基于 TimesTransformer 的农业设施环境预测与调控研究

Research on environmental prediction and regulation of agricultural facility environment based on TimesTransformer

摘要

[目的]设施农业高投入、高产出，如何提高设施环境调控效益、降低能源消耗已成为其能否可持续发展的关键。构建精准的农业设施环境时序变化预测模型并据此开发设施环境智能调控算法，是实现农业低碳生产的可行途径。然而，现有预测模型未充分利用时序数据中存在的多周期性特征，导致多步预测精度提升受限，从而制约了设施环境的精准调控。本文提出了一种用于农业设施环境预测的模型及相应的调控方案。[方法]本研究在 Transformer 架构的基础上设计了农业设施环境预测模型TimesTransformer，采用傅里叶变换获取多个时序周期特征，将其进行二维转换，随后利用局部-全局 Transformer 对二维数据实现基于时序依赖性的长期预测。基于已知环境参数，利用环境变化预测模型结合粒子群算法，搜索能耗较低且符合预期效果的环境调控方案，并与使用其他预测模型的调控方案进行对比。[结果]相比于 Crossformer 等模型，TimesTransformer 模型对农业设施环境温度的预测结果最贴合真实趋势。以溧水鸭舍数据集为例，在 12h 预测 12h 时，TimesTransformer 模型的决定系数(R²)、平均绝对误差(mean absolute error，MAE)和均方根误差(root mean squared error，RMSE)3 个评价指标分别为 0.918、0.501℃和0.621℃；在12h 预测6h 时，上述3 个评价指标分别为0.934、0.456℃和0.516℃，优于 Crossformer 等时序预测模型。环境调控方案在 1 周内的能源消耗为 341.4kW·h^-1，相比阈值控制法(504.6kW·h^-1)降低了约 32.3%。[结论]本研究提出的环境变化预测模型 TimesTransformer 可以用于实现农业设施环境的长期精准预测，其配套调控方案能够显著降低调控过程中产生的能源消耗。

Abstract

[Objectives] Facility agriculture has high input and high output. How to improve the efficiency of facility environmental regulation and reduce energy consumption has become the key to its sustainable development. Constructing an accurate predictive model of the temporal changes in the agricultural facility environment and developing intelligent control algorithms for the facility environment based on this model is a feasible way to achieve low-carbon agricultural production. However, existing predictive models have not fully utilized the multi-periodicity characteristics in the temporal data, which limits the improvement of multi-step prediction accuracy and thus restricts the precise control effect of the facility environment. This paper proposed a model for the environmental prediction of agricultural facilities and the corresponding regulation scheme. [Methods] In this study, an agricultural facilities environmental prediction model, TimesTransformer, was designed based on the Transformer architecture. The Fourier transform was adopted to obtain the characteristics of multiple time series periods, which were then transformed into two-dimensional data. Subsequently, the local-global transformer was utilized to achieve long-term prediction based on the temporal dependence for the two-dimensional data. Based on the known environmental parameters, the environmental change prediction model proposed in this paper was combined with the particle swarm optimization algorithm to search for an environmental regulation scheme with low energy consumption and meeting the expected effects, and this scheme was compared with the regulation schemes using other prediction models. [Results] Compared with models such as Crossformer, the prediction results of the TimesTransformer model for the environmental temperature of agricultural facilities were the most consistent with the real trends. Taking the Lishui duck house dataset as an example, when predicting 12h with a prediction horizon of 12h, the three evaluation indices of the TimesTransformer model, namely the coefficient of determination (R²), the mean absolute error(MAE), and the root mean squared error(RMSE), were 0.918, 0.501℃, and 0.621℃, respectively. When predicting 6h with a prediction horizon of 12h, the above three evaluation indices were 0.934, 0.456℃, and 0.516℃, respectively, which were superior to time series prediction models such as Crossformer. Moreover, the energy consumption of the environmental regulation scheme proposed in this study within one week was 341.4kW·h^-1, which was approximately 32.3% lower than the 504.6kW·h^-1 of the threshold control method. [Conclusions] The environmental change prediction model TimesTransformer proposed in this study can be used to achieve long-term accurate prediction of agricultural facility environments. Its accompanying control scheme can significantly reduce the energy consumption generated in the control process.

关键词

农业设施环境 / 温度预测 / 环境调控 / 快速傅里叶变换 / TimesTransformer / 模型预测控制

Key words

agricultural facility environment / temperature prediction / environment regulation / fast Fourier transform / TimesTransformer / model predictive control

参考文献

[1] Maraveas C, Karavass CS, Loukatos D, et al. Agricultural greenhouses: resource management technologies and perspectives for zero greenhouse gas emissions[J]. Agriculture, 2023, 13(7): 1464-1509.

[2] Goh YS, Hum YC, Lee YL, et al. A meta-analysis: food production and vegetable crop yields of hydroponics[J]. Scientia Horticulturae, 2023, 321(15): 112339.

[3] Gu DN, Andreev K, Dupre ME. Major trends in population growth around the world[J]. China CDC Weekly, 2021, 3(28): 604-613.

[4] Mittal A, Brajpuriya R, Gupta R. Solar steam generation using hybrid nanomaterials to address global environmental pollution and water shortage crisis[J]. Materials Today Sustainability, 2023, 21(1): 100319.

[5] Alabi TM, Aghimien EI, Agbajor FD, et al. A review on the integrated optimization techniques and machine learning approaches for modeling, prediction, and decision making on integrated energy systems[J]. Renewable Energy, 2022, 194(14): 822-849.

[6] 黄银锋, 孙国祥, 汪小旵, 等. 设施农业自主移动装备导航技术研究进展[J]. 智能化农业装备学报(中英文), 2025, 6(4): 1-16.

[7] Huang YF, Sun GX, Wang XC, et al. Research progress in navigation technology for autonomous mobile equipment in facility agriculture[J]. Journal of Intelligent Agricultural Mechanization(in Chinese and English), 2025, 6(4): 1-16(in Chinese with English abstract).

[8] 谢苗苗, 李华龙, 詹凯. 基于多元数据的夏季鸡舍环境质量评价及其对产蛋性能的影响[J]. 农业工程学报, 2024, 40(8): 188-197.

[9] Xie MM, Li HL, Zhan K. Evaluating the environmental quality of layer house in summer based on multivariate data and its impact on production performance[J]. Transactions of the Chinese Society of Agricultural Engineering, 2024, 40(8): 188-197(in Chinese with English abstract).

[10] Javaid M, Haleem A, Khan IH, et al. Understanding the potential applications of artificial intelligence in agriculture sector[J]. Advanced Agrochem, 2023, 2(1): 15-30.

[11] 朱德兰, 涂泓滨, 王瑞心, 等. 基于分段多区间的温室夏季温湿度智能控制策略[J]. 农业机械学报, 2022, 53(9): 334-341.

[12] Zhu DL, Tu HB, Wang RX, et al. Intelligent control strategy of temperature and humidity in greenhouse in summer based on subsection and multi-interval[J]. Transactions of the Chinese Society for Agricultural Machinery, 2022, 53(9): 334-341(in Chinese with English abstract).

[13] Chen JD, Zhu DD, Ding SJ, et al. Government environmental concerns and corporate green innovation: evidence from heavy-polluting enterprises in China[J]. Business Strategy and the Environment, 2024, 33(3): 1920-1936.

[14] Li H, Zhang L, Zhang J, et al. Impact of technological innovation on agricultural green total factor productivity in China: considering the threshold effect of environmental regulation[J]. Journal of Environmental Planning and Management, 2025, 68(5): 1057-1081.

[15] 陈安琪, 杨春彦, 董丙强, 等. 家禽肉品质评价指标及其影响因素研究进展[J]. 东北农业大学学报, 2024, 55(5): 90-100.

[16] Chen AQ, Yang CY, Dong BQ, et al. Research progress on evaluation indices of poultry meat quality and its influencing factors[J]. Journal of Northeast Agricultural University, 2024, 55(5): 90-100(in Chinese with English abstract).

[17] Vandana GD, Sejain V, Lees AM, et al. Heat stress and poultry production: impact and amelioration[J]. International Journal of Biometeorology, 2021, 65(2): 163-179.

[18] Chen SL, Liu AL, Tang F, et al. A review of environmental control strategies and models for modern agricultural greenhouses[J]. Sensors, 2025, 25(5): 1388-1410.

[19] Peng B, Guan KY, Tang JY, et al. Towards a multiscale crop modelling framework for climate change adaptation assessment[J]. Nature Plants, 2020, 6(4): 338-348.

[20] Sun N, Fan BB, Ding YH, et al. Analysis of the development status and prospect of China's agricultural sensor market under smart agriculture[J]. Sensors, 2023, 23(6): 3307-3321.

[21] Nandi A, De A, Mallick A, et al. Attention based long-term air temperature forecasting network: ALTFTNet[J]. Knowledge-Based Systems, 2022, 252(18): 109442.

[22] Zhang KN, Qiang SY, Liang ZX, et al. A greenhouse temperature prediction framework based on transformer neural network[C]// 2021 International Conference on Computer Information Science and Artificial Intelligence(CISAI). September 17-19, 2021, Kunming, China. Piscataway: IEEE, 2021: 232-236.

[23] Mao XJ, Ren N, Dai PY, et al. A variable weight combination prediction model for climate in a greenhouse based on BiGRU-Attention and LightGBM[J]. Computers and Electronics in Agriculture, 2024, 219(4): 108818.

[24] 陈冲, 刘星桥, 刘超吉, 等. 哺乳母猪舍环境舒适度评价预测模型优化[J]. 农业机械学报, 2020, 51(8): 311-319.

[25] Chen C, Liu XQ, Liu CJ, et al. Optimization of evaluation and prediction model of environmental comfort in lactating sow house[J]. Transactions of the Chinese Society for Agricultural Machinery, 2020, 51(8): 311-319(in Chinese with English abstract).

[26] Xu LY, Jiang J, Du JG. How do environmental regulations and financial support for agriculture affect agricultural green development? The mediating role of agricultural infrastructure[J]. Journal of Environmental Planning and Management, 2025, 68(3): 557-584.

[27] Ragaveena S, Shirly EA, Surendran U. Smart controlled environment agriculture methods: a holistic review[J]. Reviews in Environmental Science and Biotechnology, 2021, 20(4): 887-913.

[28] Venkataramanan V, Kavitha G, Joel MR, et al. Forest fire detection and temperature monitoring alert using IoT and machine learning algorithm[C]// 2023 5th International Conference on Smart Systems and Inventive Technology(ICSSIT). January 23-25, 2023, Tirunelveli, India. Piscataway: IEEE, 2023: 1150-1156.

[29] Tran D, Schouteten JJ, Degieter M, et al. European stakeholders' perspectives on implementation potential of precision weed control: the case of autonomous vehicles with laser treatment[J]. Precision Agriculture, 2023, 24(6): 2200-2222.

[30] 谢秋菊, 王圣超, MUSABIMANA J, 等. 基于深度强化学习的猪舍环境控制策略优化与能耗分析[J]. 农业机械学报, 2023, 54(11): 376-384, 430.

[31] Xie QJ, Wang SC, Musabimana J, et al. Pig building environment optimization control and energy consumption analysis based on deep reinforcement learning[J]. Transactions of the Chinese Society for Agricultural Machinery, 2023, 54(11): 376-384, 430(in Chinese with English abstract).

[32] Siskandar R, Santosa SH, Wiyoto W, et al. Control and automation: in smoa f(integrated smart modern agriculture and fisheries) on the greenhouse model[J]. Jurnal Ilmu Pertanian Indonesia, 2022, 27(1): 141-152.

[33] Hu JF, Shan YH, Guerrero JM, et al. Model predictive control of microgrids: an overview[J]. Renewable and Sustainable Energy Reviews, 2021, 136: 110422.

[34] Ghanbari T, Mehraban A. Stator winding fault detection of induction motors using fast fourier transform on rotor slot harmonics and least square analysis of the Park's vectors[J]. IET Electric Power Applications, 2024, 18(3): 356-366.

基金资助

国家自然科学基金项目(62072247)

江苏省碳达峰碳中和技术创新专项(BE2022310)