KR101545335B1 - Pump control system and method for rainfall water discharge of reservoir using artificial neural network and fuzzy logic - Google Patents

Abstract

The present invention relates to an artificial-intelligent rain water pump control system in real time. The rain water pump control system comprises: a measurement unit for measuring and collecting a water level of a water reservoir in real time; a data storage unit for storing measurement data received from the measurement unit and previous water level data; a calculation unit for learning water level data stored in the data storage unit through an artificial neural network technique, and predicting a future water level value based on the learned result; and an automatic control unit for controlling an output value of a pump unit in accordance with a fuzzy control logic based on the future water level value.

Description

TECHNICAL FIELD [0001] The present invention relates to an excellent pump control system using an artificial neural network and a fuzzy technique, and more particularly, to an excellent pump control system using an artificial neural network and a fuzzy technique.

The present invention relates to an artificial intelligent superior pump real time control system, and more particularly, to an artificial intelligent superior pump real time control system that predicts an amount of storm generated using an artificial neural network and a fuzzy technique, calculates a discharge amount of a pump, The control system is provided with an artificial intelligent excellent pump capable of effectively controlling the storm.

Due to the recent global climate change, the frequency of local eruptions or intensive rainfalls is increasing and the damage caused by flooding accidents in urban areas is increasing.

As an example, in July 2006, part of the bank of the Anyang Stream near Yangpyeong-dong, Seoul was collapsed due to the heavy rain of Gyeonggi-do members, resulting in enormous damage.

Inundation damages in urban areas are mainly caused by various causes such as lowlands, poor pipelines and flooding, but the flood damage caused by concentration of lowlands in the rain is the highest.

In order to prevent flooding of such low-lying areas, various measures are being taken such as improvement of sewer pipes, expansion of drainage pumps, removal of soil and soil.

However, improvement of sewerage system and removal of soil are merely long-term countermeasures, so it is not appropriate for countermeasures against rapid increase in rainfall due to heavy rain during the summer season. In order to effectively cope with heavy rainfall occurring over several hours, very important.

Anyang River flooding is a representative example of urban flooding. The Anyangcheon watershed is a typical urban river located in Seoul (District 7) and Gyeonggi-do (7:00). It is the national river, the first tributary of the Han River, with a basin area of 281.6㎢ and an extension of 32.2㎞. , About 110 billion won in flood damage, and 30 deaths (deaths).

The succession of urban flooding shows the limit of urban flood damage prevention measures. Even though the urban flood has been maximized in the low economic growth and urban expansion process, As a result.

In order to prevent flooding in the city due to heavy rain, river embankments and water gates should be installed to prevent flooding of rivers and to install stormwater pumping stations to exclude urban stormwater from outside the city.

In recent years, rapid urbanization has increased the number of impervious areas, and the importance of drainage pumping stations to effectively handle overflowing urban areas has been highlighted.

By the end of 2002, 1,028 pumping stations were installed nationwide. Of these, 368 pumping stations with electric capacity of 200kW or more managed by the municipality, and the annual operating cost reached 328 billion won.

In general, the excellent pump is controlled by the interlocking of the water level meter and the pump, and two or more pumps are operated in combination. Inverter control and algebraic control are simultaneously performed.

However, the above-described control method has a disadvantage in that it can not cope with a sudden increase in the inflow amount due to heavy rain or the like in a rapid manner.

That is, in the conventional control method, as shown in FIG. 3, when the amount of abundance increases sharply, the pump discharge amount is not effectively controlled according to the amount of water, and the pump discharge amount is repeatedly increased and decreased repeatedly. It also repeats rapidly between high water level (HWL) and low water level (LWL).

This result is caused by the fact that the existing excellent pump control method controls the pump by measuring the water level only without estimating the inflow water quantity.

Also, in the logarithmic control method, when the change rate of the inflow flow rate is large, the discharge amount control by the drain pump may not be effectively performed and the pump life may be shortened due to frequent pump control.

In order to solve the above-described problems, the present invention predicts a stormwater inflow amount in the next few minutes or several hours by using the measured water level of the reservoir and past rainfall pattern data, and controls the stormwater pump in real time based on the prediction result To provide an excellent pump control system.

Also, the excellent pump control system aims to provide an excellent pump control system that learns past data through a time-and-water pattern analysis technique and predicts a future water level value based on the past data to minimize an error of an artificial neural network learning result.

According to an aspect of the present invention, there is provided an apparatus for measuring a water level of a reservoir, A data storage unit for storing measurement data and past water level data received from the measurement unit; A calculating unit that learns the water level data stored in the data storage unit through an artificial neural network technique and predicts a future water level value based on the learning result; And an automatic control unit for controlling the output value of the pump unit according to the purge control logic based on the future water level value.

According to an embodiment, the calculation unit may classify the past water level data to assign weights, and compare the water level data with the measured water level data to determine a weight for the output value control.

According to an embodiment, the information providing unit may further include an information providing unit for providing rainfall information of the weather station to the calculation unit.

According to another aspect of the present invention, there is provided a method for measuring a water level of a reservoir in real time; Collecting the measured water level value and past water level data; Learning the collected data through a learning artificial neural network technique and predicting future water level values based on the learning results; And controlling the output value of the pump section according to the purge control logic based on the future water level value.

According to one embodiment, the step of predicting the future water level value may include analyzing the past water level data, and classifying the past water level data, and assigning a predetermined weight value to each type of water level data; And comparing the measured water level value with the typed past water level data to determine a corresponding weight value, wherein the output value can be controlled based on the determined weight value.

The excellent pump control system learns past data using an artificial neural network technique and determines the amount of control of the pump unit based on the predicted future water level, So that the service life of the pump is prolonged and the inflow water quantity which changes rapidly can be responded to more quickly.

Also, since the excellent pump control system learns past data through time-and-water pattern analysis technique, the error of the learning result is minimized and the learning time is shortened and the prediction accuracy is improved.

It should be understood, however, that the effects of the present invention are not limited to the above-described effects, but include all effects that can be deduced from the detailed description of the present invention or the configuration of the invention described in the claims.

1 is a block diagram of an excellent pump control system according to an embodiment of the present invention.
2 is a schematic diagram of an artificial neural network structure for learning past water level data.
3 shows the water level and the pump discharge amount by the excellent pump control method using only the existing water level data.
4 is a graph showing a water level and a pump discharge amount according to an excellent pump control method according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described with reference to the accompanying drawings. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and similar parts are denoted by like reference characters throughout the specification.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "indirectly connected" . Also, when an element is referred to as "comprising ", it means that it can include other elements, not excluding other elements unless specifically stated otherwise.

As used herein, the terminology used herein is intended to encompass all commonly used generic terms that may be considered while considering the functionality of the present invention, but this may vary depending upon the intent or circumstance of the skilled artisan, the emergence of new technology, and the like. Also, in certain cases, there may be a term selected arbitrarily by the applicant, in which case the meaning thereof will be described in detail in the description of the corresponding invention. Therefore, the term used in the present invention should be defined based on the meaning of the term, not on the name of a simple term, but on the entire contents of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of an excellent pump control system according to an embodiment of the present invention.

The excellent pump control system includes a metering unit 100 for measuring and collecting the water level of a reservoir in real time; A data storage unit 110 for storing measurement data and past water level data received from the measurement unit 100; A calculation unit 120 for learning water level data stored in the data storage unit 110 through an artificial neural network technique and predicting a future water level value based on the learning result; And an automatic control unit 150 for controlling an output value of the pump unit 160 according to the fuzzy control logic based on the future water level value.

According to another aspect of the present invention, there is provided a method for measuring a water level of a reservoir in real time; Collecting the measured water level value and past water level data; Learning the collected data through a learning artificial neural network technique and predicting future water level values based on the learning results; And controlling an output value of the pump unit 160 according to the fuzzy control logic based on the future water level value.

The measuring unit 100 may include a water level sensor and a rainfall sensor. The water level sensor measures the water level of the liquefied water in real time and transmits it to the data storage unit 110. The rainfall sensor is installed in a specific area of the excellent pump control system to measure the rainfall state in real time, To the storage unit 110.

According to one embodiment, the measurement unit 100 can transmit the measurement value to the data storage unit 110 through a wired or wireless transmission network, and can variously modify a transmission method in consideration of a design specification and a cost .

In addition to the water level sensor or the rainfall sensor, the measuring unit 100 may further include an image acquiring unit such as a video camera or a CCTV for collecting and transmitting image information about a jurisdiction.

The data storage unit 110 may store advance information for controlling an output value of the pump unit 160 and may store a series of measurement data required for controlling the pump unit 160 such as watershed data and stormwater network data. have.

The watershed data may include an area, a width, a ground slope, a Manning roughness coefficient, and the like of a watershed to which the excellent pump control system is subject, and the storm water network data includes a length of the storm drainage, Slope, section size, and other parameters required for pump control.

That is, various data measured by the measurement unit 100 are collected and stored in the data storage unit 110 in real time, and the collected data can be utilized as data for predicting future water level in the future.

The calculation unit 120 may estimate the future water level based on the preset data and the measurement data received from the data storage unit 110. [

In particular, the calculation unit 120 may learn past water level data received from the data storage unit 110 for future water level prediction through an artificial neural network algorithm.

The artificial neural network (neural network) is a mathematical model aiming at expressing some characteristics of brain function by computer simulation.

2 is a schematic diagram of an artificial neural network structure for learning past water level data.

The artificial neural network includes a processing structure as shown in FIG. 2, and a specific computation value can be calculated through a network connected to a channel (data path) having a weight.

The artificial neural network learning algorithm may be a Levenberg-Marquartdt algorithm using a least sqaure curve fitting method, but the present invention is not limited thereto. .

= dependent variableshere,

= dependent variables

= parameter vector

= parameter vector

As a result of learning by the artificial neural network, the calculation unit 120 can calculate the target control amount of the pump unit 160 using the current water level and the weather station rainfall data.

The excellent pump control system may further include an information providing unit 140 for providing the weather station rainfall information.

In addition, the excellent pump control system may further include a communication unit 130 for transferring data between the components.

The communication unit 130 may be connected to the data storage unit 110, the calculation unit 120, the information providing unit 140 and the automatic control unit 150. In addition, Data can be transmitted.

The communication unit 130 may include a wired data transmission network for transmitting various result values, rain data, and level data for controlling the pump unit 160, or may include a separate transmission / reception unit for wireless transmission, It is not particularly limited as long as it is a structure or means for transmission.

The information providing unit 140 may provide the data storage unit 110 or the calculation unit 120 with rainfall information associated with geographical information and geographical information and may transmit the rainfall information to the communication unit 130 and / Can be connected.

The information providing unit 140 may provide rainfall information of a weather station through a communication network such as the Internet, but it is not particularly limited as long as it provides rainfall information.

Among the new data used in the learning algorithm, two sigma (mean + standard deviation 占 2) upper and lower levels can be regarded as outliers and can be filtered out from the list of data for learning. Perform artificial neural network learning using stored operation data.

According to one embodiment, the automatic control unit 150 may control the output value of the pump unit 160 through the fuzzy control logic based on the future water level value calculated by the calculation unit 120.

The pump unit 160 may include a plurality of pumps for adjusting the water level of the liquefied water, and the plurality of pumps may be connected to the purge control logic Can be controlled individually or simultaneously.

That is, the automatic control unit 150 can determine the number of pumps to be started and the inverter control amount based on the following fuzzy logic.

The current output value represents an output value for starting the pump at the programmable logic controller (PLC), and the current water level value represents a value obtained by measuring the water level of the liquefied water supplied from the measurement unit 100. The water level increase rate is a rate of water level increase per unit time And a future water level value calculated by the programmable logic controller, and the future water level value indicates a future water level calculated by the calculation unit 120. [

The automatic control unit 150 determines the amount of output of the pump and the inverter through the purge method and transmits the output to the starter.

However, the parameter for the fuzzy control is not particularly limited as long as it is a factor that can be considered in the control of the output amount of the pump unit 160. [

According to one embodiment, the calculation unit 120 may classify the past water level data to assign weights, compare the type of the past water level data with the measured water level values, and determine a weight for the output value control .

The step of predicting the future water level value may include analyzing the past water level data, and classifying the past water level data, and assigning a predetermined weight value for each type; And comparing the measured water level value with the typed past water level data to determine a corresponding weight value, wherein the output value can be controlled based on the determined weight value.

 In other words, the artificial neural network learning generally performs independent variables for the entire past period, but if the variation of the variation characteristics of the past data greatly fluctuates with time, it may cause an error of the learning result of the artificial neural network. This error can be further increased in accordance with the variation of the concentrated rainfall and the rainfall occurrence period depending on the rainfall.

In particular, the amount of data generated during a single rainfall, such as rainfall and water level data, is vast, and over time, data to be analyzed may become excessive.

In the case of a large amount of data, the time required to learn the neural network necessarily increases, and in a computer in which the performance of a CPU such as a field control device is insufficient, actual control may become impossible due to an excessive load.

Accordingly, in order to minimize the error of the learning of the artificial neural network and shorten the learning time, the calculation unit 120 classifies the water level variation pattern according to the occurrence of rainfall (time-water level pattern analysis technique) .

That is, the calculation unit 120 may compare the water level data collected in real time with the water level data of the most similar water level generation pattern group according to the rainfall, and then use the data corresponding to the most similar water level occurrence pattern for learning.

The water level pattern analysis using the past rainfall data and the water level data can be performed on the operation data of the pumping station in operation or the data collected and accumulated in the data storage unit 110 in real time.

According to one embodiment, the calculation unit 120 may analyze the past water level occurrence pattern according to the following equation.

= 해당 시각에서의 단위시간당 수위변화율here,

= Rate of change in water level per unit time at the corresponding time

= 시간 변화율

= Time rate of change

= 강우발생 후 각 시간대별 수위변화율의 합

= Sum of rate of change of water level for each hour after rainfall

= 빈도

= Frequency

= 한계수심(제어수심)

= Critical water depth (control water depth)

= 기준수심(유효수심)

= Reference depth (effective depth)

는 강우발생 후 각 시간대별 단위 시간당 수위 증가율의 합을 의미한다. 특히, 한계수위(

)/기준수위(

)는 해당 펌프부(160)의 기준수위(유효수심) 대비 한계수심(제어수심)으로써 적용대상 펌프장의 특성 및 제어목적을 고려한 변수이다. remind

Means the sum of the rate of increase in water level per unit time per hour after rainfall. In particular,

) / Reference water level (

(Control depth) with respect to the reference level (effective depth) of the pump unit 160 is a parameter considering the characteristics of the pump station to be applied and the control purpose.

According to one embodiment, the calculator 120 may classify the past water level generation patterns into five water level patterns as shown in Table 1, and then store the results of the artificial neural network learning for each of the clusters.

pattern
(association)

(m/min)Classification criteria

(m / min) Fuzzy control weight Water level increase rate One > 5 1.0 Largest 2 4 to 5 0.8 greatness 3 3 to 4 0.6 middle 4 2-3 0.4 production 5 <2 0.2 Smallest

Table 1 shows the pattern of past water level change patterns in the water level pattern analysis according to an embodiment. The pattern 1 corresponds to the case where the rate of change of the water level per unit time is the greatest, and the fuzzy control weight is set to the highest level, so that the water level can cope with the sudden rise of the water level.

In the pattern 5, when the water level change rate per unit time is the lowest, the fuzzy control weight is set to a low value, so that a minimum water level for controlling the pump unit 160 can be secured.

4 is a graph showing a water level and a pump discharge amount according to an excellent pump control method according to an embodiment of the present invention.

According to FIG. 4, when the drainage water level control is performed according to the excellent pump control method, the change rate of the water level of the reservoir is drastically reduced as compared with the existing method based on the hydraulic model (FIG. 3) It can be seen that

In addition, the discharge amount of the drain pump controlled by the excellent pump control system shows a less variation in the discharge amount as compared with the conventional control method.

That is, according to the excellent pump control method, it is possible to control the water level of the reservoir stably in case of a sudden increase in the amount of water, minimize the life of the pump by operating the drain pump constantly, Can be prevented.

It will be understood by those skilled in the art that the foregoing description of the present invention is for illustrative purposes only and that those of ordinary skill in the art can readily understand that various changes and modifications may be made without departing from the spirit or essential characteristics of the present invention. will be. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

The scope of the present invention is defined by the appended claims, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included within the scope of the present invention.

Claims (5)

A measuring unit for measuring and collecting water level and rainfall of a reservoir in real time;
A data storage unit for storing measurement data and past water level data received from the measurement unit;
A calculation unit that learns the collected rainfall and water level data according to the Levenberg-Marquartdt algorithm and predicts a future water level value based on the learning result; And
And an automatic control unit for controlling an output value of the pump unit according to a fuzzy control logic based on a plurality of independent variables including a current output value, a current water level, a water level increase rate, and a future water level value,
Wherein the calculation unit classifies the past water level data according to the following equation to assign weights, and compares the typed past water level data with the measured water level value to determine a weight for the output value control.
[Mathematical Expression]

here,

The water level change rate per unit time at that time,

A time change rate,

Is the sum of the water level change rates for each hour after rainfall,

However,

Is the critical depth (control depth)

Means the reference depth (effective depth).
delete The method according to claim 1,
And an information providing unit for providing rainfall information of the weather station to the calculation unit. Measuring and collecting water level and rainfall in real time;
Storing the collected measurement data and past water level data;
Learning the collected data according to the Levenberg-Marquartdt algorithm and predicting a future water level value based on the learning result; And
Controlling an output value of the pump unit according to a fuzzy control logic based on a plurality of independent variables including a current output value, a current water level, a water level increase rate, and a future water level value,
The step of predicting the future water level value may include analyzing the past water level data and classifying the past water level data according to the following equation and assigning a predetermined weight to each type: And comparing the measured water level value with the typed historical water level data to determine a corresponding weight value, wherein the output value is controlled based on the determined weight.
[Mathematical Expression]

here,

The water level change rate per unit time at that time,

A time change rate,

Is the sum of the water level change rates for each hour after rainfall,

However,

Is the critical depth (control depth)

Means the reference depth (effective depth). delete

Images