Measurement analysis of the security characteristics of the water-energy-food coupling system based on the BP neural network optimized by Sailfish algorithm

This study aims to accurately quantify the security status of regional water-energy-food (WEF) coupling systems. A back propagation neural network model was improved by the sailfish optimization algorithm (SFO-BPNN). A systematic analysis was carried out to determine the security levels of the WEF coupling system in Harbin of Northeast China from 2000 to 2022. A security evaluation index system was established for the WEF coupling system. The parameters were then optimized to combine the amalgamated principal component analysis, R-cluster analysis, Pearson correlation coefficient, and coefficient of variation (PCA-RCA-PCC-CV). This multi-faceted approach facilitated the precise relationships within the WEF coupling system. The temporal evolution was determined for the security of the coupling system. It was found that the security index of the WEF coupling system shared the unique trends from the historical data. Specifically, the index first decreased and then increased in the period from 2000 to 2002 and from 2010 to 2013, respectively. There was a decrease year by year from 2002 to 2004, whereas the index climbed sharply from 2004 to 2010. A stable upward trend was observed from 2013 to 2022. Among them, Bayan County shared the lowest average security index among the regions, indicating more weaknesses in its WEF coupling system. By contrast, the urban area had the highest average security index. A more stable and reliable system was then obtained in this area. The key driving factors were identified. Precipitation was then regarded as one of the key factors. Because the precipitation directly affected the availability of the water resources, there was a cascading effect on energy and food production. The number of mechanized wells per hectare was closely related to agricultural water extraction and potential energy applications. Per capita grain production was a significant indicator of food security in the WEF coupling system. The total power of agricultural machinery is dominated by both food production and energy consumption. Compared with the traditional back-propagation neural networks (BPNN) and their optimized by genetic algorithms (GA-BPNN), the mean absolute errors of the SFO-BPNN model decreased significantly by 16.94% and 3.36%, respectively. The mean squared error decreased by 26.40% and 16.93%, respectively, compared with the BPNN and GA-BPNN, while the mean absolute percentage error decreased by 22.89% and 2.66%, respectively. In addition, the single-run time of the SFO-BPNN model was significantly shortened by 31.6% and 30.5%, respectively, whereas, the coefficient of determination increased by 0.98% and 0.15%, respectively. The better performance of the SFO-BPNN model was achieved in the accuracy and efficiency. In summary, the findings can provide a highly efficient model for the security characteristics of the WEF coupling systems. The valuable ideas can also offer to effectively alleviate the regional security risks. A solid foundation can be gained for more scientific and reasonable resource decision-making on the WEF connection.