| CPC G06F 30/27 (2020.01) | 3 Claims |
|
1. A prediction method for water quality biotoxicity based on an artificial intelligence neural network, comprising:
step S100, simulating, by a water quality simulation unit, a plurality of simulated to-be-predicted waters, and putting a fixed number of test fish into the plurality of simulated to-be-predicted waters;
step S200, collecting, by a water quality collection unit, water quality parameters of each of the plurality of simulated to-be-predicted waters, and sending, by the water quality collection unit, the water quality parameters of each of the plurality of simulated to-be-predicted waters to a water quality determination unit and a model construction unit; and collecting, by a fish collection unit, fish response parameters of the test fish in each of the plurality of simulated to-be-predicted waters, and sending, by the fish collection unit, the fish response parameters of the test fish in each of the plurality of simulated to-be-predicted waters to a fish analysis unit;
wherein the water quality parameters comprise: initial water quality parameters and real-time water quality parameters; the initial water quality parameters comprise an initial dissolved oxygen content, an initial power of hydrogen (pH) value and an initial concentration of heavy metal ions in each of the plurality of simulated to-be-predicted waters; and the real-time water quality parameters comprise a real-time dissolved oxygen content, a real-time pH value and a real-time concentration of heavy metal ions in each of the plurality of simulated to-be-predicted waters;
wherein the fish response parameters comprise: initial fish response parameters and real-time fish response parameters; the initial fish response parameters comprise an initial total survival of the test fish; and the real-time fish response parameters comprise a real-time total survival of the test fish, a real-time abnormal total number of the test fish and a real-time total number of reproduction individuals of the test fish, and the real-time abnormal total number of the test fish is a number of test fish with growth retardation or limited reproduction;
step S300, analyzing, by a water quality analysis unit, a water quality situation in each of the plurality of simulated to-be-predicted waters to obtain a variation value of water pollution, and sending, by the water quality analysis unit, the variation value of water pollution to the water quality determination unit; and analyzing, by the fish analysis unit, a fish situation of the test fish in each of the plurality of simulated to-be-predicted waters to obtain a fish abnormal value, and sending, by the fish analysis unit, the fish abnormal value to the water quality determination unit;
wherein analysis steps of the water quality analysis unit comprise:
obtaining the initial water quality parameters in each of the plurality of simulated to-be-predicted waters to obtain the initial dissolved oxygen content, the initial pH value and the initial concentration of heavy metal ions in each of the plurality of simulated to-be-predicted waters;
obtaining the real-time water quality parameters in each of the plurality of simulated to-be-predicted waters to obtain the real-time dissolved oxygen content, the real-time pH value and the real-time concentration of heavy metal ions in each of the plurality of simulated to-be-predicted waters;
calculating an initial water pollution value and a real-time water pollution value in each of the plurality of simulated to-be-predicted waters based on the initial dissolved oxygen content, the initial pH value and the initial concentration of heavy metal ions in each of the plurality of simulated to-be-predicted waters and the real-time dissolved oxygen content, the real-time pH value and the real-time concentration of heavy metal ions in each of the plurality of simulated to-be-predicted waters;
performing no-operation in response to the initial water pollution value being greater than or equal to the real-time water pollution value; and
calculating, in response to the initial water pollution value being smaller than the real-time water pollution value, a difference between the real-time water pollution value and the initial water pollution value to obtain the variation value of water pollution of each of the plurality of simulated to-be-predicted waters;
wherein analysis steps of the fish analysis unit comprise:
obtaining the real-time fish response parameters of the test fish in each of the plurality of simulated to-be-predicted waters to obtain the real-time total survival of the test fish, the real-time abnormal total number of the test fish and the real-time total number of reproduction individuals of the test fish in each of the plurality of simulated to-be-predicted waters;
obtaining the initial total survival of the test fish in each of the plurality of simulated to-be-predicted waters;
comparing the initial total survival of the test fish with the real-time total survival of the test fish;
calculating, in response to the real-time total survival of the test fish being smaller than or equal to the initial total survival of the test fish, a fish death number of the test fish in each of the plurality of simulated to-be-predicted waters by subtracting the real-time total survival of the test fish from the initial total survival of the test fish;
calculating, in response to the real-time total survival of the test fish being greater than the initial total survival of the test fish, the fish death number of the test fish in each of the plurality of simulated to-be-predicted waters by adding the initial total survival of the test fish to the real-time total number of reproduction individuals of the test fish and subtracting the real-time total survival of the test fish; and
calculating the fish abnormal value of the test fish in each of the plurality of simulated to-be-predicted waters based on the real-time abnormal total number of the test fish and the fish death number of the test fish;
step S400, determining, by the water quality determination unit, water quality of each of the plurality of simulated to-be-predicted waters to obtain an actual water quality level of each of the plurality of simulated to-be-predicted waters, and sending, by the water quality determination unit, the actual water quality level of each of the plurality of simulated to-be-predicted waters to a data comparison unit;
wherein a work process of the water quality determination unit comprises:
obtaining the variation value of water pollution of each of the plurality of simulated to-be-predicted waters;
obtaining the fish abnormal value of the test fish in each of the plurality of simulated to-be-predicted waters;
calculating a water quality determination value of each of the plurality of simulated to-be-predicted waters based on the variation value of water pollution and the fish abnormal value of the test fish; and
comparing the water quality determination value of each of the plurality of simulated to-be-predicted waters with a water quality determination threshold to determine the actual water quality level of each of the plurality of simulated to-be-predicted waters;
step S500, introducing the variation value of water pollution of each of the plurality of simulated to-be-predicted waters and the fish abnormal value of the test fish in each of the plurality of simulated to-be-predicted waters into a water quality prediction model to obtain a predicted water quality level of each of the plurality of simulated to-be-predicted waters, and sending the predicted water quality level of each of the plurality of simulated to-be-predicted waters to the data comparison unit; and
wherein a data comparison process of the data comparison unit comprises:
obtaining the predicted water quality level of each of the plurality of simulated to-be-predicted waters output by the water quality prediction model;
obtaining the actual water quality level of each of the plurality of simulated to-be-predicted waters calculated by the water quality determination unit;
comparing the actual water quality level of each of the plurality of simulated to-be-predicted waters with the predicted water quality level of the simulated to-be-predicted water;
counting, in response to the actual water quality level being equal to the predicted water quality level, a correctly predicted number once;
obtaining a number of the plurality of simulated to-be-predicted waters, recording the number of the plurality of simulated to-be-predicted waters as a number of waters, and comparing the correctly predicted number with the number of waters to obtain prediction accuracy of the plurality of simulated to-be-predicted waters;
generating, in response to the prediction accuracy being smaller than preset prediction accuracy, a signal indicating false prediction; and
generating, in response to the prediction accuracy being greater than or equal to the preset prediction accuracy, a signal indicating correct prediction;
wherein a process for obtaining the water quality prediction model comprises:
obtaining the initial water quality parameters and the real-time water quality parameters in each of the plurality of simulated to-be-predicted waters and the initial fish response parameters and the real-time fish response parameters of the test fish in each of the plurality of simulated to-be-predicted waters;
constructing the water quality prediction model by the model construction unit using a deep learning framework, inputting the initial water quality parameters and the real-time water quality parameters in each of the plurality of simulated to-be-predicted waters and the initial fish response parameters and the real-time fish response parameters of the test fish in each of the plurality of simulated to-be-predicted waters into the water quality prediction model to obtain the predicted water quality level of each of the plurality of simulated to-be-predicted waters, and outputting the predicted water quality level of each of the plurality of simulated to-be-predicted waters by the water quality prediction model; and
performing learning and training on the water quality prediction model, and outputting the predicted water quality level of each of the plurality of simulated to-be-predicted waters to the data comparison unit; and
step S600, comparing the actual water quality level of each of the plurality of simulated to-be-predicted waters with the predicted water quality level of each of the plurality of simulated to-be-predicted waters to generate the signal indicating false prediction or the signal indicating correct prediction.
|