KR20100104118A - Method for predicting water demand using group method of data handling algorithm - Google Patents

Abstract

PURPOSE: A method for predicting a water demand using a GMDH(Group Method of Data Handling) algorithm is provided to calculate optimum prediction demand amount through the GMDH algorithm, thereby accomplishing precision goal of output prediction. CONSTITUTION: A water operating system drives an intelligent model of water demand prediction. Demand prediction software is connected through a LAN network such as host computer of the water system and database and other related systems. The demand prediction software established a demand prediction GDMH model through a back-propagation learning algorithm according to an operation data of the water operating system and a data of forecast DB. A predicted demand is determined to current demand and forecast state through model-based prediction control algorithm from the intelligent model.

Description

Method for predicting water demand using Group Method of Data Handling algorithm

The present invention relates to a water demand prediction method using the GM H algorithm.

In general, water supply facilities are used by all people living in urban life and are important facilities to maintain daily activities and daily activities of the city.The lifeline corresponds to the population concentration of the city, improvement of living standards, and economic growth. line), it is required to stably supply safe and high quality water.

To maintain this, the water management operation (hereinafter referred to as "water operation"), which is a comprehensive maintenance plan for waterworks based on drainage control, remains an important task.

The system supporting water operation is to control and monitor all the water facilities from the water purification plant to the consumer to efficiently supply safe water in terms of quantity, water pressure and water quality. This forecast makes it possible to establish an operational plan for water supply to facilitate the operation of rational water pumps, the adjustment of valves, and the setting of reservoir levels.

Therefore, for optimal operation of drainage facilities, it is necessary to understand the demand structure of water demand considering the characteristics of the target area and to accurately predict demand.

Domestic water management technology (hereinafter referred to as "water management system") has been developed, installed and operated in earnest from the 1990s, based on the systems of developed countries such as Japan, with reference to large corporations. The steps to utilize the results are as follows. ① Calculate daily demand by system using various demand forecasting techniques. ② Calculate the forecasted demand by time by analyzing demand patterns using relative time coefficient (aka: demand pattern simulation of data granulization method). ③ The calculated demand forecast by hour calculates the amount of water delivered by branch point to provide economic operation, and provides the operation plan information. ④ Prepare the pump switching table for each hour by considering the electricity charges in the forecasted demand and provide it to the operator. .

Generally, the data granulization technique used to calculate the forecasted demand by time is generally used, but the multiple regression, ARIMA, and Kalman filter model techniques are used as the models used to estimate the daily demand.

Among these various demand forecasting models, the development of the most reliable and highly available demand forecasting technique is urgently needed.

SUMMARY OF THE INVENTION An object of the present invention is to provide a water demand prediction method that calculates an optimal demand amount by using a GM algorithm that shows strength in modeling and predicting a nonlinear system for stable operation of a water operation system. The present invention provides a method for predicting demand of water using the GMD algorithm that can achieve an output prediction accuracy target that sets an average% error between the actual demand and the actual demand.

Another object of the present invention is to select the past operational data of the target operation target branch, which is indirect information, as the criteria for demand forecasting and stabilization targets, and to perform the operation by online setting of optimal demand forecasting, thereby enabling the manpower by automation of demand forecasting. The present invention is to provide a water demand prediction method using a GM algorithm that performs reduction and efficiency of water operation.

In order to achieve the above object, the present invention is a method of predicting water demand in general, by incorporating and operating water intake, water purification, and pressurization stations in a water operation system by reflecting typical changes due to demand fluctuations, weather conditions, and seasonal factors. Building an intelligent model of water demand forecasting using GMD algorithm to secure stable supply, and demand forecasting software that calculates demand for forecasting target using GMD model-based predictive control method based on intelligent model Is configured on the host computer, the host computer of the operating system, the database having weather information, and the like through the LAN network, the demand prediction software is based on the operation data collected from the system, the data from the weather DB back-propagation learning Demand Using Algorithms Side if MD is characterized in that the step of building a model H, and using a model-based predictive control algorithm, not H. MD from the intelligent model consisting of determining a prediction of the demand in the current demand and the weather conditions.

According to another feature of the present invention, the optimal prediction demand is determined by using intelligent modeling and nonlinear model predictive control technique for stable operation of the manual operation system, but using intelligent modeling techniques such as GM algorithm in addition to the host computer. Based on the construction of an optimal demand forecasting model and an intelligent process model, a process simulator that determines the optimal forecasting demand using a nonlinear model predictive control technique is constructed on a higher level computer.

According to the water demand prediction method using the GM H algorithm of the present invention made as described above, the following effects can be obtained.

First, we provide a water demand forecasting method that calculates the optimal forecast demand by using GM algorithm which shows strength in modeling and forecasting any nonlinear system for stable operation of the water operation system. There is an effect that the output prediction precision target can be attained so that the average% error of is set to 10% or less.

Second, the past operation data of the target branch of manual operation, which is indirect information, is selected as the standard of demand forecasting and stabilization target, so that the operation can be carried out by online setting of optimal demand forecasting. There are other effects that can lead to efficiency.

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

<Examples>

1 is a view showing an overview of the manual operation system, Figure 2 is an operation flow chart for the main functions of one demand forecast.

3 is a conceptual diagram of a Kalman filter model sequentially estimating the state of the system, FIG. 4 is a diagram illustrating a discrete dynamic linear system of the Kalman filter, and FIG. 5 is a configuration diagram of the GM H structure.

First of all, water operation refers to plant control and water sources for individual facilities such as water intake, water purification plant and drainage. It is the same as the system operation control in "There is a system operation control that integrates and manages the whole water supply facilities from the water supply to the drainage."

In other words, it does not mean that each plant or facility is operated from intake to water purification and drainage separately, but the integrated operation and management of all facilities of waterworks.

In the water operation system for realizing such water operation, the operation data is generated and various policies are generated to satisfy the three main goals of water supply, water quality, quantity, and pressure. In order to respond to changes in demand, operate facilities scientifically and economically, and ensure equal water supply in each water supply area by early detection of water supply accidents, speedy recovery, and linkage of water intakes. To ensure and maintain water quality within drinking water quality standards by comprehensively managing individual water quality such as water treatment plant, pressurization plant, and drainage basin.

In other words, with the scientific and reasonable operation, the ultimate goal is to improve the quality of work for the entire water supply business, such as demand forecasting and stable water supply through equalized water.

As shown in Figures 1 to 4, in order to build an efficient water supply control system, it is important to estimate the water consumption of the consumer, and it is important to predict the water consumption of the consumer, and to plan a stable supply and control the supply amount based on this.

In other words, it is the basis of economic operation that determines the amount of supply by utilizing the facility capacity of the drainage or purified water without limiting the use of the water of the consumer, and to control only the determined flow rate. Therefore, it is necessary to adjust the branch valve so that the flow rate to each customer (the branch point) is supplied in consideration of the operation status of the recipient facility (drainage level or water level). When the supply flow rate is determined, the inflow and outflow volume is determined in consideration of the characteristics of the facility (water treatment plant, water intake plant, pressurization plant), power rate, and operating conditions. Considering the characteristics, the optimum pump operation number is determined.

Water Demand Forecasting Daily (short-term forecast) water demand forecasting by water supply zone is performed, and based on this, the demand forecasting by hour is calculated by drainage pattern analysis. However, long-term forecasting is performed on a monthly basis, and the calculated hourly demand forecast is based on the electricity bill, drainage level, water level, production volume, flow rate for each water supply zone, and inflow smoothing for economical operation in consideration of inflow. To calculate the two-day operation plan and transmit it to the host computer.

A drainage operation planning function should be performed to compensate for cases where the reservoir level is out of operating range or if the planned level is out of tolerance. In addition, external data (weather, climate, etc.) for demand forecasting should be entered automatically.

Appropriate techniques must be applied to build a demand forecast model, which typically includes: 1) historical performance drains, 2) the next day's highest expected daytime temperatures, 3) weather, and 4) days of the week (weekdays / holidays / special days). Calculate the demand forecast amount by considering Drainage pattern analysis uses appropriate techniques that take into account 1) seasons, 2) days of the week (weekdays / holidays / special days), and 3) weather items.

The operation flow for the main functions of daily demand forecasting is as shown in Figure 2, and the function predicts 1-2 days of daily demand by water supply area, and establishes the operation plan and establishment of water supply plan for pressurized water, water purification plant, and intake station. It is a function that calculates the quantity demanded. Historical performance volume collected from the host computer's database by daily time triggering or manual start of the driver, weather information for the next day (climate, expected peak temperature during the day, etc.) and date information (xxx, xx, xx, xx, xx) Collect it to predict the next day's drainage. Prediction uses algorithms such as computational models such as regression model, Kalman filter method and fuzzy prediction model, which are adaptive techniques that change the parameters of the prediction model according to the prediction error. Corresponds.

Although there are many methods for estimating daily demand in daily water supply systems, the multiple regression model, transition function ARIMA, and Kalman filter are generally used. The multiple regression model is the most common linear model that formulates the direct effects of external factors (such as past multiple days of performance, the next day's peak temperature, weather, and days of the week (weekdays / holidays / special days)). It is a linear model that considers the direct effects of external factors (such as the past multiple days of performance, the next day's expected maximum temperature, the weather, and the days of the week (weekdays / holidays / special days)). The Kalman filter model is a function that expresses non-linear relations of external factors such as past multiple days of performance, the next day's peak temperature, weather, and day of week (weekdays / holidays / special days).

The multiple regression model develops a daily water demand forecasting model using the factors considered to affect daily water demand as independent variables, and water demand as a dependent variable. In the case of a linearly significant effect, the linear combination of the output sequence and multiple input sequences is called a regression model. Here, the output series is called the objective variable (dependent variable), and the input series is called the explanatory variable (independent variable). This is expressed as equation (1).

&Quot; (1) &quot;

The multiple regression model undergoes three stages of verification to establish the model in a general way that is applied to a wide variety of fields.

The first is about the independent variables, and it is verified whether they are valid as independent variables and whether they are statistically independent from each other. The validity of the independent variable is verified by the 95% significant t-test.

To be verified. The second is about whether the residual generated by the model is an error that occurs randomly, and is verified through the Dublin Watson statistic that indicates the autocorrelation of the residual. Thirdly, the predicted value of the model is correct and verified by correlation coefficient and absolute mean error.

Transition Function The ARIMA model is a function in which the output sequence is linearly affected by the input sequence in a dynamic system. It consists of two parts. One is the transition function model. The influence of the input sequence is modeled through cross-correlation analysis and pre-whitening. The other is the mouth

A term representing the effects other than the force series and the delay of the input / output series. Modeled with ARIMA. After the cross-correlation analysis, the process leads to the order estimation and parameter estimation of the model, and the basic equation is shown in Equation (2).

<Equation 2>

The transition function model and the ARIMA model can be expressed as Equation (3).

<Equation 3>

Research using transition function ARIMA model in the field of environment is diverse, such as traffic noise and daily water demand prediction. The transition function ARIMA model statistically verifies the significance of the model by selecting each parameter, determining the order, and estimating the parameters.

The index that selects the most optimal model is verified by Chi-square test, AIC and SBC as follows. The chi-square test verifies the autocorrelation of the model residuals, and the AIC and SBC allow the most accurate prediction with the minimum parameters.

The chi-square test value determines whether the residual follows the white noise assumed in the model, and is determined as the Portmanto test value of Equation (4).

<Equation 4>

The Portmanto test above is approximate to the chi-square test of degrees of freedom (Kpq; p: AR course order, q: MA course order), and the Portmanto test Q above is 95% of the chi-square test. A judgment is made that meets the significance level. Therefore, it is significant if the value of the chi-square statistic is 0.05 or more.

The values of Akaike information criterion (AIC) and Schwartz's Bayesian criterion (SBC) are shown in Equations (5) and (6), respectively.

<Equation 5>

AIC = -2 × maximum log likelihood + 2 × number of fit parameters

<Equation 6>

Where n is the effective length of the time series

AIC and SBC are in two parts. One is the maximum, indicating the accuracy of the model and having a negative value. The other is to have a positive value as the number of fit parameters. This is a compromise between the model so that it has an accurate predictive value with a minimum fit parameter. The smaller the value, the better the model.

In general, the system is subject to control inputs and disturbances, and the state of the system is observed through the observation device. Observations are often scattered by noise and are not directly observed. In this situation, the Kalman filter model is used to determine the state of the system from the observed data.

It's about extracting information. This method assumes the statistical properties of noise, taking note of the dynamic characteristics of the system generating the time series output, and sequentially estimates the state of the system using the initial value information and the data observed at each time. same.

This Kalman filter is a method of optimally estimating the amount of state at t based on observation data of input and output variables at time t in Discrete Linear Dynamical System. It is described by a differential equation such as

<Equation 7>

<Equation 8>

Here, Equation (7) is called a state equation, and Equation (8) is called an observation equation, and is expressed as shown in FIG. Where Xk represents the system state at time tk and is called the state vector. Z-1 of the system represents the step delay operator and converts Xk + 1 one point forward (Xk = z-1Xk + 1). Xk + 1 is expressed as a dynamic linear system that takes as input data the system matrix Bk at time k and the captured Wk of the system.

&Quot; (9) &quot;

<Equation 10>

<Equation 11>

In the above Kalman filter model, the matrix Bk indicating the change of the state value Xk and the observation matrix Ak representing the input / output relationship of the system are set based on the physical structure of the object to which the model is applied. Estimate the state quantity Qk.

The GM H algorithm of the present invention exhibits superior characteristics in efficient use and accuracy of data as compared to the conventional prediction method. Unlike neural network, GMD algorithm has the advantage that it can process more data because modeling only necessary cells by applying the survival principle of deficit to every cell.

GMD has independent variables X ₁ , X ₂ ,... , It is to find the function _{Y = f (X 1, X} 2, ..., X m) representing the approximation of the dependent variable Y ^{^} as X _m as exploratory.

Table 1 shows (X ₁ , X ₂ ,…, X _m : Y) It shows how to use N sets of data rack into training set and test set. The training set is used to find the appropriate function, and the test set is used to evaluate the found function.

The GMH method first creates a function model that approximates the value of the dependent variable by combining two independent variable pairs. The fit model applies the fittest survives, where the pair that estimates the dependent variable values (ie, the function model) survives and the pair that does not survive. The criteria for survival of the fittest are set separately. The function represented by each surviving pair forms the next generation (or simply descendants). In the next iteration, we create a function that estimates the value of the dependent variable by combining each of these two pairs again with these descendants. As the generation continues, survival and selection continue to estimate the dependent variable, but when no further improvement is made, the generational replacement is stopped. This process is explained by dividing the generation, the criterion of survival of the fittest, the Stopping Rule, and the Ivakhnenko polynomial.

TABLE 1

In order to make the difference from the current generation to the next generation, the dependent variable Y is

Primitive equations) '. Where X _g , X _h Is an arbitrary two variables combined in the current generation, and the coefficients A, B, C, D, E, and F are obtained by the Least Square Method that minimizes the sum of error squares (this is called the Ibanenko coefficient). When applying the least-squares method, use the data in the training set.

Y ^{^} = A + BX _g CX _h + DX ² _g + EX ² _h + FX _g * X _h

It can be inferred that the values of Y ^{^} by the estimated equation will be closer to the Y value than the X _g and X _h values of the parent generation. Therefore, if we find a function that estimates Y again as Y ^{^} , we can expect to get closer. So we form a new generation (the second generation) as Y ^{^} and repeat the generational replacement for the next generation (the third generation). In Table 2, Y ^{^} is expressed as Z _i in the next generation.

If there are m variables in the first generation, the second generation generates as many descendants as many as mC2 = m (m-1) / 2. Thus, as generations continue to change, the offspring increase exponentially. If there is no improvement after repeating the generational change, the generational change is stopped and the best one is selected from the descendants to date.

In the course of the generational repetition, inappropriate offspring are eliminated and no longer generate offspring. In this case, in order to evaluate the survival and culling of descendants (Regularity Criterion), the following equation is used. If the r ² _i ² _i a r ≤ R, calculated at the pre-set to any standard value R and the resulting offspring Z _i and survival, the selection is r ² _i> R.

TABLE 2

Change of Generation from GM1H to Generation 2

<Equation 12>

j = 1, 2,... ,: Subscript representing each descendant in the newly created generation

I = nt + 1, nt + 2,... … N: Test set subscript

z _ij : The i-th element of the descendant of the jth generation within the same generation

On the other hand, the evaluation value r _j calculated from the descendants of the current generation k The minimum value of RMINk decreases gradually as shown in <Graph 1> and then increases again. R _{j in} generations where RMINk is at least The lowest descendant is the optimal descendant.

By dividing the data into a training set and a test set, the RMINk does not continue to decrease, but increases again at some point to prevent over-fitting or over-specification.

<Graph 1>

Breakpoints in Generation Replacement for Optimal Descendants

The optimal offspring have several overlapping basic formulas. As shown in Table 3, when X _i and X _j produce one descendant U, X _k and X _i produce another descendant V, and then the descendants U and V pair to produce descendant W, Let W be the optimal descendant. W is then represented by U and V, but in reality we have a polynomial consisting of X _i , X _j , X _k , and X _l . The expression of multiple overlapping basic expressions as the polynomial of the independent variable is called Ivakhnenko Polynomial.

TABLE 3

Nesting of Basic Expressions and Ibanenko Polynomials

GMD is very similar to neural networks. The primitive equation described above can be expressed as a neural network with two input nodes and one output node. (The number of intermediate nodes can be any number.) The model by GMD becomes a form of compressing neural network model. In addition, all functions represented by neural networks can be constructed by GM H.

In general, the state equation of a nonlinear dynamic system based on a difference equation is

x (k + 1) = Ф [x (k), u (k)]

y (k) = ψ [x (k)]

In this case, x (k), y (k), u (k) denote states, outputs, and inputs, respectively, and Ф [·] and ψ [·] denote nonlinear functions. Reconstructing such a nonlinear difference equation can be expressed as the following equation. Where f [·] is a nonlinear function.

y (k + 1) = f [y (k), y (k-1),. , y (k-n + 1), u (k),... u (k-m + 1)]

이고,

이다. 이 부등식들이 성립하는 이유는 뒤에서 언급할 데이터 선택 기준에 의해 영향이 적은 데이터를 버리기 때문이다. 최종적으로 하나의 출력을 얻게 되는 과정도 데이터 선택 기준의 최소값에 의해 결정된다. 모든층과 각각의 노드들의 출력은 다음 식과 같이 2차 방정식의 형태를 취하게 된다.GMH consists of a set of input nodes, which go through the middle layers to get output from these input nodes. The nodes in each layer take two outputs from the outputs of the nodes in the previous layer and produce an output. The configuration diagram of the GM H is shown in FIG. 5. In FIG. 5, z _{s and t} represent the outputs of the s-th layer and the t-th node, and m is the number of inputs. The second tier consists of i nodes,

ego,

to be. These inequalities hold because they discard data with less impact by the data selection criteria described later. The process of finally obtaining one output is also determined by the minimum value of the data selection criteria. The outputs of all layers and their respective nodes take the form of quadratic equations:

z _{s , t} = a _{s, t} z ² _{(s-1), u} + b _{s , t} z ² _{(s-1), v} + c _{s , t} z _{(s-1), u} + d _{s , t} z _{(s-1), v} + e _{s , t} z _(s-1) , u ^z _{(s-1), v} + f _{s , t}

Where a _{s, t} , b _{s , t} , c _{s , t} , d _{s , t} , e _{s , t} , f _{s , t} are the node's connection weights, and the subscripts u and v are arbitrary Point to the u and v nodes.

In other words, the principle of GMD is that, for example, the number of nodes in the next layer is determined by the number of combinations of i inputs (ci2) coming into the third layer, and then the outputs of these nodes are calculated by an equation. In other words, the suitability of the calculated output is determined by performance criteria. In order to determine the unsuitable nodes in each layer like this, the error between the desired final output and the actual node outputs is calculated as follows.

Where (z _{s , t} ) _n is the n th element of the z _{s , t} vector _, which is the output of the s th layer and the t th node. The size of the error is determined so that nodes with large errors are eliminated and the nodes with small errors are repeated, or the above steps are repeated until the desired output is obtained. And, unlike the neural network structure, which is generally used, the number of intermediate layers and the number of nodes are not predetermined.

Typical support MD H. algorithm compared to a predetermined value R determined in advance as a reference to an eclectic data and error r _{s, t} r _{s, t} is less than R taking the output of that node r _{s, t} is greater than R The node's output is discarded. However, using this approach, the number of nodes remaining increases gradually as the output of the node gets closer to the actual output as the layer increases. As a result, the computational complexity of the algorithm increases. Therefore, the present invention uses a method of preventing excessive computation by decreasing the reference R sequentially.

First, the method of terminating the algorithm is described, and the minimum value IN _s (s = 1, 2, ...) of errors in each layer is obtained, and the algorithm is terminated at the layer where the IN _s is minimum. This process is shown in Graph 2. The dots in <Graph 2> show IN _s to be. Therefore, the output of the node with the least error in layer s is the estimate of the actual output.

<Graph 2>

How the algorithm ends

As can be seen from Graph 2, in general GM H algorithm, all nodes whose error is smaller than R are selected and IN _s as the layer increases. As the spacing between and R increases, the number of nodes increases significantly. Therefore, in the present invention, in order to prevent a large increase in the number of nodes, as shown in <graph 2>, a criterion as shown in Equation 13 below is selected.

<Equation 13>

Rs = IN _s + δ

Where δ is a predetermined constant value. By setting the selection criteria as shown above, it can be seen that not only does the number of nodes increase significantly but also the data can be selected evenly for IN _s .

The GM H algorithm for modeling the nonlinear system represented by Equation 13 can be described in three stages as follows.

Step 1. Construct the Variable Vector

According to Equation 13, the output z _{s , t} (k) of the t-th node of the s layer at the moment k can be simply expressed as follows.

z _{s , t} (k) = X _{s -1} θ _s

Where X _{s −1} = [z ² _{(s-1), u} (k) z ² _{(s-1), v} (k) z _{(s-1), u} (k) z _{(s-1), v} (k) z _{(s-1), u} (k) z _{(s-1), v} (k)]

θ _s = [a _{s , t} b _{s , t} c _{s , t} d _{s , t} e _{s , t} f _{s , t} ] ^T

X _{s −1} is a data vector of order n × 6, which is a combination of two vector inputs z _{(s-1), u} (k) and z _{(s-1), v} (k). At the moment (k + 1), the input and output of the first layer are shown in Table 4. When modeling a nonlinear dynamic system with GMD, there are (m + n) input vectors applied. Each vector has p elements and is arbitrarily determined. The amount of data applied depends on the size of P. At this time, in order to obtain z _{s , t} (k), it is necessary to obtain the connection strength θ _s which is a coefficient vector. Since the purpose of this algorithm is to make the output z _{s , t} (k) of each node finally the output y (k) of the system, the general formula is:

θ _s = (X ^T _{s -1} X _{s -1} ) ^-1 X ^T _{s -1} y (k)

Step 2. Elimination of Low Impact Data

The outputs of all nodes obtained in step 1 are not passed to the next layer. In other words, the principle of survival of the fittest should be applied. In general GM H, the output and newly generated data should be divided into training and inspection parts. Thus, we eliminate unnecessary nodes by means of the mean square error obtained during the training portion of the output and generated data. Various methods have been proposed. However, in the present invention, since the nonlinear system is to be modeled dynamically, all data of Table 4 are used in the training part and the last element of the data vector is used as the test part.

Reorder all z _{s , t} (k) in layer s in increasing order of r _{s , t} . Then, nodes having a value larger than the reference value R _s are removed. Then use the unremoved variables z _{s , t} (k) as input to the (s + 1) layer.

Step 3. Test of Optimality

In this step, to determine whether to continue training _, the minimum value IN _s of r _{s , t} calculated in step 2 is compared with the previous step IN _{s -1} . The minimum value IN _s is obtained and R _s is obtained. If IN _s <IN _{s - 1} as shown in <Graph 2>, repeat steps 1 and 2, and if IN _s ≥ IN _{s - 1} , output z _{(s-1), t} (k) corresponding to RNIN _{s -1} ^{^} Y (k) is estimated as the end of all training. Then z _{(s-1), t} (k) becomes the Ivakhnenko polynomial.

TABLE 4

GMH input and output

As described above, according to the water demand prediction method using the GM algorithm according to the present invention, it is effective to determine the optimal forecasting demand using intelligent modeling and nonlinear model predictive control technique for stable operation of the water operation system. have.

In addition, the average error between the forecasted demand and the actual demand of the intelligent model should be set to 10% or less, and the past operational data of the target quarters, which are indirect information, are selected as the criteria for demand forecasting and stabilization targets. By performing the operation by online setting, it is possible to reduce manpower and efficiency of water operation and economic efficiency by automation of demand forecasting.

In addition, in the present invention, the optimal prediction demand is determined by the GM modeling of the demand forecasting process and the nonlinear model predictive control technique, and the nonlinear process optimization and the auto tuning of the manual operation system where the high efficiency and stability of the demand forecasting process are essential. It implements the optimal structure and parameter search of the manual operation software and helps to improve the reliability and maintenance of the system, standardization, and secure the technical skills. There is an effect that an improvement can be achieved.

Those skilled in the art to which the present invention pertains will understand that the present invention can be implemented in other specific forms without changing the technical spirit or essential features. Therefore, it should be understood that the above-described embodiments are to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than the foregoing description, It is intended that all changes and modifications derived from the equivalent concept be included within the scope of the present invention.

1 is a view showing an outline of a water operating system.

2 is an operational flowchart of the main functions of one demand forecast.

3 is a conceptual diagram of a Kalman filter model sequentially estimating the state of a system.

4 shows a discrete dynamic linear system of a Kalman filter.

5 is a configuration diagram of the GM H structure.

Claims (2)

In the normal water demand prediction method, Intelligent operation of water demand prediction using GMD algorithm to secure stable supply by rationally operating water intake, water purification plant and pressurization station by reflecting the usual changes due to demand fluctuation, weather condition, seasonal factors, etc. Building a model; Configuring demand prediction software on a higher level computer that calculates demand for prediction targets using a GM model-based predictive control technique based on the intelligent model; Connecting to a host computer of the manual operation system, a database having weather information, and the like through a LAN network; The demand forecasting software includes: building a demand forecasting GM model using a backpropagation learning algorithm based on operational data collected from a system and data of a weather DB; And The water demand prediction method using the GM H algorithm, characterized in that for determining the current demand and forecast demand amount in the weather conditions using the GM model-based predictive control algorithm from the intelligent model. The method of claim 1, For the stable operation of the manual operation system, the optimal forecasting demand is determined by using intelligent modeling and nonlinear model predictive control technique, and the optimal demand forecasting model is built using intelligent modeling techniques such as GMD algorithm in addition to the host computer. And water demand prediction method using GMD algorithm, which comprises a process simulator that determines the optimal demand for prediction using nonlinear model predictive control based on intelligent process model.

Images