KR20100124511A - Apparatus and method for predicting the production of disinfection by-product using a artificial neural network - Google Patents

Abstract

PURPOSE: An apparatus and method for predicting by-product after disinfection using an artificial neural is provided to select optimal artificial neural network. CONSTITUTION: A method for predicting generation of by-product using artificial neural network comprises: a data collection part(110) for collecting process factors and water quality factors; an artificial neural net model part(120) for applying one or more artificial net model(122); a feed back part(130) for predicting the by-product after giving feed back to select optimal artificial neural network.

Description

APAPATUS AND METHOD FOR PREDICTING THE PRODUCTION OF DISINFECTION BY-PRODUCT USING A ARTIFICIAL NEURAL NETWORK}

The present invention relates to an apparatus and method for predicting the production result of disinfection by-products using an artificial neural network, and more particularly, to an error in predicted and actual values of disinfection by-products (for example, trihalomethane) in a water treatment system. Generation results of disinfection byproducts using artificial neural networks to predict the results of disinfection byproducts (for example, concentration and generation) by selecting an optimal neural network model according to the nonlinear form according to the comparison result between A prediction apparatus, a method thereof, and a computer-readable recording medium having recorded thereon a program for realizing the method.

In general, water treatment plants disinfect raw water and purified water to inactivate pathogenic microorganisms that can cause waterborne diseases such as viruses, bacteria, protozoa, and the like. In the water treatment plant, 'chlorine' is the most used disinfectant because of its excellent sterilization, residual and economical efficiency. In other words, the disinfection method using chlorine provides effects such as sterilization of pathogenic microorganisms causing waterborne diseases, removal of ammonia nitrogen, killing algae, removal of iron and manganese, and maintenance of residual chlorine in the water supply line. Suppresses the regrowth of bacteria.

However, the disinfection method using chlorine disinfects by-products (for example, humic substances) that are harmful to human body by addition and substitution reaction with bromine in chlorine or seawater used as disinfectant (Disinfection By-Products, DBPs). Is generated. In particular, chlorinated disinfection by-products include trihalomethanes (THMs), and trihalomethanes are compounds in which hydrogen atoms of methane are replaced by halogen atoms (mainly chlorine, bromine, iodine). (CHCl ₃ ), bromodichloromethane (CHBrCl ₂ ), dibromochloromethane (CHBr ₂ Cl), bromoform (CHBr ₃ ), and the like. These trihalomethanes are affected by chlorine concentration, temperature, reaction time, bromide, iodine concentration, precursor concentration and properties.

In general, trihalomethane is not detected in the source water or is present at very low concentrations, but shows almost 100% detection frequency in the final treated water treated with chlorine.

On the other hand, conventionally, by changing the raw water storage and purified water treatment method, or by changing the chlorine injection point such as intermediate chlorine treatment, a study for reducing the trihalomethane in the final treated water treated with chlorine has been progressed. In particular, when the water quality of the raw water is deteriorated, there is a complex disinfection method using a high-purity water treatment facility or using chlorine dioxide or ozone in combination with chlorine during water purification treatment.

In addition, conventionally, a method of removing trihalomethane using an adsorbent such as activated carbon or reducing the possibility of generating trihalomethane through pretreatment with ozone has been proposed. Here, the method using activated carbon is known to improve the biodegradability of organic matter after ozone treatment, and the method through pretreatment with ozone is not only effective to stabilize the organic matter to a certain level, but also to improve the biological resolution of organic matter. . In addition, the change in activated carbon adsorption performance in the method using activated carbon may vary depending on the water quality to be treated, ozone injection concentration and activated carbon.

Recently, a prediction model to reduce the use of disinfectants by predicting the trihalomethane generation result in the disinfection process has been proposed. The prediction model includes a linear regression analysis such as empirical and kinetic-based models.

Such a conventional prediction model has a limitation in that the prediction performance is low due to a large number of errors in a site having a high probability of variability such as a state in which chlorine is injected very high or very low, or a condition of a network of transmission and distribution networks.

In particular, the conventional predictive model is a result of the generation of trihalomethane generated through a compound disinfection process such as ozone and chlorine combined disinfection, chlorine dioxide and chlorine combined disinfection, AOP (Advanced Oxidation Process) and chlorine combined disinfection technology. For example, it is difficult to predict the generation concentration, the generation amount, and the like. This is because, unlike the single sterilization process, the complex sterilization process exhibits a non-linear characteristic that forms a relationship with a very complex reaction. Therefore, it is difficult to predict the non-linear characteristic using a conventional linear method.

Therefore, in order to predict disinfection by-products such as trihalomethane generated through a single disinfection process as well as a single disinfection process, a prediction model that can explain the nonlinear characteristics of complex reactions and relationships needs to be proposed.

Therefore, the prior art as described above has a problem that it is difficult to predict the results of disinfection by-products because the composite disinfection process exhibits a non-linear characteristic, it is a problem of the present invention to solve these problems.

Therefore, the present invention selects the optimal neural network model according to the non-linear type according to the result of comparing the error value and the threshold for the predicted value and the actual value of the disinfection by-product (for example, trihalomethane, etc.) in the water treatment system. A computer-readable recording apparatus for generating a result of disinfection by-products using an artificial neural network and a method thereof, and a program for realizing the method, for predicting the generation result (e.g., concentration and amount of generation) in advance. The purpose is to provide a recording medium.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention which are not mentioned can be understood by the following description, and will be more clearly understood by the embodiments of the present invention. Also, it will be readily appreciated that the objects and advantages of the present invention may be realized by the means and combinations thereof indicated in the claims.

The present invention for achieving the above object, the data collection unit for collecting the process factor and water quality factor of the disinfection process; An artificial neural network model unit configured to receive the process factor and the water quality factor, and apply any one of at least one artificial neural network model object previously provided as an artificial neural network model; And selecting an optimal neural network model based on a result of comparing an error between a predicted value of disinfection by-product output through the artificial neural network model and an actual value of disinfection by-product actually confirmed in a water treatment system and a preset threshold. It includes a feedback unit for predicting the generation result of the disinfection by-product after performing the feedback for.

In addition, the data collection unit in the present invention, characterized in that to build a database by collecting the process factor and the water quality factor in a water treatment system.

In addition, in the present invention, the artificial neural network model unit, when applying the artificial neural network model, characterized in that using a multi-layer perceptron for training data in a multi-layer structure having a hidden layer between the input layer and the output layer.

In addition, the artificial neural network model unit in the present invention, characterized in that for adjusting the connection strength of the artificial neural network model in the direction of reducing the error of the prediction value and the actual value.

In addition, in the present invention, if the error between the predicted value and the actual value is less than a predetermined threshold, the feedback unit is to select the artificial neural network model as the optimal artificial neural network model considering the case where the prediction value and the actual value match It is characterized by.

In addition, the feedback unit in the present invention, if the error between the prediction value and the actual value is more than a predetermined threshold value, characterized in that to apply another artificial neural network model in addition to the artificial neural network model.

In addition, the present invention further includes a user interface unit connected to a user terminal equipped with a disinfection by-product prediction program to provide a user environment for controlling the disinfection by-product prediction process.

In addition, the production result of the disinfection by-product in the present invention is characterized in that the concentration or the amount of occurrence of the disinfection by-product.

On the other hand, the present invention is a collection step of collecting process factors and water quality factors of the disinfection process; Receiving an input of the process factor and the water quality factor, and applying one of at least one artificial neural network model object provided in advance to an artificial neural network model; A method for selecting an optimal neural network model based on a result of comparing an error between 'prediction values of disinfection by-products output through the artificial neural network model' and 'actual values of disinfection by-products actually confirmed in a water treatment system' and a preset threshold value Performing a feedback step; And a prediction step of predicting the generation result of the disinfection by-product using the selected optimal neural network model.

In the present invention, the application step is characterized in that when applying the artificial neural network model, using a multi-layer perceptron for training data in a multi-layer structure having a hidden layer between the input layer and the output layer.

In the present invention, the applying step, characterized in that for adjusting the connection strength of the artificial neural network model in the direction of reducing the error of the prediction value and the actual value.

Further, in the present invention, if the error between the predicted value and the actual value is less than or equal to a predetermined threshold value, the artificial neural network model is selected as the optimal artificial neural network model by considering that the predicted value and the actual value match. Characterized in that.

The performing of the present invention may include reapplying another artificial neural network model in addition to the artificial neural network model if an error between the predicted value and the actual value is greater than or equal to a preset threshold.

On the other hand, the present invention is a user terminal having a processor, the function of collecting the process factor and water quality factor of the disinfection process; A function of receiving the process factor and the water quality factor, and applying any one of at least one artificial neural network model object provided as an artificial neural network model; A method for selecting an optimal neural network model based on a result of comparing an error between 'prediction values of disinfection by-products output through the artificial neural network model' and 'actual values of disinfection by-products actually confirmed in a water treatment system' and a preset threshold value The ability to perform feedback; And a computer-readable recording medium having recorded thereon a program for realizing a function of predicting a result of disinfection by-product using the selected optimal neural network model.

As described above, the present invention selects and disinfects an optimal neural network model according to a non-linear form according to a comparison result of an error and a threshold of a predicted value and an actual value of a disinfection by-product (for example, trihalomethane) in a water treatment system. There is an effect that can predict the production of by-products in advance.

In addition, the present invention by applying an artificial neural network model that can explain the non-linear characteristics of complex reactions and relationships, it is possible to predict the disinfection by-products such as trihalomethane generated through the compound disinfection process as well as single disinfection process There is.

In addition, the present invention has the effect of reducing the disinfectant by controlling the amount of disinfectant injection by using the result of the production of antiseptic by-products (for example, trihalomethane) accurately predicted by the artificial neural network.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings, It can be easily carried out. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The present invention provides various kinds of disinfection by-products [eg, trihalomethane (THMs), haloacetonitrile (HANs), haloacetic acid (HAAs), halo-ketones, haloacetates (Halo-acetate). acetates, halo-aldehydes, halo-aromatics, halo-amines, halo-ethers, chloralates, etc.) Although not interpreted, the present invention will be described in detail by applying 'trihalomethane (THMs)', which is one of the representative disinfection by-products.

1 is a configuration diagram of an embodiment of the apparatus for predicting the production result of disinfection by-product using the artificial neural network according to the present invention.

As shown in FIG. 1, the apparatus for predicting the generation result of disinfection by-products using the artificial neural network according to the present invention (hereinafter referred to as “the sterilization by-product prediction apparatus”) is a disinfectant (ie, chlorine) using an artificial neural network according to a non-linear type. The anticipated results of disinfection by-products (ie trihalomethane) (eg, concentration, amount, etc.) that can occur according to the amount of Through this, the water treatment system predicts the result of trihalomethane generated by the reaction of influent water and disinfectant in a single disinfection process or multiple disinfection process in advance, thereby minimizing disinfection byproducts while maintaining disinfection effect of the disinfection process. The optimal disinfectant dosage can be determined. That is, the antiseptic by-product predictor predicts the trihalomethane generation result according to the disinfectant type, the disinfectant combination, and the disinfectant injection amount in the single disinfection process or the multiple disinfection process.

In general, the compound disinfection process first proceeds with a disinfection process using ozone or chlorine dioxide, and then performs a second chlorine disinfection process. Therefore, in the case of the compound disinfection process, the disinfection by-product predicting device proceeds with a disinfection process using ozone or chlorine dioxide (not shown in FIG. 1) first, and then proceeds with a chlorine disinfection process by a second process. Predict the result. Here, since the first disinfection process corresponds to a process that can be easily understood by those skilled in the art, a detailed description thereof will be omitted. In this way, the antiseptic by-product prediction device can predict the generation of trihalomethane for each disinfectant combination even in the compound disinfection process, thereby minimizing the generation of trihalomethane by adjusting the disinfectant injection amount according to the disinfectant combination. It can provide an optimal state.

To this end, the antiseptic by-product predictor is provided with at least one 'artificial neural network target model' for selecting an optimal artificial neural network model. Here, the artificial neural network target model may be set differently for each process and water quality factors included in the input conditions, but is preferably the same process for other factors other than the disinfectant injection amount because it is a process of selecting an optimal artificial neural network model in one water treatment system. Keep it.

First, the antiseptic by-product prediction device is applied to each model of the artificial neural network, and predicts the trihalomethane generation result (hereinafter referred to as "prediction value") and the measurement of the trihalomethane actually confirmed in the water treatment system. The error, which is the difference between the result (hereinafter referred to as "actual value"), is compared with a predetermined threshold (for example, 1.2). In this case, the disinfection by-product prediction apparatus repeats the above-described process by selecting and applying another artificial neural network target model provided in advance when the error by the predicted value and the actual value is greater than or equal to the threshold value. On the other hand, the apparatus for predicting disinfection by-products may select the currently applied artificial neural network target model as an optimal artificial neural network model when the error due to the predicted value and the actual value is less than or equal to the threshold.

In contrast, the antiseptic by-product predictor calculates the error based on the predicted value and the actual value by applying all the neural network target models collectively. Thereafter, the apparatus for predicting disinfection by-products may compare each calculated error with a threshold value and select an artificial neural network target model having a minimum error while the corresponding error is less than or equal to the threshold value as an optimal artificial neural network model.

Here, the apparatus for predicting disinfection by-products is preferably performed again after selecting an optimal neural network model after selecting the optimal neural network model as described above. This is because the actual value actually measured in the water treatment system may vary, so that the optimal artificial neural network model can be selected in consideration of the error due to the predicted value and the actual value accordingly.

On the other hand, the disinfection by-product prediction device includes a data collection unit 110, artificial neural network model unit 120, feedback unit 130.

The data collection unit 110 collects the processes and water quality factors necessary for predicting any disinfection by-products using an artificial neural network in the water treatment system to build a database (not shown in FIG. 1). Here, the process factors include flow rate, reaction tank size, water purification distance, disinfectant type and injection amount, facility specifications, drip wall condition, and water quality factors include acidity (pH), temperature, dissolved organic carbon (DOC), and absorbance. Value (UV), reaction time and residual chlorine. These processes and water quality factors are factors that affect the generation of any disinfection by-products, may be collected in a predetermined cycle or real time in the water treatment system, or may be collected by a user's operation.

The artificial neural network model unit 120 applies the artificial neural network model object to predict the generation result of a specific disinfectant by-product (that is, trihalo methane). That is, the artificial neural network model unit 120 applies an artificial neural network model target necessary for generating trihalomethane to be predicted by the data collecting unit 110, and an artificial neural network model capable of predicting the trihalomethane generation result. Build it. In this case, the artificial neural network model unit 120 performs mechanical learning using a multilayer perceptron that trains data in a multilayer structure having a hidden layer (intermediate layer) between the input layer and the output layer.

Specifically, the artificial neural network model unit 120 is a model input unit 121 for allowing the process and water quality factors belonging to the artificial neural network model target required when generating trihalomethane from the data collection unit 110 can be selected and input. ), The tree predicted by the artificial neural network model 122, the artificial neural network model 122, which is constructed with an artificial neural network capable of predicting the trihalomethane generation result using the process and water quality factors input from the model input unit 121 And a model output unit 123 for outputting a generation result (ie, a predicted value) of halomethane. As described above, the artificial neural network model 122 includes an input layer, a hidden layer, and an output layer, and the input layer is connected to the model input unit 121, and the output layer is connected to the model output unit 123. Here, the artificial neural network model unit 120 updates the number of hidden layers and the connection weight of the artificial neural network model 122 in a direction of reducing an error that is a difference between the predicted value and the error value. For example, when fixing the number of hidden layers, in order to adjust the connection strength, the connection strength is adjusted in the direction of reducing the error, the back layer is propagated in the upper layer, and the lower layer is used to adjust the connection strength again based on this. Exit (see Figure 2 below).

The feedback unit 130 calculates an error, which is a difference between the predicted value output through the artificial neural network model unit 120 and the actual value actually confirmed by the water treatment system, compares it with a threshold, and currently applies the artificial neural network according to the comparison result. Feedback whether the model object is applied to the artificial neural network model unit 120 or another artificial neural network model object to the artificial neural network model unit 120 is fed back.

Specifically, the feedback unit 130 determines whether to apply the artificial neural network model target according to the error calculation unit 131 for calculating the error between the predicted value and the actual value, the error and the threshold value, the artificial neural network model 122 And a feedback control unit 132 for updating the number of hidden layers and connection strength. At this time, if the error is less than the preset threshold, the feedback control unit 132 does not learn the artificial neural network model 122 and does not learn the artificial neural network model target, considering that the predicted value and the actual value match. Judging by the optimal artificial neural network model. On the other hand, the feedback control unit 132 applies another artificial neural network model so that learning about the artificial neural network model 122 can be performed again if the corresponding error is greater than or equal to a preset threshold.

In addition, the disinfection byproduct prediction apparatus is provided with a user interface unit (not shown in FIG. 1) which is connected to a user terminal equipped with a disinfection byproduct prediction program to provide a user environment for controlling the disinfection byproduct prediction process.

On the other hand, the antiseptic by-product predictor predicts the trihalomethane generation result by applying the optimal artificial neural network model as described above. That is, the antiseptic by-product predictor predicts the prediction value when the optimal artificial neural network model is applied to the artificial neural network model 122 as a result of generating trihalomethane.

In particular, the apparatus for predicting disinfection by-products according to the present invention applies an artificial neural network according to a non-linear form, and continuously adjusts and compares predicted variables with water quality factors as in the case of applying an artificial neural network according to a linear equation. It does not apply to limitation. That is, the disinfection by-product predictor may apply the predictor as well as the process factor to the water quality factor.

2 is a diagram illustrating an embodiment of an artificial neural network model applied to the present invention.

As shown in FIG. 2, the artificial neural network model 122 applied to the present invention applies a neural network structure and a learning algorithm. Here, the case where the structure of the multilayer perceptron is applied will be described.

As described above, the artificial neural network model 122 of the present invention is made through mechanical learning using a multilayer perceptron that includes an algorithm for training data in a multilayer structure having a hidden layer between an input layer and an output layer. Accordingly, the neural network model 122 includes an input layer 210, a hidden layer 220, and an output layer 230. Here, the input layer 210 may adjust at least one or more input nodes, the hidden layer 220 has a hidden node for calculating the artificial neural network model 122 according to the number of input nodes, the output layer 230 is an input condition According to the operation of the hidden layer 220 has at least one output result. Preferably, the number of hidden layers 220 is set in advance.

In the input layer 210, a predetermined input condition is input from the model input unit 121, and the output layer 230 outputs the classified results to the model output unit 123. That is, in the case of outputting trihalomethane as a disinfection by-product from the model output unit 123, the disinfectant injection amount and acidity (pH) corresponding to factors influencing the generation of trihalomethane as predetermined input conditions , Reaction time, dissolved organic carbon, temperature, residual chlorine, and absorbance values (UV) are provided. The hidden layer 220 may have at least one layer between the input layer 210 and the output layer 230.

The artificial neural network model 122, when an input condition is input to the input layer 210, the operation result performed by the node in the hidden layer 220 becomes the input value of the node at the next level, and this process is referred to as the output layer ( 230) to get the final result. At this time, it is the connection strength that serves to connect the nodes in each floor. This connection cannot connect nodes on the same floor, but can connect nodes on different floors. Typically, one node is connected to all nodes on the next floor.

In the artificial neural network model 122, each node may be modeled as an artificial neuron. Each neuron sums input external stimuli and reacts accordingly. This can be expressed as Equation 1 below, and a weighted sum of inputs is transferred to an activation function (or transfer function) together with a bias to output a result.

Here, net _j is a weighted sum of inputs (x ₁ , x ₂ ,..., X _p ), b _j represents a bias, and y _j represents an output result. w _ji represents the strength of the connection between the i th input and the j th neuron, and f (*) _j is the activation function of the j th neuron. In particular, f (*) _j is a nonlinear function, and a hard limiter, a threshold logic, a sigmoid function, and the like may be applied, but the sigmoid function will be applied here. This is to apply the sigmoid function, which is a nonlinear function, so that the crystal region is bounded by a gentle curve rather than a normal straight line, so that the hidden layer 230 can be learned through the derivative although the analysis of the behavior is slightly complicated. .

On the other hand, the artificial neural network model 122 compares the predicted value and the actual value to adjust the connection strength between nodes in the direction of the small error. In this case, the artificial neural network model 122 may apply a delta rule. That is, in the artificial neural network model 122, when an input condition is input to each node of the input layer 210, it is converted at each node and transferred to the hidden layer 220, and finally, the output layer 230 outputs a predicted value. In addition, the artificial neural network model 122 adjusts the connection strength in the direction of reducing the difference by comparing the prediction value and the actual value, and back propagation from the upper layer to the lower layer to adjust the connection strength of the magnetic layer based on this again. .

Specifically, the artificial neural network model 122 adjusts the connection strength from node i to node j when the p-th input pattern and the output pattern are presented using Equation 2 below.

Here, t _pj is the j component of the p-th target pattern, o _pj is the j component of the output pattern calculated by the artificial neural network model 122 in the p-th input pattern, i _pj is the i component of the p-th input pattern, δ _pj = t _pj -o _pj represents the difference (error) between the target pattern and the actual pattern.

In addition, in the neural network model 122, an input pattern and a desired output pattern may be presented. As described above, the input pattern given to the input layer 210 is propagated to the output layer 230, that is, the output pattern changed (ie, Predicted value) is compared with a target pattern (ie, an actual value). In this case, in the artificial neural network model 122, learning does not occur when the output pattern matches the target pattern. On the other hand, in the artificial neural network model 122, when the output pattern does not match the target pattern, as described above, by adjusting the connection strength between nodes in the direction to reduce the difference between the output pattern and the target pattern to learn.

Figure 3 is an embodiment flow diagram for a method for predicting the production result of disinfection by-product using the artificial neural network according to the present invention.

First, the disinfection by-product prediction device collects process factors and water quality factors of the disinfection process (S301).

Subsequently, the disinfection byproduct prediction apparatus receives the process factor and the water quality factor, and applies any one of at least one artificial neural network model object provided in advance to the artificial neural network model (S302).

Then, the antiseptic by-product prediction device selects an optimal neural network model based on a comparison of the error between 'prediction output through artificial neural network model' and 'actual value actually confirmed in water treatment system' and preset threshold. To perform the feedback (S303).

At this time, the antiseptic by-product prediction apparatus predicts the generation result of the disinfection by-product using the optimal artificial neural network model selected as described above (S304).

4 is a view showing a disinfection by-product prediction program according to the present invention, Figure 5 is a graph showing the correlation between the prediction value and the actual value by the disinfection by-product prediction program of FIG.

Here, in order to construct an artificial neural network, in the input layer, the disinfectant injection amount (that is, the chlorine injection amount) is 5-10 mg / L, the acidity is pH 5-10, the reaction time is 0.1-168 hours (hour), and the temperature is 10-25. The input conditions are set at 0.00 DEG C, an absorbance value (UV) of 0.002 to 0.021abs./cm, and dissolved organic carbon (DOC) of 0.244 to 1.21 mg / L. As described above, since the error of the predicted value and the actual value is directly proportional to the threshold value, the disinfection by-product predicting apparatus of the present invention converges so that the predicted value and the actual value have no error, that is, the predicted value and the actual value coincide.

Through this, the disinfection by-product predictor can predict the process that can minimize the trihalomethane generation due to the compound disinfection process, it is possible to reduce the disinfectant injection amount.

On the other hand, the method of the present invention as described above can be written in a computer program. And the code and code segments constituting the program can be easily inferred by a computer programmer in the art. In addition, the written program is stored in a computer-readable recording medium (information storage medium), and read and executed by a computer to implement the method of the present invention. The recording medium may include any type of computer readable recording medium.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the technical spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited by the drawings.

1 is a configuration diagram of an embodiment of the apparatus for predicting the production result of disinfection by-product using an artificial neural network according to the present invention;

2 is a diagram illustrating an embodiment of an artificial neural network model applied to the present invention;

Figure 3 is an embodiment flow diagram for a method for predicting the production result of disinfection by-product using the artificial neural network according to the present invention,

4 is a view showing a disinfection by-product prediction program according to the present invention,

5 is a graph showing the correlation between the predicted value and the actual value by the antiseptic by-product prediction program of FIG.

* Explanation of symbols for the main parts of the drawings

110: data collection unit

120: artificial neural network model unit

121: model input unit

122: artificial neural network model

123: model output unit

130: feedback unit

131: error calculation unit

132: feedback control unit

Claims (23)

A data collection unit for collecting process and water quality factors of the disinfection process; An artificial neural network model unit configured to receive the process factor and the water quality factor, and apply any one of at least one artificial neural network model object previously provided as an artificial neural network model; And A method for selecting an optimal neural network model according to a comparison result between an error of 'prediction values of disinfection by-products outputted through the artificial neural network model' and 'actual values of disinfection by-products actually confirmed in a water treatment system' and a predetermined threshold value After the feedback includes a feedback unit for predicting the production of disinfection by-products An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 1, The data collection unit, Collecting the process factors and water quality factors in the water treatment system to build a database An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 1, The artificial neural network model unit, When applying the artificial neural network model, using a multi-layered perceptron to train data in a multi-layered structure having a hidden layer between the input layer and the output layer An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 3, wherein The artificial neural network model unit, Adjusting the connection strength of the artificial neural network model in the direction of reducing the error between the prediction value and the actual value An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 1, The feedback unit, When the error between the predicted value and the actual value is less than or equal to a predetermined threshold value, the artificial neural network model is selected as an optimal artificial neural network model by considering that the predicted value and the actual value match. An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 1, The feedback unit, If the error between the predicted value and the actual value is greater than or equal to a preset threshold, another artificial neural network model is applied again in addition to the artificial neural network model. An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 1, The disinfection byproduct prediction program is connected to a user terminal equipped with a user interface for providing a user environment that can control the disinfection byproduct prediction process further comprises: An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 1, The disinfection step, Firstly, disinfection process using ozone or chlorine dioxide and secondary disinfection process using chlorine. An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 1, The process factors are flow rate, reaction tank size, water purification distance, disinfectant type and injection amount, facility specifications, drip wall conditions, The water quality factor is acidity, temperature, dissolved organic carbon, absorbance value, reaction time, residual chlorine An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 9, The artificial neural network model object, As it is provided to select the optimal neural network model in one water treatment system, it is characterized in that the same for the other factors other than the disinfectant injection amount in the process factor and the water quality factor. An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 1, The disinfection byproduct, Any one of trihalomethane, haloacetonitrile, haloacetic acid, haloketone, haloacetate, haloaldehyde, haloaromatic, haloamine, haloether, and chloral chelate An apparatus for predicting the production result of disinfection by-product using artificial neural network. The method of claim 1, The production result of the disinfection by-product, The concentration or the generation amount of the disinfection by-product An apparatus for predicting the production result of disinfection by-product using artificial neural network. A collection step of collecting process and water quality factors of the disinfection process; Receiving an input of the process factor and the water quality factor, and applying one of at least one artificial neural network model object provided in advance to an artificial neural network model; A method for selecting an optimal neural network model based on a result of comparing an error between 'prediction values of disinfection by-products output through the artificial neural network model' and 'actual values of disinfection by-products actually confirmed in a water treatment system' and a preset threshold value Performing a feedback step; And A prediction step of predicting a result of disinfection by-product using the selected optimal neural network model; Method of Predicting the Result of Disinfection Byproduct Using Artificial Neural Network. The method of claim 13, The applying step, When applying the artificial neural network model, using a multi-layered perceptron to train data in a multi-layered structure having a hidden layer between the input layer and the output layer Method of Predicting the Result of Disinfection Byproduct Using Artificial Neural Network. The method of claim 14, The applying step, Adjusting the connection strength of the artificial neural network model in the direction of reducing the error between the prediction value and the actual value Method of Predicting the Result of Disinfection Byproduct Using Artificial Neural Network. The method of claim 13, The performing step, When the error between the predicted value and the actual value is less than or equal to a predetermined threshold value, the artificial neural network model is selected as an optimal artificial neural network model by considering that the predicted value and the actual value match. Method of Predicting the Result of Disinfection Byproduct Using Artificial Neural Network. The method of claim 13, The performing step, If the error between the predicted value and the actual value is greater than or equal to a preset threshold, another artificial neural network model is applied again in addition to the artificial neural network model. Method of Predicting the Result of Disinfection Byproduct Using Artificial Neural Network. The method of claim 13, The disinfection step, Firstly, disinfection process using ozone or chlorine dioxide and secondary disinfection process using chlorine. Method of Predicting the Result of Disinfection Byproduct Using Artificial Neural Network. The method of claim 13, The process factors are flow rate, reaction tank size, water purification distance, disinfectant type and injection amount, facility specifications, drip wall conditions, The water quality factor is acidity, temperature, dissolved organic carbon, absorbance value, reaction time, residual chlorine Method of Predicting the Result of Disinfection Byproduct Using Artificial Neural Network. The method of claim 19, The artificial neural network model object, As it is provided to select the optimal neural network model in one water treatment system, it is characterized in that the same for the other factors other than the disinfectant injection amount in the process factor and the water quality factor. Method of Predicting the Result of Disinfection Byproduct Using Artificial Neural Network. The method of claim 13, The disinfection byproduct, Any one of trihalomethane, haloacetonitrile, haloacetic acid, haloketone, haloacetate, haloaldehyde, haloaromatic, haloamine, haloether, and chloral chelate Method of Predicting the Result of Disinfection Byproduct Using Artificial Neural Network. The method of claim 13, The production result of the disinfection by-product, The concentration or the generation amount of the disinfection by-product Method of Predicting the Result of Disinfection Byproduct Using Artificial Neural Network. In a user terminal having a processor, Collecting process and water quality factors in the disinfection process; A function of receiving the process factor and the water quality factor, and applying any one of at least one artificial neural network model object provided as an artificial neural network model; A method for selecting an optimal neural network model based on a result of comparing an error between 'prediction values of disinfection by-products output through the artificial neural network model' and 'actual values of disinfection by-products actually confirmed in a water treatment system' and a preset threshold value The ability to perform feedback; And A function of predicting the result of disinfection by-product using the selected optimal neural network model A computer-readable recording medium having recorded thereon a program for realizing this.

Images