Neuro-analytical forecasting of yield in programmed сultivation of agricultural crops taking into account аgrotechnological factors

Modeling the processes of bio-productivity of agricultural land is a complex multisectoral problem. The combination of a variety of biological, soil-climatic, agrotechnological and organizational-economic factors for crop forecasting requires the use, adaptation and improvement of intersectoral methods and approaches, including those based on artificial intelligence (AI).

Introduction. Agricultural production planning involves the improvement of methods for forecasting the yield and productivity of agricultural crops. In the field of agricultural production, various forecasting methods are used, including trend-seasonal statistical models, as well as artificial intelligence and machine learning methods. However, the practical implementation of such approaches constrains its wide application, in particular, the use of an iterative algorithm of variationally weighted approximations.

Object. The object is the BP of long–term yield levels of various agricultural crops.

Methods. Agrophysical processes are modeled by linear and nonlinear ordinary differential equations (ODES), as well as partial differential equations. For statistical evaluation and approximation of empirical results, the method of least modules and its weighted or generalized modifications were used. The fundamental assumption of modeling formulated by A. P. Likhatsevich (Belarus), it is accepted that each of the crop-forming factors, for example, mineral nutrition and heat and moisture availability, informs him of a change that does not depend on the effects of other factors. This can be used in mathematical modeling of the forecast yield by means of differential equations. Numerical studies for the preliminary analysis of the obtained results were carried out in the MS Excel v. 2016 environment.

Results and discussion. If the modeling of the crop yield level of factors is not linked to the biological characteristics of crops and climatic conditions, then such mathematical models will be quite universal and can adapt to different conditions. Based on the factorial approach of A. P. Likhatsevich, the authors obtained an analytical dependence that provides an operational adjustment of the forecast yield taking into account the results of segmentation of the state of crops. It is shown that it is possible to estimate the level of yield reduction using the analytical dependence obtained by the authors, taking into account the numerical processing of the results of segmentation of the state of agricultural fields, in particular the value of R_defectn obtained in the process of intelligent segmentation of the sowing area, with a given R_i,add.

Conclusions. The calculations carried out showed, in particular, that the decrease in the estimated yield on the example of barley from the maximum value of 71.4 c/ha, with values of R_defectn = 0.2 and the accepted maximum permissible R_i,dop = 0.6 will be up to 55.6 c/ ha, and the reduction in yield will be 22%.

1. Safronova T. I., Stepanov V. I. Mathematical models in reclamation problems: monograph. International Journal of Experimental Education. 2015. № 10 (2). Pp. 165-166.

2. Khvorova L. A., Topazh A. G. Building models of agroecosystems and their adaptation to specific conditions. Scientific and technical sheets SPU. 2011. № 1 (115).

3. Sukhanov P. A., Komarov A. A., Poluektov R. A. Conceptual database model for automated collection of information on the state of fields and crops. News of St. Petersburg State Agrarian University. 2013. № 31. Pp. 91-95.

4. Khvorova L. A. Optimization of the process of structural and parametric identification of productivity models of agroecosystems. News of Altai State University. 2012. № 1-1 (73). Pp. 171-175.

5. Poluektov R. A., Topazh A. G., Yakushev V. P., Medvedev S. A. Using a dynamic model of the agroecosystem to assess the impact of climatic changes on crop productivity. Bulletin of the Russian Academy of Agricultural Sciences. 2012. № 2. Pp. 7.

6. Chetyrbotsky V. A., Chetyrbotsky A. N., Levin B. V. Mathematical modeling of the dynamics of mineral nutrition of plants in the fertilizer-soil-plant system. Biophysics. 2020. V. 65. № 6. Pp. 1219-1229.

7. Likhatsevich A.P. Mathematical model of crop yield. Weight. Nats. Acad. navuk Belarusі. Ser. agrarian. navuk. 2021. V. 59. № 3. Pp. 304-318.

8. Rogachev A. F. Parametrization of econometric dependencies by the method of least modules. Management of economic systems: electronic scientific journal. 2011. № 3. P. 0421100034.

9. Rogachev A. F., Belousov I. S. Modeling of the process of training the neural network DeepLabv3 for segmentation of agricultural fields. Bulletin of Dagestan State Technical University. Technical sciences. 2023. № 50 (3). Pp. 142-149.

10. Melikhova E. V., Melikhov D. A. The use of unmanned aerial vehicles in agricultural production. International Journal of Applied Sciences and Technology Integral. 2019. № 3. P. 29.

11. Semenenko N. N. Peat-swamp soils of Polesie: transformation and ways of effective use. Minsk: Belarus. navuka, 2015. 282 p.

12. Borovoy E. P. The use of satellite images to identify unused lands in the Bykovsky district of the Volgograd region. Strategic development of the agro-industrial complex and rural territories of the Russian Federation in modern international conditions: materials of the International Scientific and Practical Conference dedicated to the 70th anniversary of Victory in the Great Patriotic War of 1941-1945. Volgograd: Volgograd State Agrarian University, 2015. V. 3. Pp. 369-373.

13. Melikhova E. V., Rogachev A. F., Skiter N. N. Information system and database for simulation of irrigated crop growing. Studies in Computational Intelligence. 2019. V. 826. Pp. 1185-1191.