Water quality abnormal data monitoring method and system based on deep learning generation error

Abstract

The invention discloses a water quality abnormal data monitoring method and system based on a deep learning generation error. The water quality abnormal data monitoring method based on the deep learning generation error comprises the following steps: acquiring water quality monitoring data; determining a water quality monitoring data class according to the obtained water quality monitoring data; performing first-class verification on the water quality monitoring data class, and judging whether the first-class verification can be passed or not; if the first-class verification is not passed, performing second-class verification on the water quality monitoring data class, and judging whether the second-class verification can be passed or not; and if the second-class verification is passed, performing deep learning, determining a deep learning generation error, and outputting a water quality anomaly detection result according to the generation error. According to the water quality abnormity monitoring method, the water quality data are monitored to different degrees, then the water quality data which are not detected to be abnormal are further monitored in a deep learning mode, and the accuracy and reliability of water quality monitoring are improved to a great extent through the water quality abnormity monitoring method.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular, to a method and system for monitoring abnormal water quality data based on deep learning generated errors. Background Art

[0002] At present, the development of the environment has received more and more attention from people, and more and more people have begun to pay attention to the protection of water quality. In the protection of water quality, abnormal monitoring of water quality is often timely and accurate. Through abnormal monitoring of water quality, changes in water quality can be quickly discovered. For example, when a river is subject to sudden pollution events (such as illegal discharge of pollutants by factories, chemical leakage, etc.), the abnormal water quality monitoring system can detect abnormal fluctuations in water quality parameters (such as chemical oxygen demand, heavy metal content, etc.) in a short time through continuous monitoring data. This enables relevant departments to take countermeasures at the early stage of pollution spread, reduce the harm of pollution to the water ecosystem and the water safety of surrounding residents, and is also crucial for the protection of drinking water sources. Taking the city's drinking water reservoir as an example, real-time abnormal water quality monitoring can timely detect situations such as excessive microorganisms in the water and the invasion of harmful substances, thereby preventing polluted water from entering the water supply system and ensuring that residents can drink safe drinking water. However, the traditional way of water quality monitoring is often carried out through on-site sampling laboratory analysis or on-site rapid testing, but this method may have the problem of reduced monitoring accuracy due to manual sampling.

[0003] Therefore, how to provide a method that can improve the accuracy of water quality abnormality monitoring has become an urgent problem to be solved in this field. Summary of the invention

[0004] The present application proposes a method for monitoring abnormal water quality data based on deep learning generated errors, comprising the following steps: acquiring water quality monitoring data; determining a water quality monitoring data class based on the acquired water quality monitoring data; performing a first-class verification on the water quality monitoring data class to determine whether the first-class verification can pass; if the first-class verification fails, performing a second-class verification on the water quality monitoring data class to determine whether the second-class verification can pass; if the second-class verification passes, performing deep learning to determine a deep learning generated error, and outputting a water quality anomaly detection result based on the generated error.

[0005] As described above, in the water quality anomaly data monitoring method based on deep learning generated errors, if one type of verification passes, deep learning is performed to determine the deep learning generated errors, and the water quality anomaly detection results are output based on the generated errors.

[0006] The water quality anomaly data monitoring method based on deep learning generated errors as described above outputs the water quality anomaly detection result if the second type of verification fails.

[0007] As described above, the method for monitoring abnormal water quality data based on deep learning error generation, obtaining water quality monitoring data includes obtaining water quality monitoring data from multiple sensors; the water quality monitoring data includes water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity data; the multiple sensors include: temperature sensor, pH sensor, dissolved oxygen sensor, turbidity sensor and conductivity sensor.

[0008] As described above, the method for monitoring abnormal water quality data based on deep learning generated errors, determining the water quality monitoring data class according to the acquired water quality monitoring data includes taking the water quality monitoring data of a city acquired from different sensors as a water quality monitoring data class.

[0009] A water quality anomaly data monitoring system based on deep learning generated errors, specifically comprising: a water quality monitoring data acquisition unit, a water quality monitoring data class determination unit, a first-class verification unit, a second-class verification unit and a deep learning unit; the water quality monitoring data acquisition unit is used to acquire water quality monitoring data; the water quality monitoring data class determination unit is used to determine the water quality monitoring data class according to the acquired water quality monitoring data; the first-class verification unit is used to perform a first-class verification on the water quality monitoring data class to determine whether the first-class verification can pass; if the first-class verification fails, a second-class verification is performed on the water quality monitoring data class of the second-class verification unit to determine whether the second-class verification can pass; if the second-class verification passes, the deep learning unit performs deep learning, determines the deep learning generated errors, and outputs water quality anomaly detection results according to the generated errors.

[0010] As described above, in the water quality anomaly data monitoring system based on deep learning generated errors, if one type of verification passes, the deep learning unit performs deep learning, determines the deep learning generated errors, and outputs the water quality anomaly detection results based on the generated errors.

[0011] The water quality anomaly data monitoring system based on deep learning to generate errors as described above outputs a water quality anomaly detection result if the second-class verification of the second-class verification unit fails.

[0012] As described above, in the water quality anomaly data monitoring system based on deep learning error generation, the water quality monitoring data acquisition unit acquires water quality monitoring data including acquiring water quality monitoring data from multiple sensors; the water quality monitoring data includes water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity data; the multiple sensors include: temperature sensor, pH sensor, dissolved oxygen sensor, turbidity sensor and conductivity sensor.

[0013] As described above, in the water quality abnormality data monitoring system based on deep learning generated errors, the water quality monitoring data class determination unit determines the water quality monitoring data class based on the acquired water quality monitoring data, including taking the water quality monitoring data of a city acquired from different sensors as a water quality monitoring data class.

[0014] This application has the following beneficial effects:

[0015] This application proposes a method of performing type I and type II verification on water quality data. First, the water quality data is monitored to different degrees, and then the water quality data that has not been detected with abnormalities is further monitored through deep learning. The water quality anomaly monitoring method of this application greatly improves the accuracy and reliability of water quality monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a flow chart of a method for monitoring abnormal water quality data based on deep learning generated errors according to an embodiment of the present application;

[0018] Figure 2 It is a schematic diagram of the internal structure of a water quality abnormality data monitoring system based on deep learning generated errors provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] The following is a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0020] Embodiment 1

[0021] like Figure 1 As shown, this embodiment provides a method for monitoring abnormal water quality data based on deep learning generated errors, which specifically includes the following steps:

[0022] Step S110: Acquire water quality monitoring data.

[0023] Specifically, water quality monitoring data is obtained from each sensor. The water quality monitoring data includes water temperature, pH value, dissolved oxygen concentration, turbidity, conductivity and other data. Each sensor includes: temperature sensor, pH sensor, dissolved oxygen sensor, turbidity sensor and conductivity sensor.

[0024] Temperature sensors usually use technologies such as thermistors, thermocouples, or semiconductor temperature sensors. The temperature is determined by measuring the heat conduction or resistance change of the water body. Water temperature has an important impact on the survival and reproduction of aquatic organisms, chemical reaction rates, and the physical properties of the water body. For example, different species of fish have a specific range of adaptation to water temperature, and water temperatures that are too high or too low may cause fish death.

[0025] pH sensors measure the concentration of hydrogen ions in water using glass electrodes or other types of electrodes. The potential generated by the electrodes is logarithmically related to the hydrogen ion concentration and can be calibrated to determine the pH value. The pH value reflects the acidity or alkalinity of the water. The proper pH value is critical to the health of aquatic life and also affects the solubility and toxicity of chemicals in the water. For example, acidic water will lead to increased dissolution of metal ions, which can be toxic to aquatic life.

[0026] Dissolved oxygen sensors determine dissolved oxygen concentrations by electrochemical and optical methods. Electrochemical dissolved oxygen sensors determine dissolved oxygen concentrations by measuring the current between electrodes, while optical dissolved oxygen sensors use the principles of fluorescence or optical absorption for measurement. Dissolved oxygen is one of the important indicators of water quality. It is essential for the respiration and survival of aquatic organisms, and also affects the decomposition of organic matter and the circulation of nutrients in the water. For example, low dissolved oxygen levels may cause fish to suffocate and die, while high dissolved oxygen levels are conducive to the activity of aerobic microorganisms and promote the decomposition of organic matter.

[0027] Turbidity sensors determine turbidity by measuring the scattering or absorption of light by suspended particles in water. Common turbidity sensors include scattering turbidity sensors and transmission turbidity sensors. Turbidity reflects the content of suspended particles in water. Highly turbid water can affect the vision and breathing of aquatic organisms, and may also carry pathogens and pollutants. For example, in drinking water treatment, reducing turbidity is one of the important treatment steps.

[0028] Conductivity sensors measure conductivity based on the ability of water to conduct electricity. Conductivity is proportional to the concentration of dissolved ions in water. Conductivity can reflect the content of dissolved salts in water. It is important for assessing the hardness, salinity, and ion composition of water. For example, water with high conductivity may not be suitable for certain industrial uses or agricultural irrigation.

[0029] Step S120: Determine the water quality monitoring data class according to the acquired water quality monitoring data.

[0030] The data obtained by different sensors are different water quality monitoring data. Since the water quality monitoring data obtained may be water quality data from multiple provinces and cities, if you want to monitor the water quality data of a specific city in a province, you need to further classify the obtained water quality data and treat the water quality monitoring data of a city obtained from different sensors as a water quality monitoring data class.

[0031] The water quality data in a city is regarded as a category of water quality data. For example, the water quality monitoring data such as water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity of City A collected from different sensors is regarded as a water quality monitoring data category, and the water quality monitoring data such as water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity of City B collected from different sensors is regarded as a water quality monitoring data category. By analogy, the water quality monitoring data such as water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity collected from different cities can be regarded as multiple water quality monitoring data categories.

[0032] Step S130: Perform a type-one verification on the water quality monitoring data class to determine whether the type-one verification can pass.

[0033] Performing a one-class check on multiple water quality monitoring data classes includes the following sub-steps:

[0034] Step S1301: Perform a type-one check on each water quality monitoring data in the acquired water quality monitoring data class to determine a type-one fluctuation value corresponding to each water quality monitoring data.

[0035] Taking a water quality monitoring data class as an example, a type of verification is performed on each water quality monitoring data in the water quality monitoring data class, and a type of fluctuation value P1 of each water quality monitoring data is determined specifically according to the following formula.

[0036]

[0037] Where t represents the time when the current water quality monitoring data (such as water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity) is obtained, and x t represents the data value of the current water quality monitoring data obtained at time t, T represents the entire period of obtaining the current water quality monitoring data, K represents the entire period of obtaining the current water quality monitoring data under a certain record in the history, k represents the time when the current water quality monitoring data (such as water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity) is obtained under a certain record in the history, x k Represents the data value of the current water quality monitoring data obtained at time k.

[0038] Step S1302: Determine whether all the fluctuation values ​​corresponding to each water quality monitoring data can pass the verification.

[0039] If the type of fluctuation value P1 corresponding to each water quality monitoring data does not exceed the specified threshold, it is considered that the current water quality monitoring data in the water quality monitoring data monitored within the current time T is normal, the verification passes, and step S150 is executed.

[0040] If there is a type I fluctuation value P1 that exceeds the specified threshold, it is considered that the current water quality monitoring data in the water quality monitoring data class acquired within time T may be abnormal, and step S140 is executed.

[0041] Through the above method, multiple Class I fluctuation values ​​can be obtained, and each water quality monitoring data corresponds to a Class I fluctuation value. For example, if the current water quality monitoring data is water temperature data, and its Class I fluctuation value P1 has exceeded the specified threshold, then the water temperature data obtained during the entire period may be abnormal. At this time, there is no need to confirm the Class I fluctuation values ​​of other water quality monitoring data, and directly perform Class II verification on them. Through the Class II verification, it can be directly determined whether the water quality is abnormal.

[0042] Step S140: Perform a second-class verification on the water quality monitoring data class to determine whether the second-class verification can pass.

[0043] The current water quality monitoring data that may be abnormal (such as water temperature, pH value, dissolved oxygen concentration, turbidity and one of the conductivity) is subjected to a second-class verification.

[0044] The second type fluctuation value P2 of the current water quality monitoring data (such as water temperature, pH value, dissolved oxygen concentration, turbidity and one of conductivity) is determined according to the following formula.

[0045] P2=x t -x t-1 / x t-1

[0046] x t represents the data value of the current water quality monitoring data obtained at time t, x t-1 Indicates the data value of the current water quality monitoring data obtained at time t-1,

[0047] According to the determined second-class fluctuation value P2, if the second-class fluctuation value P2 is greater than the specified threshold, it may indicate that a drastic biological or chemical change has occurred in the water body, and the second-class verification fails, and the monitoring result of abnormal water quality is directly output. If the second-class fluctuation value P2 is less than the specified threshold, it is considered that the current water quality is still in a normal state, the second-class verification passes, and step S150 is executed.

[0048] Through the second-class verification, the abnormal second-class fluctuation value P2 of water temperature data, for example, can be confirmed. If the fluctuation value P2 is greater than the specified threshold, the monitoring result of water quality abnormality is directly output.

[0049] The above-mentioned monitoring, which confirms whether there is any abnormality in water quality through type I calibration and type II calibration respectively, can first conduct a simple monitoring of the water quality, providing further guarantee for subsequent water quality monitoring.

[0050] Step S150: perform deep learning, determine the deep learning generated error, and output the water quality anomaly detection result according to the generated error.

[0051] After the first and second type verifications do not detect any abnormality, deep learning is performed based on the acquired water quality monitoring data, wherein step S150 specifically includes the following sub-steps:

[0052] Step S1501: Build a deep learning model.

[0053] The deep learning model constructed in this embodiment is a recurrent neural network (RNN), wherein RNN has a cyclic structure, can process time series data, and can capture the time dependency in the data.

[0054] Specifically, RNN can monitor the time series data in water quality monitoring, such as the changes in water temperature, dissolved oxygen, pH value, etc. over time. It can learn the changing trends and periodicity of water quality parameters, predict future water quality conditions, and detect abnormalities in a timely manner. For continuously monitored water quality data, RNN and its variants can be used for real-time anomaly detection to improve the timeliness and accuracy of monitoring.

[0055] Step S1502: Input multiple water quality data monitoring data classes into the deep learning model respectively to determine the predicted values ​​of multiple water quality monitoring data in each water quality monitoring data class.

[0056] Before determining the predicted value of the water quality monitoring data according to the deep learning model, it also includes training the deep learning model. After the training, the water quality monitoring data in each water quality monitoring data class is input into the deep learning model.

[0057] The deep learning model establishes a mapping relationship between the input water quality monitoring data and the output predicted water quality status by learning a large amount of normal water quality data. When new water quality monitoring data is input into the trained model, the deep learning model will generate corresponding prediction values ​​based on the input water quality monitoring data.

[0058] Step S1503: Determine the deep learning error based on the predicted value of the water quality monitoring data.

[0059] The difference between the actual water quality monitoring data value and the corresponding predicted value output by the model based on the water quality monitoring data is the error generated by deep learning. Suppose a deep learning model has been trained on a large amount of normal water quality parameter data such as water temperature, pH value, dissolved oxygen, etc. When a new set of water quality parameter data is input into the model, the model will predict the corresponding water quality state or parameter value. If there is a difference between the actual water quality parameter and the model prediction value, this difference is the error.

[0060] The errors include mean square error, mean absolute error, etc.

[0061] The mean square error gives higher weight to larger errors, so it is more sensitive to outliers. Therefore, this embodiment uses the mean square error as the deep learning error, where the deep learning error MSE is expressed as Where n is the number of data points of the current water quality monitoring data, Y i Indicates the actual value of the current water quality monitoring data. Represents the predicted value of the current water quality monitoring data.

[0062] Step S1504: Output the abnormal water quality monitoring result according to the deep learning error.

[0063] By analyzing the size, distribution and trend of the error, it is possible to determine whether the water quality data is abnormal. If the deep learning error exceeds the preset threshold range, it means that the water quality is abnormal, and the abnormal water quality monitoring result is output.

[0064] Among them, the present application first monitors the water quality data to different degrees through the above-mentioned first and second type verification methods, and then further monitors the water quality data without detected abnormalities through deep learning. The water quality anomaly monitoring method of the present application greatly improves the accuracy and reliability of water quality monitoring.

[0065] Embodiment 2

[0066] like Figure 2 As shown, this embodiment provides a water quality abnormality data monitoring system based on deep learning error generation, which specifically includes: a water quality monitoring data acquisition unit 210, a water quality monitoring data class determination unit 220, a first type verification unit 230, a second type verification unit 240, and a deep learning unit 250.

[0067] The water quality monitoring data acquisition unit 210 is used to acquire water quality monitoring data.

[0068] Specifically, water quality monitoring data is obtained from each sensor. The water quality monitoring data includes water temperature, pH value, dissolved oxygen concentration, turbidity, conductivity and other data. Each sensor includes: temperature sensor, pH sensor, dissolved oxygen sensor, turbidity sensor and conductivity sensor.

[0069] Temperature sensors usually use technologies such as thermistors, thermocouples, or semiconductor temperature sensors. The temperature is determined by measuring the heat conduction or resistance change of the water body. Water temperature has an important impact on the survival and reproduction of aquatic organisms, chemical reaction rates, and the physical properties of the water body. For example, different species of fish have a specific range of adaptation to water temperature, and water temperatures that are too high or too low may cause fish death.

[0070] pH sensors measure the concentration of hydrogen ions in water using glass electrodes or other types of electrodes. The potential generated by the electrodes is logarithmically related to the hydrogen ion concentration and can be calibrated to determine the pH value. The pH value reflects the acidity or alkalinity of the water. The proper pH value is critical to the health of aquatic life and also affects the solubility and toxicity of chemicals in the water. For example, acidic water will lead to increased dissolution of metal ions, which can be toxic to aquatic life.

[0071] Dissolved oxygen sensors determine dissolved oxygen concentrations by electrochemical and optical methods. Electrochemical dissolved oxygen sensors determine dissolved oxygen concentrations by measuring the current between electrodes, while optical dissolved oxygen sensors use the principles of fluorescence or optical absorption for measurement. Dissolved oxygen is one of the important indicators of water quality. It is essential for the respiration and survival of aquatic organisms, and also affects the decomposition of organic matter and the circulation of nutrients in the water. For example, low dissolved oxygen levels may cause fish to suffocate and die, while high dissolved oxygen levels are conducive to the activity of aerobic microorganisms and promote the decomposition of organic matter.

[0072] Turbidity sensors determine turbidity by measuring the scattering or absorption of light by suspended particles in water. Common turbidity sensors include scattering turbidity sensors and transmission turbidity sensors. Turbidity reflects the content of suspended particles in water. Highly turbid water can affect the vision and breathing of aquatic organisms, and may also carry pathogens and pollutants. For example, in drinking water treatment, reducing turbidity is one of the important treatment steps.

[0073] Conductivity sensors measure conductivity based on the ability of water to conduct electricity. Conductivity is proportional to the concentration of dissolved ions in water. Conductivity can reflect the content of dissolved salts in water. It is important for assessing the hardness, salinity, and ion composition of water. For example, water with high conductivity may not be suitable for certain industrial uses or agricultural irrigation.

[0074] The water quality monitoring data class determination unit 220 is used to determine the water quality monitoring data class according to the acquired water quality monitoring data.

[0075] The data obtained by different sensors are different water quality monitoring data. Since the water quality monitoring data obtained may be water quality data from multiple provinces and cities, if you want to monitor the water quality data of a specific city in a province, you need to further classify the obtained water quality data and treat the water quality monitoring data of a city obtained from different sensors as a water quality monitoring data class.

[0076] The water quality data in a city is regarded as a category of water quality data. For example, the water quality monitoring data such as water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity of City A collected from different sensors is regarded as a water quality monitoring data category, and the water quality monitoring data such as water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity of City B collected from different sensors is regarded as a water quality monitoring data category. By analogy, the water quality monitoring data such as water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity collected from different cities can be regarded as multiple water quality monitoring data categories.

[0077] The first type verification unit 230 is used to perform a first type verification on the water quality monitoring data class to determine whether the first type verification can be passed.

[0078] A type of verification unit 230 includes the following submodules: a fluctuation value determination module and a judgment module.

[0079] The fluctuation value determination module performs a type-one check on each water quality monitoring data in the acquired water quality monitoring data class to determine the type-one fluctuation value corresponding to each water quality monitoring data. Taking a water quality monitoring data class as an example, a type-one check is performed on each water quality monitoring data in the water quality monitoring data class, and the type-one fluctuation value P1 of each water quality monitoring data is determined specifically according to the following formula.

[0080]

[0081] Where t represents the time when the current water quality monitoring data (such as water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity) is obtained, and x t represents the data value of the current water quality monitoring data obtained at time t, T represents the entire period of obtaining the current water quality monitoring data, K represents the entire period of obtaining the current water quality monitoring data under a certain record in the history, k represents the time when the current water quality monitoring data (such as water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity) is obtained under a certain record in the history, x k Represents the data value of the current water quality monitoring data obtained at time k.

[0082] The judgment module judges whether a type of fluctuation value corresponding to each water quality monitoring data can all pass the verification.

[0083] If the first-class fluctuation value P1 corresponding to each water quality monitoring data does not exceed the specified threshold, it is considered that the current water quality monitoring data in the water quality monitoring data monitored within the current T time is normal, the verification is passed, and the deep learning unit 250 is executed. If there is a first-class fluctuation value P1 that exceeds the specified threshold, it is considered that the current water quality monitoring data in the water quality monitoring data class acquired within the T time may be abnormal, and the second-class verification unit 240 is executed.

[0084] Through the above method, multiple Class I fluctuation values ​​can be obtained, and each water quality monitoring data corresponds to a Class I fluctuation value. For example, if the current water quality monitoring data is water temperature data, and its Class I fluctuation value P1 has exceeded the specified threshold, then the water temperature data obtained during the entire period may be abnormal. At this time, there is no need to confirm the Class I fluctuation values ​​of other water quality monitoring data, and directly perform Class II verification on them. Through the Class II verification, it can be directly determined whether the water quality is abnormal.

[0085] The second type verification unit 240 is used to perform a second type verification on the water quality monitoring data class to determine whether the second type verification can be passed.

[0086] Perform a second-class check on the current water quality monitoring data (such as water temperature, pH value, dissolved oxygen concentration, turbidity, and conductivity) that may be abnormal. Determine the second-class fluctuation value P2 of the current water quality monitoring data (such as water temperature, pH value, dissolved oxygen concentration, turbidity, and conductivity) according to the following formula.

[0087] P2=x t -x t-1 / x t-1

[0088] x t represents the data value of the current water quality monitoring data obtained at time t, x t-1 Indicates the data value of the current water quality monitoring data obtained at time t-1. According to the determined second-class fluctuation value P2, if the second-class fluctuation value P2 is greater than the specified threshold, it may indicate that a drastic biological or chemical change has occurred in the water body, then the second-class verification fails, and the monitoring result of abnormal water quality is directly output. If the second-class fluctuation value P2 is less than the specified threshold, it is considered that the current water quality is still in a normal state, then the second-class verification passes, and the deep learning unit 250 is executed.

[0089] Through the second-class verification, the abnormal second-class fluctuation value P2 of water temperature data, for example, can be confirmed. If the fluctuation value P2 is greater than the specified threshold, the monitoring result of water quality abnormality is directly output.

[0090] The above-mentioned monitoring, which confirms whether there is any abnormality in water quality through type I calibration and type II calibration respectively, can first conduct a simple monitoring of the water quality, providing further guarantee for subsequent water quality monitoring.

[0091] The deep learning unit 250 is used to perform deep learning, determine the deep learning generated error, and output the water quality anomaly detection result according to the generated error.

[0092] When the first and second type of verifications do not detect any abnormalities, deep learning is performed based on the acquired water quality monitoring data. The deep learning unit 250 includes the following submodules: a construction module, a prediction value determination module, a learning error determination module, and a monitoring result output module.

[0093] Building blocks for constructing deep learning models.

[0094] The deep learning model constructed in this embodiment is a recurrent neural network (RNN), in which RNN has a cyclic structure, can process time series data, and can capture the time dependency in the data. Specifically, RNN can monitor the time series data in water quality monitoring, such as changes in water temperature, dissolved oxygen, pH value, etc. over time. The changing trend and periodicity of water quality parameters can be learned, the future water quality status can be predicted, and abnormal conditions can be discovered in time. For continuously monitored water quality data, RNN and its variants can be used for real-time anomaly detection to improve the timeliness and accuracy of monitoring.

[0095] The prediction value determination module is used to input multiple water quality data monitoring data classes into the deep learning model respectively to determine the prediction values ​​of multiple water quality monitoring data in each water quality monitoring data class. Before determining the prediction value of the water quality monitoring data according to the deep learning model, it also includes training the deep learning model. After the training, the water quality monitoring data in each water quality monitoring data class is input into the deep learning model.

[0096] The deep learning model establishes a mapping relationship between the input water quality monitoring data and the output predicted water quality status by learning a large amount of normal water quality data. When new water quality monitoring data is input into the trained model, the deep learning model will generate corresponding prediction values ​​based on the input water quality monitoring data.

[0097] The learning error determination module is used to determine the deep learning error based on the predicted value of the water quality monitoring data.

[0098] The difference between the actual water quality monitoring data value and the corresponding predicted value output by the model based on the water quality monitoring data is the error generated by deep learning. Assume that a deep learning model has been trained on a large amount of normal water quality parameter data such as water temperature, pH value, dissolved oxygen, etc. When a new set of water quality parameter data is input into the model, the model will predict the corresponding water quality state or parameter value. If there is a difference between the actual water quality parameters and the model predicted values, this difference is the error. The error includes mean square error, mean absolute error, etc. The mean square error gives higher weights to larger errors, so it is more sensitive to outliers. Therefore, this embodiment uses the mean square error as the deep learning error, and the deep learning error MSE is expressed as Where n is the number of data points of the current water quality monitoring data, Y i Indicates the actual value of the current water quality monitoring data. Represents the predicted value of the current water quality monitoring data.

[0099] The monitoring result output module is used to output abnormal water quality monitoring results based on deep learning errors.

[0100] By analyzing the size, distribution and trend of the error, it is possible to determine whether the water quality data is abnormal. If the deep learning error exceeds the preset threshold range, it means that the water quality is abnormal, and the abnormal water quality monitoring result is output.

[0101] Among them, the present application first monitors the water quality data to different degrees through the above-mentioned first and second type verification methods, and then further monitors the water quality data without detected abnormalities through deep learning. The water quality anomaly monitoring method of the present application greatly improves the accuracy and reliability of water quality monitoring.

[0102] The present application also provides a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute the water quality abnormality data monitoring method based on deep learning to generate errors.

[0103] The embodiments disclosed in the present invention provide a computer-readable storage medium, in which computer program instructions are stored. When the computer program instructions are executed on a computer, the computer executes the above-mentioned method for monitoring abnormal water quality data based on deep learning to generate errors.

[0104] An embodiment of the present invention provides a processor for processing the above-mentioned method for monitoring abnormal water quality data based on deep learning generated errors.

[0105] In the embodiment of the present invention, the processor may be an integrated circuit chip having the ability to process signals. The processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0106] The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor for execution, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The processor reads the information in the storage medium and completes the steps of the above method in combination with its hardware.

[0107] The storage medium may be a memory, which may be, for example, a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memory.

[0108] Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct RAM bus random access memory (DRRAM).

[0109] This application has the following beneficial effects:

[0110] This application proposes a method of performing type I and type II verification on water quality data. First, the water quality data is monitored to different degrees, and then the water quality data that has not been detected with abnormalities is further monitored through deep learning. The water quality anomaly monitoring method of this application greatly improves the accuracy and reliability of water quality monitoring.

[0111] Although the present application is described with reference to examples, this is for illustrative purposes only and is not intended to limit the present application, and changes, additions and / or deletions to the embodiments may be made without departing from the scope of the present application.

[0112] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for monitoring abnormal water quality data based on deep learning generation errors, characterized in that: The following steps are involved: Obtain water quality monitoring data; Determine the water quality monitoring data category based on the acquired water quality monitoring data; Perform a type of verification on the water quality monitoring data class to determine whether the type of verification can pass; If the first-class verification fails, the water quality monitoring data class is subjected to the second-class verification to determine whether the second-class verification can pass; If the second-class verification passes, deep learning is performed to determine the error generated by deep learning, and the water quality anomaly detection result is output based on the generated error.

2. The method for monitoring abnormal water quality data based on deep learning generation error according to claim 1, characterized in that: If one type of verification passes, deep learning is performed to determine the error generated by deep learning, and the water quality anomaly detection result is output based on the generated error.

3. The method for monitoring abnormal water quality data based on deep learning generation error according to claim 1, characterized in that: If the second type of verification fails, the abnormal water quality detection result will be output.

4. The method for monitoring abnormal water quality data based on deep learning generation error according to claim 1, characterized in that: Acquiring water quality monitoring data includes acquiring water quality monitoring data from a plurality of sensors; Water quality monitoring data include water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity data; The multiple sensors include: temperature sensor, pH sensor, dissolved oxygen sensor, turbidity sensor and conductivity sensor.

5. The method for monitoring abnormal water quality data based on deep learning generation error according to claim 1, characterized in that: Determining the water quality monitoring data class according to the acquired water quality monitoring data includes taking the water quality monitoring data of a city acquired from different sensors as a water quality monitoring data class.

6. A water quality abnormal data monitoring system based on deep learning generated errors, characterized in that: Specifically include: A water quality monitoring data acquisition unit, a water quality monitoring data class determination unit, a first-class verification unit, a second-class verification unit, and a deep learning unit; A water quality monitoring data acquisition unit, used for acquiring water quality monitoring data; A water quality monitoring data class determination unit, used to determine the water quality monitoring data class according to the acquired water quality monitoring data; A type-one verification unit is used to perform a type-one verification on the water quality monitoring data class to determine whether the type-one verification can pass; If the first-class verification fails, the second-class verification is performed on the water quality monitoring data of the second-class verification unit to determine whether the second-class verification can pass; If the second type of verification passes, the deep learning unit performs deep learning, determines the deep learning generated error, and outputs the water quality anomaly detection result based on the generated error.

7. The water quality abnormal data monitoring system based on deep learning generated errors as claimed in claim 6, characterized in that: If one type of verification passes, the deep learning unit performs deep learning, determines the deep learning generated error, and outputs the water quality anomaly detection result based on the generated error.

8. The water quality abnormal data monitoring system based on deep learning generated errors as claimed in claim 6, characterized in that: If the second-class verification of the second-class verification unit fails, the abnormal water quality detection result is output.

9. The water quality abnormal data monitoring system based on deep learning generated errors as claimed in claim 6, characterized in that: The water quality monitoring data acquisition unit acquires the water quality monitoring data, including acquiring the water quality monitoring data from a plurality of sensors; Water quality monitoring data include water temperature, pH value, dissolved oxygen concentration, turbidity and conductivity data; The multiple sensors include: temperature sensor, pH sensor, dissolved oxygen sensor, turbidity sensor and conductivity sensor.

10. The water quality abnormal data monitoring system based on deep learning generated errors according to claim 6, characterized in that: The water quality monitoring data class determination unit determines the water quality monitoring data class according to the acquired water quality monitoring data, including taking the water quality monitoring data of a city acquired from different sensors as a water quality monitoring data class.

Images