Safety protection early warning method based on neural network

Abstract

The invention relates to the technical field of safety early warning, in particular to a safety protection early warning method based on a neural network, and the method comprises the following steps: collecting the pH value, dissolved oxygen concentration, turbidity, temperature and flow of effluent quality, carrying out data cleaning and abnormal value elimination, carrying out time sequence sorting, and generating water quality time sequence data. According to the method, the neural network model is applied to predict the water quality parameters such as the pH value, the dissolved oxygen concentration, the turbidity, the temperature and the flow, the prediction error is combined with the short-term trend value, and the water quality risk is dynamically evaluated. And meanwhile, the parameter weight is dynamically adjusted to match the real-time importance of each parameter, so that the prediction accuracy is improved. The analysis of the long-period trend and the fluctuation range enables the early warning system to perform effective early warning before the risk reaches a threshold value, and allows timely countermeasures to be taken, thereby enhancing the perspectiveness and initiative of water quality management.

Description

Technical Field

[0001] The present invention relates to the technical field of safety early warning, and in particular to a safety protection early warning method based on neural network. Background Art

[0002] The field of safety early warning technology involves the development and implementation of preventive measures to predict and reduce potential risks and threats. This field focuses on identifying risks in advance through real-time data monitoring, risk assessment models and the establishment of early warning mechanisms. In the sewage treatment process, prediction and early warning are required.

[0003] Existing technologies often take reactive rather than preventive measures, that is, they only start to deal with problems after they have already occurred, rather than warning of impending risks in advance. In addition, fixed weight allocation and parameter evaluation mechanisms show their limitations in the face of sudden changes, such as lack of sufficient flexibility to adjust risk assessment models when unpredicted pollution events occur. These limitations lead to inadequate risk management, which may pose potential threats to public health and environmental safety. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a safety protection early warning method based on a neural network.

[0005] In order to achieve the above object, the present invention adopts the following technical solution, a safety protection early warning method based on neural network, comprising the following steps: Collect the pH value, dissolved oxygen concentration, turbidity, temperature and flow rate of the effluent water quality, perform data cleaning and outlier removal, perform time series sorting, and generate water quality time series data; Based on the water quality time series data, a neural network model is trained to predict pH value, dissolved oxygen concentration, turbidity, temperature and flow parameters, and the short-term trend value and prediction error of each parameter are calculated to obtain the parameter prediction trend value. According to the importance of the parameter, the parameter weight is dynamically adjusted, the parameter weight is combined with the parameter prediction trend value, and the parameter prediction value is updated to obtain a water quality prediction parameter set; Based on the water quality prediction parameter set, select parameters of the prediction error, analyze them, calculate the long-term trend value and fluctuation range of the parameters, and merge them with the water quality prediction parameter set, update all parameter prediction values, obtain comprehensive water quality prediction data, compare the comprehensive water quality prediction data with the safety threshold, calculate the risk index of each parameter, and generate a water quality risk assessment index; Based on the water quality risk assessment indicators, the risk index of each parameter is summarized, the overall risk level is calculated, and compared with the preset warning value to generate a safety warning signal.

[0006] Preferably, the steps for acquiring the water quality time series data are: Monitor the pH value, dissolved oxygen concentration, turbidity, temperature and flow rate of the effluent water quality, analyze the data range of each parameter, identify and eliminate abnormal data points, and generate preliminary processed water quality data; Based on the preliminarily processed water quality data, the data is sorted according to the collected timestamps to obtain water quality data sorted in time series; Based on the water quality data after the time series is sorted, the changing trends of pH value, dissolved oxygen concentration, turbidity, temperature and flow are analyzed, the time series characteristics are extracted, and the water quality time series data are generated.

[0007] Preferably, the step of obtaining the parameter prediction trend value is: Using the water quality time series data, taking pH value, dissolved oxygen concentration, turbidity, temperature and flow rate as input variables, training a neural network model to obtain a trained neural network model; Specifically, the modeling statement of the neural network model is: net=newff (input samples in the training sample set, [number of neurons in the hidden layer, number of neurons in the output layer], {hidden layer activation function, output layer activation function}, 'training function for error back propagation').

[0008] It is a three-layer neural network structure, with 7 nodes in the input layer and 2 nodes in the output layer; The maximum number of iterations is set to 10000 and the precision is controlled to 10 -3 ; The prediction model parameters are obtained: the number of neurons in the hidden layer is 12, the learning rate is 0.3, the activation function of the hidden layer is the tansig function, the activation function of the output layer is the purelin function, and the training function is the Levenberg-Marquardt algorithm.

[0009] Based on the trained neural network model, the short-term trend value of each water quality parameter is predicted, and the prediction error is calculated. The calculation formula is: ; in, is the prediction error of the ith parameter, is the observed value, is the predicted value, is the length of the time series, is the weight of each time point, is the standard deviation at time point t; Based on the prediction error, a prediction trend value of each water quality parameter is generated to obtain a parameter prediction trend value.

[0010] Preferably, the steps of obtaining the water quality prediction parameter set are: According to the past data of each water quality parameter and the current environmental changes, the importance is judged, and the weight of each parameter is adjusted to obtain an adjusted parameter weight set; Based on the adjusted parameter weight set, the updated parameter prediction value is calculated using the following formula: ; in, is the updated parameter prediction value, is the weight of the i-th parameter, is the predicted trend value of the i-th parameter, is the total number of parameters; Based on the updated parameter prediction values, the parameters are merged to generate a water quality prediction parameter set.

[0011] Preferably, the steps for obtaining the comprehensive water quality prediction data are: Based on the water quality prediction parameter set, identifying and selecting parameters of the prediction error to obtain a selected error parameter set; Based on the selected error parameter set, the long-term trend value of each parameter is calculated, and the calculation formula is: ; in, is the long-term trend value of the i-th parameter, is the number of observation periods, is the observed value of the i-th parameter in the m-th period, is the average value of the i-th parameter in all cycles; Based on the long-term trend values, they are merged and updated into a water quality prediction parameter set to form comprehensive water quality prediction data.

[0012] Preferably, the steps for obtaining the water quality risk assessment index are: Obtain the comprehensive water quality prediction data, compare it with the preset safety threshold, identify the parameters exceeding the threshold, and obtain a list of over-threshold parameters; Based on the above-threshold parameter list, the risk index of each parameter is calculated using the following formula: ; in, is the risk index of the i-th parameter, is the predicted value of the parameter, is the corresponding safety threshold, is the standard deviation of the corresponding parameter; Based on the risk index, the risk index is summarized to generate a water quality risk assessment indicator.

[0013] Preferably, the steps of obtaining the safety warning signal are: Summarize the risk index of each parameter in the water quality risk assessment index to obtain an overall risk index summary; Based on the overall risk index summary, the overall risk level is calculated, the safety status of the entire water body is analyzed, and the overall risk level is formed.

[0014] Preferably, the step of acquiring the safety warning signal further comprises: comparing the overall risk level with a preset warning value, and generating a safety warning signal according to the comparison result.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are: The present invention predicts water quality parameters such as pH value, dissolved oxygen concentration, turbidity, temperature and flow rate by applying a neural network model, and combines the prediction error with the short-term trend value to achieve a dynamic assessment of water quality risks. At the same time, the parameter weights are dynamically adjusted to match the real-time importance of each parameter, thereby improving the accuracy of the prediction. The analysis of long-term trends and fluctuation ranges enables the early warning system to provide effective early warnings before the risk reaches the threshold, allowing timely response measures to be taken, thereby enhancing the foresight and initiative of water quality management. This not only improves the accuracy of risk management, but also enhances the overall adaptability and flexibility, ensuring that water treatment facilities can maintain good operation in a changing environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] See also Figure 1 The present invention provides a technical solution, a safety protection early warning method based on a neural network, comprising the following steps: Collect the pH value, dissolved oxygen concentration, turbidity, temperature and flow rate of the effluent water quality, perform data cleaning and outlier removal, perform time series sorting, and generate water quality time series data; Based on water quality time series data, the neural network model is trained to predict pH value, dissolved oxygen concentration, turbidity, temperature and flow parameters, and the short-term trend value and prediction error of each parameter are calculated to obtain the parameter prediction trend value. According to the importance of the parameter, the parameter weight is dynamically adjusted, and the parameter weight is combined with the parameter prediction trend value to update the parameter prediction value to obtain the water quality prediction parameter set. Based on the water quality prediction parameter set, select the parameters of the prediction error, analyze them, calculate the long-term trend value and fluctuation range of the parameters, and merge them with the water quality prediction parameter set to update all parameter prediction values ​​to obtain comprehensive water quality prediction data. Compare the comprehensive water quality prediction data with the safety threshold, calculate the risk index of each parameter, and generate water quality risk assessment indicators; Based on the water quality risk assessment indicators, the risk index of each parameter is summarized, the overall risk level is calculated, and compared with the preset warning value to generate a safety warning signal.

[0019] The steps for obtaining water quality time series data are as follows: Monitor the pH value, dissolved oxygen concentration, turbidity, temperature and flow rate of the effluent water quality, analyze the data range of each parameter, identify and eliminate abnormal data points, and generate preliminary processed water quality data; Based on the water quality data that has been preliminarily processed, the data is sorted according to the collected timestamps to obtain the water quality data sorted in time series; Based on the water quality data sorted by time series, the changing trends of pH value, dissolved oxygen concentration, turbidity, temperature and flow are analyzed, the time series characteristics are extracted, and water quality time series data are generated.

[0020] Specifically, the pH value, dissolved oxygen concentration, turbidity, temperature and flow rate of the effluent water quality are monitored, and the continuous data of each parameter are recorded using real-time sensors. The data set of each parameter is preliminarily screened, and the abnormal points in the data are identified through the set threshold system, such as any readings that deviate from the set normal operating range. By comparing the data changes at each time point, the continuity and accuracy of the data are confirmed, thereby eliminating data deviations caused by sensor errors or external interference. The data of each parameter is standardized to ensure the consistency of subsequent analysis. Based on these precise data, a comprehensive and preliminarily processed water quality data set is generated. This data set includes all filtered and verified water quality monitoring data, which provides a basis for the next step of time series analysis. Based on the water quality data that has been preliminarily processed, the entire data set is sorted according to the timestamp of each data point, and the timestamp is verified and corrected to ensure that all data are arranged in the order of the actual collection time. By comparing the time intervals between consecutive data points, the time consistency of the data is checked to eliminate time label errors caused by system clock errors or synchronization problems. The data is processed using a sorting algorithm to ensure the time sequence of the data. Through this step of sorting, the water quality data after time series sorting is obtained. This data set will be directly used for further trend analysis and feature extraction; Based on the water quality data sorted by time series, the changing trends of pH value, dissolved oxygen concentration, turbidity, temperature and flow were analyzed. Statistical software was used to perform regression analysis of the data. The pattern of changes of each parameter over time was identified by calculating the correlation between data points. The sliding window technology was implemented for each parameter to capture short-term and long-term change trends, and time series characteristics such as seasonal fluctuations, mutation points and periodic changes were extracted from these trends.

[0021] The steps to obtain the parameter prediction trend value are: Using water quality time series data, pH value, dissolved oxygen concentration, turbidity, temperature and flow rate as input variables, the neural network model is trained to obtain a trained neural network model; Based on the trained neural network model, the short-term trend value of each water quality parameter is predicted and the prediction error is calculated. The calculation formula is: ; in, is the prediction error of the ith parameter, is the observed value, is the predicted value, is the length of the time series, is the weight of each time point, is the standard deviation at time point t; Based on the prediction error, the prediction trend value of each water quality parameter is generated to obtain the parameter prediction trend value.

[0022] Specifically, when using water quality time series data to train the neural network model, the input data set needs to be normalized first to eliminate the influence of different dimensions and enhance the generalization ability of the model. During the normalization process, each parameter (such as pH value, dissolved oxygen concentration, turbidity, temperature and flow) will be adjusted to the same scale. For example, the data is converted into z-scores by calculating the mean and standard deviation of each parameter. This processing step ensures the numerical stability of the data during training. Then, according to the characteristics of the water quality parameters and the time dependence of the data, the number of network layers and the number of nodes in each layer are set. The model is trained through the back propagation algorithm, and the weights and biases are continuously updated to minimize the difference between the predicted values ​​and the actual values. The training process of the model is completed to obtain the trained neural network model.

[0023] The benefit of the formula is that, through the square root form of the weighted square error, it can enhance the sensitivity to sample points with large prediction errors, so that the model training focuses more on those difficult-to-predict outliers and improves the accuracy of the overall prediction. Represents the prediction error of the i-th parameter, which is obtained by calculating the weighted difference between the model prediction value and the actual observation value. is the actual observed value at time t, the water quality parameter value directly measured by the sensor, is the predicted value of the model at time t, calculated by the neural network model, is the length of the time series, indicating the total number of time points contained in the entire data set. The weight of each time point reflects the importance of data at different time points in the error calculation. The weight setting can be determined based on the representativeness of the time point or the data quality. is the standard deviation at time point t, which is used to standardize the prediction error to ensure that data of different scales and dimensions can contribute fairly to the total error; Calculation process: A data set with observations at three time points is [5.0,7.0,6.0], and the corresponding model prediction value is [4.5,7.2,5.9], standard deviation The weights of all time points are [0.5, 0.5, 0.5]. All are 1, substitute into the formula to calculate the term at the first time point: , the item at the second time point: , the item at the third time point: , the total error is: , which shows that the average error of the prediction is 1.095 after standardization.

[0024] The predicted trend value of each water quality parameter is generated based on the prediction error. The generation of the predicted trend value is based on the short-term prediction error obtained from the neural network model. These error data will be used to analyze the future change trend of each parameter. For each parameter, the moving average or other statistical measures of its prediction error will be calculated to evaluate its performance stability in the short term. In this way, potential trends or cyclical changes can be revealed, providing early warnings for possible future conditions.

[0025] The steps to obtain the water quality prediction parameter set are: According to the past data of each water quality parameter and the current environmental changes, the importance is judged, and the weight of each parameter is adjusted to obtain an adjusted parameter weight set; Based on the adjusted parameter weight set, the updated parameter prediction value is calculated using the formula: ; in, is the updated parameter prediction value, is the weight of the i-th parameter, is the predicted trend value of the i-th parameter, is the total number of parameters; Based on the updated parameter prediction values, the parameters are merged to generate a water quality prediction parameter set.

[0026] Specifically, an evaluation is conducted based on the historical data of each water quality parameter and current environmental changes, the model feedback is analyzed, and the weight of each parameter is adjusted to obtain an adjusted parameter weight set. Trends and mutation points are identified through historical data, and the magnitude of the weight change is determined through moving average or exponential smoothing. The importance of each parameter is evaluated based on its sensitivity to water quality changes and the rate of change.

[0027] The benefit of the formula is that it reflects the real-time importance of different water quality parameters by dynamically adjusting the weights of each parameter, thereby more accurately predicting changes in water quality; represents the updated parameter prediction value, is the weight of the ith parameter, determined based on the statistical analysis of the impact of the parameter on water quality in a specific time period. is the predicted trend value of the parameter, which is obtained by linear regression of past data or other time series models. is the total number of parameters involved in the calculation; Calculation process: Setting the total number of parameters , corresponding to pH, dissolved oxygen, turbidity, temperature and flow, determined by actual monitoring during the monitoring period They are 0.2, 0.3, 0.1, 0.25, 0.15 respectively. The predicted trend values ​​are 7.0, 6.0, 4.0, 8.0, and 5.0 respectively, so the calculation process is: ; The results show that the updated comprehensive parameter prediction value is 6.35, which reflects the weighted trend of current water quality parameters. Higher values ​​may indicate significant changes in certain parameters, requiring further analysis or intervention.

[0028] Based on the updated parameter prediction values, all relevant water quality parameters are integrated to generate an overall water quality prediction parameter set; this process involves integrating the predicted values ​​of each independent parameter into a unified framework. The data of each parameter is first standardized to ensure data consistency, and then the weighted average method is used to merge the parameters according to their weights to form a comprehensive data set.

[0029] The steps to obtain comprehensive water quality prediction data are as follows: Based on the water quality prediction parameter set, identifying and selecting parameters of the prediction error to obtain a selected error parameter set; Based on the selected error parameter set, the long-term trend value of each parameter is calculated using the following formula: ; in, is the long-term trend value of the i-th parameter, is the number of observation periods, is the observed value of the i-th parameter in the m-th period, is the average value of the i-th parameter in all cycles; Based on the long-term trend values, they are merged and updated into the water quality prediction parameter set to form comprehensive water quality prediction data.

[0030] Specifically, identifying parameters with significant prediction errors involves comparing past prediction data with actual monitoring data. During this process, for each parameter, the deviation between its actual value and the predicted value is recorded to evaluate the accuracy of the prediction model of each parameter; for parameters whose errors exceed the safety threshold (the safety threshold is determined based on past data analysis, such as when the error continues to exceed 10%, it is defined as exceeding the safety range), they are marked as error parameters that need to be analyzed in detail; these parameters may indicate that the current model fails to adapt to data changes under specific conditions, such as data fluctuations caused by seasonal fluctuations or environmental mutations, thus forming a targeted error parameter set, which contains all parameters that exceed the preset safety threshold for subsequent model adjustment and retraining; The benefit of the formula is that by calculating the fluctuation range and long-term trend of each parameter, we can observe the stability of the parameter over time and its behavior pattern under different conditions, thus making more accurate predictions about future water quality changes; Represents the long-term trend value of the i-th parameter, reflecting the overall trend of the parameter over time. is the number of observation periods, indicating the time range for trend analysis. is the observed value of the i-th parameter in the m-th period, showing the actual measured value of the parameter in each period. is the average value of the i-th parameter in all periods, which is used to calculate the deviation and overall trend of each period; Calculation process: There are 5 periods of data, and the observed values ​​of a certain parameter are: 10, 12, 11, 13, 14, and the average value is 12. Apply the formula to calculate the long-term trend value, the process is: ; The result shows that the parameter has a small fluctuation range during the observation period, indicating that the parameter is relatively stable during the observed period, with a fluctuation range of 1.2, indicating that the parameter may maintain a similar fluctuation trend in the future; The parameter set is merged and updated based on the calculated long-term trend values. In this process, the long-term trend values ​​of each parameter are first compared with the original water quality prediction parameter set to check which parameters' long-term trends show changes. Subsequently, the prediction model is adjusted based on the actual observed values ​​of the fluctuation range of each parameter. The fluctuation range is determined by calculating the difference between the maximum and minimum values ​​within the period, reflecting the stability and variation range of the parameter. After integrating all the data, a comprehensive water quality prediction data set is formed, which will include the adjusted parameter prediction values.

[0031] The steps for obtaining water quality risk assessment indicators are as follows: Obtain comprehensive water quality prediction data, compare it with the preset safety threshold, identify the parameters exceeding the threshold, and obtain a list of over-threshold parameters; Based on the list of over-threshold parameters, the risk index of each parameter is calculated using the following formula: ; in, is the risk index of the i-th parameter, is the predicted value of the parameter, is the corresponding safety threshold, is the standard deviation of the corresponding parameter; Based on the risk index, the risk index is summarized to generate water quality risk assessment indicators.

[0032] Specifically, after obtaining the comprehensive water quality prediction data, a comparison was made between each parameter and the safety threshold in the environmental protection standards, especially for key water quality indicators such as pH, turbidity, and chemical oxygen demand. The thresholds of these indicators are set according to national and local environmental protection standards. For example, the pH value is usually set at a safety range of 6.5 to 8.5, and the turbidity threshold may be set at different standards according to the type and use of the water body. These parameters are continuously tracked through real-time data monitoring. Once any parameter is found to exceed the threshold, an alarm will be triggered and recorded in the list of over-threshold parameters; The formula is beneficial in that it not only considers the absolute deviation of the parameter from the threshold, but also adds consideration of the volatility of the parameter by introducing the standard deviation ratio, making the risk index more sensitive to the fluctuation of the parameter value, thus providing a more accurate risk assessment; Calculation process: Setting water quality parameters 8.5 (pH value), the safety threshold is 7.0, standard deviation is 0.3, according to the formula: ; Calculated: ; This shows that the risk index of pH is 22.44, which means that the current pH value is 22.44% higher than the safety threshold. The adjustment factor 1.0429 indicates that this parameter has a small volatility, but it still needs attention. This result shows that the current level of pH may have a slight adverse effect on water quality. If the risk index continues to exceed 20%, further water quality treatment or preventive measures are recommended. The overall safety status of the water body is assessed by summarizing the calculated risk index of each parameter. This process involves comparing the risk index of each parameter with the set risk level thresholds, which are set based on previous environmental studies and statistical data. For example, a risk index of more than 20% may indicate a medium risk, while a risk index of more than 50% indicates a high risk. This method can determine which parameters may cause water quality problems, thereby providing a basis for water quality improvement measures.

[0033] The steps to obtain safety warning signals are as follows: Summarize the risk index of each parameter in the water quality risk assessment index to obtain the overall risk index summary; Based on the overall risk index summary, calculate the overall risk level, analyze the safety status of the entire water body, and form an overall risk level; Based on the overall risk level, it is compared with the preset warning value and a safety warning signal is generated according to the comparison result.

[0034] Specifically, the safety threshold is set based on historical data and environmental standards. It marks the critical point where water quality goes from safe to dangerous. Through this comparison, risk points that exceed the safety threshold are identified and included in the high-risk category, thus forming a comprehensive risk index view.

[0035] After obtaining the summarized risk index, the overall risk level is calculated by analyzing the weight distribution of each index. The weight distribution is based on the importance of each parameter in safety monitoring. The importance assessment is based on the potential harm of the parameter to water quality. For example, the risk index of heavy metal content will be assigned a higher weight than the risk index of color. The weighted average calculation multiplies each risk index by its corresponding weight and then accumulates them. This calculation result represents the risk level of the overall water body, thereby providing a quantitative and comprehensive water quality safety assessment.

[0036] The overall risk level is compared with the set warning value, which is developed based on industry standards and experience from previous risk events and defines the maximum acceptable risk limit. Through comparative analysis, it is determined whether the current risk level exceeds this warning threshold. If it does, the early warning mechanism will be automatically triggered, which includes generating a safety early warning signal and notifying relevant managers. The early warning signal is customized according to the severity of the risk level, enabling managers to quickly identify potential high-risk areas and take appropriate countermeasures to achieve timely intervention in water quality safety management.

[0037] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A safety protection early warning method based on neural network, characterized in that: The following steps are involved: Collect the pH value, dissolved oxygen concentration, turbidity, temperature and flow rate of the effluent water quality, perform data cleaning and outlier removal, perform time series sorting, and generate water quality time series data; Based on the water quality time series data, a neural network model is trained to predict pH value, dissolved oxygen concentration, turbidity, temperature and flow parameters, and the short-term trend value and prediction error of each parameter are calculated to obtain the parameter prediction trend value. According to the importance of the parameter, the parameter weight is dynamically adjusted, the parameter weight is combined with the parameter prediction trend value, and the parameter prediction value is updated to obtain a water quality prediction parameter set; Based on the water quality prediction parameter set, select parameters of the prediction error, analyze them, calculate the long-term trend value and fluctuation range of the parameters, and merge them with the water quality prediction parameter set, update all parameter prediction values, obtain comprehensive water quality prediction data, compare the comprehensive water quality prediction data with the safety threshold, calculate the risk index of each parameter, and generate a water quality risk assessment index; Based on the water quality risk assessment indicators, the risk index of each parameter is summarized, the overall risk level is calculated, and compared with the preset warning value to generate a safety warning signal.

2. The neural network-based safety protection early warning method according to claim 1 is characterized in that: The steps for obtaining the water quality time series data are as follows: Monitor the pH value, dissolved oxygen concentration, turbidity, temperature and flow rate of the effluent water quality, analyze the data range of each parameter, identify and eliminate abnormal data points, and generate preliminary processed water quality data; Based on the preliminarily processed water quality data, the data is sorted according to the collected timestamps to obtain water quality data sorted in time series; Based on the water quality data after the time series is sorted, the changing trends of pH value, dissolved oxygen concentration, turbidity, temperature and flow are analyzed, the time series characteristics are extracted, and the water quality time series data are generated.

3. The neural network-based safety protection early warning method according to claim 1 is characterized in that: The steps for obtaining the parameter prediction trend value are as follows: Using the water quality time series data, taking pH value, dissolved oxygen concentration, turbidity, temperature and flow rate as input variables, training a neural network model to obtain a trained neural network model; Based on the trained neural network model, the short-term trend value of each water quality parameter is predicted, and the prediction error is calculated. The calculation formula is: ； in, is the prediction error of the ith parameter, is the observed value, is the predicted value, is the length of the time series, is the weight of each time point, is the standard deviation at time point t; Based on the prediction error, a prediction trend value of each water quality parameter is generated to obtain a parameter prediction trend value.

4. The neural network-based safety protection early warning method according to claim 1 is characterized in that: The steps for obtaining the water quality prediction parameter set are: According to the past data of each water quality parameter and the current environmental changes, the importance is judged, and the weight of each parameter is adjusted to obtain an adjusted parameter weight set; Based on the adjusted parameter weight set, the updated parameter prediction value is calculated using the following formula: ； in, is the updated parameter prediction value, is the weight of the i-th parameter, is the predicted trend value of the i-th parameter, is the total number of parameters; Based on the updated parameter prediction values, the parameters are merged to generate a water quality prediction parameter set.

5. The neural network-based safety protection early warning method according to claim 1 is characterized in that: The steps for obtaining the comprehensive water quality prediction data are as follows: Based on the water quality prediction parameter set, identifying and selecting parameters of the prediction error to obtain a selected error parameter set; Based on the selected error parameter set, the long-term trend value of each parameter is calculated, and the calculation formula is: ; in, is the long-term trend value of the i-th parameter, is the number of observation periods, is the observed value of the i-th parameter in the m-th period, is the average value of the i-th parameter in all cycles; Based on the long-term trend values, they are merged and updated into a water quality prediction parameter set to form comprehensive water quality prediction data.

6. The neural network-based safety protection early warning method according to claim 1 is characterized in that: The steps for obtaining the water quality risk assessment indicators are as follows: Obtain the comprehensive water quality prediction data, compare it with the preset safety threshold, identify the parameters exceeding the threshold, and obtain a list of over-threshold parameters; Based on the above-threshold parameter list, the risk index of each parameter is calculated using the following formula: ; in, is the risk index of the i-th parameter, is the predicted value of the parameter, is the corresponding safety threshold, is the standard deviation of the corresponding parameter; Based on the risk index, the risk index is summarized to generate a water quality risk assessment indicator.

7. The neural network-based safety protection early warning method according to claim 1 is characterized in that: The steps of obtaining the safety warning signal are: Summarize the risk index of each parameter in the water quality risk assessment index to obtain an overall risk index summary; Based on the overall risk index summary, the overall risk level is calculated, the safety status of the entire water body is analyzed, and the overall risk level is formed.

8. The neural network-based safety protection early warning method according to claim 7 is characterized in that: The step of acquiring the safety warning signal also includes: comparing the overall risk level with a preset warning value, and generating a safety warning signal according to the comparison result.

Images