Distributed water quality monitoring system for small-scale rural water supply

Abstract

The present invention provides a distributed water quality monitoring system for small-scale rural water supply, comprising: a measurement module, comprising a plurality of sensors distributed at different positions and used for measuring water quality to obtain water quality data; an adaptive weighted fusion module, used for performing weighted fusion on the water quality data obtained at a same time point; and a prediction module, used for predicting water quality on the basis of water quality data obtained by means of the weighted fusion.

Description

A distributed water quality monitoring system for small-scale rural water supply Technical Field

[0001] This invention relates to the field of water quality monitoring, and more particularly to a distributed water quality monitoring system for small-scale rural water supply. Background Technology

[0002] Currently, for small-scale rural water supply systems, testing water quality in only one location is insufficient to accurately represent the water quality situation. Therefore, it's necessary to test water quality at different locations. However, the current method simply involves direct data comparison between tests at different locations. While this meets some needs, excessive data comparison makes it difficult for general staff to differentiate between different sources. Furthermore, different locations may have different water quality priorities, making this method inadequate for meeting the overall requirements.

[0003] Summary of the Invention

[0004] The main objective of this invention is to provide a distributed water quality monitoring system for small-scale rural water supply, which can integrate water quality data from different locations and of different types, and weight and fuse these data to obtain a more accurate water quality evaluation value.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: a distributed water quality monitoring system for small-scale rural water supply, comprising:

[0006] The detection module includes multiple sensors distributed in different locations to detect water quality and obtain water quality data;

[0007] An adaptive weighted fusion module is used to perform weighted fusion of the water quality data obtained at the same time.

[0008] The prediction module is used to predict water quality based on the water quality data obtained by weighted fusion.

[0009] The weighted fusion of the water quality data obtained at the same time specifically includes the following steps:

[0010] Step 11: The number of sensors is m, and the water quality data obtained at a certain moment is X1, X2, K, X m , for X i Perform a data consistency check, i = 1, 2, ..., m, with the check criteria being X1, X2, K, X m The difference between any two adjacent values ​​cannot exceed a given threshold ε, as shown in the equation:

[0011] Obtain n water quality data points that meet the data consistency requirements, where n ≤ m;

[0012] Step 12: The fusion value obtained after weighted fusion Expressed using formula (2):

[0013] Among them, W p Let p be the weighting factor for the p-th water quality data, satisfying

[0014] Step 13: Based on the theory of finding extrema of multivariate functions, find the weighting factor Wp that minimizes the total mean square error. * ,

[0015] Where, σp 2 Let be the variance of the p-th water quality data point.

[0016] Preferably, σp is calculated. 2 and σi 2 The following steps are adopted:

[0017] Step 131: Let there be two different sensors h and j, whose measured values ​​are X and J respectively. h and X j The corresponding observation errors are V h and V j That is: X h =X+V h X j =X+V j X is the theoretical value detected by the sensor, where V h and V j For zero-mean stationary noise, V h and V j They are uncorrelated and have a mean of zero, V h and V j It is also unrelated to X, then X h The variance is

[0018] Step 132, X h and X j cross-correlation coefficient R hj =E[X h X j ] = E[X 2 The autocorrelation coefficient R of X h =E[X h X h ] = E[X 2 ]+E[Vh 2 ], thus obtaining: σh 2 =E[Vh 2 ] = R hh -R hj(4),

[0019] Where R hh and R hj The values ​​are obtained from the time-domain estimates, as follows:

[0020] Step 132: Assume that the number of water quality data points detected by sensors h and j is k, and R hh The time-domain estimate is R hh (k), R hj Time domain estimate R hj (k), then we get formula (5):

[0021] Similarly, we obtain formula (6):

[0022] Step 133: Extend the calculation process in Step 132 to include n sensors (n≥2), and estimate the time-domain correlation coefficient R between each pair of the n sensors. hj mean of (k) As an estimate, it is expressed in the following formula:

[0023] Step 134, obtain R hh R hj Substituting into formula (4) yields σp 2 and σi 2 Then the obtained σp 2 and σi 2 Substituting into formula (3) yields The result As W p Substituting into formula (2), we obtain the fusion value.

[0024] Preferably, the prediction module specifically includes the following steps during prediction:

[0025] Step 21: Decompose the fusion value obtained by the adaptive weighted fusion module, specifically including the following steps:

[0026] Step 211: Remove obviously erroneous data and derive a usable sample set based on this, where the data is the fusion value;

[0027] Step 212: Based on the improved overall average empirical mode decomposition, the time series of the fused value is decomposed into decomposition terms IMF1, MF2, Λ, IMF based on different frequencies. K and remaining items RS K K is the number of decomposition terms;

[0028] Step 213: Test the randomness of the decomposition terms and remainder terms based on permutation entropy;

[0029] Step 22: Based on the waveform characteristics of the decomposition terms and remainder terms at different frequencies, different prediction methods are selected for prediction, and the prediction results are verified. The prediction methods are prediction using a bidirectional long short-term memory network and prediction using an ABC-BP neural network, and the prediction results are obtained.

[0030] Step 23: Add the obtained prediction results to get the predicted value, and judge the quality of the water based on the predicted value.

[0031] Preferably, step 212 specifically includes the following steps:

[0032] Step 2121: Let the data after processing in step 211 be s(t). Determine the maximum and minimum points on s(t), and connect all the maximum and minimum points with two curves respectively.

[0033] Step 2122: Take the two curves as the upper and lower envelopes of s(t) respectively. The mean of each point on the upper and lower envelopes is denoted as m, and the difference between s(t) and m is denoted as h. Then h = s(t) - m.

[0034] Step 2123: Treat h as a new s(t), repeat the above operation. When the conditions of m being zero or close to zero, h being less than the set value, or h being one of the monotonic functions are met, denote c1 = h, treat c1 as IMF1(t), and then denote s(t) - c1 = r1(t), where r1(t) is the remainder term.

[0035] Step 2124: Treat r1(t) as the new s(t), and repeat the above process until c K or r K (t) satisfies the given termination condition, and IMF2(t), c2,r2(t), IMF3(t), c3,r3(t), ΛΛ are obtained in sequence;

[0036] After steps 2125 and 2124 are completed, the following decomposition is obtained:

[0037] Where, r K (t) is called the remainder term.

[0038] This invention also provides an integrated intelligent water purification device for small-scale rural water supply projects, comprising:

[0039] Membrane pool;

[0040] Multiple membrane modules placed in a membrane pool;

[0041] The water pump has its inlet connected to the outlets of multiple membrane modules via pipes.

[0042] The flocculant dosing tank is used to add flocculant to the membrane tank.

[0043] The product water tank is connected to the outlet of the product water pump via a pipe.

[0044] In the aforementioned distributed water quality monitoring system, multiple sensors are installed at different locations on the water purification equipment to detect water quality.

[0045] Preferably, the system also includes a blower connected to the bottom of each membrane module via a pipe, the blower being able to aerate the membrane module.

[0046] Preferably, the device further includes a backwash water pump, the inlet of which is connected to the product water tank, and the outlet of which is connected to the outlet of the membrane module.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] This invention performs weighted fusion of data detected by multiple sensors to obtain more accurate detection values; at the same time, based on the weighted fusion detection values, multiple prediction models are used for prediction, and the prediction results are combined to obtain more accurate prediction values. Attached Figure Description

[0049] Figure 1 is a prediction flowchart according to Embodiment 1 of the present invention;

[0050] Figure 2 is a structural schematic diagram of the water purification equipment. Detailed Implementation

[0051] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0052] Example 1

[0053] A distributed water quality monitoring system for small-scale rural water supply includes a detection module, an adaptive weighted fusion module, and a prediction module. The detection module comprises multiple sensors distributed at different locations. These sensors can be of the same type or different types and are used to detect water quality parameters to obtain water quality data. These parameters include one or more of the following: pH value, residual chlorine, turbidity, and conductivity. The detection module performs measurements at regular intervals, acquiring multiple sets of water quality data.

[0054] The adaptive weighted fusion module is used to perform weighted fusion of water quality data acquired at the same time. The specific fusion method is as follows:

[0055] Step 11: The number of sensors is m, and the water quality data obtained at a certain moment is X1, X2, K, X m , for X i Perform a data consistency check on (i = 1, 2, ..., m), with the check criterion being X1, X2, K, X... m The difference between any two adjacent values ​​cannot exceed a given threshold ε, as shown in the equation:

[0056] Finally, n water quality data points that meet the data consistency requirements are obtained, where n ≤ m;

[0057] Step 12: The fusion value obtained after weighted fusion Expressed using formula (2):

[0058] Among them, W p Let p be the weighting factor for the p-th water quality data, satisfying

[0059] Step 13: Based on the theory of finding extrema of multivariate functions, find the weighting factor corresponding to the minimum total mean square error.

[0060] Where, σp 2 Calculate Wp to represent the variance of the p-th water quality data point. * The following steps are adopted:

[0061] Step 131: Let there be two different sensors h and j, whose measured values ​​are X and J respectively. h and X j The corresponding observation errors are V h and V j That is: X h =X+V h X j =X+V j X is the theoretical value detected by the sensor, where V h and V j For zero-mean stationary noise, V h and V j They are uncorrelated and have a mean of zero, V h and V j It is also unrelated to X, then X h The variance is σh 2 =E[Vh 2 ];

[0062] Step 132, X h and X j cross-correlation coefficient R hj =E[X hX j ] = E[X 2 The autocorrelation coefficient R of X hh =E[X h X h ] = E[X 2 ]+E[Vh 2 ], thus obtaining: σh 2 =E[Vh 2 ] = R hh -R hj (4),

[0063] Where R hh and R hj The values ​​are obtained from the time-domain estimates, as follows:

[0064] Step 132: Let k be the number of water quality data points detected by sensors h and j, and R hh The time-domain estimate is R hh (k), R hj Time domain estimate R hj (k), then we get formula (5):

[0065] Similarly, we obtain formula (6):

[0066] Step 133: Extend the calculation process in Step 132 to include n sensors (n≥2), and estimate the time-domain correlation coefficient R between each pair of the n sensors. hj mean of (k) As an estimate, it is expressed in the following formula:

[0067] Step 134, obtain R hh R hj Substituting into formula (4) yields σp 2 and σi 2 Then the obtained σp 2 and σi 2 Substituting into formula (3) yields The result As W p Substituting into formula (2), we obtain the fusion value.

[0068] The prediction module employs a time-series data classification and prediction model based on MEEMD-BiLSTM-improved BP neural network, specifically including the following steps:

[0069] Step 21, decomposition, specifically includes the following steps:

[0070] Step 211, Data Preprocessing: Complete the storage and management of nonlinear and non-stationary signal sequence data, the basic analysis of the relationship between the mean and variance of the data, the calculation of the correlation coefficient between the data, etc., remove obviously erroneous data through analysis, and obtain a usable sample set based on this, the data being the fused value;

[0071] Step 212: Based on the improved total mean empirical mode decomposition (MEEMD), the time series of fusion values ​​satisfying randomness is decomposed into decomposition terms IMF1, MF2, Λ, IMF based on different frequencies. K and remaining items RS K K is the number of terms in the decomposition;

[0072] Step 212 specifically includes the following steps:

[0073] Step 2121: Let the data after step 212 be s(t). Determine the maximum and minimum points on s(t), and connect all the maximum and minimum points with two curves respectively.

[0074] Step 2122: Take the two curves as the upper and lower envelopes of s(t) respectively. The mean of each point on the upper and lower envelopes is denoted as m, and the difference between s(t) and m is denoted as h. Then h = s(t) - m.

[0075] Step 2123: Treat h as a new s(t), repeat the above operation, and when the conditions of m being zero or close to zero, h being sufficiently small, or h being one of the monotonic functions are met, denote c1 = h. Treat c1 as IMF1(t), and denote it as s(t) - c1 = r1(t), where r1(t) is the remainder term;

[0076] Step 2124: Treat r1(t) as the new s(t), and repeat the above process until c n or r n (t) satisfies the given termination condition (the remainder is small enough or becomes a monotonic function); IMF2(t), c2,r2(t), IMF3(t), c3,r3(t), ΛΛ are obtained in sequence;

[0077] After steps 2125 and 2124 are completed, the following decomposition is obtained:

[0078] Where, r K (t) is called the remainder term.

[0079] Step 213: Based on the permutation entropy, test the randomness of the decomposition terms and remainder terms, specifically using existing techniques;

[0080] Step 22: If the decomposition terms and remainder terms meet the requirement of randomness, select different prediction methods to make predictions according to the waveform characteristics of decomposition terms and remainder terms of different frequencies, and verify the prediction results. The prediction methods are prediction using a bidirectional long short-term memory network (BiLSTM) and prediction using an ABC-BP neural network.

[0081] Step 24: Add the prediction results of the bidirectional long short-term memory network and the prediction results of the ABC-BP neural network together, and the combined value is the final prediction value.

[0082] Example 2

[0083] This embodiment is an integrated intelligent water purification device for small-scale rural water supply projects, including a membrane tank, multiple membrane modules placed in the membrane tank, a product water pump, a flocculant dosing tank, and a product water tank. The water to be treated first enters the membrane tank. The inlet of the product water pump is connected to the outlet of the multiple membrane modules through pipes. The flocculant dosing tank is used to add flocculant to the membrane tank. The membrane modules are used to filter the water in the membrane tank. The outlet of the product water pump is connected to the product water tank through pipes. The product water tank is used to store the treated water.

[0084] The water purification equipment also includes a blower, which is connected to the bottom of each membrane module via a pipe. The blower can aerate the membrane module, enabling the membrane module to achieve a better treatment effect.

[0085] The water purification equipment also includes a backwash pump, the inlet of which is connected to the product water tank, and the outlet of which is connected to the outlet of the membrane module. The backwash pump is used to clean the membrane module by reverse flushing it with water when cleaning is required.

[0086] Furthermore, the water purification equipment also includes the distributed water quality monitoring system in Embodiment 1, with multiple sensors distributed at multiple locations on the equipment to detect water quality and control the production water pump, flocculant dosing tank, blower and backwash water pump to perform corresponding actions based on the prediction results.

[0087] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A distributed water quality monitoring system for small-scale rural water supply, comprising: The detection module includes multiple sensors distributed in different locations to detect water quality and obtain water quality data; An adaptive weighted fusion module is used to perform weighted fusion of the water quality data obtained at the same time. The prediction module is used to predict water quality based on the water quality data obtained by weighted fusion. The weighted fusion of the water quality data obtained at the same time specifically includes the following steps: Step 11: The number of sensors is m, and the water quality data obtained at a certain moment is X1, X2, K, X m , for X i Perform a data consistency check, i = 1, 2, ..., m, with the check criterion being X. 1, X2,K,X m The difference between any two adjacent values ​​cannot exceed a given threshold ε, as shown in the equation: Obtain n water quality data points that meet the data consistency requirements, where n ≤ m; Step 12, the weighted fusion value obtained after fusion Equation (2) is used to express: wherein W p is the weighting factor of the pth water quality data, satisfying Step 13, according to the extremum theory of multivariate function, the corresponding weighted factor Wp is solved when the total mean square error is minimum * , wherein σp 2 is the variance of the pth water quality data.

2. The system of claim 1, wherein, Compute σp 2 and σi 2 The following steps are employed: Step 131: Let there be two different sensors h and j, whose measured values ​​are X and J respectively. h and X j The corresponding observation errors are V h and V j X h =X+V h X j =X+V j X is the theoretical value detected by the sensor, where V h and V j for zero-mean stationary noise, V h and V j are uncorrelated and have zero mean, V h and V j are uncorrelated and have zero mean, V h and V 2 are uncorrelated and have zero mean, V 2 and V Step 132, X h and X j cross-correlation coefficient R hj =E[X h X j ] = E[X 2 The autocorrelation coefficient R of X hh =E[X h X h ] = E[X 2 ]+E[Vh 2 ], thus obtaining: σh 2 = E[Vh 2 ] = R hh - R hj (4), where R hh and R hj are obtained from the time domain estimates as follows: Step 133, set the number of water quality data detected by sensors h and j as k, R hh (k) is the time domain estimation value of R hh (k), R hj (k) is the time domain estimation value of R hj (k), then formula (5) is obtained: By the same token, equation (6) is obtained: Step 133, extend the calculation process in step 132 to n sensors, n≥2, and estimate the time-domain correlation coefficient R hj the mean of (k) As an estimated value, the following formula is used: Step 134, substituting the obtained R hh , R hj into formula (4) to obtain σp 2 and σi 2 , and then substituting the obtained σp 2 and σi 2 into formula (3) to obtain The resulting As W p Substituting into equation (2), the fusion value is obtained 3. The system of claim 2, wherein, The prediction module specifically includes the following steps during prediction: Step 21: Decompose the fusion value obtained by the adaptive weighted fusion module, specifically including the following steps: Step 211: Remove obviously erroneous data and derive a usable sample set based on this, where the data is the fusion value; Step 212, decomposing the time series of the fusion value into decomposition terms IMF1, MF2, Λ, IMF of different frequencies based on the improved ensemble average empirical mode decomposition K and a residual term RS K , K is the number of decomposition terms; Step 213: Based on the permutation entropy test, examine the randomness of the decomposition terms and the remainder term to determine whether the decomposition terms satisfy randomness; Step 22: Based on the waveform characteristics of the decomposition terms and remainder terms at different frequencies, different prediction methods are selected for prediction, and the prediction results are verified. The prediction methods are prediction using a bidirectional long short-term memory network and prediction using an ABC-BP neural network, and the prediction results are obtained. Step 23: Add the obtained prediction results to get the predicted value, and judge the quality of the water based on the predicted value.

4. The system according to claim 3, step 212 specifically includes the following steps: Step 2121: Let the data after processing in step 211 be s(t). Determine the maximum and minimum points on s(t), and connect all the maximum and minimum points with two curves respectively. Step 2122: Take the two curves as the upper and lower envelopes of s(t) respectively. The mean of each point on the upper and lower envelopes is denoted as m, and the difference between s(t) and m is denoted as h. Then h = s(t) - m. Step 2123: Treat h as a new s(t), repeat the above operation. When the conditions of m being zero or close to zero, h being less than the set value, or h being one of the monotonic functions are met, denote c1 = h, treat c1 as IMF1(t), and then denote s(t) - c1 = r1(t), where r1(t) is the remainder term. Step 2124, take r1(t) as new s(t), repeat the above process until c K or r K (t) satisfies the given termination condition, in turn, get IMF2(t), c2, r2(t), IMF3(t), c3, r3(t), ΛΛ; Step 2125, after step 2124 is stopped, the following decomposition formula is obtained: wherein r K (t) is called the remainder term.

5. An integrated intelligent water purification device for rural small water supply projects, characterized in that, include: Membrane pool; Multiple membrane modules placed in a membrane pool; The water pump has its inlet connected to the outlets of multiple membrane modules via pipes. The flocculant dosing tank is used to add flocculant to the membrane tank. The water production tank is connected to the outlet of the water production pump via a pipe. The system according to any one of claims 1-4, wherein a plurality of the sensors are disposed at different locations in the water purification device for detecting water quality.

6. The integrated intelligent water purification device for rural small water supply projects according to claim 5, characterized in that, It also includes a blower, which is connected to the bottom of each membrane module via a pipe and is capable of aerating the membrane module.

7. The integrated intelligent water purification device for rural small water supply projects according to claim 6, characterized in that, It also includes a backwash water pump, the inlet of which is connected to the product water tank, and the outlet of which is connected to the outlet of the membrane module.

Images