Analysis Of Factor Affecting Chlorophyll A Concentration And Study On Prediction Model

With the continuous progress of human society,water environmental pollution and governance issues have received widespread attention.The content of chlorophyll a（Chl-a）is an important parameter to measure the eutrophication status of the water body and the content of algae in the water body,by analyzing the changes in the concentration of Chl-a,the dynamic changes of the water quality can be grasped more clearly,and it can provide significant help for the early warning of the algae outbreak in the water body and the governance of the water environment.This paper uses the Internet of Things technology to obtain water quality data in the study area from 2020/5 to 2020/12,based on this,analyze the temporal and spatial characteristics of water quality data and related influencing factors of Chl-a,at the same time,to a certain extent,the water pollution in the study area was evaluated,and a Chl-a concentration prediction model was constructed to realize the prediction of the Chl-a concentration in the study area.The main conclusions of this paper are as follows:（1）By analyzing the temporal and spatial characteristics of the water quality monitoring factors in the study area,it was concluded that June and September were the peaks of total nitrogen（TN）concentration,and TN in November was the minimum;the maximum value of total phosphorus（TP）concentration was in November and the minimum value was in July;the ammonia nitrogen（NH₃-N）concentration was the least polluted in July,and the pollution degree gradually increases with time;the Permanganate Index(COD_Mn)generally showed a trend of gradually decreasing with the change of months;the content of dissolved oxygen（DO）in the water body showed a certain upward trend with the monthly changes;the monthly changes of p H were mainly multi-peak distribution,with multi-peak points and valley points;the time characteristic of water turbidity was roughly distributed in an inverse proportional function curve,that was,with the change of the month,the concentration value gradually decreases.The TN concentration at the monitoring sites at both ends of the study area was lower than that at the center monitoring site;the spatial characteristics of TP were generally distributed in a multi-peak form;the spatial characteristics of NH₃-N and p H were that the concentration of NH₃-N in the middle station was relatively high,and the concentration of NH₃-N at the two ends was relatively low;During the first 3 months,the COD_Mn of each monitoring site increased slightly with water flow,There was no significant fluctuation in CODMn at each monitoring site in the next 3 months;the spatial distribution characteristics of DO and water turbidity did not change significantly.（2）Based on the four index data of Chl-a,TP,TN and COD_Mn,the comprehensive nutritional status index method is used to evaluate the eutrophication degree of the water body in the study area.The results showed that the water quality of the study area was betweenⅣandⅡfor a long time,and it was in a mild eutrophication state from May to September,and changed to a mesotrophic state after September.（3）By analyzing the distribution characteristics of Chl-a concentration and its correlation with the monitored water quality factors and meteorological factors,it was shown that the time characteristics of Chl-a concentration increase at a certain rate before August,with the change of time,the concentration of Chl-a began to gradually decrease with a small amplitude;its spatial characteristic was roughly that the concentration of Chl-a at the middle monitoring station was slightly higher than the two ends.there was a significant positive correlation between Chl-a concentration and nitrate,TP,DO,p H,water temperature,and sunshine hours.There was a significant negative correlation between Chl-a concentration,COD_Mn,and rainfall,and a significant negative correlation between ln（Chl-a）and ln（TN）.（4）Take the water quality data of the study area as the original sample data,the maximum correlation minimum redundancy（MRMR）algorithm was used to select a subset of features with better effects from the original sample data.As the input sample data for the prediction model,Then the combination of elite genetic algorithm（EGA）and simulated annealing algorithm（SA）was used to optimize the initial parameters in the extreme learning machine（ELM）network,effectively avoided the parameter falling into a local optimum during the optimization process,and finally the MRMR-SA-EGA-ELM predicts the Chl-a concentration model.The experimental results showed that the mean absolute error（MAE）,mean square error（MSE）and coefficient of determination（R²）of the MRMR-SA-EGA-ELM model predicting the concentration of chlorophyll a are 0.704,0.714,0.903,respectively,while the ELM model predicted the results MAE,MSE,and R² are 5.092,38.671,and 0.472 respectively.The effect of the MRMR-SA-EGA-ELM model has been significantly improved,and the accurate prediction of Chl-a concentration can be achieved.