结合经济生态约束的“区域-农户”种植规划研究

CASIA OpenIR > 毕业生 > 硕士学位论文

	结合经济生态约束的“区域-农户”种植规划研究
	许钰林
	2023-05
页数	80
学位类型	硕士
中文摘要	随着科技的快速发展和国家对农业的大力支持，我国人均粮食产量已超过国际上公认的 400 公斤粮食安全线。但在高产的同时也存在一些问题，其中大豆自给率较低问题突出。中央农村工作会议和中央一号文件均明确指出“扩种大豆”。因此进行种植规划，合理划分每种作物的种植面积对保障国家粮食安全具有重要意义。然而，现有的种植面积规划模型大多使用作物当前价格或去年同期价格作为参考，误差较大。另一方面，求解农户作物分配时，传统方法存在灵活性较差的问题。因此，本文以黑龙江省为背景，构建多目标种植规划模型并求解，分别探索种植面积规划问题和多农户作物分配问题，旨在提升模型构建的灵活性和求解结果的可靠性。 1. 基于 LSTM 的作物预期价格的预测。经济效益是种植规划模型的主要目标函数，作物预期价格是其中的重要参数。本文使用大豆、玉米和小麦三种农作物历史月度价格数据，应用 LSTM 模型进行训练，对三者收获时期的预期价格进行预测。除此之外，本文通过分析玉米和大豆两者价格相关性，发现两者具有较强的正向相关性，且玉米价格是大豆价格的格兰杰原因，因此本文进行了结合玉米价格进行大豆价格预测的实验，大豆价格预测结果的均方根误差（Root Mean Square Error，RMSE）提升了 9.6%。 2. 基于作物预期价格的多目标种植面积规划模型。以黑龙江省为例，求解大豆、玉米和小麦三种旱地作物的最优种植面积。黑龙江省作为我国重要的粮食种植区域，进行省层面种植规划时，目标需要同时考虑到经济效益、生态效益和国家政策。其中，经济效益设定为农户的总收益；生态效益设定为作物总耗水量；国家政策目标设定为“扩种大豆”，即最大化黑龙江省的大豆种植面积。另一方面，种植收益会显著影响农户决策，由于大豆亩利润低于玉米亩利润，为了促进农户增收，提升农户种植大豆的积极性，本文以黑龙江省为例，构建了加入大豆额外补贴1的种植面积规划模型，添加农户的亩平均收益不低于去年亩平均收益的约束，为相关部门提供决策支持。 3. 基于强化学习的多农户作物分配模型的构建与求解。构建并求解多目标种植面积规划模型，可以确定三种作物的总体最优种植面积。为了将种植需求落实到具体土地上，本文构建了多农户作物分配模型，同时为多个农户提供决策支持，引导农户种植合适的作物。这不仅更好地落实了各作物的种植需求，还可以避免农户生产的盲目性，降低农户面临的市场风险。模型的构建需要考虑到：农户公平性，即：考虑到社会效益，每个农户的平均利润差距需要在合理范围之内；作物的轮作要求，即：尽量满足轮作约束保护土壤。本文使用强化学习求解农户层面的作物分配问题，在与模拟环境的互动中学习满足作物轮作约束并提升农户间的公平性，提升了模型构建的灵活性。综上，本文构建了区域水平的种植面积规划模型和农户水平的作物分配模型。在实际应用中，同时构建两个模型，可以将区域级别的调控目标落实到具体的作物生产中。
英文摘要	With the rapid development of science and technology and the State’s strong support for agriculture, per capita grain output of our country has exceeded the internationally recognized 400 kg grain safety line. However, there are some problems with high yield, among which the problem of low self-sufficiency rate of soybean is prominent. Central Rural Work Conference and No.1 Central Document clearly pointed out that ”expanding the cultivation of big beans”. Therefore, it is of great significance to carry out crop planning and reasonably divide the planting area of each crop to ensure national food security. However, most of the existing studies use the current prices of crops or the prices of the same period last year as a reference, with large errors. On the other hand, the traditional methods have the problem of poor flexibility in solving the problem of crop allocation. Therefore, this thesis takes Heilongjiang Province as the background, constructs a multi-objective planting planning model and solves it, explores the crop planning problem and multi-farmer crop allocation problem respectively, aimed at improves the flexibility of model construction and the reliability of solution results. 1. Prediction of expected crop prices. Economic benefit is the main objective function of crop planning model, and the expected price of crops is an important parameter. This thesis uses the historical monthly price data of soybean, corn and wheat to train the LSTM model and forecast the expected price of the three crops at harvest time. In addition, through the analysis of corn and soybean price correlation, the results show that they have a strong positive correlation, and corn price is the Granger cause of soybean price. Therefore this thesis uses corn price together with soybean price for soybean price forecasting. Experimental results show that root mean square error (RMSE) of soybean price forecasting results was improved by 0.6%. 2. Multi-objective planting area planning model based on crop expected price. Taking Heilongjiang Province as an example, the optimal planting areas of soybean, corn and wheat are solved. Heilongjiang Province is an important food planting area in china. When planning planting at provincial level, the objectives should take into account economic benefits, ecological benefits and national policies. Among them, the economic benefit is set as the total income of farmers; The ecological benefit is set as the total water consumption of crops; The national policy goal is to ”expand soybean planting”, that is, to maximize the soybean planting area in Heilongjiang Province. On the other hand, the planting income will significantly affect the decision-making of farmers. Since the profit per mu of soybean is lower than that of corn, in order tpromote the income increase of farmers and enhance the enthusiasm of farmers to plant soybean, taking Heilongjiang Province as an example, this thesis constructs the planting area planning model with soybean additional subsidy, and adds constraint that the average income per mu of farmers is not lower than the average income per mu of last year, which provides decision support for relevant departments. 3. Construction and solution of multi-farmer crop allocation model based on reinforcement learning. The multi-objective acreage planning model is constructed and solved, and the overall optimal acreage of three crops can be determined. In order to realize the planting demand on the specific land, this thesis constructs a multi-farmer crop allocation model, and provides decision support for multiple farmers to guide farmers to plant appropriate crops. This not only better fulfills the planting needs of each crop, but also avoids the blindness of farmers’production and reduces the market risks faced by farmers. The construction of the model needs to take into account the equity of farmers, namely: Considering the social benefits, the average profit gap of each farmer needs to be within a reasonable range; Crop rotation requirements, namely, try to meet rotation constraints and protect soil. This thesis uses reinforcement learning to solve the crop allocation problem at the farmer level, and learn to satisfy the crop rotation constraints and improve the fairness among farmers in the interaction with the simulation environment, which improves the flexibility of model construction. To sum up, this thesis constructs a regional-level crop acreage planning model and a farmer-level crop allocation model. In practical application, two models can be constructed at the same time, so that the regional-level regulatory objectives can be implemented into specific crop production.
关键词	种植规划扩种大豆农作物价格预测粒子群算法强化学习
语种	中文
七大方向——子方向分类	人工智能+农业
国重实验室规划方向分类	智能计算与学习
是否有论文关联数据集需要存交	否
文献类型	学位论文
条目标识符	http://ir.ia.ac.cn/handle/173211/52301
专题	毕业生_硕士学位论文
推荐引用方式 GB/T 7714	许钰林. 结合经济生态约束的“区域-农户”种植规划研究[D],2023.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
毕业论文-许钰林.pdf（2700KB）	学位论文		限制开放	CC BY-NC-SA