JP2012079229A - Demand prediction device and water operation monitoring system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a demand prediction device capable of predicting demand amount with high accuracy even in a prediction condition, appearance frequency of which is small in the past.SOLUTION: A demand prediction device 2 according to the invention includes an appearance frequency determination unit 21, a similar condition search unit 23, and a demand prediction calculation unit 22. When it is determined that appearance frequency in the past of a prediction condition is equal to or less than a predetermined frequency by the appearance frequency determination unit 21, the similar condition search unit 23 obtains a similar condition corresponding to the prediction condition from a similar condition database 24c and obtains data on a demand model pattern corresponding to the similar condition obtained from a demand pattern database 24d. Then, the demand prediction calculation unit 22 performs demand prediction on the basis of the data on the demand model pattern corresponding to the similar condition obtained by the similar condition search unit 23 and an actual value of water demand in the past obtained from an actual value database 24b.

Description

  The present invention relates to a demand prediction device used for operation and monitoring of a waterworks plant and a water operation monitoring system including the same.

  Conventionally, in an operation monitoring system of a waterworks plant, each facility has independently made an operation plan using a computer. In recent years, with the widening of the water supply business, a water operation monitoring and control system has been established that monitors the water treatment facilities such as water purification plants, water distribution plants, and pumping stations that are scattered throughout the central management room. .

  The operation plan by the water operation monitoring control system is mainly divided into two, namely, a demand forecasting plan and an allocation plan. In the demand prediction planning, the demand amount on the day is predicted in advance, and in the allocation planning, the ratio of the demand amount in each water distribution area is calculated based on the predicted demand amount. In addition, since the demand for water varies depending not only on weather conditions such as weather and temperature, but also on social life conditions such as days of the week, holidays, New Years, and Bon Festivals, it is necessary to make a demand prediction every day.

  As a method for demand prediction as described above, a model for forecasting demand using a statistical method based on actual demand volume data (demand forecast model) has been used to predict the daily demand volume and changes over time. A technique has been proposed (see, for example, Patent Document 1). Patent Document 1 proposes a method and a demand prediction device that perform demand prediction using a neural network model.

JP-A-10-219758

  As described above, conventionally, the amount of water demand is affected by weather conditions and social life conditions. Therefore, in the demand prediction calculation by the conventional statistical method, conditions such as weather, temperature, day of the week, and special day (May holidays, Bon Festival, New Year, etc.) are used as the prediction conditions. Then, past planning conditions (demand patterns) corresponding to the prediction conditions are extracted from the performance database, and demand prediction is performed using the extracted demand patterns.

  However, the frequency of appearance of a single forecast condition such as rare weather or unusual day, or a forecast condition that combines unusual day and rare temperature or weather, etc. is about several times a year. If the condition is met, it will only appear once every few years. Therefore, the demand pattern corresponding to such a condition with a low appearance frequency is stored only about several times in the performance database of the computer.

  Therefore, when the demand prediction is performed under the prediction condition with such a low appearance frequency, the demand pattern created based on the old actual value is used. However, there arises a problem that the demand pattern cannot be used when the current demand situation does not match the past situation due to, for example, a review of the water distribution area. In this case, it is necessary for the operator to intervene and to set demand forecasts and set forecast values again. In addition, even if a demand prediction is performed using a demand pattern calculated based on an old actual value, there is a problem that only a demand prediction value with poor accuracy can be calculated.

  The present invention has been made to solve the above problems, and an object of the present invention is to provide a demand prediction apparatus capable of predicting a demand amount with higher accuracy even in a prediction condition with a low frequency of occurrence in the past, and a demand prediction apparatus thereof Is to provide a water operation monitoring system.

  In order to solve the above problems, the demand prediction apparatus of the present invention is configured to include an appearance frequency determination unit, a similar condition search unit, and a demand prediction calculation unit, and the functions of the units are as follows. The appearance frequency determination unit determines whether or not the past appearance frequency of the prediction condition is equal to or lower than a predetermined frequency based on the prediction condition of the water demand and the actual value indicating the past water demand situation stored in the actual value database. To determine. The similar condition search unit, when the appearance frequency determination unit determines that the past appearance frequency of the prediction condition is equal to or lower than the predetermined frequency, creates a similar condition that is created based on the past actual value and correlates with the prediction condition. Demand model pattern data corresponding to the acquired similar condition from the demand pattern database storing the demand model pattern data of the water demand by searching the similar condition database to be stored and acquiring the similar condition corresponding to the prediction condition To get. Then, the demand prediction calculation unit acquires the actual value of the past water demand from the actual value database, and based on the acquired actual value and the data of the demand model pattern corresponding to the similar condition acquired by the similar condition search unit Demand forecast.

  In addition, the water operation monitoring system of the present invention includes an input device unit, a result value database, a similar condition database, a demand pattern database, an appearance frequency determination unit, a similar condition search unit, a demand prediction calculation unit, a result The configuration includes a value acquisition unit, and the function of each unit is as follows. The input device unit receives water demand prediction conditions. The actual value database stores actual values indicating past water demand conditions. The similar condition database is created based on past performance values and stores similar conditions correlated with the prediction conditions. The demand pattern database stores data on the demand model pattern of water demand. The appearance frequency determination unit is configured such that the past appearance frequency of the prediction condition is less than or equal to a predetermined frequency based on the prediction condition input to the input device unit and the actual value indicating the past water demand situation stored in the actual value database. Judge whether there is. The similar condition search unit searches the similar condition database to obtain a similar condition corresponding to the prediction condition when the appearance frequency determination unit determines that the past appearance frequency of the prediction condition is equal to or lower than a predetermined frequency, and Then, demand model pattern data corresponding to the obtained similar condition is obtained from the demand pattern database. The demand prediction calculation unit acquires the actual value of the past water demand from the actual value database, and based on the acquired actual value and the data of the demand model pattern corresponding to the similar condition acquired by the similar condition search unit, Make demand forecasts. And a track record value acquisition part acquires the data regarding the track record value stored in a track record value database.

  As described above, in the demand prediction apparatus of the present invention, when the input prediction condition is a rare condition, the demand is generated using a similar condition that is created based on the past actual value and has a correlation with the prediction condition. Make a prediction. That is, according to the present invention, it is possible to perform demand prediction considering more past actual values, particularly recent actual values, even under rare prediction conditions. Therefore, according to the present invention, it is possible to predict demand with high accuracy even under rare prediction conditions.

1 is a schematic configuration diagram of a demand prediction device and a water operation monitoring system according to an embodiment of the present invention. It is a figure which shows an example of the data structure of the total demand amount model pattern stored in a demand pattern database. It is a figure which shows an example of a data structure of the daily demand amount model pattern for every water distribution area stored in a demand pattern database. It is a figure which shows an example of a data structure of the unit time demand model pattern at the time of planning for every water distribution area stored in a demand pattern database. It is a flowchart which shows the procedure of a demand prediction process. It is a flowchart which shows the procedure of a similar condition search process.

  Hereinafter, an embodiment of a demand prediction device and a water operation monitoring system according to the present invention will be described with reference to the drawings. However, the present invention is not limited to the examples shown below.

[Configuration of water operation monitoring system]
FIG. 1 shows a schematic block configuration of a water operation monitoring system according to an embodiment of the present invention. FIG. 1 shows only the main part related to the demand prediction process.

  In FIG. 1, the structural example of the water operation | use monitoring system 100 provided in the middle of each water supply path | route which goes to the water distribution area 205a, 205b from the distribution reservoir 201 is shown. In FIG. 1, a pump 202 and a flow meter 203 are provided in the middle of the water supply path from the distribution reservoir 201 to the distribution area 205a, and a flow meter 204 is provided in the middle of the water supply path from the distribution reservoir 201 to the distribution area 205b. An example is shown.

  The water operation monitoring system 100 includes an input / output display device 1 (input device unit), a demand prediction device 2, and a process data interface device 3 (result value acquisition unit).

  The input / output display device 1 has an input unit (not shown) that allows an operator to input demand prediction conditions. The input / output display device 1 also includes a display unit (not shown) that displays information such as the demand prediction result planned by the demand prediction device 2 and the current water distribution status.

  The input / output display device 1 is connected to a demand prediction calculation unit 22 to be described later in the demand prediction device 2, and the demand prediction conditions (for example, planning date, input through the input unit in the input / output display device 1). The planning period, the day before, the day's weather, the maximum temperature, etc.) are output to the demand prediction calculation unit 22. The demand prediction result calculated by the demand prediction calculation unit 22 is output to the input / output display device 1 and displayed on the display unit in the input / output display device 1.

  The demand prediction device 2 makes a demand prediction based on a demand prediction condition (hereinafter simply referred to as a prediction condition) input from the operator via the input / output display device 1, and the prediction result is input to the input / output display device 1. Output to. The internal configuration of the demand prediction device 2 will be described in detail later.

  The process data interface device 3 is connected to flow meters 204 and 205 provided in the middle of each water supply path from the distribution reservoir 201 to the distribution areas 205a and 205b. Acquire the flow rate of the supply path (data related to actual values) That is, the process data interface device 3 acquires data indicating the current state of water demand in each water supply path.

  In addition, the process data interface device 3 is connected to an actual value database 24b described later in the demand prediction device 2, and based on the acquired flow rate, various actual values (total demand amount, water distribution) of water demand in each water supply route. The daily demand amount and the predetermined hourly demand amount for each area are calculated. And the process data interface apparatus 3 outputs the various actual value of the calculated water demand to the actual value database 24b.

[Configuration of demand forecasting device]
As shown in FIG. 1, the demand prediction apparatus 2 includes an appearance frequency determination unit 21, a demand prediction calculation unit 22, a similar condition search unit 23, a database unit 24, a similar condition creation unit 25, and a demand pattern creation unit. 26. The database unit 24 includes a predicted value database 24a, a performance value database 24b, a similar condition database 24c, and a demand pattern database 24d. The data structure in each database will be described in detail later.

  In addition, although this embodiment demonstrated the example which provides the database part 24 in the demand prediction apparatus 2, this invention is not limited to this, You may provide outside the demand prediction apparatus 2. FIG. For example, the database unit 24 may be provided in a storage unit (not shown) of the water operation monitoring system 100.

  The appearance frequency determination unit 21 is connected to the demand prediction calculation unit 22 and the actual value database 24b. The appearance frequency determination unit 21 acquires the prediction condition via the input / output display device 1 and the demand prediction calculation unit 22 when planning the demand prediction. In addition, the appearance frequency determination unit 21 compares the acquired prediction condition with the actual value stored in the actual value database 24b to determine whether the prediction condition is a rare condition. Specifically, it is determined whether or not the appearance frequency of the prediction condition is equal to or less than a predetermined ratio in the past and the predetermined period. Then, the appearance frequency determination unit 21 outputs the determination result of the appearance frequency of the prediction condition to the demand prediction calculation unit 22.

  The appearance frequency threshold and the search period in the determination process are appropriately set according to, for example, the predicted region and environment. Also, in this determination process, when the appearance frequency threshold is set to 0%, it is determined whether or not a condition that matches the prediction condition has appeared in the past predetermined period.

  The demand prediction calculation unit 22 is connected to the input / output display device 1, the appearance frequency determination unit 21, the similar condition search unit 23, the predicted value database 24a, the actual value database 24b, and the demand pattern database 24d. The demand prediction calculation unit 22 performs demand prediction by calculating various demand prediction parameters described later based on a prediction condition or a similar condition described later (a condition highly correlated with the prediction condition) corresponding to the prediction condition. Then, the demand prediction calculation unit 22 outputs the calculated demand prediction result to the input / output display device 1 and the predicted value database 24a. Various demand prediction parameters calculated by the demand prediction calculation unit 22 and the calculation method thereof will be described in detail later.

  The similar condition search unit 23 is connected to the demand prediction calculation unit 22, the similar condition database 24c, and the demand pattern database 24d. The similar condition search unit 23 responds to the prediction condition based on a command signal input from the demand prediction calculation unit 22 when the appearance frequency determination unit 21 obtains a determination result that the prediction condition is a rare condition. The similar condition is searched in the similar condition database 24c.

  Specifically, the similar condition search unit 23 compares the prediction condition with similar condition data (described later, element correspondence table) stored in the similar condition database 24c, and the similar condition described later corresponding to the prediction condition. It is determined whether or not exists. And when the similar condition corresponding to a prediction condition exists, the similar condition search part 23 acquires the similar condition (element mentioned later) from the similar condition database 24c.

  Moreover, the similar condition search part 23 searches the demand pattern database 24d based on the acquired similar conditions, and acquires the data of the various demand model patterns corresponding to the similar conditions. Then, the similar condition search unit 23 outputs the acquired data of various demand model patterns to the demand prediction calculation unit 22.

  The similar condition creating unit 25 is connected to the result value database 24b, the similar condition database 24c, and the demand pattern database 24d. The similar condition creating unit 25 generates similar condition data (described later) based on actual values updated on other days as well as rare forecast conditions and various demand model patterns created based on the actual values. Create (update) element correspondence table). Then, the similar condition creating unit 25 outputs the created similar condition data to the similar condition database 24c. Note that the timing of generating (updating) the data of similar conditions is arbitrary, but it is preferable to perform it periodically (for example, about once a day).

  The demand pattern creation unit 26 is connected to the result value database 24b and the demand pattern database 24d. The demand pattern creation unit 26 acquires the actual value from the actual value database 24b, and generates a data set of the demand model pattern based on the acquired actual value. Then, the demand pattern creation unit 26 stores the generated demand model pattern data set in the demand pattern database 24d. Thereby, the data of various demand model patterns are updated every day, and the newly updated demand model pattern can be used for the next and subsequent demand forecasts. The demand model pattern can be updated (created) at an arbitrary timing as long as the demand prediction is completed.

[Configuration of database section]
Next, the structure of data stored in each database constituting the database unit 24 will be described.

  The predicted value database 24a includes, for example, planning date, forecasting conditions, and various demand forecasting parameters calculated by a demand forecasting method to be described later (total demand for the entire distribution area, daily demand for each distribution area, and planning. A data set made up of various data such as hourly demand) is stored.

  The actual value database 24b stores actual data of water demand for a predetermined period in the past from the planning date. Specifically, for example, date, day of the week, weather, maximum temperature, information on whether or not it is a special day, and actual values of various demands (total demand for the entire distribution area, daily unit of each distribution area A data set composed of various kinds of performance data such as demand amount and demand amount at the time of planning) is stored. Note that the retention period of the data stored in the record value database 24b is appropriately set according to, for example, the predicted region and environment.

  The similar condition database 24c stores (registers) data of an element correspondence table that is referred to when the prediction condition is a rare condition.

  An element is classification information (section) set in advance for each prediction condition. Specifically, when the prediction condition is a weather condition, for example, weather information such as clear, clear, cloudy, rain, snow, and sleet is an element. When the prediction condition is a temperature condition, for example, temperature classification information such as less than 0 ° C., 0 ° C. or more and less than 10 ° C., 10 ° C. or more and less than 20 ° C., 20 ° C. or more is an element. Further, when the prediction condition is a day condition, each day information from Monday to Sunday is an element. Furthermore, when the prediction condition is a peculiar day, peculiar day information such as Golden Week (GW), Bon Festival, year-end and New Year is an element. It should be noted that the setting of such elements can be appropriately changed according to, for example, the predicted region and environment.

  The element correspondence table is table data indicating a correspondence relationship between elements having similar demand model patterns, and is a data indicating a correlation between the elements. In the present embodiment, an element correspondence table is created for each type of prediction condition (for example, weather condition, temperature condition, day condition, etc.). For example, an element correspondence table of weather conditions indicates a correlation between elements such as clear, clear, cloudy, rain, snow, and sleet.

  In the present embodiment, when the prediction condition is a rare condition, a similar condition corresponding to the prediction condition is extracted from the element correspondence table. More specifically, in the element correspondence table of the weather conditions, for example, when the element “cloudy” is registered as an element having high correlation with the element “sunny” (similar condition), “sunny” is input as the prediction condition. In this case, the element “cloudy” is output as the similar condition. The element correspondence table is created (updated) by comparing demand model patterns between elements using a demand model pattern created (updated) from the actual values measured daily. A specific method for generating the element correspondence table will be described in detail later.

  The demand pattern database 24d stores demand model pattern data used when calculating predicted values (various demand prediction parameters) of various demand quantities.

2A to 2D show data configuration examples of the total demand model pattern used when calculating the predicted value of the total demand (Q _A ) of the entire water distribution area. In the present embodiment, the total demand model pattern (J _a ) is collected as table data for each element as shown in FIGS. That is, since the total demand model pattern (J _a ) varies depending on the element f, the element f is treated as a variable and the total demand model pattern is denoted as J _a (f) below.

FIG. 2A is a data configuration example of the total demand model pattern J _a (f) for each weather element. In the example shown in FIG. 2A, A1 (%) to A4 (%) in FIG. 2A indicate the total demand model pattern J when the weather element f is clear, cloudy, rainy, and snow, respectively. _a is the value of (f).

FIG. 2B is a data configuration example of the total demand model pattern J _a (f) for each temperature element. In the example shown in FIG. 2 (b), B1 (%) to B5 (%) in FIG. 2 (b) have temperature elements f of 10 ° C. or less, 11 to 15 ° C., 16 to 20 ° C., 21 to 21, respectively. This is the value of the total demand model pattern J _a (f) when the temperature is 25 ° C. and 26 ° C. or higher.

FIG. 2C is a data configuration example of the total demand model pattern J _a (f) for each day element. In the example shown in FIG. 2 (c), C1 (%) to C7 (%) in FIG. 2 (c) indicate the total demand model pattern J _a (f) when the day component f is Monday to Sunday, respectively. Is the value of

FIG. 2D is a data configuration example of the total demand model pattern J _a (f) for each specific day element. In the example shown in FIG. 2 (d), D1 (%) to D4 (%) in FIG. 2 (d) indicate the total demand when the singular day elements f are Golden Week (GW), Bon Festival, year-end and New Year respectively. This is the value of the quantity model pattern J _a (f).

And when calculating the predicted value of the total demand amount (Q _A ) of the entire water distribution area, the demand prediction calculation unit 22 or the similar condition search unit 23 sets the search condition (prediction condition or similar condition) of the demand model pattern. Based on the element f, the corresponding total demand model pattern J _a (f) is extracted from various table data as shown in FIGS.

FIGS. 3A to 3D show data configuration examples of a daily demand amount model pattern used when calculating a predicted value of the daily demand amount (Q _HD ) in each water distribution area. In the present embodiment, the daily demand model pattern (J _hd ) of each water distribution area is compiled as table data for each element as shown in FIGS. 3 (a) to 3 (d). That is, the daily demand model pattern (J _hd ) of each water distribution area varies depending on the water distribution area d (= d1 to dn) and the element f (= weather element, temperature element, day element or specific day element). Hereinafter, the water distribution area d and the element f are treated as variables, and the daily demand model pattern of each water distribution area d is denoted as J _hd (d, f).

FIG. 3A is a data configuration example of the daily demand model pattern J _hd (d, f) for each water distribution area d in each weather element f. In the example shown in FIG. 3A, AX11 (%) to AX4N (%) in FIG. 3A are supplied to each water distribution area d when the weather element f is clear, cloudy, rainy, and snow, respectively. This is a value indicating the ratio of the daily water distribution amount, and is the value of the daily demand model pattern J _hd (d, f) for each weather element.

FIG. 3B is a data configuration example of the daily demand model pattern J _hd (d, f) of each water distribution area d in each temperature element f. In the example shown in FIG. 3 (b), BX11 (%) to BX5N (%) in FIG. 3 (b) have temperature elements f of 10 ° C. or less, 11 to 15 ° C., 16 to 20 ° C., 21 to 25, respectively. It is a value indicating the ratio of the daily water distribution amount supplied to each water distribution area d when it is at ℃ and 26 ℃ or more, and is the value of the daily demand model pattern J _hd (d, f) for each temperature element is there.

FIG.3 (c) is a data structural example of the daily unit demand model pattern _Jhd (d, f) of each water distribution area d in each day element f. In the example shown in FIG. 3 (c), CX11 (%) to CX7N (%) in FIG. 3 (c) indicate the water distribution amount supplied to each water distribution area d when the day element f is Monday to Sunday, respectively. Is a value of the daily demand model pattern J _hd (d, f) for each day of the week element.

FIG. _3D is a data configuration example of the daily unit demand model pattern J _hd (d, f) of each water distribution area d in each specific day element f. In the example shown in FIG. 3 (d), DX11 (%) to DX4N (%) in FIG. 3 (d) indicate the distribution of water when the singular day elements f are Golden Week (GW), Bon Festival, year-end and New Year respectively. This is a value indicating the ratio of the amount of water supplied to the area d, and is a value of the daily unit demand model pattern J _hd (d, f) for each specific day element.

And when calculating the predicted value of the daily demand amount (Q _HD ) in each water distribution area d, the demand prediction calculating unit 22 or the similar condition searching unit 23 searches the demand model pattern search condition (prediction condition or similar). The corresponding daily demand model pattern J _hd (d, f) is shown in FIGS. 3A to 3D based on the element f of the condition) and the information (d1 to dn) of the water distribution area d. Extract from various table data.

4A to 4D, a planning unit demand amount model used when calculating a predicted value of a demand amount (Q _HT ) in a planning time unit in a predetermined distribution area d (any of d1 to dn). The example of a data structure of a pattern is shown.

In the present embodiment, the planning unit demand model pattern (J _ht ) for each water distribution area is compiled as table data for each element as shown in FIGS. 4 (a) to (d). In other words, the planning unit demand model pattern (J _ht ) of each water distribution area varies depending on the element f (= weather element, temperature element, day element or specific day element) and the planning time t (= t1 to tm). Further, the planning unit demand model pattern (J _ht ) varies depending on the water distribution area d (= d1 to dn). Therefore, hereinafter, the distribution area d, the element f, and the planning time t are treated as variables, and the planning unit demand model pattern in each distribution area d is described as J _ht (d, f, t).

  Moreover, in the example shown to Fig.4 (a)-(d), it is set to 24 hours (1 day) in one cycle of planning time t = t1-tm. Specifically, when the demand forecast planning interval Δt is set to one hour, for example, the planning time t = 1, 2,..., 24 [hours] can be set. Furthermore, in this embodiment, the data set shown to Fig.4 (a)-(d) is set for every water distribution area.

FIG. 4A is a data configuration example of a unit demand model pattern J _ht (d, f, t) at the time of planning for a predetermined water distribution area d in each weather element f. In the example shown in FIG. 4A, AT11 (%) to AT4M (%) in FIG. 4A are supplied at each planning time t when the weather element f is clear, cloudy, rainy, and snow, respectively. This is a value indicating the ratio of the amount of water distribution, and is a value of the unit demand model pattern J _ht (d, f, t) at the time of planning for each weather element.

FIG. 4B is a data configuration example of a planning unit demand model pattern J _ht (d, f, t) for a predetermined water distribution area d in each temperature element f. In the example shown in FIG. 4B, BT11 (%) to BT5M (%) in FIG. 4B have temperature elements f of 10 ° C. or less, 11 to 15 ° C., 16 to 20 ° C., 21 to 25, respectively. It is a value indicating the proportion of the water distribution supplied at each planning time t when the temperature is at or above 26 ° C., and the value of the planning unit demand model pattern J _ht (d, f, t) by temperature element is there.

FIG. 4C is a data configuration example of a unit demand model pattern J _ht (d, f, t) at the time of planning for a predetermined water distribution area d in each day element f. In the example shown in FIG. 4 (c), CT11 (%) to CT7M (%) in FIG. 4 (c) indicate the water distribution amount supplied at each planning time t when the day element f is Monday to Sunday, respectively. Is a value of the planning unit demand model pattern J _ht (d, f, t) for each day of the week element.

FIG. 4D is a data configuration example of a unit demand model pattern J _ht (d, f, t) at the time of planning for a predetermined water distribution area d in each specific day element f. In the example shown in FIG. 4 (d), DT11 (%) to DT4M (%) in FIG. 4 (d) are the planning times when the singular day elements f are Golden Week (GW), Bon Festival, year-end and New Year respectively. This is a value indicating the proportion of the water distribution supplied every t, and is a value of the unit demand model pattern J _ht (d, f, t) at the time of planning for each specific day element.

And when calculating the predicted value of the demand amount (Q _HT ) in the planning time unit in each water distribution area d, the demand prediction calculating part 22 or the similar condition searching part 23 is the information (d1 to dn) of the water distribution area d. Based on the element f of the demand model pattern search condition (forecast condition or similar condition) and the planning time t (t1 to tm), the corresponding planning unit demand model pattern J _ht (d, f, t) Are extracted from various table data as shown in FIGS.

  In the present embodiment, as described above, an example in which a weather element, a temperature element, a day element, and a specific day element are used as elements has been described. However, the present invention is not limited to this and affects demand forecast. Any parameter (condition) that gives can be used as an element.

[Calculation method of demand forecast parameters]
Next, a method for calculating various demand prediction parameters in the demand prediction calculation unit 22 will be described. In the present embodiment, when the prediction condition input to the demand prediction device 2 via the input / output display device 1 is not a rare condition, the demand prediction parameter is calculated based on the prediction condition. On the other hand, when the forecast condition is a rare condition, the demand forecast parameter is calculated based on the similar condition corresponding to the forecast condition stored in the similar condition database 24c.

In the present embodiment, the demand prediction parameters calculated by the demand prediction calculation unit 22 at the time of demand prediction are the following three. Note that the various demand forecast parameters calculated are stored in the forecast value database 24a.
(1) The total water demand per day Q _A (i) [m ³ / day] predicted in all water distribution areas on the demand forecasting date i (forecast date)
(2) Daily demand Q _HD (i, d) [m ³ / day] predicted in each distribution area d on the planning date i
(3) The water demand Q _HT (i, d, t) [m ³ / Δt] predicted in each water distribution area d at every planning time t (for each planning period Δt) on the planning date i.

Note that by calculating the demand amount Q _HT (i, d, t) for each planning time in (3) above, the demand amount (demand prediction) change characteristic for each planning time t can be obtained. Further, the demand forecast planning interval Δt can be set arbitrarily.

First, a method for calculating various demand parameters when a specific day is not included in the prediction condition will be described. First, the demand forecasting device 2 has table data for each element (f1: weather element, temperature element or day of the week element) stored in the demand pattern database 24d based on a forecast condition or a similar condition corresponding to the forecast condition (figure 1). 2-4). The demand prediction calculation unit 22 then calculates the total demand model pattern J _a (f1) of the element f1 corresponding to the prediction condition or the similar condition, and the daily demand model pattern J _hd (d, f1) of each water distribution area d. ), And the planning unit demand amount model pattern model pattern J _ht (d, f1, t) for each water distribution area d.

  Next, the demand prediction calculation unit 22 searches the actual value database 24b, and the total demand amount Q (i−j) (j) for the past predetermined period (here, an example for the past k days will be described) from the planning date i. = 1 to k) is acquired. In addition, the period (k days) of the past actual value acquired here is suitably set according to the area, environment, etc. to predict, for example.

Then, the demand prediction calculation unit 22 uses the acquired total demand model pattern J _a (f1) and the actual value of the total demand Q (ij) for the past predetermined period, on the planning date i The total amount of water demand Q _A (i) per day predicted in step 1 is calculated by the following formula 1.

That is, on the planning date i, the total water demand Q _A (i) per day predicted in all water distribution areas is the average value of the total demand (Q) for the past k days. This is a value obtained by multiplying J _a (f1).

Next, the demand forecast calculation unit 22 calculates the daily demand model pattern J _hd (d, f1) for each acquired water distribution area d and the daily forecast for the planning date i calculated by the above formula 1. Using the demand amount Q _A (i), the daily demand amount Q _HD (i, d) of water predicted in each distribution area d on the planning date i is calculated by the following formula 2.

Next, the demand prediction calculation unit 22 predicts the acquired unit demand amount model pattern J _ht (d, f1, t) of each water distribution area d and each water distribution area d calculated by the above formula 2. Using the daily demand amount Q _HD (i, d), the water demand amount Q _HT (i, d, t) predicted for each planning time t in each water distribution area d is calculated by the following Equation 3.

  When the forecast condition does not include a specific day, various demand prediction parameters are calculated as described above. When the forecast condition includes a specific day (when the planning date i is a specific day), Various demand prediction parameters are calculated as follows.

First, the demand prediction device 2 uses table data (FIG. 2 (d) to 4) according to specific day elements (f2: for example, GW, Obon, year-end or new year) stored in the demand pattern database 24d based on the conditions of the specific days. Search (see (d)). Then, the demand prediction device 2 includes the total demand model pattern J _a (f2) corresponding to the singular day element f2, the daily demand model pattern J _hd (d, f2) of each distribution area d, and each distribution area. A unit demand amount model pattern J _ht (d, f2, t) at the time of planning is acquired.

Next, the demand prediction calculation unit 22 calculates the total demand amount model of the singular day element f2 to the total demand amount Q _A (i) per day of all the water distribution areas predicted on the planning date i, which is calculated by the above formula 1. By multiplying the pattern J _a (f2), the total amount of water demand Q _A (i) per day predicted in all the water distribution areas on a specific day is calculated. That is, the total amount of water demand Q _A (i) per day predicted when the planning date i is a peculiar day is calculated based on the following Equation 4.

In addition, the demand prediction calculation unit 22 adds the daily demand amount Q _HD (i, d) predicted in each water distribution area d calculated by Equation 2 to the daily unit of each water distribution area d in the specific day element f2. The demand amount model pattern J _hd (d, f2) is multiplied to calculate the daily demand amount Q _HD (i, d) of water predicted in each water distribution area d on a specific day. That is, the daily water demand Q _HD (i, d) predicted in each water distribution area d on a specific day is calculated based on the following Equation 5.

Furthermore, the demand prediction calculation unit 22 uses the water demand Q _HT (i, d, t) for each planning time t predicted in each water distribution area d calculated by the above mathematical formula 3 for each water distribution in the singular day element f2. Multiplying the unit demand model pattern J _ht (d, f2, t) at the time of planning for the zone d, the water demand Q _HT (i, d) at each planning time t predicted in each water distribution zone d on a specific day , T). That is, the water demand Q _HT (i, d, t) for each planning time t predicted in each water distribution area d on a specific day is calculated based on the following Equation 6.

In the present embodiment, as described above, various demand prediction parameters (Q _A (i), Q _HD (i, d), and Q _HT (i, d, t)) are calculated based on the formulas 1 to 6. .

  In the present embodiment, the example in which the demand prediction is performed based on the demand prediction parameters calculated by the above mathematical formulas 1 to 6 has been described. However, the present invention is not limited to this, and any demand prediction parameter other than the above is separately calculated. Then, the demand prediction may be performed. In addition, any parameter can be used as long as it is a parameter having the same meaning as the demand forecast parameter calculated in the above formulas 1 to 6. In this case, the formula for calculating the demand forecast parameter is a formula other than the formulas 1 to 6. A calculation formula can be used.

[Similar correspondence table creation method]
Next, a method for creating an element correspondence table stored in the similar condition database 24c will be described. In the present embodiment, as described above, an element correspondence table is created for each type of prediction condition.

As described above, the element correspondence table is created (updated) using a demand model pattern created (updated) based on the actual values measured every day, not limited to the days when the prediction conditions are rare. Specifically, in the present embodiment, the error of the demand model pattern is calculated between the predetermined element and each of the other elements, and the other element having the smallest error is registered as the similarity condition of the predetermined element. Here, an example will be described in which a similar correspondence table is created (updated) using the planning unit demand model pattern J _ht (d, f, t).

First, the similar condition creation unit 25 acquires a planning unit demand model pattern J _ht (d, f, t) for each planning time t in a predetermined water distribution area d from the demand pattern database 24d. Then, at each planning time t, an error is calculated by comparing the planning-time unit demand model pattern J _ht (d, f, t) between the predetermined element and each of the other elements.

  Next, an error for each planning time t calculated between the predetermined element and each of the other elements is added to calculate an error for one day. Then, the error for one day calculated between the predetermined element and each of the other elements is compared, and the other element having the smallest error is registered as a similarity condition. In the present embodiment, the similarity correspondence table is created by performing the similar condition registration process for each element. When the error between the elements does not fall within the predetermined range, information indicating that the similarity condition for the predetermined element is “none” is registered in the element correspondence table.

As described above, when the element correspondence table is created by comparing the unit demand model pattern J _ht (d, f, t) at the time of planning, the element correspondence table can be created for each distribution area. The present invention is not limited to this, and the element correspondence table may be created by comparing the total demand model pattern J _a (f) and the daily demand model pattern J _hd (d, f).

In the above-described similar condition registration process, an example has been described in which the error of the planning unit demand model pattern J _ht (d, f, t) between the predetermined element and each of the other elements is directly calculated. Is not limited to this. For example, the error ΔJ _HT (distribution area planning unit error) of the planning unit demand model pattern J _ht (d, f, t) between the predetermined element P and each other element S in each water distribution area is expressed by the following formula 7 You may calculate by. Note that the following formula 7 is derived from the above formula 3.

Q _HDP (i, d) and Q _HTP (i, d, t) in Equation 7 are respectively the daily demand amount Q _{HD of} the water distribution area d calculated using the demand model pattern of the predetermined element P. (I, d) and the demand amount Q _HT (i, d, t) for each planning time t. In addition, Q _HDS (i, d) and Q _HTS (i, d, t) in Equation 7 are respectively calculated for the daily distribution area d calculated using the demand model pattern of each other element S. The demand amount Q _HD (i, d) and the demand amount Q _HT (i, d, t) for each planning time t. Note that the various demand amounts in Equation 7 are calculated by the similar condition creating unit 25 based on the actual values updated daily and the various demand model patterns.

  In the above embodiment, the example in which the element correspondence table is created for each prediction condition (for example, the weather condition, the temperature condition, or the day condition) has been described, but the present invention is not limited to this. For example, the element correspondence table may be created by calculating an error of the demand model pattern between elements regardless of the prediction conditions. Specifically, not only does the demand model pattern compare between a given element in the weather condition and each other element in the weather element, but also the given element in the weather element and each element in the temperature element and day of the week element The element correspondence table may be created by comparing the demand model pattern with.

[Processing when forecasting demand]
Next, the operation | movement of the demand prediction process in the demand prediction apparatus 2 of this embodiment is demonstrated concretely, referring FIG. In addition, FIG. 5 is a flowchart which shows the process sequence of the demand prediction method of this embodiment.

  First, the operator performs a predetermined operation on the input unit of the input / output display device 1, and inputs prediction conditions such as planning date, planning time, weather, maximum temperature, and day of the week (step S1).

  Next, the appearance frequency determination unit 21 in the demand prediction device 2 acquires the input prediction condition and determines whether the prediction condition is a rare condition (step S2). Specifically, the appearance frequency determination unit 21 searches the actual value database 24b, and whether the application frequency under the condition equivalent to the prediction condition is equal to or lower than a predetermined frequency (including a case where it does not appear) in the past certain period. Judge whether or not.

  And when the application frequency of the conditions equivalent to prediction conditions is below predetermined frequency, the appearance frequency determination part 21 determines with the input prediction conditions being rare conditions. The condition equivalent to the prediction condition here includes not only the condition that the numerical value matches but also the condition that the numerical value is within a predetermined range when the prediction condition is a numerical value such as an air temperature condition. Meaning.

  In step S2, when the appearance frequency determination unit 21 determines that the prediction condition is not a rare condition, step S2 is NO. In this case, the appearance frequency determination unit 21 outputs the determination result to the demand prediction calculation unit 22, and the demand prediction calculation unit 22 performs the processing after step S <b> 5 described later based on the determination result.

  On the other hand, when the appearance frequency determination unit 21 determines in step S2 that the prediction condition is a rare condition, step S2 is YES. In this case, based on the determination result, the demand prediction calculation unit 22 outputs a command signal for searching for a similar condition to the similar condition search unit 23.

  Next, the similar condition search unit 23 acquires a prediction condition based on the command signal input from the demand prediction calculation unit 22, and searches the similar condition database 24c based on the prediction condition. Then, the similar condition search unit 23 compares the acquired prediction condition with the data (element correspondence table) in the similar condition database 24c, and determines whether or not there is a similar condition (element) corresponding to the prediction condition. (Step S3). Note that the processing procedure of the search processing of the similar condition database 24c in step S3 will be described in detail later.

  In step S3, when there is no similar condition corresponding to the prediction condition, step S3 is NO. In this case, the similar condition search unit 23 outputs the determination result to the demand prediction calculation unit 22, and the demand prediction calculation unit 22 performs the processing after step S <b> 5 described later based on the determination result.

  On the other hand, if there is a similar condition corresponding to the prediction condition in step S3, step S3 is YES. In this case, the similar condition search unit 23 acquires the similar condition (element) (step S4).

Next, the demand prediction device 2 searches the demand pattern database 24d based on the inputted prediction condition or similar condition (step S5), and various demand model patterns (total demand) corresponding to the prediction condition or similar condition. The quantity model pattern J _a (f), the daily unit demand model pattern J _hd (d, f), and the planning unit demand model pattern J _ht (d, f, t)) are acquired.

  If the prediction condition is not a rare condition, or if the similar condition does not exist in the similar condition database 24c, in step S5, the demand prediction calculation unit 22 searches the demand pattern database 24d based on the prediction condition, and responds to it. Acquire various demand model patterns. On the other hand, when the prediction condition is a rare condition and the similar condition exists in the similar condition database 24c, in step S5, the similar condition search unit 23 searches the demand pattern database 24d on the basis of the similar condition, and copes with it. Acquire various demand model patterns. Then, the similar condition search unit 23 outputs the acquired various demand model patterns to the demand prediction calculation unit 22.

  Subsequently, the demand prediction calculation part 22 searches the performance value database 24b, and acquires the performance value of the total demand (Q) for the past predetermined period from the planning date. And the demand prediction calculating part 22 calculates various demand prediction parameters according to Numerical formulas 1-6 mentioned above using the acquired actual value and various demand model patterns, and performs demand prediction (step S6).

  Next, the demand prediction calculation unit 22 outputs the calculated demand prediction result to the input / output display device 1. The input / output display device 1 displays the calculated demand prediction result on the display unit (step S7).

  Next, the operator confirms the demand prediction result displayed on the display unit of the input / output display device 1, and determines whether or not the prediction result is suitable for actual water operation (step S8).

  In Step S8, when the operator determines that the calculated demand prediction result is inappropriate for the actual water operation, Step S8 is NO. In this case, the operator performs a predetermined operation on the input unit of the input / output display device 1 to correct the prediction condition (step S9). After that, it returns to step S5 and repeats the process after step S5.

  In the present embodiment, an example has been described in which, when step S8 is NO, the process returns to step S5 via step S9, but the present invention is not limited to this. For example, when step S8 is NO, the process may return to step S2 via step S9 to determine again whether the corrected prediction condition is a rare condition. Moreover, when step S8 becomes NO determination, you may return to step S1 directly from step S8.

  On the other hand, when the operator determines in step S8 that the calculated demand prediction result is appropriate for the actual water operation, step S8 is YES. In this case, the demand prediction calculation unit 22 stores the calculated demand prediction result in the predicted value database 24a (step S10), and ends the demand prediction process. In this embodiment, the demand for water distribution is predicted in this way.

[Similar condition search processing method]
Here, an example of the similar condition search process performed in step S3 when the input prediction condition is a rare condition will be described with reference to FIG. FIG. 6 is a flowchart showing the procedure of the search process for similar conditions.

  In the present embodiment, as will be described later, it is determined whether the input prediction condition is a single condition or a compound condition. The single condition is that the prediction condition is a weather condition, a temperature condition, or the like. , Day-of-the-week conditions, and special day conditions. The composite condition is a condition obtained by combining a plurality of these single conditions (for example, a combination of weather conditions and temperature conditions, etc.), and is a condition that is established as a rare prediction condition.

  In the present embodiment, when the prediction condition is a composite condition, a search is performed by preferentially selecting a condition that has a greater influence on the demand prediction from the composite condition. That is, in this embodiment, when the prediction condition is a composite condition, a condition having a high degree of influence on the demand prediction is selected and a similar condition is searched. Note that this priority can be changed as appropriate according to, for example, the predicted region and environment. Here, it is assumed that the degree of influence on the demand prediction is high in the order of the weather condition, the temperature condition, and the day of the week condition.

  Furthermore, in this embodiment, when a specific day is included in the composite condition, the search is performed using a condition other than the specific day. Singular days are May consecutive holidays, Obon, New Year, local events, and so on, so the conditions are very low in appearance frequency (appearance frequency about once a year). Therefore, it is unlikely that the element correspondence table updated with the recent actual value reflects the influence of the target specific day, and the priority of this condition is low in terms of prediction accuracy. Therefore, although the search process described below searches for similar conditions within a certain past period, in this embodiment, from the viewpoint of prediction accuracy, it is assumed that a singular day does not appear within the past certain period. Is not selected as a search condition for similar conditions.

  First, the similar condition search unit 23 determines whether or not the input prediction condition (rare prediction condition) is a single condition (step S11).

  In step S11, when the prediction condition is a single condition, step S11 is YES. In this case, the similar condition search unit 23 searches for a similar condition using a single condition (step S20).

  On the other hand, if the prediction condition is a composite condition in step S11, step S11 is NO. In this case, the similar condition search unit 23 determines whether the weather condition is included in the composite condition (step S12).

  In step S12, when the weather condition is included in the composite condition, step S12 is YES. In this case, the similar condition search unit 23 selects a weather condition from the composite conditions (step S13). Thereafter, the similar condition search unit 23 searches for similar conditions using the weather conditions (step S20).

  On the other hand, when the weather condition is not included in the composite condition in step S12, step S12 is NO. In this case, the similar condition search unit 23 determines whether or not the temperature condition is included in the composite condition (step S14).

  In step S14, when the temperature condition is included in the combined condition, step S14 is YES. In this case, the similar condition search unit 23 selects an air temperature condition from the composite conditions (step S15). Thereafter, the similar condition search unit 23 searches for the similar condition using the temperature condition (step S20).

  On the other hand, if the temperature condition is not included in the composite condition in step S14, step S14 is NO. In this case, the similar condition search unit 23 determines whether or not the day condition is included in the composite condition (step S16).

  If the day condition is included in the composite condition in step S16, step S16 is YES. In this case, the similar condition search unit 23 selects a day of the week condition from the complex conditions (step S17). Thereafter, the similar condition search unit 23 searches for similar conditions using the day of the week conditions (step S20).

  On the other hand, if the day condition is not included in the composite condition in step S16, step S16 is NO. In this case, the similar condition search unit 23 determines whether or not a unique day is included in the composite condition (step S18).

  In step S18, when a peculiar day is included in the composite condition, step S18 is YES. In this case, the similar condition search unit 23 selects a condition other than the specific day from the composite conditions (step S19). At this time, when there are two or more conditions other than the specific day, the similar condition search unit 23 selects an arbitrary condition from the conditions. Thereafter, the similar condition search unit 23 searches for the similar condition using the condition selected in step S19 (step S20).

  On the other hand, if the unique condition is not included in the composite condition in step S18, step S18 is NO. In this case, the similar condition search unit 23 selects an arbitrary condition from the complex conditions. And the similar condition search part 23 searches for a similar condition using the selected conditions (step S20).

  In the present embodiment, when the prediction condition is a rare condition, the similar condition search unit 23 searches for the similar condition as described above. In the above-described example, when the weather condition, the temperature condition, and the day of the week condition are not included in the composite condition, the search condition is arbitrarily selected from the composite condition. However, the present invention is not limited to this. If the priority (impact on the demand forecast) is set for the forecast conditions other than the weather conditions, temperature conditions, and day conditions, and the combined conditions do not include the weather conditions, temperature conditions, and day conditions The search condition may be selected based on the priority order.

  As described above, in the demand prediction device 2 of the present embodiment, when the prediction condition is a rare condition, an element correspondence table is created using a demand model pattern created (updated) based on the actual values measured every day ( Update), and the demand is predicted based on the similar condition extracted from the element correspondence table.

  That is, in the demand prediction apparatus 2 of the present embodiment, even if the prediction condition is rare, demand prediction can be made based on more past actual values, particularly recent actual values. Accuracy can be improved. As a result, it is possible to calculate a highly accurate plan value even when an allocation plan based on the demand forecast is made. Therefore, in the water operation monitoring system 100 of the present embodiment, more stable water supply can be realized.

  Further, in the present embodiment, it is possible to promote automation of processing that has been conventionally involved by an operator, and therefore, it is possible to reduce the number of system management personnel.

  In the above-described embodiment, an example in which one prediction condition or similar condition is used when searching for and obtaining a demand model pattern has been described, but the present invention is not limited to this. The demand model pattern may be searched using two or more prediction conditions or similar conditions (elements). In this case, an average value of a plurality of acquired demand model patterns may be used as the demand model pattern, or a demand model pattern having elements close to the update date may be adopted.

  DESCRIPTION OF SYMBOLS 1 ... Input / output display apparatus, 2 ... Demand prediction apparatus, 3 ... Process data interface apparatus, 21 ... Appearance frequency determination part, 22 ... Demand prediction calculation part, 23 ... Similar condition search part, 24 ... Database part, 24a ... Predicted value Database, 24b ... Actual value database, 24c ... Similar condition database, 24d ... Demand pattern database, 25 ... Similar condition creation unit, 26 ... Demand pattern creation unit, 100 ... Water operation monitoring system

Claims (7)

Appearance frequency determination for determining whether or not the past appearance frequency of the prediction condition is equal to or lower than a predetermined frequency based on the prediction condition of the water demand and the actual value indicating the past water demand situation stored in the actual value database And
When the appearance frequency determining unit determines that the past appearance frequency of the prediction condition is equal to or lower than a predetermined frequency, the similar condition that is created based on the past actual value and correlates with the prediction condition is stored. The similar condition database is searched to acquire the similar condition corresponding to the forecast condition, and the demand model pattern data corresponding to the acquired similar condition is obtained from the demand pattern database storing the demand model pattern data of water demand. A similar condition search unit to be acquired;
Acquire the actual value of past water demand from the actual value database, based on the acquired actual value and the data of the demand model pattern corresponding to the similar condition acquired in the similar condition search unit, A demand prediction apparatus comprising a demand prediction calculation unit that performs demand prediction. Further, the actual value database;
The similar condition database;
The demand prediction apparatus according to claim 1, further comprising the demand pattern database.   The demand prediction calculation unit, when the appearance frequency determination unit determines that the past appearance frequency of the prediction condition is not equal to or less than a predetermined frequency, stores the demand model pattern data corresponding to the prediction condition in the demand pattern database The demand prediction device according to claim 1, wherein the demand prediction device is obtained from the following.   When the prediction condition is a complex condition, the similar condition search unit selects the prediction condition having the highest degree of influence on the demand prediction from the prediction conditions when searching the similar condition database. It performs, The demand prediction apparatus as described in any one of Claims 1-3 characterized by the above-mentioned.   The similar condition search unit, when the prediction condition is a compound condition and a specific date is included in the prediction condition, the search is performed based on conditions other than the specific date when searching the similar condition database. It performs, The demand prediction apparatus as described in any one of Claims 1-4 characterized by the above-mentioned.   The demand prediction calculation unit includes a total water demand on a demand forecast planning date, a total water demand on each water distribution area on the planning date, and a water demand at predetermined time intervals on each water distribution area on the planning date. The demand prediction device according to claim 1, wherein an amount is calculated. An input unit for inputting water demand prediction conditions;
An actual value database for storing actual values indicating past water demand conditions;
A similar condition database that is created based on the past actual values and stores similar conditions correlated with the prediction conditions;
A demand pattern database for storing water demand pattern data;
Based on the prediction condition input to the input device unit and the actual value indicating the past water demand situation stored in the actual value database, the past appearance frequency of the prediction condition is a predetermined frequency or less. An appearance frequency determination unit that determines whether or not
When the appearance frequency determination unit determines that the past appearance frequency of the prediction condition is equal to or lower than a predetermined frequency, the similarity condition database is searched to acquire the similarity condition corresponding to the prediction condition, and A similar condition search unit that acquires data of the demand model pattern corresponding to the acquired similar condition from the demand pattern database;
Acquire the actual value of past water demand from the actual value database, based on the acquired actual value and the data of the demand model pattern corresponding to the similar condition acquired in the similar condition search unit, A water operation monitoring system comprising: a demand prediction calculation unit that performs demand prediction; and a result value acquisition unit that acquires data related to the result value stored in the result value database.

Images