KR-20260067873-A - LSTM-based Public Treatment Facility Management Method

KR20260067873AKR 20260067873 AKR20260067873 AKR 20260067873AKR-20260067873-A

Abstract

One embodiment of the present invention comprises: a sensor unit for measuring influent water in real time; a database unit for storing data collected by the sensor unit; a prediction model generation unit for predicting the total nitrogen concentration after time t based on the data stored in the database unit; a process control unit for determining the next process by comparing the total nitrogen concentration predicted by the prediction model generation unit with a designated reference concentration; and a plasma treatment unit for treating the influent water with low-temperature plasma. The above prediction model generation unit provides a public treatment facility management technique characterized by being a smart prediction model generation unit combined with a deep learning-based LSTM (Long Short-Term Memory) model.

Inventors

김현우
이연아

Assignees

전북대학교산학협력단

Dates

Publication Date: 20260513
Application Date: 20241106

Claims (8)

Sensor unit for measuring inflow water in real time; A database unit that stores data collected by the sensor unit above; A prediction model generation unit that predicts the total nitrogen concentration after time t based on data stored in the above database unit; A process control unit that determines the next process by comparing the total nitrogen concentration predicted by the above prediction model generation unit with a specified reference concentration; and A plasma treatment unit that treats low-temperature plasma in the influent water; comprising a configuration A public treatment facility management technique characterized in that the above prediction model generation unit is a smart prediction model generation unit combined with a deep learning-based LSTM (Long Short-Term Memory) model.
In paragraph 1, the prediction model generation unit is, A data input processing unit that processes time-series data collected from a sensor unit and converts it into LSTM (Long Short-Term Memory)-based training data; An LSTM network component including an LSTM neural network that predicts total nitrogen concentration based on past data and current data; and A public treatment facility management technique characterized by including a model learning and prediction unit that predicts the total nitrogen concentration in real time after future t hours through learning from past data.
In paragraph 2, in the LSTM network configuration part, The above LSTM neural network is composed of an input gate; a forget gate; and an output gate, A public treatment facility management technique characterized by having the ability to learn the time-series dependence on total nitrogen concentration.
In paragraph 1, the sensor unit is, A public treatment facility management method characterized by being composed of a multi-sensor capable of simultaneously measuring pH, SS, TOC, TN, TP, COD, BOD, and total coliform data of influent water in real time.
In paragraph 1, the prediction model generation unit is, A public treatment facility management technique characterized by improving the accuracy of total nitrogen concentration prediction in real time by using pH, SS, TOC, TN, TP, COD, BOD, and total coliform data of influent collected through the sensor unit as training data and updating it in real time.
In paragraph 1, the process control unit is, A public treatment facility management technique characterized by automatically activating a plasma treatment unit to reduce the nitrogen concentration of the influent when the predicted total nitrogen concentration exceeds a standard concentration.
In paragraph 6, the process control unit is, A public treatment facility management method characterized by automatically controlling the operating time and plasma treatment intensity of the plasma treatment unit according to the extent that the total nitrogen concentration predicted by the prediction model generation unit exceeds a specified standard concentration.
In claim 1, the plasma treatment unit is, A public treatment facility management technique characterized by maintaining an activated state until the predicted total nitrogen concentration stabilizes below a reference concentration, thereby continuously treating the influent with low-temperature plasma.

Description

LSTM-based Public Treatment Facility Management Method The present invention relates to a public treatment facility management technique using an LSTM technique, and more specifically, to a total nitrogen management technique that applies an LSTM (Long Short-Term Memory) model among deep learning techniques to protect the performance of public treatment facilities and maximize operational stability. Various monitoring and control systems have been introduced in wastewater treatment facilities to monitor the quality of influent and effluent in real time. These systems primarily monitor water quality parameters such as pH, Chemical Oxygen Demand (COD), Biochemical Oxygen Demand (BOD), Total Nitrogen (TN), and Total Phosphorus (TP), and have controlled treatment processes manually or automatically based on this data. However, most current treatment facilities use a fixed-rule control method based on physical/chemical models, and there are limitations in dynamically adjusting the state of influent and effluent in real time. Existing treatment facilities primarily use a method of collecting data through water quality sensors and issuing warnings or manually adjusting processes when certain standards are exceeded. Although some automated control systems have been introduced, these systems mainly operate based on predefined algorithms and have limitations in that they lack the ability to learn or predict real-time data changes. As a result, treatment facilities are operated in a reactive manner after problems occur, making it difficult to implement proactive response or prediction-based control. In addition, some treatment facilities have attempted to predict changes in water quality using statistical regression analysis techniques, but these regression models suffer from poor prediction accuracy because they fail to adequately account for the temporal dependency of the data. In particular, chemical treatment is carried out when total nitrogen (TN) or total phosphorus (TP) concentrations exceed standard limits, and there are limitations in that it is difficult to take preemptive measures before such abnormal situations occur, and this approach is not cost-effective. In summary, most existing control systems for public treatment facilities respond only after a problem occurs, and there are limitations to proactive response through real-time data analysis and prediction. To overcome these limitations, there is a need for new techniques that can learn long-term data patterns and optimize treatment processes in real time based on them. FIG. 1 is a summary diagram briefly illustrating the operation process of a public treatment facility management technique according to one embodiment of the present invention. Figure 2 is a graph and table visually showing the results of data analysis on the influent parameters of a public treatment facility. Figure 3 is a correlation matrix that visually shows the correlation between water quality factor data of public treatment facilities. Figure 4 is a graph showing data in which a smart prediction model generation unit combined with a deep learning-based LSTM (Long Short-Term Memory) model predicted the total nitrogen concentration in the future through learning from past data. FIG. 5 is a set of equations summarizing evaluation indicators for evaluating LSTM-based prediction data in a public treatment facility management technique according to one embodiment of the present invention. Figure 6 is a graph showing the results of evaluating the prediction performance of LSTM and RNN models according to various evaluation metrics, and a table summarizing the results. The present invention will be described below with reference to the attached drawings. However, the present invention can be implemented in various different forms and is not limited to the embodiments described herein, and should be understood to include all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention. In addition, to clearly explain the invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification have been given similar reference numerals. Throughout the specification, when it is stated that a part is "connected (connected, in contact, joined)" to another part, this includes not only cases where they are "directly connected," but also cases where they are "indirectly connected" with other members in between. Furthermore, when a part such as a layer, film, region, or plate is described as being “on” another part, this includes not only cases where it is “immediately above” the other part, but also cases where there is another part in between. Additionally, in this specification, when a part such as a layer, film, region, or plate is described as being formed “on” another part, the direction in which it is formed is not limited to the upward direction only, but includes cases where it is formed in the