fifth fertilizer adjustment optimization space is performed to generate the fertilizer ratio optimization result. Specifically, for each scheme in the fifth fertilizer adjustment optimization space, a comprehensive quality score is calculated, and the scheme with the highest score is selected as the final optimization result.

[0065] This application's embodiments utilize big data mining to determine the scores of the current polypeptide fertilizer ratio in terms of fruit color, sweetness, and other properties. Then, a target plant growth regulator is introduced to adjust the current ratio using multiple parameters to construct a fertilizer ratio adjustment space. Optimization is then performed within this adjustment space based on the current color, sweetness, and other performance scores, establishing corresponding color-promoting, sweetness-promoting, and loss-optimizing fertilizer adjustment domains. Finally, these three adjustment domains are combined to guide multi-domain joint optimization within the adjustment space, thereby generating an optimized fertilizer ratio scheme. These technical methods solve the technical problem of insufficient precision in fertilizer ratio optimization in existing methods for promoting fruit color change and sweetness, leading to poor overall results. This approach improves the precision of optimization, thereby enhancing the overall quality of the fruit.

[0066] In the above text, refer to Figure 1 This paper describes in detail a method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness according to embodiments of the present invention. Next, we will refer to... Figure 2 This invention describes a peptide fertilizer formulation optimization system for promoting fruit color change and sweetness according to an embodiment of the present invention.

[0067] A polypeptide fertilizer ratio optimization system for promoting fruit color change and sweetness according to an embodiment of the present invention addresses the technical problem of insufficient precision in fertilizer ratio optimization in existing methods for promoting fruit color change and sweetness, leading to poor overall results. This system improves the precision of optimization, thereby enhancing the overall quality of the fruit. The polypeptide fertilizer ratio optimization system for promoting fruit color change and sweetness includes: a big data mining module 10, a multi-parameter adjustment module 20, a color promotion optimization module 30, a sweetness promotion optimization module 40, a performance loss optimization module 50, and a multi-domain guided joint optimization module 60.

[0068] The big data mining module 10 is used to perform big data mining based on the current peptide fertilizer ratio scheme of the target fruit to determine the current fertilizer application color score, current fertilizer application sweetness score, and other performance scores of the current fertilizer application; the multi-parameter adjustment module 20 is used to introduce target plant growth regulators to adjust the current peptide fertilizer ratio scheme through multiple parameters to obtain the fertilizer ratio adjustment space; the color promotion optimization module 30 is used to optimize the color promotion of the fertilizer ratio adjustment space through the current fertilizer application color score to establish a color promotion fertilizer regulation domain; the sweetness promotion optimization module... Block 40 is used to optimize the fertilizer ratio adjustment space based on the sweetness score of the current fertilizer application, and establish a sweetness-enhancing fertilizer adjustment domain; performance loss optimization module 50 is used to optimize the fertilizer ratio adjustment space based on other performance scores of the current fertilizer application, and obtain a loss-optimized fertilizer adjustment domain; multi-domain guided joint optimization module 60 is used to perform multi-domain guided joint optimization based on the color-enhancing fertilizer adjustment domain, the sweetness-enhancing fertilizer adjustment domain, and the loss-optimized fertilizer adjustment domain, and generate fertilizer ratio optimization results.

[0069] The detailed description of the specific configuration of the big data mining module 10 is explained as follows: As mentioned above, big data mining is performed based on the current polypeptide fertilizer ratio scheme of the target fruit. The big data mining module 10 may further include: a big data retrieval unit for performing application big data retrieval on the target fruit according to the current polypeptide fertilizer ratio scheme to obtain a historical evaluation set of fertilizer application color, a historical evaluation set of fertilizer application sweetness, and a historical evaluation set of fertilizer application other performance; an attention allocation aggregation unit for performing attention allocation aggregation under support analysis based on the historical evaluation set of fertilizer application color to obtain the current fertilizer application color score; performing attention allocation aggregation under support analysis based on the historical evaluation set of fertilizer application sweetness to obtain the current fertilizer application sweetness score; and performing attention allocation aggregation under support analysis based on the historical evaluation set of fertilizer application other performance to obtain the current fertilizer application other performance scores.

[0070] Specifically, the attention allocation aggregation based on support analysis of the historical evaluation set of fertilizer application color is used to obtain the current fertilizer application color score. The attention allocation aggregation unit may further include: a support evaluation subunit for performing support evaluation on each historical evaluation score of fertilizer application color in the historical evaluation set of fertilizer application color to obtain a color evaluation support sequence; a credible attention allocation subunit for performing credible attention allocation on the historical evaluation set of fertilizer application color based on the color evaluation support sequence to obtain a color evaluation attention configuration sequence; and a weighted aggregation subunit for performing weighted aggregation on the historical evaluation set of fertilizer application color based on the color evaluation attention configuration sequence to generate the current fertilizer application color score.

[0071] The detailed description of the specific configuration of the color promotion optimization module 30 is as follows: As mentioned above, the color promotion optimization is performed on the fertilizer ratio adjustment space through the current fertilizer application color score to establish a color promotion fertilizer adjustment domain. The color promotion optimization module 30 may further include: a fertilizer ratio adjustment scheme extraction unit for extracting the Dth fertilizer ratio adjustment scheme according to the fertilizer ratio adjustment space, where D is a positive integer; a fertilizer application color evaluation record retrieval unit for retrieving fertilizer application color evaluation records according to the target fruit to obtain a fertilizer ratio sample set and a fruit color evaluation record set; and a multidimensional perturbation aggregation learning unit for learning based on the fertilizer ratio sample set. A multidimensional perturbation aggregation learning process is performed on the fruit color evaluation record set to establish a fertilizer application color score prediction channel. The fertilizer application color score prediction unit inputs the Dth fertilizer ratio adjustment scheme into the fertilizer application color score prediction channel and outputs the Dth fertilizer application color prediction score. The color promotion evaluation unit evaluates the Dth fertilizer application color prediction score based on the current fertilizer application color score to obtain the Dth fertilizer color promotion coefficient. The color promotion fertilizer regulation domain generation unit adds the Dth fertilizer ratio adjustment scheme to the color promotion fertilizer regulation domain if the Dth fertilizer color promotion coefficient is greater than or equal to the fertilizer color promotion threshold.

[0072] The process involves performing multidimensional perturbation aggregation learning based on the fertilizer ratio sample set and the fruit color evaluation record set to establish a fertilizer application color score prediction channel. The multidimensional perturbation aggregation learning unit may further include: a first color evaluation training set acquisition subunit for aligning the fertilizer ratio sample set and the fruit color evaluation record set to obtain a first color evaluation training set; an SVR model supervised learning subunit for supervising the SVR model based on the first color evaluation training set to obtain a first color score prediction model and a first model loss characteristic corresponding to the first color score prediction model; a multidimensional adversarial perturbation injection subunit for injecting multidimensional adversarial perturbation into the first color evaluation training set based on the first model loss characteristic to obtain a second color evaluation training set; a GRU model supervised learning subunit for supervising the GRU model based on the second color evaluation training set to obtain a second color score prediction model; and a federated aggregation learning subunit for performing federated aggregation learning based on the first color score prediction model and the second color score prediction model to generate the fertilizer application color score prediction channel.

[0073] The detailed description of the specific configuration of the multi-domain guided joint optimization module 60 is explained as follows: As mentioned above, the fertilizer ratio adjustment space is subjected to multi-domain guided joint optimization based on the color-promoting fertilizer adjustment domain, the sweetness-promoting fertilizer adjustment domain, and the loss-optimizing fertilizer adjustment domain to generate fertilizer ratio optimization results. The multi-domain guided joint optimization module 60 may further include: a differential-guided reproductive optimization unit used to perform differential-guided reproductive optimization on the fertilizer ratio adjustment space based on the color-promoting fertilizer adjustment domain to obtain a first fertilizer adjustment optimization space; and to perform differential-guided reproductive optimization on the fertilizer ratio adjustment space based on the sweetness-promoting fertilizer adjustment domain. The fertilizer ratio adjustment space is subjected to differential-guided propagation optimization to obtain a second fertilizer adjustment optimization space; based on the loss-optimized fertilizer adjustment domain, the fertilizer ratio adjustment space is subjected to differential-guided propagation optimization to obtain a third fertilizer adjustment optimization space; the fruit quality promotion evaluation model construction unit is used to construct a fruit quality promotion evaluation model based on the fruit quality promotion evaluation conditions; the joint optimization unit is used to perform joint optimization on the first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, and the third fertilizer adjustment optimization space according to the fruit quality promotion evaluation model to obtain the fertilizer ratio optimization result.

[0074] The method involves using the color-enhancing fertilizer regulation domain to guide differential propagation optimization within the fertilizer ratio regulation space, thereby obtaining a first fertilizer regulation optimization space. The differential propagation optimization unit may further include: a differential identification subunit for identifying differences in the fertilizer ratio regulation space based on the color-enhancing fertilizer regulation domain, thereby obtaining a first fertilizer regulation differential distribution; a mutation propagation subunit for guiding mutation propagation within the fertilizer ratio regulation space based on the first fertilizer regulation differential distribution, thereby obtaining a first fertilizer regulation propagation group; a color-enhancing optimization subunit for performing color-enhancing optimization on the first fertilizer regulation propagation group using the current fertilizer application color score, thereby establishing a second fertilizer regulation propagation group; a sweetness-enhancing optimization subunit for performing sweetness-enhancing optimization on the second fertilizer regulation propagation group using the current fertilizer application sweetness score, thereby establishing a third fertilizer regulation propagation group; and a performance loss optimization subunit for performing performance loss optimization on the third fertilizer regulation propagation group using the current fertilizer application other performance scores, thereby obtaining the first fertilizer regulation optimization space.

[0075] Specifically, the first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, and the third fertilizer adjustment optimization space are jointly optimized according to the fruit quality promotion evaluation model to obtain the fertilizer ratio optimization result. The joint optimization unit may further include: an intersection analysis subunit for performing intersection analysis on the color promotion fertilizer adjustment domain, the sweetness promotion fertilizer adjustment domain, and the loss optimization fertilizer adjustment domain to obtain a fourth fertilizer adjustment optimization space; an optimization space merging subunit for merging the first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, the third fertilizer adjustment optimization space, and the fourth fertilizer adjustment optimization space to obtain a fifth fertilizer adjustment optimization space; and a fruit quality promotion evaluation iterative optimization subunit for performing fruit quality promotion evaluation iterative optimization on the fifth fertilizer adjustment optimization space according to the fruit quality promotion evaluation model to generate the fertilizer ratio optimization result.

[0076] The fruit quality promotion evaluation model construction unit may further include: the fruit quality promotion evaluation conditions include color promotion weight, sweetness promotion weight, and other performance loss weight.

[0077] The polypeptide fertilizer ratio optimization system for promoting fruit color change and sweetness provided in this embodiment of the invention can execute the polypeptide fertilizer ratio optimization method for promoting fruit color change and sweetness provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0078] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0079] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness, characterized in that, include: Based on the current peptide fertilizer formulation scheme for the target fruit, big data mining was conducted to determine the color score, sweetness score, and other performance scores of the current fertilizer application. By introducing a target plant growth regulator, the current polypeptide fertilizer formulation scheme is adjusted in multiple parameters to obtain the fertilizer formulation adjustment space; By using the color score of the current fertilizer application, the fertilizer ratio adjustment space is optimized for color promotion, and a color-promoting fertilizer adjustment domain is established. By using the sweetness score of the current fertilizer application, the sweetness-promoting optimization of the fertilizer ratio adjustment space is carried out, and a sweetness-promoting fertilizer adjustment domain is established. By applying other performance scores to the current fertilizer, the performance loss of the fertilizer ratio adjustment space is optimized to obtain the loss-optimized fertilizer adjustment domain. Based on the color-promoting fertilizer regulation domain, the sweetness-promoting fertilizer regulation domain, and the loss-optimizing fertilizer regulation domain, the fertilizer ratio regulation space is subjected to multi-domain guided joint optimization to generate fertilizer ratio optimization results.

2. The method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness as described in claim 1, characterized in that, Big data mining was applied based on the current peptide fertilizer formulation for the target fruit, including: Based on the current polypeptide fertilizer formulation scheme, a big data retrieval was performed on the target fruit to obtain historical evaluation sets of fertilizer application color, fertilizer application sweetness, and other fertilizer application performance. Based on the historical evaluation set of fertilizer application color, attention allocation aggregation under support analysis is performed to obtain the current fertilizer application color score; Based on the historical evaluation set of fertilizer application sweetness, attention allocation aggregation under support analysis is performed to obtain the current fertilizer application sweetness score; Based on the historical evaluation set of other performance of the fertilizer application, attention allocation aggregation under support analysis is performed to obtain the current fertilizer application other performance score.

3. The method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness as described in claim 2, characterized in that, Based on the historical evaluation set of fertilizer application color, attention allocation aggregation under support analysis is performed to obtain the current fertilizer application color score, including: Support evaluation is performed on each fertilizer application color history evaluation score in the fertilizer application color history evaluation set to obtain a color evaluation support sequence. Based on the color evaluation support sequence, a reliable attention allocation is performed on the fertilizer application color history evaluation set to obtain a color evaluation attention configuration sequence; The fertilizer application color history evaluation set is weighted and aggregated according to the color evaluation attention configuration sequence to generate the current fertilizer application color score.

4. The method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness as described in claim 1, characterized in that, By using the current fertilizer application color score to optimize the fertilizer ratio adjustment space for color enhancement, a color-enhancing fertilizer adjustment domain is established, including: Based on the fertilizer ratio adjustment space, extract the Dth fertilizer ratio adjustment scheme, where D is a positive integer; Based on the target fruit, fertilizer application color evaluation records are retrieved to obtain a fertilizer ratio sample set and a fruit color evaluation record set; Based on the fertilizer ratio sample set and the fruit color evaluation record set, multidimensional perturbation aggregation learning is performed to establish a fertilizer application color score prediction channel. Input the Dth fertilizer ratio adjustment scheme into the fertilizer application color score prediction channel, and output the Dth fertilizer application color prediction score; Based on the current fertilizer application color score, the color promotion evaluation of the Dth fertilizer application color prediction score is performed to obtain the Dth fertilizer color promotion coefficient. If the color-promoting coefficient of fertilizer D is greater than or equal to the color-promoting threshold, the fertilizer ratio adjustment scheme of fertilizer D is added to the color-promoting fertilizer adjustment domain.

5. The method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness as described in claim 4, characterized in that, Based on the fertilizer ratio sample set and the fruit color evaluation record set, multidimensional perturbation aggregation learning is performed to establish a fertilizer application color score prediction channel, including: Align the fertilizer ratio sample set and the fruit color evaluation record set to obtain the first color evaluation training set; The SVR model is subjected to supervised learning based on the first color evaluation training set to obtain the first color score prediction model and the first model loss characteristics corresponding to the first color score prediction model. Based on the loss characteristics of the first model, a multidimensional adversarial perturbation is injected into the first color evaluation training set to obtain a second color evaluation training set. The GRU model is supervised learning based on the second color evaluation training set to obtain the second color score prediction model; Federated aggregation learning is performed based on the first color score prediction model and the second color score prediction model to generate the fertilizer application color score prediction channel.

6. The method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness as described in claim 1, characterized in that, Based on the color-promoting fertilizer regulation domain, the sweetness-promoting fertilizer regulation domain, and the loss-optimizing fertilizer regulation domain, a multi-domain guided joint optimization is performed on the fertilizer ratio regulation space to generate fertilizer ratio optimization results, including: Based on the color-promoting fertilizer regulation domain, the fertilizer ratio regulation space is differentially guided to seek optimal reproduction, thereby obtaining the first fertilizer regulation optimization space; Based on the sweetness-promoting fertilizer regulation domain, differential guidance and breeding optimization are performed on the fertilizer ratio regulation space to obtain a second fertilizer regulation optimization space; Based on the loss-optimized fertilizer regulation domain, the fertilizer ratio regulation space is differentially guided to perform breeding optimization to obtain a third fertilizer regulation optimization space; Based on the evaluation criteria for promoting fruit quality, a fruit quality promotion evaluation model is constructed. Based on the fruit quality promotion evaluation model, the first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, and the third fertilizer adjustment optimization space are jointly optimized to obtain the fertilizer ratio optimization result.

7. The method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness as described in claim 6, characterized in that, Based on the color-promoting fertilizer regulation domain, differential-guided breeding optimization is performed on the fertilizer ratio regulation space to obtain the first fertilizer regulation optimization space, including: Based on the color-promoting fertilizer regulation domain, the fertilizer ratio regulation space is differentiated to obtain the first fertilizer regulation difference distribution; Based on the differential distribution of the first fertilizer regulation, the fertilizer ratio regulation space is guided to undergo mutation and reproduction to obtain the first reproductive group of fertilizer regulation; The color score of the current fertilizer application is used to promote color optimization in the first fertilizer regulation reproductive group, and a second fertilizer regulation reproductive group is established. The sweetness score of the current fertilizer application is used to optimize the sweetness of the second fertilizer-regulated reproductive group and establish a third fertilizer-regulated reproductive group. The first fertilizer regulation optimization space is obtained by optimizing the performance loss of the third reproductive group of the fertilizer by applying other performance scores of the current fertilizer application.

8. The method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness as described in claim 6, characterized in that, Based on the fruit quality promotion evaluation model, the first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, and the third fertilizer adjustment optimization space are jointly optimized to obtain the fertilizer ratio optimization results, including: The intersection of the color-promoting fertilizer regulation domain, the sweetness-promoting fertilizer regulation domain, and the loss-optimizing fertilizer regulation domain is analyzed to obtain a fourth fertilizer regulation optimization space. The first fertilizer adjustment optimization space, the second fertilizer adjustment optimization space, the third fertilizer adjustment optimization space, and the fourth fertilizer adjustment optimization space are merged to obtain the fifth fertilizer adjustment optimization space; Based on the fruit quality promotion evaluation model, the fifth fertilizer adjustment optimization space is iteratively optimized to generate the fertilizer ratio optimization result.

9. The method for optimizing the formulation of polypeptide fertilizers to promote fruit color change and sweetness as described in claim 6, characterized in that, The fruit quality improvement evaluation criteria include weights for color improvement, sweetness improvement, and other performance loss.

10. A polypeptide fertilizer formulation optimization system for promoting fruit color change and sweetness, characterized in that, The system is used to implement the method for optimizing the ratio of polypeptide fertilizers to promote fruit color change and sweetness as described in any one of claims 1-9, the system comprising: The big data mining module is used to perform big data mining based on the current peptide fertilizer formulation scheme of the target fruit to determine the color score, sweetness score, and other performance scores of the current fertilizer application. A multi-parameter adjustment module is used to introduce a target plant growth regulator to adjust the current polypeptide fertilizer ratio scheme in multiple parameters, thereby obtaining the fertilizer ratio adjustment space. The color-enhancing optimization module is used to enhance the color of the fertilizer ratio adjustment space by using the color score of the current fertilizer application, and to establish a color-enhancing fertilizer adjustment domain. The sweetness-enhancing optimization module is used to enhance the sweetness of the fertilizer ratio adjustment space by using the current fertilizer application sweetness score, and to establish a sweetness-enhancing fertilizer adjustment domain. The performance loss optimization module is used to perform performance loss optimization on the fertilizer ratio adjustment space by applying other performance scores to the current fertilizer, and obtain the loss optimization fertilizer adjustment domain. The multi-domain guided joint optimization module is used to perform multi-domain guided joint optimization on the fertilizer ratio adjustment space based on the color-promoting fertilizer adjustment domain, the sweetness-promoting fertilizer adjustment domain, and the loss-optimizing fertilizer adjustment domain, and generate fertilizer ratio optimization results.

Images