Online monitoring method and system for water quality in coal mine working faces based on deep learning

Abstract

The present invention discloses an online monitoring method for water quality of a coal mine working face based on deep learning, comprising: obtaining several types of ion concentration data of a first water area collected by several sensors; inputting the ion concentration data into a first judgment model, a second judgment model, and a third judgment model respectively to obtain a first judgment probability, a second judgment probability, and a third judgment probability, wherein each judgment probability includes at least one target judgment result, and each target judgment result corresponding to each judgment probability has a corresponding weight factor, and the target judgment result is the probability of being judged as a preset water quality; calculating the target judgment result corresponding to each judgment probability according to the weight factor to obtain a comprehensive target judgment probability, and the comprehensive target prediction probability is the comprehensive probability that each target prediction result is judged as a preset water quality; and selecting the result with the largest comprehensive probability of the preset water quality in the target judgment results as the current water quality type.

Description

Technical Field

[0001] The present invention belongs to the field of water quality detection, and specifically relates to a method and system for online monitoring of water quality in coal mine working faces based on deep learning. Background Art

[0002] Online monitoring of water quality at coal mine working faces is a key control measure to ensure coal mine production safety and protect the environment. The prior art has many documents publicly available on water quality evaluation. For example, application number 201710106308.8 discloses a water quality evaluation and classification method based on a neural network, which trains the neural network to identify water quality. However, since the water quality at the coal mine working face is different from the surface water quality, the water quality at the coal mine working face contains more elements, and the water quality changes stably. When performing corresponding water quality detection, corresponding sensors are required to collect concentrations, resulting in a large number of parameters. Therefore, when identifying water quality, if the content measurement error is large, it is easy to make an incorrect identification. In addition, a lot of data needs to be read before identification, and it takes a long time to read and store the data, which affects the real-time nature of water quality monitoring. Summary of the Invention

[0003] In order to solve the above problems existing in the prior art, the present invention provides a method and system for online monitoring of water quality in coal mine working faces based on deep learning. The technical problems to be solved by the present invention are achieved through the following technical solutions:

[0004] A method for online monitoring of water quality in coal mine working faces based on deep learning, comprising:

[0005] Acquiring several ion concentration data of a first water area collected by several sensors;

[0006] Inputting the ion concentration data into the first judgment model, the second judgment model, and the third judgment model respectively to obtain a first judgment probability, a second judgment probability, and a third judgment probability, wherein each judgment probability includes at least one target judgment result, and each target judgment result corresponding to each judgment probability has a corresponding weight factor, and the target judgment result is the probability of being judged as a preset water quality;

[0007] Calculate the target judgment result corresponding to each judgment probability according to the weight factor to obtain a comprehensive target judgment probability, wherein the comprehensive target prediction probability is the comprehensive probability that each target prediction result corresponds to the preset water quality;

[0008] The result with the largest comprehensive probability of the preset water quality in the target judgment result is selected as the current water quality type.

[0009] In a specific embodiment, the first judgment model is a BP God General Network model, the second judgment model is a Bayesian prediction model, and the third judgment model is a three-line graph model.

[0010] In a specific embodiment, it also includes:

[0011] Querying all sensor identifiers in the first water area to read historical data of each ion in a preset time period from a database corresponding to the sensor identifier to obtain historical data of several ion concentrations;

[0012] The ion concentration at a future preset moment is predicted according to the several types of ion concentration historical data, so as to obtain the water quality type at the future preset moment based on the ion concentration at the future preset moment.

[0013] In a specific embodiment, when it is determined that the water quality type at the future preset moment is abnormal, the future preset moment is marked and an early warning is issued.

[0014] In a specific embodiment, obtaining the concentration data of several ions in the first water area collected by several sensors further includes:

[0015] Selecting one sensor from the plurality of sensors as a mother sensor and the remaining sensors as child sensors, acquiring historical ion concentration data of the mother sensor and each child sensor to obtain a mother sequence and a plurality of child sequences;

[0016] Performing correlation analysis on the parent sequence and the child sequence, selecting the sub-sensors corresponding to the sub-sequences with correlation greater than a correlation threshold as associable sensors, and selecting the sub-sensors corresponding to the sub-sequences with correlation less than the correlation threshold as unassociable sensors;

[0017] Calculating the mother sequence and the associable sequence of each associable sensor according to an absolute correlation algorithm to obtain a corresponding weighted vector according to the mother sequence and each associable sequence;

[0018] Acquire ion concentration data of a first water area collected by a mother sensor and an unassociated sensor, process the ion concentration data of the mother sensor according to a corresponding weighted vector to obtain ion concentration data of an associable sensor associated with the mother sensor, and determine the water quality type based on the ion concentration data of the mother sensor, the ion concentration data of the associable sensor calculated by the mother sensor, and the ion concentration data of the unassociated sensor.

[0019] The present invention also discloses a coal mine working face water quality online monitoring system based on deep learning, comprising:

[0020] A data acquisition module, configured to acquire data on concentrations of several ions in a first water area collected by several sensors;

[0021] a probability judgment module, configured to input the ion concentration data into a first judgment model, a second judgment model, and a third judgment model, respectively, to obtain a first judgment probability, a second judgment probability, and a third judgment probability, wherein each judgment probability includes at least one target judgment result, each target judgment result corresponding to each judgment probability has a corresponding weight factor, and the target judgment result is a probability of being judged as a preset water quality;

[0022] A probability prediction module is used to calculate the target judgment result corresponding to each judgment probability according to the weight factor to obtain a comprehensive target judgment probability, wherein the comprehensive target prediction probability is the comprehensive probability that each target prediction result corresponds to the preset water quality;

[0023] The water quality determination module is used to select the result with the largest comprehensive probability of the preset water quality in the target judgment result as the current water quality type.

[0024] In a specific embodiment, the first judgment model is a BP God General Network model, the second judgment model is a Bayesian prediction model, and the third judgment model is a three-line graph model.

[0025] In a specific embodiment, it also includes:

[0026] A historical concentration acquisition module is used to query all sensor identifiers in the first water area to read historical data of each ion in a preset time period from a database corresponding to the sensor identifier to obtain historical data of several ion concentrations;

[0027] The water quality type prediction module is used to predict the ion concentration at a future preset moment based on the historical data of the several ion concentrations, so as to obtain the water quality type at the future preset moment based on the ion concentration at the future preset moment.

[0028] In a specific embodiment, when it is determined that the water quality type at the future preset moment is abnormal, the future preset moment is marked and an early warning is issued.

[0029] In a specific embodiment, the data acquisition module further includes:

[0030] a mother sensor selection unit, configured to select one sensor from the plurality of sensors as a mother sensor and the remaining sensors as child sensors, and obtain historical ion concentration data of the mother sensor and each child sensor to obtain a mother sequence and a plurality of child sequences;

[0031] an association determination unit, configured to perform an association analysis on the parent sequence and the child sequence, and select the sub-sensors corresponding to the sub-sequences having an association degree greater than an association degree threshold as associable sensors, and select the sub-sensors corresponding to the sub-sequences having an association degree less than the association degree threshold as unassociable sensors;

[0032] a weighted vector determining unit, configured to calculate the mother sequence and the associable sequence of each associable sensor according to an absolute correlation algorithm, so as to obtain a corresponding weighted vector according to the mother sequence and each associable sequence;

[0033] A data processing unit is used to obtain ion concentration data of the first water area collected by the mother sensor and the unassociated sensor, process the ion concentration data of the mother sensor according to the corresponding weighted vector to obtain ion concentration data of the associable sensor associated with the mother sensor, and determine the water quality type based on the ion concentration data of the mother sensor, the ion concentration data of the associable sensor calculated by the mother sensor, and the ion concentration data of the unassociated sensor.

[0034] Beneficial effects of the present invention:

[0035] The deep learning-based online monitoring method for coal mine working face water quality of the present invention comprehensively considers three judgment models to accurately obtain the water quality type. In addition, through correlation analysis, the corresponding element concentration can be calculated based on the relevant concentration without collecting the concentration of certain elements, thereby reducing the complexity of data collection and analysis and improving measurement efficiency.

[0036] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of a method for online monitoring of water quality in a coal mine working face based on deep learning provided by an embodiment of the present invention;

[0038] Figure 2 This is a module block diagram of a coal mine working face water quality online monitoring system based on deep learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The present invention will be further described in detail below with reference to specific examples, but the embodiments of the present invention are not limited thereto.

[0040] Example 1

[0041] See Figure 1 , Figure 1 This is a flow chart of a method for online water quality monitoring at a coal mine working face based on deep learning provided by an embodiment of the present invention, including:

[0042] S1. Acquire data on the concentration of several ions in a first water area collected by several sensors. Coal mine working face water quality includes at least the concentrations of eight basic ions, including K, Na, Ca, Mg, Cl, SO4, HCO3, and CO3. Because it differs from surface water quality and is exposed to different coal mine working conditions, it may also contain other ions, such as Fe, Cu, and Li. Appropriate sensors must be used to collect the concentration of each ion that may affect water quality measurement.

[0043] S2. Input the ion concentration data into the first judgment model, the second judgment model, and the third judgment model respectively to obtain the first judgment probability, the second judgment probability, and the third judgment probability, wherein each judgment probability includes at least one target judgment result, and each target judgment result corresponding to each judgment probability has a corresponding weight factor, and the target judgment result is the probability of being judged as the preset water quality; in this embodiment, for example, the first judgment model is the BP God General network model, the second judgment model is the Bayesian prediction model, and the third judgment model is the three-line graph model.

[0044] S3. Calculate the target judgment result corresponding to each judgment probability according to the weight factor to obtain a comprehensive target judgment probability, where the comprehensive target prediction probability is the comprehensive probability that each target prediction result corresponds to the preset water quality;

[0045] S4. Select the result with the largest comprehensive probability of the preset water quality in the target judgment result as the current water quality type.

[0046] Because different judgment models operate on different principles, their accuracy and focus vary. The high number of ions present in coal mine environments can reduce the accuracy and speed of each model. The model determines several possible water quality scenarios, and the target judgment result is the probability of a particular water quality. This probability may include multiple target judgment results, each with varying probabilities. The result with the highest probability is generally selected as the final result. The judgment probabilities and target judgment results obtained by the three judgment models are different. This is because each model has corresponding detection defect errors. Therefore, this embodiment assigns a corresponding weight factor to each target judgment result obtained by each model. The weight factor is used to represent the probability that the target judgment result obtained by the model is accurate. For example, the probability that the water quality obtained by the model is HCO3·SO4-Ca·Na·(Cu) is 80%, and the probability that the model is accurate for the water quality judgment result is 80%. Therefore, the probability that the water quality obtained by the model is HCO3·SO4-Ca·Na·(Cu) is 80%*80%=64%. Accordingly, different models will give corresponding water quality judgment probabilities and the probability that the water quality judgment results are accurate. In this embodiment, each target judgment result obtained by the first judgment model, the second judgment model, and the third judgment model is calculated according to the weight factor and then added to obtain a comprehensive target judgment probability. It should be noted that the comprehensive target judgment probability is obtained by adding the target judgment result of each model and the corresponding weight factor. If the three models are relatively accurate in judging the water quality, the comprehensive target judgment probability may exceed 100% after adding. Therefore, we only need to select the result with the largest comprehensive target judgment probability among the target judgment results as the current water quality type.

[0047] In a specific embodiment, it also includes:

[0048] Querying all sensor identifiers in the first water area to read historical data of each ion in a preset time period from a database corresponding to the sensor identifier to obtain historical data of several ion concentrations;

[0049] Based on the historical data of the several types of ion concentrations, the ion concentration at a preset future time is predicted to determine the water quality type at that preset future time based on the ion concentration at the preset future time. This embodiment uses a K-nearest neighbor regression algorithm for prediction, which is trained on data from the six months preceding the selected time to predict data at any time in the next seven days. Of course, other prediction algorithms can also be selected for prediction, such as the AACNet prediction algorithm, which uses a convolutional network with an attention mechanism and an autoregressive model.

[0050] In one embodiment, when the water quality at a predetermined future time is determined to be abnormal, the predetermined future time is marked and an alert is issued. Preferably, an alert is issued one day before the predetermined future time, and the water quality is assessed through real-time data collection. If the assessment result is normal, the mark is removed.

[0051] In a specific embodiment, obtaining the concentration data of several ions in the first water area collected by several sensors further includes:

[0052] Selecting one sensor from the plurality of sensors as a mother sensor and the remaining sensors as child sensors, acquiring historical ion concentration data of the mother sensor and each child sensor to obtain a mother sequence and a plurality of child sequences;

[0053] Performing correlation analysis on the parent sequence and the child sequence, selecting the sub-sensors corresponding to the sub-sequences with correlation greater than a correlation threshold as associable sensors, and selecting the sub-sensors corresponding to the sub-sequences with correlation less than the correlation threshold as unassociable sensors;

[0054] Calculating the mother sequence and the associable sequence of each associable sensor according to an absolute correlation algorithm to obtain a corresponding weighted vector according to the mother sequence and each associable sequence;

[0055] Acquire ion concentration data of a first water area collected by a mother sensor and an unassociated sensor, process the ion concentration data of the mother sensor according to a corresponding weighted vector to obtain ion concentration data of an associable sensor associated with the mother sensor, and determine the water quality type based on the ion concentration data of the mother sensor, the ion concentration data of the associable sensor calculated by the mother sensor, and the ion concentration data of the unassociated sensor.

[0056] It should be noted that the historical water quality data measured by different sensors are generally correlated with each other in adjacent time and space. Assume that the length of the time series is T. If the slopes of the curves of the two groups of water quality time series are closer in the corresponding time periods, the greater the correlation coefficient between the two and the higher the degree of correlation. On the contrary, the smaller the correlation coefficient, the lower the degree of correlation. Since the degree of correlation of different data is different, if the degree of correlation is low, the error of the result obtained by the formula will be greater and it cannot be used for practical use. Therefore, this embodiment sets a correlation threshold and only performs correlation analysis on data greater than the correlation threshold. As mentioned above, the correlation threshold can be determined based on the slope of the curve. For example, the correlation threshold is determined based on the ratio of the slopes of the parent sequence and the child sequence fitted on the curve, and the range of the slope ratio is guaranteed to be between 0-1. The closer to 1, the greater the correlation. The average value of the slope ratio of multiple points can be selected for comparison with the correlation threshold.

[0057] For associable sensors, analysis and calculation are performed based on the absolute correlation algorithm.

[0058] Let X i is a data sequence, denoted as X i =(x i (1),x i (2),...,x i (n)), i=0,1,2,...,m, the data at sequence number k is x i (k), k=1,2,...,n. The parent sequence is X0=(x0(1),x0(2),...,x0(n)), and the remaining sequences are subsequences.

[0059] Both the parent sequence and the child sequence are initialized to zero, which is recorded as Right now make:

[0060]

[0061] After the initial point of the mother sequence is zeroed, it can be expressed as The subsequence is Then we can deduce that:

[0062]

[0063]

[0064] Calculate X0 and X i Regarding the correlation coefficient of the k-th value, the calculation formula is as follows:

[0065]

[0066] Calculate the mean correlation coefficient between the i-th subsequence and the parent sequence as the absolute correlation between the parent sequence and the subsequence. Since this method reflects the degree of correlation between factors through the changes in the time series of each factor, and there is no rate of change at time t = 1, only n-1 time points are involved in the calculation. The formula is as follows:

[0067]

[0068] The correlation coefficient matrix γM is constructed by performing absolute correlation analysis between the parent sequence and the child sequence.

[0069]

[0070] where γ 01 (x0,x1) is the absolute correlation between the parent sequence x0 and the child sequence x1.

[0071] Assume that the water quality history data of the mother sensor g0 is consistent with the surrounding four sub-sensors (g i ,i=1,...,4) is related to the historical data of water quality. is the weighted vector of the water quality historical data of sensor g0 at time t and the water quality historical data of the four surrounding sensors:

[0072]

[0073] Among them, w 0i The sensor g0 and the surrounding four sensors (g i ,i=1,...,4)’s association weight. 0i It is composed of sensor g0 and sensor (g i , i=1,...,4) to calculate the absolute correlation between them. 0i =γ 0i , represents the i-th sensor g at time t i Measured historical water quality data.

[0074] Considering the time correlation, assuming that the water quality history data at time t+1 is related to the water quality history data at time b, then b is called the look-back time. At time t+1, the spatiotemporal input matrix of the water quality history data on sensor g0 is

[0075]

[0076] Thus, the ion concentration data of the associable sensor associated with the mother sensor is obtained according to the ion concentration data of the mother sensor.

[0077] It's worth noting that in some larger coal mines, there may be multiple water areas, namely, a first water area, a second water area, and other related water areas. Since the ions in multiple water areas within a mine are strongly correlated, the above-mentioned absolute correlation algorithm can also be used to perform correlation analysis on the ion concentrations of other water areas. For example, based on the ion concentration data of the first water area, the absolute correlation algorithm can be used to obtain the ion concentration data of the second water area in the same mine. Of course, it is also possible to perform correlation analysis on only the concentrations of certain ions in the first water area and other second water areas. The specific analysis method can be referred to above and will not be repeated here.

[0078] The deep learning-based online monitoring method for coal mine working face water quality of the present invention comprehensively considers three judgment models to accurately obtain the water quality type. In addition, through correlation analysis, the corresponding element concentration can be calculated based on the relevant concentration without collecting the concentration of certain elements, thereby reducing the complexity of data collection and analysis and improving measurement efficiency.

[0079] Figure 2 This is a module block diagram of a deep learning-based coal mine working face water quality online monitoring system provided by an embodiment of the present invention, including:

[0080] Data acquisition module 1, used to obtain several ion concentration data of a first water area collected by several sensors;

[0081] Probability judgment module 2, used to input the ion concentration data into the first judgment model, the second judgment model, and the third judgment model respectively to obtain a first judgment probability, a second judgment probability, and a third judgment probability, wherein each judgment probability includes at least one target judgment result, and each target judgment result corresponding to each judgment probability has a corresponding weight factor, and the target judgment result is the probability of being judged as a preset water quality;

[0082] Probability prediction module 3, used to calculate the target judgment result corresponding to each judgment probability according to the weight factor to obtain a comprehensive target judgment probability, wherein the comprehensive target prediction probability is the comprehensive probability that each target prediction result corresponds to the preset water quality;

[0083] The water quality determination module 4 is configured to select the result with the largest comprehensive probability of the preset water quality in the target judgment result as the current water quality type.

[0084] In a specific embodiment, the first judgment model is a BP God General Network model, the second judgment model is a Bayesian prediction model, and the third judgment model is a three-line graph model.

[0085] In a specific embodiment, it also includes:

[0086] A historical concentration acquisition module is used to query all sensor identifiers in the first water area to read historical data of each ion in a preset time period from a database corresponding to the sensor identifier to obtain historical data of several ion concentrations;

[0087] The water quality type prediction module is used to predict the ion concentration at a future preset moment based on the historical data of the several ion concentrations, so as to obtain the water quality type at the future preset moment based on the ion concentration at the future preset moment.

[0088] In a specific embodiment, when it is determined that the water quality type at the future preset moment is abnormal, the future preset moment is marked and an early warning is issued.

[0089] In a specific embodiment, the data acquisition module further includes:

[0090] a mother sensor selection unit, configured to select one sensor from the plurality of sensors as a mother sensor and the remaining sensors as child sensors, and obtain historical ion concentration data of the mother sensor and each child sensor to obtain a mother sequence and a plurality of child sequences;

[0091] an association determination unit, configured to perform an association analysis on the parent sequence and the child sequence, and select the sub-sensors corresponding to the sub-sequences having an association degree greater than an association degree threshold as associable sensors, and select the sub-sensors corresponding to the sub-sequences having an association degree less than the association degree threshold as unassociable sensors;

[0092] a weighted vector determining unit, configured to calculate the mother sequence and the associable sequence of each associable sensor according to an absolute correlation algorithm, so as to obtain a corresponding weighted vector according to the mother sequence and each associable sequence;

[0093] A data processing unit is used to obtain ion concentration data of the first water area collected by the mother sensor and the unassociated sensor, process the ion concentration data of the mother sensor according to the corresponding weighted vector to obtain ion concentration data of the associable sensor associated with the mother sensor, and determine the water quality type based on the ion concentration data of the mother sensor, the ion concentration data of the associable sensor calculated by the mother sensor, and the ion concentration data of the unassociated sensor.

[0094] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification.

[0095] Although the present application is described herein in conjunction with various embodiments, in the process of implementing the claimed application, those skilled in the art can understand and implement other changes to the disclosed embodiments by reviewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple situations. A single processor or other unit can implement several functions listed in the claims. Certain measures are recorded in different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0096] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A method for online monitoring of water quality in coal mine working faces based on deep learning, characterized in that: include: Acquiring several ion concentration data of a first water area collected by several sensors; Inputting the ion concentration data into the first judgment model, the second judgment model, and the third judgment model respectively to obtain a first judgment probability, a second judgment probability, and a third judgment probability, wherein each judgment probability includes at least one target judgment result, and each target judgment result corresponding to each judgment probability has a corresponding weight factor, and the target judgment result is the probability of being judged as a preset water quality; Calculate the target judgment result corresponding to each judgment probability according to the weight factor to obtain a comprehensive target judgment probability, wherein the comprehensive target judgment probability is the comprehensive probability that each target prediction result corresponds to the preset water quality; Selecting the result with the largest comprehensive probability of the preset water quality in the target judgment result as the current water quality type; Acquiring several ion concentration data of a first water area collected by several sensors, further comprising: Selecting one sensor from the plurality of sensors as a mother sensor and the remaining sensors as child sensors, acquiring historical ion concentration data of the mother sensor and each child sensor to obtain a mother sequence and a plurality of child sequences; Performing correlation analysis on the parent sequence and the child sequence, selecting the sub-sensors corresponding to the sub-sequences with correlation greater than a correlation threshold as associable sensors, and selecting the sub-sensors corresponding to the sub-sequences with correlation less than the correlation threshold as unassociable sensors; Calculating the mother sequence and the associable sequence of each associable sensor according to an absolute correlation algorithm to obtain a corresponding weighted vector according to the mother sequence and each associable sequence; Acquire ion concentration data of a first water area collected by a mother sensor and an unassociated sensor, process the ion concentration data of the mother sensor according to a corresponding weighted vector to obtain ion concentration data of an associable sensor associated with the mother sensor, and determine the water quality type based on the ion concentration data of the mother sensor, the ion concentration data of the associable sensor calculated by the mother sensor, and the ion concentration data of the unassociated sensor.

2. The method for online monitoring of water quality in coal mine working faces based on deep learning according to claim 1 is characterized in that: The first judgment model is a BP God General network model, the second judgment model is a Bayesian prediction model, and the third judgment model is a three-line graph model.

3. The method for online monitoring of water quality in coal mine working faces based on deep learning according to claim 1, characterized in that: Also includes: Querying all sensor identifiers in the first water area to read historical data of each ion in a preset time period from a database corresponding to the sensor identifier to obtain historical data of several ion concentrations; The ion concentration at a future preset moment is predicted according to the several types of ion concentration historical data, so as to obtain the water quality type at the future preset moment based on the ion concentration at the future preset moment.

4. The method for online monitoring of water quality in coal mine working faces based on deep learning according to claim 3 is characterized in that: When it is determined that the water quality type at the future preset time is abnormal, the future preset time is marked and an early warning is issued.

5. A coal mine working face water quality online monitoring system based on deep learning, characterized by: include: A data acquisition module, configured to acquire data on concentrations of several ions in a first water area collected by several sensors; a probability judgment module, configured to input the ion concentration data into a first judgment model, a second judgment model, and a third judgment model, respectively, to obtain a first judgment probability, a second judgment probability, and a third judgment probability, wherein each judgment probability includes at least one target judgment result, each target judgment result corresponding to each judgment probability has a corresponding weight factor, and the target judgment result is a probability of being judged as a preset water quality; A probability prediction module is used to calculate the target judgment result corresponding to each judgment probability according to the weight factor to obtain a comprehensive target judgment probability, wherein the comprehensive target judgment probability is the comprehensive probability that each target prediction result corresponds to the preset water quality; A water quality determination module is used to select the result with the largest comprehensive probability of the preset water quality in the target judgment result as the current water quality type; The data acquisition module also includes: a mother sensor selection unit, configured to select one sensor from the plurality of sensors as a mother sensor and the remaining sensors as child sensors, and obtain historical ion concentration data of the mother sensor and each child sensor to obtain a mother sequence and a plurality of child sequences; an association determination unit, configured to perform an association analysis on the parent sequence and the child sequence, and select the sub-sensors corresponding to the sub-sequences having an association degree greater than an association degree threshold as associable sensors, and select the sub-sensors corresponding to the sub-sequences having an association degree less than the association degree threshold as unassociable sensors; a weighted vector determining unit, configured to calculate the mother sequence and the associable sequence of each associable sensor according to an absolute correlation algorithm, so as to obtain a corresponding weighted vector according to the mother sequence and each associable sequence; A data processing unit is used to obtain ion concentration data of the first water area collected by the mother sensor and the unassociated sensor, process the ion concentration data of the mother sensor according to the corresponding weighted vector to obtain ion concentration data of the associable sensor associated with the mother sensor, and determine the water quality type based on the ion concentration data of the mother sensor, the ion concentration data of the associable sensor calculated by the mother sensor, and the ion concentration data of the unassociated sensor.

6. The coal mine working face water quality online monitoring system based on deep learning according to claim 5 is characterized in that: The first judgment model is a BP God General network model, the second judgment model is a Bayesian prediction model, and the third judgment model is a three-line graph model.

7. The coal mine working face water quality online monitoring system based on deep learning according to claim 5 is characterized in that: Also includes: A historical concentration acquisition module is used to query all sensor identifiers in the first water area to read historical data of each ion in a preset time period from a database corresponding to the sensor identifier to obtain historical data of several ion concentrations; The water quality type prediction module is used to predict the ion concentration at a future preset moment based on the historical data of the several ion concentrations, so as to obtain the water quality type at the future preset moment based on the ion concentration at the future preset moment.

8. The coal mine working face water quality online monitoring system based on deep learning according to claim 7 is characterized in that: When it is determined that the water quality type at the future preset time is abnormal, the future preset time is marked and an early warning is issued.

Images