Water quality monitoring method and device based on multi-source data

Abstract

This invention discloses a water quality monitoring method and device based on multi-source data, relating to the field of water quality monitoring. The method includes: acquiring dissolved oxygen concentration and sensor internal temperature sequences; applying equally spaced short-duration pulse excitation; extracting temperature and dissolved oxygen response segments caused by the pulses; determining the temperature rise time constant and dissolved oxygen fall time constant; constructing response characteristic ratios and arranging them into a sequence; calculating the moving median and moving interquartile range of the sequence; when three consecutive response characteristic ratios are all greater than the moving median plus twice the moving interquartile range, a membrane aging or contamination fault is determined, and the fault initiation time is recorded. This invention utilizes the thermo-mass coupling response characteristics under pulse excitation and achieves real-time online diagnosis of membrane status through dynamic statistical thresholds, effectively eliminating hydrodynamic and turbidity interferences, and improving the sensitivity and reliability of fault detection.

Description

Technical Field

[0001] This invention belongs to the field of water quality monitoring technology, and in particular relates to a water quality monitoring method and device based on multi-source data. Background Technology

[0002] Dissolved oxygen sensors are core equipment in water quality monitoring, widely used for measuring dissolved oxygen content in surface water, groundwater, and industrial wastewater. Dissolved oxygen sensors typically employ a membrane-based electrochemical principle, using a breathable membrane to isolate dissolved oxygen in the water from the electrodes and electrolyte inside the sensor. During long-term online monitoring, contaminants such as microorganisms, sediment, and grease gradually adhere to the membrane surface, leading to a decrease in oxygen permeability and causing problems such as lower measured values ​​and slower response times.

[0003] Currently, determining whether dissolved oxygen sensor diaphragms are aging or contaminated relies primarily on regular manual inspections and offline cleaning and calibration. This method has a significant time lag, failing to detect the deterioration process of the diaphragm in a timely manner. Consequently, even after a malfunction occurs, the monitoring data may have become distorted without being noticed. Some existing technologies attempt to determine diaphragm condition by monitoring the steady-state drift of the dissolved oxygen signal; however, these methods are easily affected by environmental factors such as water temperature and flow rate, resulting in a high false alarm rate.

[0004] Therefore, how to automatically and accurately identify diaphragm aging or contamination faults by utilizing multi-source data collected by the sensors themselves during online water quality monitoring has become an urgent technical problem to be solved. Summary of the Invention

[0005] This invention provides a water quality monitoring method and device based on multi-source data to solve the problems of delayed diaphragm fault identification and susceptibility to environmental interference in the prior art.

[0006] In a first aspect, the present invention provides a water quality monitoring method based on multi-source data, comprising: Obtain the dissolved oxygen concentration measurement sequence of the dissolved oxygen sensor at the same monitoring section of the target water area within a preset time period, and the internal temperature sequence of the temperature sensor integrated inside the dissolved oxygen sensor within the same time period; During the preset time period, multiple equally spaced short-time pulse excitation signals are applied to the heating element of the dissolved oxygen sensor. The duration of each pulse excitation signal is a first preset duration, and the pulse interval is a second preset duration. In response to a certain application time of a short-time pulse excitation signal, a certain temperature response segment is extracted from the internal temperature sequence of the sensor, starting from the certain application time and extending backward for a third preset time period, and a certain dissolved oxygen response segment is extracted from the dissolved oxygen concentration measurement sequence within the same time interval. Determine a certain temperature rise time constant for a certain temperature response segment, and a certain dissolved oxygen fall time constant for a certain dissolved oxygen response segment; A response characteristic ratio of a short-time pulse excitation signal is determined based on a certain temperature rise time constant and a certain dissolved oxygen fall time constant, and the response characteristic ratios of multiple consecutive short-time pulse excitation signals are arranged in chronological order to obtain a response characteristic ratio sequence. Calculate the sliding median and sliding interquartile range of the response characteristic ratio sequence. When the response characteristic ratios of three consecutive short-time pulse excitation signals are all greater than the sliding median plus twice the sliding interquartile range, it is determined that the dissolved oxygen sensor has a membrane aging or contamination fault within the preset time period, and the fault start time is recorded as the application time corresponding to the first excessive pulse.

[0007] Secondly, the present invention provides a water quality monitoring device based on multi-source data, comprising: The acquisition module is configured to acquire the dissolved oxygen concentration measurement sequence of the dissolved oxygen sensor at the same monitoring section of the target water area within a preset time period, and the internal temperature sequence of the temperature sensor integrated inside the dissolved oxygen sensor within the same time period. The application module is configured to apply multiple equally spaced short-time pulse excitation signals to the heating element of the dissolved oxygen sensor within the preset time period, wherein the duration of each pulse excitation signal is a first preset duration and the pulse interval is a second preset duration. The interception module is configured to, in response to a certain application time of a certain short-time pulse excitation signal, intercept a certain temperature response segment from the internal temperature sequence of the sensor, starting from the certain application time and extending backward for a third preset time, and intercept a certain dissolved oxygen response segment from the dissolved oxygen concentration measurement sequence within the same time interval. The first determining module is configured to determine a certain temperature rise time constant of a certain temperature response segment and a certain dissolved oxygen fall time constant of a certain dissolved oxygen response segment; The second determining module is configured to determine a certain response characteristic ratio of a certain short-time pulse excitation signal based on a certain temperature rise time constant and a certain dissolved oxygen fall time constant, and to arrange the response characteristic ratios of multiple consecutive short-time pulse excitation signals in chronological order to obtain a response characteristic ratio sequence. The determination module is configured to calculate the sliding median and sliding interquartile range of the response characteristic ratio sequence. When the response characteristic ratios of three consecutive short-time pulse excitation signals are all greater than the sliding median plus twice the sliding interquartile range, it is determined that the dissolved oxygen sensor has a membrane aging or contamination fault within the preset time period, and the fault start time is recorded as the application time corresponding to the first excessive pulse.

[0008] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the water quality monitoring method based on multi-source data according to any embodiment of the present invention.

[0009] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the steps of the water quality monitoring method based on multi-source data according to any embodiment of the present invention.

[0010] The water quality monitoring method and device based on multi-source data of this application have the following beneficial effects: By applying a short-time pulse excitation signal, the temperature response and dissolved oxygen response are acquired simultaneously. The response characteristic ratio is constructed using the temperature rise time constant and the dissolved oxygen fall time constant, which enables active, real-time detection of membrane aging or contamination faults without manual intervention, significantly improving the timeliness of fault detection. When extracting temperature response and dissolved oxygen response segments, the natural decay and recovery process during the natural cooling relaxation period was analyzed to identify multi-order thermal relaxation modes and mass transfer recovery modes. A dynamic baseline was constructed using a multi-input multi-output state space model, which effectively removed the influence of ambient temperature drift and background oxygen consumption in the water body on the impulse response and improved the accuracy of response segment extraction. Using principal axis regression analysis, the temperature rise time constant and dissolved oxygen fall time constant corresponding to multiple consecutive pulse excitation signals are regressed on a two-dimensional plane. The slope of the regression line is used as the response characteristic ratio. Under clean membrane conditions, the response characteristics of temperature and dissolved oxygen show a stable proportional relationship. When the membrane ages or becomes contaminated, the heat conduction path and oxygen diffusion path are affected to varying degrees, causing the proportional relationship to shift and the slope to change accordingly. This method captures minute changes in membrane state from a physical mechanism perspective, exhibiting high detection sensitivity and robustness. By simultaneously collecting water flow velocity and turbidity data, abnormal response characteristics caused by hydrodynamic disturbances or particulate impacts are eliminated, avoiding misjudgments caused by instantaneous flow field changes and further improving the reliability of fault diagnosis. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating a water quality monitoring method based on multi-source data, provided in an embodiment of the present invention; Figure 2 This is a structural block diagram of a water quality monitoring device based on multi-source data, provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Please see Figure 1 The diagram shows a flowchart of a water quality monitoring method based on multi-source data according to this application.

[0015] like Figure 1 As shown, the water quality monitoring method based on multi-source data specifically includes the following steps: Step S101: Obtain the dissolved oxygen concentration measurement sequence of the dissolved oxygen sensor at the same monitoring section of the target water area within a preset time period, and the internal temperature sequence of the temperature sensor integrated inside the dissolved oxygen sensor within the same time period.

[0016] In this step, the dissolved oxygen sensor is first fixedly installed at the monitoring section of the target water area, ensuring that the membrane surface of the dissolved oxygen sensor is completely submerged in the water and maintains a preset angle with the water flow direction to guarantee sufficient contact between the water and the membrane surface. The dissolved oxygen sensor integrates a temperature sensor located between the inner side of the membrane and the electrolyte chamber. The installation distance between this temperature sensor and the inner surface of the membrane does not exceed a preset threshold distance, allowing for sensitive detection of temperature changes transmitted from the heating element through the various physical layers to the inner side of the membrane.

[0017] After the dissolved oxygen sensor is activated, it continuously collects dissolved oxygen concentration values ​​in the water at a preset sampling frequency. The temperature sensor synchronously collects internal temperature values ​​at the same sampling frequency. The preset time period is a continuous time interval during online monitoring. The start and end times of the preset time period are pre-set by the monitoring task, and its length is sufficient to cover multiple complete pulse excitation cycles. The preset sampling frequency must satisfy the Nyquist sampling theorem to ensure complete capture of the transient waveform characteristics of temperature rise and dissolved oxygen decrease caused by pulse excitation. Typically, the preset sampling frequency is set to at least 10 times the reciprocal of the first preset time period.

[0018] The dissolved oxygen concentration values ​​collected chronologically within a preset time period constitute a dissolved oxygen concentration measurement sequence, while the temperature values ​​collected at the same timestamp within the same preset time period constitute a sensor internal temperature sequence. The dissolved oxygen concentration measurement sequence and the sensor internal temperature sequence are strictly aligned in time, and each sampling point in the sequence is labeled with a corresponding acquisition time.

[0019] After completing the above data acquisition, the dissolved oxygen concentration measurement sequence and the sensor internal temperature sequence are preprocessed. The preprocessing includes removing isolated missing values ​​caused by instantaneous sensor disconnection or data transmission packet loss, and limiting abnormal peaks in the sequence that exceed the preset physical reasonable range due to electromagnetic interference. Finally, a continuous and complete dissolved oxygen concentration measurement sequence and sensor internal temperature sequence are formed for subsequent steps.

[0020] Step S102: During the preset time period, multiple equally spaced short-time pulse excitation signals are applied to the heating element of the dissolved oxygen sensor. The duration of each pulse excitation signal is a first preset duration, and the pulse interval is a second preset duration.

[0021] In this step, a miniature heating element is integrated within the dissolved oxygen sensor. This miniature heating element is embedded in the sensor's internal chamber between the membrane body layer and the electrolyte cavity. Thermal coupling between the miniature heating element and the inner surface of the membrane body layer is achieved through a thermally conductive adhesive layer. When not in operation, the miniature heating element does not heat the electrolyte or the membrane, thus not affecting the normal electrochemical measurement process of dissolved oxygen.

[0022] Within a preset time period, a series of equally spaced short-duration pulse excitation signals are applied to the micro heating element via the microcontroller built into the dissolved oxygen sensor or an external data acquisition and control unit. The short-duration pulse excitation signal is a rectangular voltage pulse or a rectangular current pulse. The duration of each pulse excitation signal is defined as a first preset duration, and the time interval between the rising edges of two adjacent pulse excitation signals is defined as a second preset duration. The first and second preset durations are pre-set and stored in the parameter register of the microcontroller before monitoring begins, based on the membrane thermal response characteristics and oxygen diffusion response characteristics of the dissolved oxygen sensor.

[0023] The first preset duration must meet the following conditions: within the duration of a single pulse excitation signal, the heat generated by the micro-heating element is sufficient to raise the temperature of the sensor's internal chamber by a preset amplitude relative to the baseline temperature before the pulse application, while preventing the overall temperature of the diaphragm body and electrolyte from overshooting and exceeding the sensor's normal operating temperature range. Typically, the first preset duration is set to a fixed value between 0.5 seconds and 2 seconds.

[0024] The second preset duration must meet the following conditions: within the interval between two adjacent pulse excitation signals, the temperature of the sensor's internal chamber and the dissolved oxygen concentration distribution in each physical layer of the diaphragm can fully recover from the disturbance caused by the previous pulse to a near-natural equilibrium state, ensuring that the initial conditions when the next pulse is applied are essentially the same as those when the previous pulse is applied. Simultaneously, the value of the second preset duration must also ensure that a sufficient number of pulse excitation signals can be generated within the preset time period to meet the sample size requirements for subsequent time constant analysis windows and sliding statistical analysis. Typically, the second preset duration is set to a fixed value between 30 seconds and 5 minutes.

[0025] While applying each short-duration pulse excitation signal to the micro heating element, the microcontroller adds a precise pulse application time tag to the data record. The pulse application time tag is synchronized with the acquisition time tags of the dissolved oxygen concentration measurement sequence and the internal temperature sequence of the sensor, so that subsequent steps can accurately correlate each pulse excitation signal with the corresponding temperature response segment and dissolved oxygen response segment in time.

[0026] Step S103: In response to a certain application time of a certain short-time pulse excitation signal, a certain temperature response segment is extracted from the internal temperature sequence of the sensor, starting from the certain application time and extending backward for a third preset time period, and a certain dissolved oxygen response segment of the same time interval is extracted from the dissolved oxygen concentration measurement sequence.

[0027] In this step, the time interval between all adjacent short-time pulse excitation signals within the preset time period is defined as the natural cooling relaxation period. During the natural cooling relaxation period, the heating element of the dissolved oxygen sensor is in a non-working state, and the dissolved oxygen sensor exchanges heat and dissolved oxygen with the water body only through natural convection and diffusion. Extract the temperature natural decay subsequence and dissolved oxygen natural recovery subsequence for each natural cooling relaxation period from the internal temperature sequence of the sensor and the dissolved oxygen concentration measurement sequence, respectively; Each temperature decay subsequence is decomposed into a linear superposition of multiple independent first-order thermal relaxation processes, where each first-order thermal relaxation process corresponds to a physical layered structure with independent thermal capacity and thermal resistance within the dissolved oxygen sensor. A global optimization algorithm based on singular value decomposition and trust region reflection is used to jointly solve the time constants and amplitude coefficients of multiple first-order thermal relaxation processes. The time constants and amplitude coefficients identified in all natural cooling relaxation periods are arranged in ascending order of time constant to form a thermal relaxation time constant spectrum. Each dissolved oxygen natural recovery subsequence is decomposed into a linear superposition of multiple independent first-order mass transfer recovery processes, wherein each first-order mass transfer recovery process corresponds to a physical layered structure with independent oxygen storage capacity and diffusion resistance within the dissolved oxygen sensor. The global optimization algorithm based on singular value decomposition and trust region reflection is used to jointly solve the time constants and amplitude coefficients of multiple first-order mass transfer recovery processes. The time constants and amplitude coefficients identified during all natural cooling relaxation periods are arranged in ascending order of time constant to form a mass transfer recovery time constant spectrum. Cluster analysis is performed on each time constant in the thermal relaxation time constant spectrum and the mass transfer recovery time constant spectrum. Multiple time constants belonging to the same cluster center are merged into the same mode. The modal time constant of the same mode is determined based on the mean of all time constants under the same mode, and the modal amplitude coefficient of the same mode is determined based on the mean of all amplitude coefficients under the same mode, thus forming a robust thermal mode parameter set and a robust mass mode parameter set, respectively. Taking the application time as the prediction starting point, the modal time constants and modal amplitude coefficients of each mode in the robust thermal modal parameter set and the robust mass modal parameter set are substituted into a multi-input multi-output state-space model containing temperature and dissolved oxygen cross-coupling terms. The temperature prediction baseline sequence and dissolved oxygen prediction baseline sequence are generated from the prediction starting point and have a length of the third preset time. The original temperature response sequence, which is truncated from the application moment and has a length of the third preset duration, is subtracted point by point from the temperature prediction baseline sequence to obtain the temperature response segment. The original dissolved oxygen response sequence, which is taken from the application time and forward for the same duration, is subtracted point by point from the dissolved oxygen prediction baseline sequence to obtain the dissolved oxygen response segment.

[0028] In one specific embodiment, step S1031 defines the natural cooling relaxation period. Within a preset time period, the time interval between all adjacent short-time pulse excitation signals is defined as the natural cooling relaxation period. The start time of a natural cooling relaxation period is the application time of the previous short-time pulse excitation signal plus the first preset duration, that is, the moment when the previous pulse just ended; the end time of the natural cooling relaxation period is the application time of the next short-time pulse excitation signal, that is, the moment when the next pulse is about to start. During the natural cooling relaxation period, the heating element of the dissolved oxygen sensor is in a non-working state and does not generate additional heat. The dissolved oxygen sensor only relies on natural convection and molecular diffusion to exchange heat and dissolved oxygen with the surrounding water. Step S1032: Extract naturally decayed and recovered subsequences Within a preset time period, temperature data corresponding to each natural cooling relaxation period is extracted from the internal temperature sequence of the sensor and arranged in chronological order to form a temperature natural decay subsequence for that natural cooling relaxation period. The temperature natural decay subsequence records the entire process of the sensor's internal temperature gradually decaying from the pulse peak to the water environment temperature after the pulse heating ends. Meanwhile, dissolved oxygen concentration data corresponding to each natural cooling relaxation period were extracted from the dissolved oxygen concentration measurement sequence and arranged in chronological order to form the dissolved oxygen natural recovery subsequence for that natural cooling relaxation period. The dissolved oxygen natural recovery subsequence recorded the entire process of the membrane oxygen permeability gradually returning to normal after the pulse heating ended and the dissolved oxygen measurement value recovering from the low state caused by the pulse to the true dissolved oxygen concentration of the water body. Step S1033: First-order thermal relaxation decomposition of the temperature-naturally decaying subsequence The dissolved oxygen sensor contains multiple physical layers, from the micro heating element outwards to the water body, including an electrolyte cavity layer, a membrane body layer, and a contaminant layer that may adhere to the outer surface of the membrane. Each physical layer has its own independent specific heat capacity, density, thickness, and contact thermal resistance with adjacent layers. Therefore, each physical layer exhibits independent first-order thermal relaxation characteristics during cooling, meaning that the heat stored in the physical layer decays exponentially, and the rate of decay is determined by the product of the heat capacity and thermal resistance of the physical layer. Based on the above physical model, each temperature natural decay subsequence is expressed as a linear superposition of multiple independent first-order thermal relaxation processes, mathematically represented as a multi-exponential decay function, where each exponential decay term corresponds to a physical layer structure. The time constant of this term depends on the product of the heat capacity and thermal resistance of the physical layer structure, and the amplitude coefficient of this term depends on the heat contained in the physical layer structure at the beginning of the natural cooling relaxation period. Step S1034: Solve for the thermal relaxation time constant and amplitude coefficient to construct the thermal relaxation time constant spectrum. For each temperature natural decay subsequence corresponding to the multi-exponential decay function, a global optimization algorithm based on singular value decomposition and trust region reflection is used to jointly solve the time constant and amplitude coefficient of all first-order thermal relaxation processes. The solution process consists of two stages. The first stage uses singular value decomposition to reduce the dimensionality of the multi-exponential decay function and perform initial value estimation, projecting the original high-dimensional nonlinear parameter space onto a low-dimensional subspace. This identifies the number of each exponential decay mode and provides initial estimates of the time constants and amplitude coefficients of each mode. The second stage, starting from these initial estimates, employs a trust-region reflection algorithm to perform constrained optimization search in the parameter space of the multi-exponential decay function. By iteratively approximating the global optimum, the final output is the time constant and amplitude coefficient of each first-order thermal relaxation process corresponding to the temperature-naturally decaying subsequence. The time constants and amplitude coefficients identified during all natural cooling relaxation periods within a preset time period are summarized and arranged in ascending order of time constant to form a thermal relaxation time constant spectrum. Each element in the time constant spectrum contains a time constant value and its corresponding amplitude coefficient value. Step S1035: First-order mass transfer recovery decomposition of the dissolved oxygen spontaneous recovery subsequence Corresponding to the thermal relaxation process, the physical layered structures in the dissolved oxygen sensor, such as the electrolyte chamber layer, the membrane body layer, and the contaminant adhesion layer, have their own independent oxygen storage capacity and diffusion resistance. The oxygen storage capacity of each physical layered structure depends on the product of the oxygen solubility and the volume of that physical layered structure, and the diffusion resistance depends on the ratio of the thickness of that physical layered structure to the oxygen diffusion coefficient. After the pulse heating ends, the concentration distribution of dissolved oxygen in each physical layered structure gradually recovers to the equilibrium state. The recovery process of each physical layered structure exhibits independent first-order mass transfer recovery characteristics. Based on the above physical model, each dissolved oxygen natural recovery subsequence is expressed as a linear superposition of multiple independent first-order mass transfer recovery processes, mathematically represented as a multi-exponential recovery function, where each exponential recovery term corresponds to a physical hierarchical structure. Step S1036: Solve for the mass transfer recovery time constant and amplitude coefficient to construct the mass transfer recovery time constant spectrum; Using the same global optimization algorithm based on singular value decomposition and trust region reflection as in step S1034, the time constants and amplitude coefficients of all first-order mass transfer recovery processes are jointly solved for the multi-exponential recovery function corresponding to each dissolved oxygen natural recovery subsequence. The mass transfer recovery time constants and amplitude coefficients identified during all natural cooling relaxation periods within a preset time period are summarized and arranged in ascending order of time constant to form a mass transfer recovery time constant spectrum. Step S1037: Cluster analysis to form a robust modal parameter set During multiple natural cooling relaxation processes, due to the slight fluctuations in the water environment and the influence of measurement noise, the time constants and amplitude coefficients identified by the same physical stratified structure in different natural cooling relaxation periods may have a certain degree of dispersion. In order to obtain more representative and robust modal parameters, cluster analysis was performed on the time constants in the thermal relaxation time constant spectrum and the mass transfer recovery time constant spectrum, respectively. The specific implementation of cluster analysis is as follows: all time constants in each time constant spectrum are arranged according to their numerical values, the distance between adjacent time constants is calculated, and natural breakpoints with a distance significantly greater than the average distance are identified as the boundaries of modes corresponding to different physical hierarchical structures. Time constants that fall between two adjacent natural breakpoints, or between the minimum time constant and the first natural breakpoint, or between the maximum time constant and the last natural breakpoint, are classified as belonging to the same cluster center. Multiple time constants belonging to the same cluster center are merged into the same mode. The arithmetic mean of all time constants under the same mode is calculated, and the arithmetic mean is determined as the mode time constant of the same mode. At the same time, the amplitude coefficients corresponding to each time constant under the same mode are extracted, and the arithmetic mean of these amplitude coefficients is calculated. The arithmetic mean is determined as the mode amplitude coefficient of the same mode. The time constants and corresponding amplitude coefficients of all thermal relaxation modes obtained after clustering and merging are summarized to form a robust thermal mode parameter set, and the time constants and corresponding amplitude coefficients of all mass transfer recovery modes obtained after clustering and merging are summarized to form a robust mass mode parameter set. Step S1038: Recursively predict and generate temperature prediction baseline sequences and dissolved oxygen prediction baseline sequences. Taking a certain application time as the prediction starting point, obtain the robust thermal mode parameter set and robust mass mode parameter set determined after the most recent natural cooling relaxation period before that application time; The modal time constants and modal amplitude coefficients of each mode in the robust thermal modal parameter set, as well as the modal time constants and modal amplitude coefficients of each mode in the robust mass modal parameter set, are substituted into a pre-constructed multi-input multi-output state-space model that includes temperature and dissolved oxygen cross-coupling terms. The multi-input multi-output state-space model takes the temperature state variable and dissolved oxygen state variable of each mode as the internal state, takes the current sensor internal temperature value and dissolved oxygen concentration value as the input, and takes the future temperature prediction value and dissolved oxygen prediction value as the output. The model has cross-coupling terms to characterize the instantaneous effect of temperature change on oxygen diffusion rate and the feedback effect of oxygen concentration change on local micro-convection heat transfer. Starting from the prediction start point, and using the sampling interval of each sampling point within the third preset time period as the recursive step size, the state-space model is used to progressively calculate and generate a temperature prediction baseline sequence and a dissolved oxygen prediction baseline sequence with a length of the third preset time period, starting from the prediction start point. The temperature prediction baseline sequence represents the trajectory of temperature change that the sensor should follow under the assumption of not applying the current pulse excitation signal; the dissolved oxygen prediction baseline sequence represents the trajectory of dissolved oxygen measurement that should follow under the assumption of not applying the current pulse excitation signal. Step S1039: Obtain the temperature response segment The original temperature response sequence with a length of three preset times is extracted from a certain application time and subtracted from the temperature prediction baseline sequence generated in step S1038 point by point at the same time to obtain a certain temperature response segment. Each data point in the certain temperature response segment represents the net temperature increment caused by pulse heating at that time, eliminating the contribution of ambient temperature trend changes and the natural cooling process inside the sensor. Step S10310: Obtain dissolved oxygen response fragment The original dissolved oxygen response sequence of the same time length is extracted from a certain application time and subtracted from the dissolved oxygen prediction baseline sequence generated in step S1038 point by point at the same time to obtain a certain dissolved oxygen response segment. Each data point in a certain dissolved oxygen response segment represents the net change in dissolved oxygen caused by pulse heating at that time, eliminating the contributions of background oxygen consumption and trend changes in dissolved oxygen concentration in the water body.

[0029] In summary, by analyzing the natural decay and recovery patterns of sensor temperature and dissolved oxygen during the natural cooling relaxation period, we identified the thermal relaxation time constant spectrum and mass transfer recovery time constant spectrum of the multi-physical layered structure inside the dissolved oxygen sensor. Based on this, we constructed a dynamic baseline prediction model, which enabled the accurate extraction of the net temperature response and net dissolved oxygen response caused by pulse heating from the original measurement sequence. Compared to conventional baseline subtraction methods that simply subtract fixed time intervals before and after pulse application, this approach offers the following significant advantages: First, it effectively eliminates interference from diurnal temperature variations, seasonal temperature changes, and the gradual trend of dissolved oxygen background concentration in the water, ensuring a comparable and unified benchmark for pulse responses applied at different times. Second, through multimodal identification and cluster analysis, it adaptively determines the true relaxation characteristics of different physical hierarchical structures within the dissolved oxygen sensor, avoiding fitting biases caused by artificially pre-setting a single time constant or a fixed baseline model. Third, the cross-coupling terms in the multi-input multi-output state-space model fully consider the dynamic interaction between temperature and oxygen concentration changes, making the baseline prediction results closer to the actual physical process and laying a reliable foundation for the accurate extraction of subsequent time constants.

[0030] Step S104: Determine a certain temperature rise time constant of a certain temperature response segment and a certain dissolved oxygen fall time constant of a certain dissolved oxygen response segment.

[0031] In this step, a continuous wavelet transform is performed on a certain temperature response segment to obtain a temperature-time-scale spectrum. The first scale band corresponding to the temperature rise caused by the pulse excitation is identified on the temperature-time-scale spectrum. The first starting inflection moment of the wavelet coefficient modulus maxima ridge on the first scale band is extracted. The time difference between the first starting inflection moment and the certain application moment is determined as the certain temperature rise time constant. A continuous wavelet transform is performed on a certain dissolved oxygen response segment to obtain a concentration-time-scale spectrum. The second scale band corresponding to the dissolved oxygen decrease caused by the pulse excitation is identified on the concentration-time-scale spectrum. The second starting inflection moment of the wavelet coefficient modulus maxima ridge on the second scale band is extracted. The time difference between the second starting inflection moment and the certain application moment is determined as the dissolved oxygen decrease time constant.

[0032] In one specific implementation, firstly, a wavelet basis function with continuous first derivative and excellent time-frequency localization characteristics, such as the Mexican hat wavelet or Morlet wavelet, is selected to perform a continuous wavelet transform on the temperature response segment. The continuous wavelet transform expands the original one-dimensional time-domain signal into a two-dimensional temperature-time-scale spectrum by translating the wavelet basis function along the time axis at different scales and convolving it with the temperature response segment. In the temperature-time-scale spectrum, the horizontal axis represents time, the vertical axis represents scale, and the brightness or color intensity of each point on the spectrum represents the magnitude of the wavelet coefficients at that time and scale.

[0033] On the temperature-time-scale spectrum, the temperature rise caused by pulse excitation manifests as significant local maxima of wavelet coefficient moduli within a specific scale range. By scaling, the scale range where these local maxima are concentrated is identified and designated as the first scale band. The scale parameters corresponding to the first scale band match the rise rate of the pulse excitation signal and the characteristic time scale of the temperature response, typically falling within the mid-to-high frequency scale range.

[0034] Within the first scale band, the positions of the wavelet coefficient modulus maxima at each scale are traced along the time axis, and these points are connected to form a wavelet coefficient modulus maxima ridge. The wavelet coefficient modulus maxima ridge characterizes the propagation trajectory of signal singularities along the scale during temperature rise. Before the pulse excitation is applied, the temperature response segment is at the baseline level, and the wavelet coefficient modulus maxima ridge remains at a stable low value; when the pulse excitation causes the temperature to rise rapidly, the wavelet coefficient modulus maxima ridge shows a significant upward inflection.

[0035] The moment corresponding to the inflection point where the wavelet coefficient modulus maxima ridge on the first scale band transitions from a stationary state to a monotonically increasing state is extracted and recorded as the first inflection point. The first inflection point represents the moment when the temperature rise caused by pulse heating begins to propagate. The time difference between the first inflection point and the applied time is calculated as a temperature rise time constant.

[0036] Using the same wavelet basis function and continuous wavelet transform method as the one used to extract the temperature rise time constant, a continuous wavelet transform is performed on a certain dissolved oxygen response segment to obtain a concentration-time-scale spectrum. In the concentration-time-scale spectrum, the horizontal axis represents time, the vertical axis represents scale, and the values ​​on the spectrum represent the wavelet coefficient modulus values ​​of the dissolved oxygen concentration signal at each time point and at each scale.

[0037] In the concentration-time-scale spectrum, the decrease in dissolved oxygen caused by pulse excitation manifests as significant local maxima of wavelet coefficient moduli within a specific scale range. By scale scanning, the scale range where these local maxima are concentrated is identified and designated as the second scale band. Since the decrease in dissolved oxygen is limited by the diffusion rate of oxygen molecules, its response speed is typically slower than that of the temperature rise process; therefore, the scale range corresponding to the second scale band is usually more biased towards the lower frequency range than that of the first scale band.

[0038] Within the second scale band, the positions of the wavelet coefficient modulus maxima at each scale are traced along the time axis, and these positions are connected to form a ridge line of wavelet coefficient modulus maxima. Before the pulse excitation is applied, the dissolved oxygen response segment is at the baseline level, and the wavelet coefficient modulus maxima ridge line remains at a stable low value; when the pulse excitation causes the dissolved oxygen concentration to begin to decrease rapidly, the wavelet coefficient modulus maxima ridge line shows a significant downward inflection.

[0039] The moment corresponding to the inflection point where the wavelet coefficient modulus maxima ridge on the second scale band transitions from a stationary state to a monotonically decreasing state is extracted and recorded as the second inflection point. The second inflection point represents the moment when the dissolved oxygen decrease caused by pulse heating begins to propagate. The time difference between the second inflection point and the applied time is calculated as the dissolved oxygen decrease time constant.

[0040] In summary, by performing independent wavelet analyses on the temperature response segment and the dissolved oxygen response segment, there is no coupling error between the temperature rise time constant and the dissolved oxygen fall time constant. When the membrane ages or becomes contaminated, the heat conduction path and the oxygen diffusion path are affected to different degrees, which is reflected in the independent changes of the temperature rise time constant and the dissolved oxygen fall time constant, respectively. This provides two reliable and physically meaningful independent variables for the subsequent construction of the response characteristic ratio.

[0041] Step S105: Determine a response characteristic ratio of a short-time pulse excitation signal based on a certain temperature rise time constant and a certain dissolved oxygen fall time constant, and arrange the response characteristic ratios of multiple consecutive short-time pulse excitation signals in chronological order to obtain a response characteristic ratio sequence.

[0042] In this step, taking the short-time pulse excitation signal as the current pulse, K consecutive short-time pulse excitation signals including the current pulse are selected forward to form a time constant analysis window, where K is an odd number and not less than 5; The temperature rise time constant and dissolved oxygen fall time constant corresponding to each short-time pulse excitation signal within the time constant analysis window are obtained to form K pairs of time constant data points; Plot the K pairs of time constant data points in a Cartesian coordinate system with the temperature rise time constant as the abscissa and the dissolved oxygen fall time constant as the ordinate. Perform principal axis regression analysis on the K pairs of time constant data points to obtain the slope of the principal axis regression line. The slope of the principal axis regression line is used as the response characteristic ratio of a certain short-time pulse excitation signal.

[0043] It should be noted that, within the preset time period, water flow velocity data and turbidity data at the monitoring section are collected simultaneously. Align the application time of each short-time pulse excitation signal with the water flow velocity data and the turbidity data in time, and extract the maximum flow velocity fluctuation value and the maximum turbidity fluctuation value within a preset time window before and after the application time of each short-time pulse excitation signal. When the maximum flow velocity fluctuation value corresponding to a certain short-time pulse excitation signal exceeds the preset flow velocity fluctuation threshold, or the maximum turbidity fluctuation value exceeds the preset turbidity fluctuation threshold, the response characteristic ratio corresponding to the certain short-time pulse excitation signal is marked as an invalid value. Remove all response characteristic ratios marked as invalid from the response characteristic ratios of multiple consecutive short-time pulse excitation signals, and arrange the remaining response characteristic ratios in the original time order; For the sequence gaps caused by the removal of invalid values, the arithmetic mean of the ratios of two consecutive valid response features is used to fill the gaps, thereby obtaining the response feature ratio sequence.

[0044] In one specific embodiment, the structure of the time constant analysis window is first determined. Taking a short-time pulse excitation signal for which the response characteristic ratio needs to be calculated as the current pulse, the process traces backward along the time axis, selecting K consecutive short-time pulse excitation signals, including the current pulse, to form a time constant analysis window. K is an odd number and not less than 5, for example, K is 7 or 9. Using an odd window length ensures that the current pulse is centered within the window, with (K-1) / 2 pulses before and after the current pulse serving as context references. The time constant analysis window slides pulse by pulse along the time axis, and each pulse, when used as the current pulse, constructs a time constant analysis window centered on itself.

[0045] Next, K pairs of time constant data points are obtained within the time constant analysis window. For each short-time pulse excitation signal within the time constant analysis window, the temperature rise time constant and dissolved oxygen fall time constant corresponding to the short-time pulse excitation signal determined in step S104 are extracted and paired together. A total of K pairs of time constant data points are obtained within the window, and each pair of data points is represented as (temperature rise time constant, dissolved oxygen fall time constant).

[0046] Then, K pairs of time constant data points were plotted in a Cartesian coordinate system, and principal axis regression analysis was performed. A scatter plot of the K pairs of time constant data points was created with the temperature rise time constant as the x-axis and the dissolved oxygen fall time constant as the y-axis. Principal axis regression analysis was performed on the K pairs of time constant data points. This analysis considers measurement errors in both the x-axis and y-axis directions, and determines the direction of the regression line by minimizing the orthogonal distance from the data points to the regression line, thus obtaining the slope of the principal axis regression line.

[0047] The slope of the principal axis regression line has a clear physical meaning. When the membrane is clean, the temperature rise time constant and dissolved oxygen fall time constant corresponding to each pulse excitation signal maintain a stable proportional relationship. The K-pair time constant data points are distributed along a relatively fixed straight line on the two-dimensional plane, and the slope of the principal axis regression line is near the baseline value. When the membrane gradually ages or becomes contaminated, the additional obstruction of the contamination layer to the oxygen molecule diffusion path is greater than its additional obstruction to the heat conduction path. This causes the increase in the dissolved oxygen fall time constant to exceed the increase in the temperature rise time constant. The distribution direction of the K-pair time constant data points on the two-dimensional plane deflects counterclockwise, and the slope of the principal axis regression line increases accordingly. Therefore, the slope of the principal axis regression line can directly reflect the degree to which the membrane state deviates from a clean state.

[0048] The slope of the principal axis regression line obtained from the above principal axis regression analysis is used as a certain response characteristic ratio of the current pulse.

[0049] Following the steps described above, the corresponding response characteristic ratio is calculated for each short-time pulse excitation signal within a preset time period, resulting in a set of response characteristic ratios for multiple consecutive short-time pulse excitation signals.

[0050] Within a preset time period, water flow velocity and turbidity data are simultaneously collected at the monitoring section. Water flow velocity data is acquired using an acoustic Doppler current profiler or a rotor-type current meter installed at the monitoring section, with the sampling frequency consistent with or time-synchronized with the dissolved oxygen sensor. Turbidity data is acquired using an optical backscattering turbidimeter installed at the monitoring section, with the sampling frequency also consistent with or time-synchronized with the dissolved oxygen sensor.

[0051] The application time of each short-time pulse excitation signal is strictly aligned with the water flow velocity and turbidity data in time. Using the application time of each short-time pulse excitation signal as the center, a preset time window is extended forward and backward. The length of the preset time window is typically set to one-quarter to one-half of a second preset duration. All flow velocity data within the preset time window are extracted, and the difference between the maximum and minimum flow velocity values ​​within the preset time window is calculated. This difference is taken as the maximum flow velocity fluctuation value corresponding to the short-time pulse excitation signal. Simultaneously, all turbidity data within the preset time window are extracted, and the difference between the maximum and minimum turbidity values ​​within the preset time window is calculated. This difference is taken as the maximum turbidity fluctuation value corresponding to the short-time pulse excitation signal.

[0052] The maximum flow velocity fluctuation value corresponding to each short-time pulse excitation signal is compared with a preset flow velocity fluctuation threshold, and the corresponding maximum turbidity fluctuation value is also compared with a preset turbidity fluctuation threshold. The flow velocity fluctuation threshold and turbidity fluctuation threshold are preset based on historical statistical values ​​of the monitoring section under normal and stable water flow conditions. Typically, the 95th percentile of historical flow velocity fluctuation values ​​and the 95th percentile of historical turbidity fluctuation values ​​are used as the flow velocity fluctuation threshold and turbidity fluctuation threshold, respectively.

[0053] When the maximum velocity fluctuation value corresponding to a short-term pulse excitation signal exceeds the velocity fluctuation threshold, or the maximum turbidity fluctuation value exceeds the turbidity fluctuation threshold, it indicates that a significant hydrodynamic disturbance or particulate impact event occurred near the application time of the short-term pulse excitation signal. Such events instantaneously alter the convective heat transfer conditions and oxygen mass transfer boundary layer thickness on the membrane surface, leading to abnormal fluctuations in the temperature rise time constant and dissolved oxygen fall time constant that are not caused by the membrane's inherent characteristics. Therefore, the response characteristic ratio corresponding to this short-term pulse excitation signal is marked as an invalid value.

[0054] Iterate through all short-time pulse excitation signals within a preset time period, remove all response feature ratios marked as invalid from the set of response feature ratios of multiple consecutive short-time pulse excitation signals, and arrange the remaining response feature ratios marked as valid in the order of the original pulse application time.

[0055] For any missing positions in the sequence after invalid values ​​have been removed, find the nearest valid response characteristic ratio before and after the missing position. Add these two valid response characteristic ratios together, divide by two, and fill the missing position with the arithmetic mean. If the missing position is at the beginning or end of the sequence, use the value of the nearest valid response characteristic ratio to fill forward or backward. After interpolation and filling, a response characteristic ratio sequence with a length equal to the total number of short-time pulse excitation signals is obtained.

[0056] In summary, by constructing a time constant analysis window and performing principal axis regression analysis on a two-dimensional plane, the coupling relationship between the two independent physical quantities, the temperature rise time constant and the dissolved oxygen fall time constant, is mapped to the slope of the regression line. Compared with the simple method of directly taking the ratio of the two time constants as the response characteristic ratio, principal axis regression analysis effectively reduces the impact of single-pulse measurement noise on the response characteristic ratio. The statistical regression of multiple data points within the window has a smoothing and suppression effect on random noise, making the response characteristic ratio more stable and robust in representing the membrane state. At the same time, principal axis regression considers the measurement errors of both variables simultaneously, avoiding the estimation bias introduced by ordinary least squares regression, which only considers the error of the dependent variable, and ensuring the unbiasedness of the slope estimation. By simultaneously collecting water flow velocity and turbidity data, and using the flow velocity and turbidity fluctuations near the pulse application time as quantitative criteria for interference identification, abnormal response characteristic ratios caused by hydrodynamic disturbances and particulate matter impacts can be accurately identified and eliminated. This interference identification method based on independent physical quantities does not rely on the statistical distribution characteristics of the response characteristic ratio itself, thus avoiding the risk of incorrectly identifying and eliminating abnormal changes caused by actual membrane degradation as interference. The gaps created after eliminating invalid values ​​are filled by linear interpolation of adjacent valid values. While preserving the overall trend continuity of the sequence, it avoids introducing false fluctuations through interpolation, providing a high-quality and continuous response characteristic ratio sequence for subsequent sliding statistical analysis and fault determination.

[0057] Step S106: Calculate the sliding median and sliding interquartile range of the response characteristic ratio sequence. When the response characteristic ratios of three consecutive short-time pulse excitation signals are all greater than the sliding median plus twice the sliding interquartile range, it is determined that the dissolved oxygen sensor has a membrane aging or contamination fault within the preset time period, and the fault start time is recorded as the application time corresponding to the first excessive pulse.

[0058] In this step, a sliding step size covering the entire length of the response feature ratio sequence is set, and the response feature ratio sequence is divided into sliding subsequences to obtain multiple continuous and partially overlapping sliding subsequences, each sliding subsequence containing an odd number of continuous response feature ratios; For each sliding subsequence, the median of all response feature ratios within the sliding subsequence is taken as the subsequence median corresponding to the sliding subsequence, and the interquartile range of all response feature ratios within the sliding subsequence is taken as the subsequence interquartile range corresponding to the sliding subsequence, wherein the interquartile range is the difference between the upper quartile and the lower quartile of the response feature ratios within the sliding subsequence. Calculate the mean of the medians of all sliding subsequences, and use the mean of the medians of all sliding subsequences as the sliding median of the response feature ratio sequence; Calculate the mean of the interquartile ranges of all sliding subsequences, and use the mean of the interquartile ranges of all sliding subsequences as the sliding interquartile range of the response feature ratio sequence.

[0059] In one specific embodiment, based on the response feature ratio sequence obtained in step S105, the global sliding median and sliding interquartile range are calculated using a sliding subsequence statistical method, and a dynamic fault determination threshold is established accordingly to achieve reliable detection of diaphragm aging or contamination faults and accurate tracing of the fault initiation time. The specific implementation method is as follows: Step 1: Set the parameters of the sliding subsequence and perform sequence partitioning.

[0060] Obtain the response characteristic ratio sequence output in step S105. Assume that the response characteristic ratio sequence contains N response characteristic ratios, numbered sequentially from the 1st to the Nth according to the order of pulse application time. The value of N is equal to the total number of actual effective short-time pulse excitation signals within the preset time period.

[0061] A sliding step size is set, which is an index interval of one response characteristic ratio. A subsequence length M is also set, where M is an odd number and M is less than N. The value of M is determined based on the total number of pulse excitation signals within a preset time period and the time scale of diaphragm fault evolution. Typically, M is set to an odd number between one-third and one-half of N; for example, when N is 60, M can be 19 or 21. Using an odd number for M ensures that the median of the subsequence can be directly taken as the single value in the middle position after sorting, without averaging adjacent values.

[0062] Starting from the beginning of the response feature ratio sequence, the sequence slides forward with a step size of 1, extracting a continuous segment of length M of response feature ratios with each slide, forming a sliding subsequence. Specifically, the first sliding subsequence consists of the first to the Mth response feature ratios, the second sliding subsequence consists of the second to the (M+1)th response feature ratios, and so on, with the i-th sliding subsequence consisting of the i-th to the (i+M-1)-th response feature ratios. The sliding process continues until the (i+M-1)-th response feature ratio equals N, i.e., the end of the sliding subsequence touches the end of the response feature ratio sequence. After partitioning, a total of N-M+1 sliding subsequences are obtained, with an overlap of M-1 response feature ratios between adjacent sliding subsequences.

[0063] Step 2: Calculate the median and interquartile range of each sliding subsequence.

[0064] For each sliding subsequence obtained from the partitioning, perform the following operations.

[0065] The M response feature ratios contained in the sliding subsequence are sorted in ascending order to obtain an ordered sequence of length M. The value at the (M+1) / 2th position in the sorted ordered sequence is taken as the median of the corresponding subsequence.

[0066] The value at the (M+1) / 4th position in the sorted sequence is taken as the lower quartile, and the value at the 3(M+1) / 4th position is taken as the upper quartile. When (M+1) / 4 or 3(M+1) / 4 is not an integer, the corresponding quartile is determined by linear interpolation of the values ​​at two adjacent sorted positions. The difference between the upper and lower quartiles is taken as the interquartile range of the corresponding sliding subsequence.

[0067] Perform the above calculations on each of the N-M+1 sliding subsequences to obtain the median and interquartile range of the N-M+1 subsequences.

[0068] Step 3: Calculate the moving median and moving interquartile range of the response feature ratio sequence.

[0069] The medians of the N-M+1 subsequences calculated in the second step are averaged arithmetically, that is, the sum of the medians of all subsequences is divided by N-M+1, and the result is used as the moving median of the response feature ratio sequence.

[0070] The interquartile ranges of the N-M+1 subsequences calculated in the second step are arithmetically averaged, that is, the sum of the interquartile ranges of all subsequences is divided by N-M+1, and the result is used as the moving interquartile range of the response feature ratio sequence.

[0071] The moving median and moving interquartile range (interquartile range) mentioned above are both single values. The moving median represents the central tendency of the response characteristic ratio sequence after removing local fluctuations over the entire preset time period. The moving interquartile range represents the overall dispersion of the response characteristic ratio sequence after removing local fluctuations over the entire preset time period. Since these two statistics are obtained by averaging the statistics of all moving subsequences covering the entire sequence length, they reflect the overall statistical characteristics of the membrane state over the entire preset time period, are not significantly affected by local abnormal fluctuations in individual subsequences, and have good robustness.

[0072] Step 4: Establish fault determination thresholds and perform fault determination.

[0073] After obtaining the moving median and moving interquartile range of the response feature ratio sequence, the fault determination threshold is calculated. The fault determination threshold is the moving median plus twice the moving interquartile range.

[0074] Following the chronological order, starting with the first short-time pulse excitation signal, each short-time pulse excitation signal's corresponding response characteristic ratio is checked sequentially to see if it is greater than the fault determination threshold. A continuous overshoot counter is set, initially set to zero. When the response characteristic ratio of a short-time pulse excitation signal is found to be greater than the fault determination threshold, the continuous overshoot counter is incremented by 1; when the response characteristic ratio of a short-time pulse excitation signal is found to be less than or equal to the fault determination threshold, the continuous overshoot counter is reset to zero.

[0075] When the count value of the continuous exceedance counter reaches 3, that is, when the response characteristic ratio of three consecutive short-time pulse excitation signals is greater than the moving median plus twice the moving interquartile range, it is determined that the dissolved oxygen sensor has a membrane aging or contamination fault within the preset time period. At this time, the first exceedance pulse among the three consecutive exceedance pulses is traced back, and the application time corresponding to the first exceedance pulse is recorded as the fault start time.

[0076] If the count value of the continuous exceedance counter fails to reach 3 after the entire response characteristic ratio sequence check is completed, it is determined that the dissolved oxygen sensor has not experienced membrane aging or contamination failure within the preset time period.

[0077] In summary, the method of this application obtains the dissolved oxygen concentration and sensor internal temperature sequence; applies equally spaced short-duration pulse excitation; extracts temperature response segments and dissolved oxygen response segments caused by the pulses; determines the temperature rise time constant and dissolved oxygen fall time constant; constructs response characteristic ratios and arranges them into a sequence; calculates the moving median and moving interquartile range of the sequence; when three consecutive response characteristic ratios are all greater than the moving median plus twice the moving interquartile range, the membrane aging or contamination fault is determined, and the fault initiation time is recorded; utilizing the thermo-mass coupling response characteristics under pulse excitation, the membrane status is diagnosed in real time online through dynamic statistical thresholds, effectively eliminating hydrodynamic and turbidity interferences, and improving the sensitivity and reliability of fault detection.

[0078] Please see Figure 2 The diagram shows a structural block diagram of a water quality monitoring device based on multi-source data according to this application.

[0079] like Figure 2 As shown, the water quality monitoring device 200 includes an acquisition module 210, an application module 220, an interception module 230, a first determination module 240, a second determination module 250, and a judgment module 260.

[0080] The acquisition module 210 is configured to acquire the dissolved oxygen concentration measurement sequence of the dissolved oxygen sensor at the same monitoring section of the target water area within a preset time period, and the internal temperature sequence of the temperature sensor integrated inside the dissolved oxygen sensor within the same time period; the application module 220 is configured to apply multiple equally spaced short-time pulse excitation signals to the heating element of the dissolved oxygen sensor within the preset time period, the duration of each pulse excitation signal being a first preset duration and the pulse interval being a second preset duration; the interception module 230 is configured to, in response to a certain application time of a certain short-time pulse excitation signal, intercept a temperature response segment from the internal temperature sequence of the sensor, starting from the certain application time and extending backward for a third preset duration, and intercept a dissolved oxygen response segment from the dissolved oxygen concentration measurement sequence within the same time interval; first The determining module 240 is configured to determine a certain temperature rise time constant of a certain temperature response segment and a certain dissolved oxygen fall time constant of a certain dissolved oxygen response segment; the second determining module 250 is configured to determine a certain response characteristic ratio of a certain short-time pulse excitation signal based on the certain temperature rise time constant and the certain dissolved oxygen fall time constant, and arrange the response characteristic ratios of multiple consecutive short-time pulse excitation signals in chronological order to obtain a response characteristic ratio sequence; the judging module 260 is configured to calculate the sliding median and sliding interquartile range of the response characteristic ratio sequence, and when the response characteristic ratios of three consecutive short-time pulse excitation signals are all greater than the sliding median plus twice the sliding interquartile range, it is judged that the dissolved oxygen sensor has a membrane aging or contamination fault within the preset time period, and the fault start time is recorded as the application time corresponding to the first excessive pulse.

[0081] It should be understood that Figure 2 The modules and references described in the document Figure 1 The steps described in the text correspond to those in the method described above. Therefore, the operations, features, and corresponding technical effects described above also apply to the method described in the text. Figure 2 The various modules in the document will not be described in detail here.

[0082] In other embodiments, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the water quality monitoring method based on multi-source data in any of the above method embodiments. In one embodiment, the computer-readable storage medium of the present invention stores computer-executable instructions, which are configured as follows: Obtain the dissolved oxygen concentration measurement sequence of the dissolved oxygen sensor at the same monitoring section of the target water area within a preset time period, and the internal temperature sequence of the temperature sensor integrated inside the dissolved oxygen sensor within the same time period; During the preset time period, multiple equally spaced short-time pulse excitation signals are applied to the heating element of the dissolved oxygen sensor. The duration of each pulse excitation signal is a first preset duration, and the pulse interval is a second preset duration. In response to a certain application time of a short-time pulse excitation signal, a certain temperature response segment is extracted from the internal temperature sequence of the sensor, starting from the certain application time and extending backward for a third preset time period, and a certain dissolved oxygen response segment is extracted from the dissolved oxygen concentration measurement sequence within the same time interval. Determine a certain temperature rise time constant for a certain temperature response segment, and a certain dissolved oxygen fall time constant for a certain dissolved oxygen response segment; A response characteristic ratio of a short-time pulse excitation signal is determined based on a certain temperature rise time constant and a certain dissolved oxygen fall time constant, and the response characteristic ratios of multiple consecutive short-time pulse excitation signals are arranged in chronological order to obtain a response characteristic ratio sequence. Calculate the sliding median and sliding interquartile range of the response characteristic ratio sequence. When the response characteristic ratios of three consecutive short-time pulse excitation signals are all greater than the sliding median plus twice the sliding interquartile range, it is determined that the dissolved oxygen sensor has a membrane aging or contamination fault within the preset time period, and the fault start time is recorded as the application time corresponding to the first excessive pulse.

[0083] Computer-readable storage media may include a stored program area and a stored data area, wherein the stored program area may store an application program required for operating the device and at least one function; and the stored data area may store data created based on the use of the water quality monitoring device based on multi-source data. Furthermore, the computer-readable storage medium may include high-speed random access memory, and may also include memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include memory remotely configured relative to a processor, which can be connected to the water quality monitoring device based on multi-source data via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0084] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention, such as... Figure 3 As shown, the device includes a processor 310 and a memory 320. The electronic device may also include an input device 330 and an output device 340. The processor 310, memory 320, input device 330, and output device 340 can be connected via a bus or other means. Figure 3Taking a bus connection as an example, the memory 320 is the computer-readable storage medium described above. The processor 310 executes various server functions and data processing by running non-volatile software programs, instructions, and modules stored in the memory 320, thereby implementing the water quality monitoring method based on multi-source data described in the above embodiment. The input device 330 can receive input digital or character information and generate key signal inputs related to user settings and function control of the water quality monitoring device based on multi-source data. The output device 340 may include a display screen or other display device.

[0085] The aforementioned electronic device can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0086] In one implementation, the above-described electronic device is used in a water quality monitoring device based on multi-source data, for a client application, and includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Obtain the dissolved oxygen concentration measurement sequence of the dissolved oxygen sensor at the same monitoring section of the target water area within a preset time period, and the internal temperature sequence of the temperature sensor integrated inside the dissolved oxygen sensor within the same time period; During the preset time period, multiple equally spaced short-time pulse excitation signals are applied to the heating element of the dissolved oxygen sensor. The duration of each pulse excitation signal is a first preset duration, and the pulse interval is a second preset duration. In response to a certain application time of a short-time pulse excitation signal, a certain temperature response segment is extracted from the internal temperature sequence of the sensor, starting from the certain application time and extending backward for a third preset time period, and a certain dissolved oxygen response segment is extracted from the dissolved oxygen concentration measurement sequence within the same time interval. Determine a certain temperature rise time constant for a certain temperature response segment, and a certain dissolved oxygen fall time constant for a certain dissolved oxygen response segment; A response characteristic ratio of a short-time pulse excitation signal is determined based on a certain temperature rise time constant and a certain dissolved oxygen fall time constant, and the response characteristic ratios of multiple consecutive short-time pulse excitation signals are arranged in chronological order to obtain a response characteristic ratio sequence. Calculate the sliding median and sliding interquartile range of the response characteristic ratio sequence. When the response characteristic ratios of three consecutive short-time pulse excitation signals are all greater than the sliding median plus twice the sliding interquartile range, it is determined that the dissolved oxygen sensor has a membrane aging or contamination fault within the preset time period, and the fault start time is recorded as the application time corresponding to the first excessive pulse.

[0087] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A water quality monitoring method based on multi-source data, characterized in that, include: Obtain the dissolved oxygen concentration measurement sequence of the dissolved oxygen sensor at the same monitoring section of the target water area within a preset time period, and the internal temperature sequence of the temperature sensor integrated inside the dissolved oxygen sensor within the same time period; During the preset time period, multiple equally spaced short-time pulse excitation signals are applied to the heating element of the dissolved oxygen sensor. The duration of each pulse excitation signal is a first preset duration, and the pulse interval is a second preset duration. In response to a certain application time of a short-time pulse excitation signal, a certain temperature response segment is extracted from the internal temperature sequence of the sensor, starting from the certain application time and extending backward for a third preset time period, and a certain dissolved oxygen response segment is extracted from the dissolved oxygen concentration measurement sequence within the same time interval. Determine a certain temperature rise time constant for a certain temperature response segment, and a certain dissolved oxygen fall time constant for a certain dissolved oxygen response segment; A response characteristic ratio of a short-time pulse excitation signal is determined based on a certain temperature rise time constant and a certain dissolved oxygen fall time constant, and the response characteristic ratios of multiple consecutive short-time pulse excitation signals are arranged in chronological order to obtain a response characteristic ratio sequence. Calculate the sliding median and sliding interquartile range of the response characteristic ratio sequence. When the response characteristic ratios of three consecutive short-time pulse excitation signals are all greater than the sliding median plus twice the sliding interquartile range, it is determined that the dissolved oxygen sensor has a membrane aging or contamination fault within the preset time period, and the fault start time is recorded as the application time corresponding to the first excessive pulse.

2. The water quality monitoring method based on multi-source data according to claim 1, characterized in that, The step of extracting a temperature response segment from the internal temperature sequence of the sensor, starting from the applied time and extending backward for a third preset time period, and extracting a dissolved oxygen response segment from the dissolved oxygen concentration measurement sequence within the same time interval includes: The time interval between all adjacent short-time pulse excitation signals within the preset time period is defined as the natural cooling relaxation period. During the natural cooling relaxation period, the heating element of the dissolved oxygen sensor is in a non-working state, and the dissolved oxygen sensor exchanges heat and dissolved oxygen with the water body only through natural convection and diffusion. Extract the temperature natural decay subsequence and dissolved oxygen natural recovery subsequence for each natural cooling relaxation period from the internal temperature sequence of the sensor and the dissolved oxygen concentration measurement sequence, respectively; Each temperature decay subsequence is decomposed into a linear superposition of multiple independent first-order thermal relaxation processes, where each first-order thermal relaxation process corresponds to a physical layered structure with independent thermal capacity and thermal resistance within the dissolved oxygen sensor. A global optimization algorithm based on singular value decomposition and trust region reflection is used to jointly solve the time constants and amplitude coefficients of multiple first-order thermal relaxation processes. The time constants and amplitude coefficients identified in all natural cooling relaxation periods are arranged in ascending order of time constant to form a thermal relaxation time constant spectrum. Each dissolved oxygen natural recovery subsequence is decomposed into a linear superposition of multiple independent first-order mass transfer recovery processes, wherein each first-order mass transfer recovery process corresponds to a physical layered structure with independent oxygen storage capacity and diffusion resistance within the dissolved oxygen sensor. The global optimization algorithm based on singular value decomposition and trust region reflection is used to jointly solve the time constants and amplitude coefficients of multiple first-order mass transfer recovery processes. The time constants and amplitude coefficients identified during all natural cooling relaxation periods are arranged in ascending order of time constant to form a mass transfer recovery time constant spectrum. Cluster analysis is performed on each time constant in the thermal relaxation time constant spectrum and the mass transfer recovery time constant spectrum. Multiple time constants belonging to the same cluster center are merged into the same mode. The modal time constant of the same mode is determined based on the mean of all time constants under the same mode, and the modal amplitude coefficient of the same mode is determined based on the mean of all amplitude coefficients under the same mode, thus forming a robust thermal mode parameter set and a robust mass mode parameter set, respectively. Taking the application time as the prediction starting point, the modal time constants and modal amplitude coefficients of each mode in the robust thermal modal parameter set and the robust mass modal parameter set are substituted into a multi-input multi-output state-space model containing temperature and dissolved oxygen cross-coupling terms. The temperature prediction baseline sequence and dissolved oxygen prediction baseline sequence are generated from the prediction starting point and have a length of the third preset time. The original temperature response sequence, which is truncated from the application moment and has a length of the third preset duration, is subtracted point by point from the temperature prediction baseline sequence to obtain the temperature response segment. The original dissolved oxygen response sequence, which is taken from the application time and has the same time length, is subtracted point by point from the dissolved oxygen prediction baseline sequence to obtain the dissolved oxygen response segment.

3. The water quality monitoring method based on multi-source data according to claim 1, characterized in that, The determination of a certain temperature rise time constant for a certain temperature response segment and a certain dissolved oxygen fall time constant for a certain dissolved oxygen response segment includes: A continuous wavelet transform is performed on a certain temperature response segment to obtain a temperature-time-scale spectrum. The first scale band corresponding to the temperature rise edge caused by pulse excitation is identified on the temperature-time-scale spectrum. The first starting inflection moment of the wavelet coefficient modulus maxima ridge line on the first scale band is extracted. The time difference between the first starting inflection moment and the certain application moment is determined as the certain temperature rise time constant. A continuous wavelet transform is performed on a certain dissolved oxygen response segment to obtain a concentration-time-scale spectrum. The second scale band corresponding to the dissolved oxygen decrease caused by the pulse excitation is identified on the concentration-time-scale spectrum. The second starting inflection moment of the wavelet coefficient modulus maxima ridge on the second scale band is extracted. The time difference between the second starting inflection moment and the certain application moment is determined as the dissolved oxygen decrease time constant.

4. The water quality monitoring method based on multi-source data according to claim 3, characterized in that, The determination of a certain response characteristic ratio of a certain short-time pulse excitation signal based on a certain temperature rise time constant and a certain dissolved oxygen fall time constant includes: Taking the aforementioned short-time pulse excitation signal as the current pulse, select K consecutive short-time pulse excitation signals including the current pulse to form a time constant analysis window, where K is an odd number and not less than 5; The temperature rise time constant and dissolved oxygen fall time constant corresponding to each short-time pulse excitation signal within the time constant analysis window are obtained to form K pairs of time constant data points; Plot the K pairs of time constant data points in a Cartesian coordinate system with the temperature rise time constant as the abscissa and the dissolved oxygen fall time constant as the ordinate. Perform principal axis regression analysis on the K pairs of time constant data points to obtain the slope of the principal axis regression line. The slope of the principal axis regression line is used as the response characteristic ratio of a certain short-time pulse excitation signal.

5. The water quality monitoring method based on multi-source data according to claim 1, characterized in that, The step of arranging the response characteristic ratios of multiple consecutive short-time pulse excitation signals in chronological order to obtain a response characteristic ratio sequence includes: During the preset time period, water flow velocity data and turbidity data at the monitoring section are collected synchronously. Align the application time of each short-time pulse excitation signal with the water flow velocity data and the turbidity data in time, and extract the maximum flow velocity fluctuation value and the maximum turbidity fluctuation value within a preset time window before and after the application time of each short-time pulse excitation signal. When the maximum flow velocity fluctuation value corresponding to a certain short-time pulse excitation signal exceeds the preset flow velocity fluctuation threshold, or the maximum turbidity fluctuation value exceeds the preset turbidity fluctuation threshold, the response characteristic ratio corresponding to the certain short-time pulse excitation signal is marked as an invalid value. Remove all response characteristic ratios marked as invalid from the response characteristic ratios of multiple consecutive short-time pulse excitation signals, and arrange the remaining response characteristic ratios in the original time order; For the sequence gaps caused by the removal of invalid values, the arithmetic mean of the ratios of two consecutive valid response features is used to fill the gaps, thereby obtaining the response feature ratio sequence.

6. The water quality monitoring method based on multi-source data according to claim 1, characterized in that, The calculation of the moving median and moving interquartile range of the response feature ratio sequence includes: Set a sliding step size that covers the entire length of the response feature ratio sequence, divide the response feature ratio sequence into sliding subsequences, and obtain multiple continuous and partially overlapping sliding subsequences, each sliding subsequence containing an odd number of continuous response feature ratios; For each sliding subsequence, the median of all response feature ratios within the sliding subsequence is taken as the subsequence median corresponding to the sliding subsequence, and the interquartile range of all response feature ratios within the sliding subsequence is taken as the subsequence interquartile range corresponding to the sliding subsequence, wherein the interquartile range is the difference between the upper quartile and the lower quartile of the response feature ratios within the sliding subsequence. Calculate the mean of the medians of all sliding subsequences, and use the mean of the medians of all sliding subsequences as the sliding median of the response feature ratio sequence; Calculate the mean of the interquartile ranges of all sliding subsequences, and use the mean of the interquartile ranges of all sliding subsequences as the sliding interquartile range of the response feature ratio sequence.

7. A water quality monitoring device based on multi-source data, characterized in that, include: The acquisition module is configured to acquire the dissolved oxygen concentration measurement sequence of the dissolved oxygen sensor at the same monitoring section of the target water area within a preset time period, and the internal temperature sequence of the temperature sensor integrated inside the dissolved oxygen sensor within the same time period. The application module is configured to apply multiple equally spaced short-time pulse excitation signals to the heating element of the dissolved oxygen sensor within the preset time period, wherein the duration of each pulse excitation signal is a first preset duration and the pulse interval is a second preset duration. The interception module is configured to, in response to a certain application time of a certain short-time pulse excitation signal, intercept a certain temperature response segment from the internal temperature sequence of the sensor, starting from the certain application time and extending backward for a third preset time, and intercept a certain dissolved oxygen response segment from the dissolved oxygen concentration measurement sequence within the same time interval. The first determining module is configured to determine a certain temperature rise time constant of a certain temperature response segment and a certain dissolved oxygen fall time constant of a certain dissolved oxygen response segment; The second determining module is configured to determine a certain response characteristic ratio of a certain short-time pulse excitation signal based on a certain temperature rise time constant and a certain dissolved oxygen fall time constant, and to arrange the response characteristic ratios of multiple consecutive short-time pulse excitation signals in chronological order to obtain a response characteristic ratio sequence. The determination module is configured to calculate the sliding median and sliding interquartile range of the response characteristic ratio sequence. When the response characteristic ratios of three consecutive short-time pulse excitation signals are all greater than the sliding median plus twice the sliding interquartile range, it is determined that the dissolved oxygen sensor has a membrane aging or contamination fault within the preset time period, and the fault start time is recorded as the application time corresponding to the first excessive pulse.

8. An electronic device, characterized in that, include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the method described in any one of claims 1 to 6.

Images