Safety early warning method based on water environment monitoring data

Abstract

The invention discloses a safety early warning method based on water environment monitoring data, and relates to the field of water quality monitoring. The method specifically comprises the following steps: S1, historical data acquisition, S2, monitoring data acquisition, S3, data calculation, S4, result analysis and S5, data storage. According to the safety early warning method based on the water environment monitoring data, a future prediction value is linear by using an AR model, the AR model comprises historical data, a future result predicted by an AR result and the historical data are input into an LSTM recurrent neural network model, and at the moment, data participating in prediction comprise future prediction data and historical prediction data, so that the prediction accuracy is improved. Therefore, the prediction result can be improved, the LSTM recurrent neural network model can comprehensively consider the influence of various factors on the future, the historical data reflect the long-term mode and trend of the time sequence, the AR model performs prediction based on the current observation value, and the characteristics of the time sequence can be more comprehensively considered by combining the LSTM recurrent neural network model and the AR model.

Description

Technical Field

[0001] The invention relates to the technical field of water quality monitoring, and in particular to a safety early warning method based on water environment monitoring data. Background Art

[0002] At present, with the continuous development of social economy, water pollution incidents caused by industrial and domestic sources occur frequently, which seriously damages the ecological environment and poses a threat to human life and health. Water environment monitoring can determine the distribution of pollutants in the water environment at a fixed time and point, basically find the source of pollutants, and preliminarily determine the impact of pollution on human health. It is the front-end work of environmental pollution and governance, and is of great significance to water environment protection.

[0003] In the Chinese patent with publication number CN114217037A, a safety warning method and system based on water environment monitoring data are mentioned. The invention can monitor the water environment of each city according to the water quality monitoring status data information and the water quality monitoring change data information. Specifically, it can monitor which city has a better water environment, or monitor which city has the largest degree of change in water environment after a specified time period, so as to obtain a more reliable safety warning result. The invention detects the dissolved oxygen in the water quality as the water quality index, and sets a threshold for warning by comparing the historical city water quality index. However, the invention is only a level warning. As long as the water quality index value reaches the threshold, the warning will be issued. It cannot be divided into registration and classification of warnings, so it is impossible to provide warning levels to decision makers in time, further reducing the effectiveness of emergency management decisions for sudden water pollution. Secondly, the invention only warns by comparing the current water quality index with the historical water quality index. The historical data has uncertainty, resulting in large warning errors. Summary of the invention

[0004] The present invention aims to solve the above-mentioned problems at least to a certain extent. By dividing water quality warnings into four levels, the warning level can be obtained according to the warning prediction information and external warnings can be issued, which helps to provide decision makers with the warning level, so that they can make appropriate decisions according to the warning level and improve the effectiveness of emergency management decisions for sudden water pollution.

[0005] To achieve the above objectives, the present invention is implemented by the following technical scheme: specifically comprising the following steps:

[0006] S1: Historical data collection: First, we collect historical data on water quality index concentrations at river sections and lakes and reservoirs in the monitored cities over the past two years to establish a database;

[0007] S2: Collection of monitoring data: The current water quality index concentration is collected through the sensor equipment to obtain the actual concentration C of the current water quality index. 实际 ;

[0008] S3: Data calculation: Then use the AR algorithm model to predict the current water quality index concentration to obtain the predicted water quality index concentration C (预测) ;

[0009] Further predicting the water quality index concentration C (预测) As the input value, it is input into the LSTM recurrent neural network model to obtain the final predicted water quality index concentration C / (预测) ;

[0010] S4: Result analysis: The actual concentration of the current water quality index C 实际 and the final predicted water quality index concentration C / (预测) By using the standard deviation method model to obtain the abnormal characteristic value T, the abnormal threshold of the abnormal characteristic value T is set to four values: K1=1, K2=2, K3=3 and K4=4; then, the abnormal characteristic value is analyzed, the warning level is determined, and a warning message is issued;

[0011] S5: Data storage: After the warning ends, the actual concentration of the current water quality index C 实际 Stored in the storage module.

[0012] Preferably, the AR model in S3 can analyze the water quality index of the target city in the historical time of the target city and predict the future water quality index value. The AR model expression is:

[0013] C (预测) =c+φ1X(t-1)+φ2X(t-2)+...+φp*X(tp)+ε(t);

[0014] Among them, X(t) represents the observation value at the current time point t; c is a constant, and the constant term is usually zero; φ1, φ2, ..., φp are autoregressive coefficients; ε(t) is the error term; p represents the order of the model, that is, the number of historical time points considered in the past two years.

[0015] Preferably, the water quality index concentration C is predicted in S3. (预测) The input value is input into the LSTM recurrent neural network model, and the LSTM model can obtain future information data predicted by using historical data from the past two years and the AR algorithm model.

[0016] Preferably, the standard deviation method model expression in S4 is:

[0017] T=IC 实际 -C / (预测) Ⅰ;

[0018] Among them C实际 is the actual concentration of the current water quality index, C / (预测) To finally predict the concentration of water quality indicators.

[0019] Preferably, the warning levels in S4 are divided into level one warning, level two warning, level three warning and level four warning, which are marked in red, orange, yellow and blue respectively.

[0020] Preferably, in S4, K1=1, K2=2, K3=3 and K4=4 correspond to the first-level warning, second-level warning, third-level warning and fourth-level warning of the warning level respectively. First, the abnormal characteristic value T is judged with the threshold K4=4. If the abnormal characteristic value T≥4, it is determined as a fourth-level warning. The abnormal characteristic value T<4 is compared with the third-level warning threshold K3=3. If the abnormal characteristic value T≥3, a third-level warning is determined. The abnormal characteristic value T<3 is compared with the second-level warning threshold K3=2. The abnormal characteristic value T≥2 is a second-level alarm. The abnormal characteristic value T<2 is compared with the first-level warning threshold K3=1. The abnormal characteristic value T≥1 is a first-level alarm. The abnormal characteristic value T<1 is determined as no warning.

[0021] Preferably, the data stored in S5 can be used as one of the historical data for the next water quality warning.

[0022] Beneficial Effects

[0023] The present invention provides a safety early warning method based on water environment monitoring data, which has the following beneficial effects compared with the prior art:

[0024] This safety early warning method based on water environment monitoring data divides water quality early warning into four levels. It can obtain early warning levels according to early warning prediction information and issue external early warnings, which helps to provide early warning levels for decision makers, so that they can make appropriate decisions according to the early warning levels, thereby improving the effectiveness of emergency management decisions for sudden water pollution.

[0025] Secondly, by first using the linear future prediction value of the AR model, and the AR model contains historical data, the AR results predict the future results and the historical data are input into the LSTM recurrent neural network model. At this time, the data involved in the prediction includes future prediction data and historical prediction data, thereby improving the prediction results. In addition, the LSTM recurrent neural network model can comprehensively consider the impact of multiple factors on the future. Historical data reflects the long-term patterns and trends of the time series, while the AR model makes predictions based on the current observations. Combining the two can more comprehensively consider the characteristics of the time series. Finally, the AR model mainly considers the linear autoregressive relationship of the time series, while the LSTM recurrent neural network model can better handle nonlinear relationships and long-term dependencies. Combining the two can make up for their respective shortcomings and improve the prediction ability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a block diagram of the method of the present invention;

[0027] Figure 2 The figure is a flow chart of data calculation and structural analysis of the present invention. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0029] Example:

[0030] See also Figure 1 as well as Figure 2 , a safety early warning method based on water environment monitoring data, specifically comprising the following steps:

[0031] S1: Historical data collection: First, we collect historical data on water quality index concentrations at river sections and lakes and reservoirs in the monitored cities over the past two years to establish a database;

[0032] S2: Collection of monitoring data: The current water quality index concentration is collected through the sensor equipment to obtain the actual concentration C of the current water quality index. 实际 ;

[0033] S3: Data calculation: Then use the AR algorithm model to predict the current water quality index concentration to obtain the predicted water quality index concentration C (预测) ;

[0034] Further predicting the water quality index concentration C (预测) As the input value, it is input into the LSTM recurrent neural network model to obtain the final predicted water quality index concentration C / (预测) ;

[0035] S4: Result analysis: The actual concentration of the current water quality index C 实际 and the final predicted water quality index concentration C / (预测) By using the standard deviation method model to obtain the abnormal characteristic value T, the abnormal threshold of the abnormal characteristic value T is set to four values: K1=1, K2=2, K3=3 and K4=4; then, the abnormal characteristic value is analyzed, the warning level is determined, and a warning message is issued;

[0036] S5: Data storage: After the warning ends, the actual concentration of the current water quality index C 实际 Stored in the storage module.

[0037] The AR model in S3 can analyze the water quality index of the target city in the historical time of the target city and predict the future water quality index value. The AR model expression is:

[0038] C (预测) =c+φ1X(t-1)+φ2X(t-2)+...+φp*X(tp)+ε(t);

[0039] Among them, X(t) represents the observation value at the current time point t; c is a constant, and the constant term is usually zero; φ1, φ2, ..., φp are autoregressive coefficients; ε(t) is the error term; p represents the order of the model, that is, the number of historical time points considered in the past two years.

[0040] The water quality index concentration C will be predicted in S3 (预测) The input value is input into the LSTM recurrent neural network model, and the LSTM model can obtain future information data predicted by using historical data from the past two years and the AR algorithm model.

[0041] The standard deviation method model expression in S4 is:

[0042] T=IC 实际 -C / (预测) Ⅰ;

[0043] Among them C 实际 is the actual concentration of the current water quality index, C / (预测) To finally predict the concentration of water quality indicators.

[0044] The warning levels in S4 are divided into level 1 warning, level 2 warning, level 3 warning and level 4 warning, which are marked in red, orange, yellow and blue respectively.

[0045] In S4, K1=1, K2=2, K3=3 and K4=4 correspond to the first-level warning, second-level warning, third-level warning and fourth-level warning of the warning levels respectively. First, the abnormal characteristic value T is judged with the threshold K4=4. If the abnormal characteristic value T≥4, it is determined as a fourth-level warning. The abnormal characteristic value T<4 is compared with the third-level warning threshold K3=3. If the abnormal characteristic value T≥3, it is determined as a third-level warning. The abnormal characteristic value T<3 is compared with the second-level warning threshold K3=2. The abnormal characteristic value T≥2 is a second-level alarm. The abnormal characteristic value T<2 is compared with the first-level warning threshold K3=1. The abnormal characteristic value T≥1 is a first-level alarm. The abnormal characteristic value T<1 is determined as no warning.

[0046] The data stored in S5 can be used as one of the historical data for the next water quality early warning.

[0047] First, the historical data of water quality index concentration in the past two years is used as a database. When monitoring the current water quality, the sensor equipment is used to detect the current water quality index concentration, and the AR model is used to capture the historical data for prediction to obtain the predicted water quality index concentration C. (预测) Then C (预测) Data is input into the LSTM recurrent neural network model, which captures historical data and C (预测) Data prediction to obtain C / (预测) , and then use the standard deviation method model to calculate the abnormal characteristic value T. The system analyzes the abnormal characteristic value T, determines whether there is an early warning and the level of the early warning when there is an early warning, and issues early warning information to the outside.

[0048] First, water quality warnings are divided into four levels. The warning level can be obtained based on the warning prediction information and external warnings can be issued, which helps to provide decision makers with warning levels, so that they can make appropriate decisions based on the warning level and improve the effectiveness of emergency management decisions for sudden water pollution.

[0049] Secondly, by first using the linear future prediction value of the AR model, and the AR model contains historical data, the AR results predict the future results and the historical data are input into the LSTM recurrent neural network model. At this time, the data involved in the prediction includes future prediction data and historical prediction data, thereby improving the prediction results. In addition, the LSTM recurrent neural network model can comprehensively consider the impact of multiple factors on the future. Historical data reflects the long-term patterns and trends of the time series, while the AR model makes predictions based on the current observations. Combining the two can more comprehensively consider the characteristics of the time series. Finally, the AR model mainly considers the linear autoregressive relationship of the time series, while the LSTM recurrent neural network model can better handle nonlinear relationships and long-term dependencies. Combining the two can make up for their respective shortcomings and improve the prediction ability of the model.

[0050] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0051] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0052] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A safety early warning method based on water environment monitoring data, characterized in that: The specific steps include: S1: Historical data collection: First, we collect historical data on water quality index concentrations at river sections and lakes and reservoirs in the monitored cities over the past two years to establish a database; S2: Collection of monitoring data: The current water quality index concentration is collected through the sensor equipment to obtain the actual concentration C of the current water quality index. 实际 ; S3: Data calculation: Then use the AR algorithm model to predict the current water quality index concentration to obtain the predicted water quality index concentration C (预测) ; Further predicting the water quality index concentration C (预测) As the input value, it is input into the LSTM recurrent neural network model to obtain the final predicted water quality index concentration C / (预测) ; S4: Result analysis: The actual concentration of the current water quality index C 实际 and the final predicted water quality index concentration C / (预测) By using the standard deviation method model to obtain the abnormal characteristic value T, the abnormal threshold of the abnormal characteristic value T is set to four values: K1=1, K2=2, K3=3 and K4=4; then, the abnormal characteristic value is analyzed, the warning level is determined, and a warning message is issued; S5: Data storage: After the warning ends, the actual concentration of the current water quality index C 实际 Stored in the storage module.

2. A safety early warning method based on water environment monitoring data according to claim 1, characterized in that: The AR model in S3 can analyze the water quality index of the target city in the historical time of the target city and predict the future water quality index value. The AR model expression is: C (预测) =c+φ1X(t-1)+φ2X(t-2)+...+φp*X(t-p)+ε(t); Among them, X(t) represents the observation value at the current time point t; c is a constant, and the constant term is usually zero; φ1, φ2, ..., φp are autoregressive coefficients; ε(t) is the error term; p represents the order of the model, that is, the number of historical time points considered in the past two years.

3. A safety early warning method based on water environment monitoring data according to claim 1, characterized in that: The water quality index concentration C will be predicted in S3 (预测) The input value is input into the LSTM recurrent neural network model, and the LSTM model can obtain future information data predicted by using historical data from the past two years and the AR algorithm model.

4. A safety early warning method based on water environment monitoring data according to claim 1, characterized in that: The standard deviation method model expression in S4 is: T=ⅠC 实际 -C / (预测) Ⅰ; Among them C 实际 is the actual concentration of the current water quality index, C / (预测) To finally predict the concentration of water quality indicators.

5. A safety early warning method based on water environment monitoring data according to claim 1, characterized in that: The warning levels in S4 are divided into level 1 warning, level 2 warning, level 3 warning and level 4 warning, which are marked in red, orange, yellow and blue respectively.

6. A safety early warning method based on water environment monitoring data according to claim 5, characterized in that: In S4, K1=1, K2=2, K3=3 and K4=4 correspond to the first-level warning, second-level warning, third-level warning and fourth-level warning of the warning levels respectively. First, the abnormal characteristic value T is judged with the threshold K4=4. If the abnormal characteristic value T≥4, it is determined as a fourth-level warning. The abnormal characteristic value T<4 is compared with the third-level warning threshold K3=3. If the abnormal characteristic value T≥3, it is determined as a third-level warning. The abnormal characteristic value T<3 is compared with the second-level warning threshold K3=2. The abnormal characteristic value T≥2 is a second-level alarm. The abnormal characteristic value T<2 is compared with the first-level warning threshold K3=1. The abnormal characteristic value T≥1 is a first-level alarm. The abnormal characteristic value T<1 is determined as no warning.

7. A safety early warning method based on water environment monitoring data according to claim 1, characterized in that: The data stored in S5 can be used as one of the historical data for the next water quality early warning.

Images