JPH05181507A - Pump plant operation controller - Google Patents

Abstract

PURPOSE:To effectively utilize the knowledge of a skillful operator and stored data and to complement the change of a control parameter with the operational result data of the skillful operator. CONSTITUTION:This device is composed of a water level predicting means 10 to predict the water level of a pump well from various measured data by providing a function to learn the result data in the past, requested pouring quantity predicting means 20 to predict the pouring quantity of the pump from the predictive water level by providing a function to learn the result data, and pump operation planning means 30 to prepare the operational plan of the pump from the predictive pump pouring quantity by providing knowledge storing data concerning respective pumps.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an operation control device applied to the operation control of a plurality of pumps in a pump plant.

[0002]

2. Description of the Related Art A pump plant is constructed as shown in FIG. That is, water from a river or reservoir is taken in from the inflow gate 1 and stored in the pump well 3 via the sand basin 2. The water in the pump well 3 is pumped up by a plurality of controlled pumps 4 and supplied to the discharge tank 5.

As a conventional control method of the pump 4 in such a pump plant, there are two methods of controlling the rainfall amount and the rainfall intensity.
Based on the three meteorological information and the three observation information of the water level of the pump well 3, the inflow amount, and the pump discharge amount, the experienced operator empirically determines the number of pumps 4 and the start timing, and the operation plan. Is to create.

Another conventional control method is shown in FIG.
As shown in, the weather control information and the observation event are input, and the operation command of the skilled operator is output, and the learning control device having the learning function is made to learn the operation control technology of the experienced operator. A method for obtaining a pump operation control command equivalent to the above has also been developed.

Further, as another conventional control method, as disclosed in Japanese Patent Application Laid-Open No. 1-113810, Japanese Patent Application Laid-Open No. 1-113811 and Japanese Patent Application Laid-Open No. 1-113812, the operation of a skilled operator is performed. A method is also known in which knowledge about control is organized and stored in the form of rules, and inference using the previous knowledge is performed according to input data such as an inflow amount and a pump well water level to obtain a pump operation control command.

[0006]

In the conventional pump plant operating method described above, it is difficult to accumulate and pass on the knowledge of the experienced operator by the method relying only on the experience of the experienced operator, and the operator's individual difference. There is a difference in the judgment criteria of control by.

Further, in the method of learning the operation control technique of the skilled operator, since the judgment depends on one learning control device, the correction and adjustment of the control variable, the selection of a plurality of pumps and the determination of the use sequence are performed. It is difficult to create such an operation plan. For example, if it is necessary to correct the pump discharge amount due to the addition of pumps or the inflow amount due to changes in the basin topography, all data will be learned again.

Further, in the method based on the inference using the above knowledge, it is difficult to extract the knowledge because the actual operation record data during heavy rain is small. In addition, since many parameters are used for control, it is difficult to maintain consistency with other knowledge when introducing new knowledge or modifying knowledge.

The present invention has been made by paying attention to such a problem, and a pump plant capable of performing equivalent operation control within a range that can be complemented by the operation record data of the experienced operator of the pump plant given at the time of learning. It is intended to provide the operation control device of

A further object of the present invention is to provide an operation control device for a pump plant in which the inflow prediction system is improved to improve the water level and pump discharge system.

[0011]

According to the present invention, a water level predicting means having a learning function for predicting the water level of a pump well of a pump plant, and a pump discharge amount by using an output of the water level predicting means as an input. Pump plant operation control including a pump discharge amount predicting unit having a learning function for operating the pump, and a pump operation planning unit having knowledge for creating an operation plan of the pump using the output of the pump discharge amount predicting unit as an input. A device is provided.

Further, according to the present invention, the water level predicting means having a learning function for predicting the water level of the pump well of the pump plant, and the learning for predicting the pump discharge amount by using the output of the water level predicting means as an input. A pump discharge amount predicting means having a function, a pump operation planning means having knowledge for creating an operation plan of the pump by inputting the output of the pump discharge amount predicting means, and an output of the pump operation planning means and weather information. As a input, predict the pump well inflow,
There is provided a pump plant operation control device comprising an inflow quantity predicting means having a function of using the output as an input of the water level predicting means at the next time.

[0013]

[Operation] In the water level prediction means, first, the past performance data is learned. Then, the control variable value is output for the plurality of input information. Similarly, the past discharge data is learned in the pump discharge amount prediction means. Next, input the input information including the previous control variable value, output another control variable value, and
The input information including the control variable value is input to the pump operation planning means having knowledge for planning, the pump operation control command is output with reference to the knowledge held by the means, and the pump plant is controlled. Further, input information including the operation control command is input to the inflow amount prediction means to output a predicted value, and the output is set as one of the input information of the water level prediction means at the next time.

[0014]

FIG. 1 is a block diagram showing a pump plant operation control device according to the present invention. In FIG. 1, reference numeral 10 is a water level predicting means having a learning function for predicting the water level of the pump well of the pump plant, and 20 is a learning function for predicting the pump discharge amount by using the output of the water level predicting means 10 as an input. The required discharge amount predicting means, 30 is an operation planning means having knowledge for creating an operation plan of the pump using the output of the required discharge amount predicting means 20 as an input.

The water level predicting means 10 and the required discharge amount predicting means 20 are neural nets realized on a computer.
Further, the pump operation planning means 30 is an expert system.

In the water level predicting means 10, the precipitation measuring means 11 supplies the time series information of the rainfall amount obtained as this output to the inflow amount predicting means 12, and the inflow amount predicting means 12 uses the input rainfall amount for the pump station. Outputs a value that predicts the amount of inflow to. The water level measuring means 13 outputs the pump well water level. Then, the predicting unit 14 uses the rainfall amount, the inflow amount predicted value that is the output from the inflow amount predicting unit 12, the current pump well water level obtained from the water level measuring unit 13, the number of pump startups, and the discharge valve opening degree. The water level at the next time is predicted from the determined current value of the pump discharge amount, and the predicted water level is output. FIG. 2 shows an embodiment showing changes in the input / output values of the predicting means 14. The predicting means 14 has a result of learning past performance data.

The required discharge amount predicting means 20 includes a pump well water level predicted value which is an output from the water level predicting means 10, and a water level change rate and a limit water level calculated by the calculating means 21 from time series data of the pump well water level. Prediction means 2 from water level deviation information
At 2, the pump discharge amount required at the next time is determined and output as the required discharge amount. FIG. 3 shows an embodiment showing changes in the input / output values of the predicting means 22. The prediction means 22
Also has the result of learning the past performance data.

The water level predicting means 10 and the predicting means 14 and 22 of the required discharge amount predicting means 20 are three-layer neural network models, and the inputs are the time series data for the past 30 minutes and the intervals for 5 minutes, and the outputs are the following. Use the water level at the time or the discharge amount. These data are standardized in the range of 0 to 1 and given as analog quantities. In addition, the learning data is the performance data of the skilled operator during heavy rain when a plurality of pumps are activated, and these are used as the BP learning method (error backpropagation learning method; Ba
ck Propagation) to learn repeatedly.

The pump operation planning means 30 determines the required number of pumps from the required discharge quantity which is the output of the required discharge quantity predicting means 20, and the inference engine 31 further causes the inference engine 31 to database 3.
The usage history of each pump stored in 2, the start-up time,
Determine the optimum operation pump and its operation plan by referring to the limiting conditions such as restart cooling time. The pump operation planning means 30 is an expert system that organizes knowledge in the form of rules. The rule is to preferentially use a pump that is rarely used at the time of starting the pump, and to use the one with a longer cooling time if the frequency is the same. Figure 4
An example showing changes in the input / output values of the pump operation planning means 30 is shown in FIG.

FIG. 5 is a block diagram showing another embodiment of the operation control device for a pump plant according to the present invention. In the figure, the same elements as those in FIG.
Detailed description thereof will be omitted. According to FIG. 5, the inflow quantity prediction value input to the water level prediction means 10 for water level prediction at the next time is input from the pump operation planning means 30 and other inflow quantity prediction means 40 using the time series information as input information. It has gained. The inflow amount predicting means 40 is a numerical analysis system.

The inflow predicting means 40 is the rainfall measuring means 4
1. A rainfall prediction unit 42 and an inflow amount analysis unit 43. The rainfall intensity is obtained as an output from the rainfall measurement unit 41, the predicted rainfall amount is obtained as an output from the rainfall prediction unit 42, and a pump operation plan. From the pump operation control command which is the output from the means 30, the inflow quantity required by the predicting means 14 in the water level predicting means 10 at the next time is calculated by the inflow quantity analyzing means 43 and output as an inflow quantity prediction value. The inflow predicting means 40 takes in real-time information from the local meteorological observation system "AMeDAS" of the Meteorological Agency through the NTT line, and analyzes the rainfall outflow model of the target area.

[0022]

As described above, according to the present invention, the skilled operation is performed by the water level predicting means 10 which has learned the judgment of the skilled operator, the required discharge amount predicting means 20 and the pump operation planning means 30 having knowledge about the operation. Operation control equivalent to that of personnel can be performed. Further, the water level prediction means 10 and the required discharge amount prediction means 20
Since three means corresponding to the three judgment processes of the skilled operator, that is, the pump operation planning means 30, are provided, when it is necessary to correct the control variables or delete the control variables of low importance from the input data, Rather than having to retrain the data, only the relevant devices need to be trained, which facilitates device tuning. Further, by giving a rule regarding pump selection in the pump operation planning means 30, it becomes possible to perform an optimum operation plan for a plurality of pumps.

Further, according to the present invention, the inflow amount predicting means 4
At 0, by inputting the operation control command, which is the output of the pump operation planning means 30, together with the meteorological information, it becomes possible to predict the inflow amount based on the current pump plant state, and it is possible to output the information with high prediction accuracy. Become.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of a pump plant operation control device according to the present invention.

FIG. 2 is a change diagram of various information showing an example of learning control of a water level prediction unit.

FIG. 3 is a change diagram of various kinds of information showing an example of learning control of a required discharge amount prediction unit.

FIG. 4 is a change diagram of input / output values showing an example of operation support of a pump operation planning unit.

FIG. 5 is a block diagram showing a second embodiment of the pump plant operation control device according to the present invention.

FIG. 6 is a conceptual diagram of a pump plant.

FIG. 7 is a block diagram showing an embodiment of a conventional pump plant operation control device.

[Explanation of symbols]

 10 Water Level Predicting Means 11 Rainfall Measuring Means 12 Inflow Predicting Means 13 Water Level Measuring Means 14 Predicting Means 20 Required Discharge Predicting Means 21 Calculating Means 22 Predicting Means 30 Pump Operation Planning Means 31 Inference Engine 32 Database 40 Inflow Predicting Means 41 Rainfall Measuring Means 42 Rainfall Prediction Means 43 Inflow Analysis Means

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification code Internal reference number FI technical display location G05D 9/12 C 7001-3H

Claims (2)

[Claims] 1. A water level predicting means having a learning function for predicting a water level of a pump well of a pump plant, and a pump discharge rate having a learning function for predicting a pump discharge rate by using an output of the water level predicting means as an input. A pump plant operation control device comprising: a prediction unit; and a pump operation planning unit that has knowledge for creating an operation plan of a pump using the output of the pump discharge amount prediction unit as an input. 2. A water level predicting unit having a learning function for predicting a water level of a pump well of a pump plant, and a pump discharge amount having a learning function for predicting a pump discharge amount by using an output of the water level predicting unit as an input. A predicting means, and a pump operation planning means having knowledge for creating an operation plan of the pump using the output of the pump discharge amount predicting means as an input;
Pump plant operation comprising an inflow rate predicting means having a function of predicting a pump well inflow rate by inputting the output of the pump operation planning means and weather information and using the output as an input of the water level predicting means at the next time. Control device.