JP2020201609A - Water demand prediction system and method thereof - Google Patents

Abstract

To provide a waterworks system capable of accurately predicting a water demand even when a change in the water demand is non-uniform.SOLUTION: A computer system in a waterworks system: acquires a water demand of a prescribed water distribution area, and weather information of the water distribution area containing forecast and day-of-week information; sets weather information and day-of-week information of a prescribed period and a water demand of a latest prescribed period as related conditions to the water demand of the prescribed period; and calculates a similarity between related information on a prediction day and the related conditions on each day of a past prescribed period, and predicts a water demand of the prediction day by preferentially using a water demand actual achievement value of a past day having a high similarity condition.SELECTED DRAWING: Figure 1

Description

The present invention relates to a water demand forecasting system for predicting the amount of water demand for each reference period for a predetermined distribution zone in which water is distributed from a water distribution facility via a water pipe network, and a method thereof.

Conventionally, control for distributing water to consumers including ordinary households has been applied to water supply facilities. Water demand fluctuates not only every day, every predetermined time, day of the week and time, but also due to external factors such as weather and temperature. Therefore, there is a conventional example of trying to accurately predict the amount of water demand.

For example, in Japanese Patent Application Laid-Open No. 2012-99049, the actual value of water demand on the same day as the predicted date is obtained from the past predetermined period, and the average for each time is calculated for 24 hours a day. A water distribution plan forecasting system that forecasts water demand using water demand pattern values is disclosed.

This water distribution plan forecasting system acquires the latest actual water demand value, and when the deviation between the water demand pattern value (forecast value) and the actual value deviates from a predetermined allowable range, the deviation between the pattern value and the actual value. In order to correct the pattern value so that the amount of water is reduced, a certain amount is added or subtracted from the pattern value in an attempt to improve the accuracy of water demand forecasting.

Japanese Unexamined Patent Publication No. 2012-99049

The water distribution plan prediction system described above can improve the prediction accuracy of water demand in cases where water demand increases or decreases uniformly, but water demand is changed, for example, according to the weather. However, in the case where the water is shifted, moved forward, or postponed in time, the accuracy of water demand forecasting is adversely deteriorated by correcting the pattern value. Therefore, an object of the present invention is to provide a system capable of accurately predicting water demand even in cases where changes in water demand are not uniform.

In order to achieve the above object, the present invention is to predict the amount of water demand for each reference period for a predetermined water distribution area in which water is distributed from the water distribution facility via the distribution pipe network. The water demand amount, the weather information of the distribution zone, and the daytime information are acquired, and as the related conditions for the water demand amount of the reference period, the weather information of the predetermined period, the daytime information, and the predetermined range before the reference period. Water demand is set, and the degree of similarity between the related conditions in each reference period of the past predetermined period and the related conditions in the reference period for which the water demand is to be predicted is calculated, for example, the degree of similarity is high. By preferentially using the actual value of water demand in the reference period with conditions, the water demand in the reference period for which the water demand is to be predicted is predicted.

According to the present invention, it is possible to provide a system that can accurately predict the water demand even in the case where the change in the water demand is not uniform.

This is an example of a block diagram of a water supply system including a computer system as a water demand forecasting system. This is an example of a water demand data management table. This is an example of a day / weather data management table. This is an example of the setting screen for water demand forecast conditions. This is an example of a forecast condition management table. It is explanatory drawing of the kernel ridge regression model. It is a graph explaining the relationship between the latest water demand data of the forecast date and the learning date. It is an explanatory graph of the relationship between the latest water demand data on the forecast date and the learning date. It is a graph explaining the relationship between the latest water demand data of the forecast date and the learning date. This is an example of a water demand forecast data management table. This is an example of the prediction result display data management table. This is an example of a prediction result display screen. This is an example of the model parameter management table.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. The present invention is not limited to the embodiments described below.

FIG. 1 shows a block configuration diagram of the water supply system 102. The water supply system includes a water distribution facility 100, a monitoring and control system (SCADA) 103 for the water distribution facility, a computer system 101 for predicting water demand, and a network 111 for connecting the monitoring and control system 103 and the computer system 101. ..

The water distribution facility includes a water purification plant 104, a pump 105, a distribution reservoir 106, a water pipe network 109, and measuring instruments 107 and 108, and the monitoring and control system 103 monitors and controls the pump 105 and measuring instruments 107 and 108. ..

The water purified at the water purification plant 104 is pumped to the distribution reservoir 106 by the pump 105, temporarily stored in the distribution reservoir 106, and then distributed to the customer 110 via the distribution pipe network 109. The pump 105, the water level meter 107, and the flow meter 108 measure the operating state (ON / OFF state) of the pump 105, the water level of the distribution reservoir 106, and the amount of water distributed to the customer 110 (water demand amount) in a 1-minute cycle, respectively. , Sends to the monitoring and control system 103.

The monitoring and control system 103 acquires the measured value of water demand (in units of 1 minute) of the target distribution area, which distributes water from the water distribution facility via the distribution pipe network 109, and takes the sum of the 1 hour to be taken every hour. The hourly demand amount is totaled and transmitted to the computer system 101 via the network 111.

The weather information providing system (weather information source) 112, which exists outside the water supply system 102, updates the actual value of the weather information in the area including the target distribution area and the forecast value of the weather information in the near future such as the next day and the day after next. Each time a value is obtained, it is distributed to the computer system 101 via the network 111.

The computer system 101 acquires water demand data from the monitoring and control system 103 and weather information data (including forecasts) from the weather information providing system 112 via the network 111, and acquires the above data every day (every reference period). Using the data, the forecast calculation of the time demand amount from a predetermined time (for example, 0 o'clock) to 24 hours ahead is performed, and the forecast result is transmitted to the monitoring control system 103 via the network 111.

The monitoring and control system 103 acquires the water demand prediction result, and an hourly water operation plan (water purification plan of the water purification plant 104) from the predetermined time to 24 hours ahead according to the water demand prediction result. , Operation plan of pump 105, water storage plan of distribution reservoir 106, etc.), and operation control of water purification equipment and pump 105 of water purification plant 104 is performed based on the above water operation plan. As described above, daily water demand forecasting and water operation planning based on it are carried out, and operation control of the water supply system is performed based on the above water operation plan so as to supply an appropriate amount of water to consumers. ing.

The computer system 101 includes general computer hardware such as a controller (CPU), a storage device (RAM, hard disk, flash memory, etc.), an input unit 121 (keyboard, mouse, etc.), and a display unit 122 (display, printer, etc.). Be prepared.

In the storage device, a data acquisition unit 123, a prediction condition setting unit 124, a water demand amount prediction unit 125, a prediction result display unit 126, and a model parameter determination unit 127 are functional modules realized by the controller executing a program. It is set. Each functional module may be paraphrased as a means, a unit, or the like.

The storage device stores water demand data management table 131, day / weather data management table 132, prediction condition management table 133, model parameter management table 134, water demand prediction data management table 135, and prediction result display data management table 136. Has been done. The controller uses these tables when executing the above modules.

The data acquisition unit 123 acquires the actual water demand data transmitted from the monitoring and control system 103 and the weather information data transmitted from the weather information providing system 112, and registers them in a predetermined table.

The forecast condition setting unit 124 is used for forecast start time input by the administrator of the computer system 101 by the input unit 121, related conditions for water demand (weather, daytime, water demand during the immediate neighborhood period, etc.), and forecast calculation. Conditions used for prediction calculation such as the number of days of the latest past day (learning day) to be used are set in a predetermined table.

The water demand forecasting unit 125 predicts the water demand on the forecast date based on the kernel ridge regression model using the above-mentioned related condition data on the forecast date and each day of the past predetermined period, and monitors and controls the forecast result. Deliver to system 103.

The forecast result display unit 126 creates a display screen of the water demand forecast result and displays it on the display unit 122.

The model parameter determination unit 127 determines the model parameters of the kernel ridge regression model used for the prediction calculation of the water demand.

The water demand data management table 131 is a data group for managing the hourly water demand measurement value.

The day / weather data management table 132 is a data group for managing day of the week information and weather information (including forecast) for each day.

The forecast condition management table 133 contains information such as the forecast start time, related information on the water demand (weather, day, water demand during the period of the immediate neighborhood, etc.), and the number of days of the latest past day (learning day) used for the forecast calculation. It is a data group for managing the conditions for predictive calculation.

The model parameter management table 134 is a data group for managing the model parameters of the kernel ridge regression model used for the prediction calculation of the water demand.

The water demand forecast data management table 135 is a data group for managing the hourly water demand forecast value.

The forecast result display data management table 136 shows the hourly water demand predicted value, measured value (actual value), water demand integrated error, and the distribution reservoir water level converted value of the integrated error from the forecast start time. , A data group for managing the display information of the prediction result.

By executing the following processes (1) to (5), the computer system 101 refers to the water demand in the past day, which has conditions close to the conditions related to the current water demand, and determines the water demand. Realize forecasts and achieve highly accurate water demand forecasts that match changes in water demand.

(1) Acquisition of water demand / weather information (2) Setting of forecast conditions (3) Forecast of water demand (4) Display of water demand forecast result (5) Determination of model parameters The following (1) to ( 5) will be described with reference to FIGS. 2 to 13. Regarding the acquisition process (1) of water demand / weather information, the monitoring control system 103 sets the latest water demand measurement value and its measurement time every time the hourly water demand of the target distribution area is totaled. It is transmitted to the computer system 101 via the network 111.

The data acquisition unit 123 of the computer system 101 sequentially acquires the transmitted water demand measurement value (actual value) and the measurement time, and registers it in the water demand data management table 131. FIG. 2 shows an example of the water demand data management table 131 registered by the data acquisition unit 123.

The weather information providing system 112 distributes the actual value of the weather information of the area including the target distribution zone and the forecast value of the weather information of the next day to the computer system once a day via the network 111.

The data acquisition unit 123 acquires the above-delivered weather information, classifies the weather information into information on the maximum temperature, the presence or absence of fine weather, the presence or absence of rain, and the presence or absence of snow, and sets the date, day (Monday to Sunday, national holiday), Register (update) in the day / weather data management table 132 along with the distinction between actual / forecast. FIG. 3 shows an example of the day / weather data management table 132 registered by the data acquisition unit 123.

The prediction condition setting process (2) will be described. In general, the daily water demand (24-hour time series) changes depending on the weather, maximum temperature, day of the day, the latest water demand (time series), etc., and the water demand changes. Days with similar related conditions tend to have similar water demand.

Therefore, the computer system 101 predicts the water demand on the forecast date using the past water demand actual data (multi-day data is possible) having the above-mentioned related conditions similar to the forecast date. The forecast condition setting process (2) is a water demand forecast calculation such as the forecast start time, the conditions related to the water demand, and the number of days in the last past day referred to in the forecast calculation, which are necessary for the calculation of the above water demand forecast. Prerequisites (prediction conditions) necessary for the above are set in advance.

The administrator of the computer system 101 uses the input unit 121 to calculate the forecast start time, the conditions related to the water demand (selected from the above weather, maximum temperature, day of day, latest water demand, etc.), and the forecast calculation. Enter the number of days in the last past day to be referenced. Then, the prediction condition setting unit 124 of the computer system registers the above-set prediction conditions in the prediction condition management table 133.

FIG. 4 is an example of a display screen for setting the prediction condition displayed on the display unit 122 by the prediction condition setting unit 124. This display screen is appropriate for the target distribution area from the weather of the day, the maximum temperature, the day of the week, the latest water demand (time series), etc. as typical conditions related to the water demand. It is configured so that the administrator can select the appropriate one.

When the latest water demand (time series) is selected, the manager also inputs how many hours of the latest water demand will be used for the forecast calculation and the period. For example, if the forecast start time is 0 o'clock and the latest water demand period is 8 hours, water from 16:00 to 00:00 (24:00) will be used for water demand time series data from 16:00 to 24:00 the next day. Demand forecasting is done.

When the administrator selects the latest water demand (time series), the water demand can be used for forecast calculation with "water demand time series data alone" or as "water demand time series data + difference data". Select either to use or to use as "water demand time series data + total amount data".

As will be described later, as a related condition for the above water demand, not only the latest water demand time series data but also the difference data or the total amount data is used for the prediction calculation, so that the prediction calculation having different characteristics (actual value trend and) It is possible to make predictions with similar waveforms, predictions with a small integration error with actual values, etc.).

When the administrator selects the day of the week information, the day of the week information is used for the prediction calculation as "weekdays / holidays (2 categories)", "each day / holiday (8 categories)", or "each day / holiday /". Select either holidays (8 categories), but change the specified day of the week for consecutive holidays of 3 days or more.

"Each day of the week / holiday (8 categories), but change the specified day of the week for consecutive holidays of 3 days or more" means changing the day before the consecutive holidays of 3 days or more to Friday, changing the first day to Saturday, and then changing it. The last day is changed to Sunday. This is because, depending on the distribution zone, changing the day of the week as described above may improve the prediction accuracy of water demand. For example, in the case of three consecutive holidays of "Friday (holiday), Saturday, and Sunday", the day before that is Thursday, but since it is a weekday before the holiday, it may show a pattern similar to the water demand on Friday. Also, for example, in the case of three consecutive holidays of "Saturday, Sunday, and Monday (holiday)", the last Monday (holiday) is a holiday, but since it is a holiday before weekdays, it shows a pattern similar to the water demand on Sunday. In some cases. Therefore, changing the above day of the week leads to improvement in prediction accuracy.

FIG. 5 shows an example of the prediction condition management table 133 registered by the prediction condition setting unit 124. The setting contents of the prediction condition management table 133 in FIG. 5 correspond to the setting contents when the prediction conditions are set as shown in FIG.

As described above, the prediction condition setting unit 124 performs the prediction condition setting process.

Next, the water demand prediction process (3) will be described. The computer system 101 includes related condition data (weather, maximum temperature, day of day, latest water demand time series, etc.) for the water demand on the forecast date, and the above related condition data for the water demand on each day in the latest predetermined period. And, using the actual water demand data corresponding to the forecast target time zone of each day, priority is given to the actual water demand value for at least one day in the most recent predetermined period having the above related condition data similar to the forecast date. Use it to forecast water demand on the day of the forecast.

For example, the similarity between the relevant condition data on the day of the forecast and the relevant condition data on each day of the most recent predetermined period is calculated, and the water in the forecast target time zone of the past day having the relevant condition data with a high degree of similarity. Forecast calculation is performed using a model formula in which the actual demand value is prioritized (for example, a weighted average of the actual water demand values is taken according to the similarity of each past day). The computer system 101 uses, for example, a kernel ridge regression model as a model for predictive calculation based on the similarity as described above.

と表すことができる。カーネルリッジ回帰モデルは上記の考え方に基づくものであり、（式１）における入力データと実績データとの近さを表す関数をガウスカーネル関数で表し、それに対する乗数を実績データのＹ座標値でなくパラメータで表したモデルであり、以下のように定式化される。 The kernel ridge regression model will be described with reference to FIG. When the actual data (learning data) indicated by ◯ is given on the XY plane, it is considered to predict the Y coordinate value Ynew corresponding to the input of the X coordinate value Xnew. At this time, Ynew can be expected to take a value close to the Y coordinate of the actual data in the vicinity of Xnew, and it is not necessary to consider the actual data far from Xnew. Therefore, when N actual data (learning data) are given, the value Ynew of Y corresponding to the input Xnew is

It can be expressed as. The kernel ridge regression model is based on the above idea. The function that expresses the closeness between the input data and the actual data in (Equation 1) is represented by the Gaussian kernel function, and the multiplier for it is not the Y coordinate value of the actual data. It is a model expressed by parameters and is formulated as follows.

ここで、Ｎ：学習データの日数
αｉ：回帰パラメータ（ｉ＝１，・・・，Ｎ）
（Ｘｔ（ｉ），Ｙｔ＋ｎ（ｉ））：直近過去Ｎ日間における第ｉ日のモデル学習データ（予測日の入出力データと同じ時刻に対応した過去日の入出力ペアデータ）を用いて予測する。ここで、ｋ（Ｘｔ，Ｘｔ（ｉ））は、予測日の入力データＸｔと過去第ｉ日の学習用入力データＸｔ（ｉ）との近さを表すガウスカーネル関数であり、

ここで、||Ｘｔ−Ｘｔ（ｉ）||：入力データ（ベクトル）Ｘｔ、Ｘｔ（ｉ）の距離（Ｌ２ノルム、各ベクトルの各成分の残差平方和の平方根をとったもの）
β：ハイパーパラメータ（β＞０）
で定義される。ガウスカーネル関数は０〜１の連続値をとり、予測日の入力データＸｔと過去第ｉ日の学習用入力データＸｔ（ｉ）との距離が近いほど１に近い値をとり、距離が遠いほど０に近い値をとる。よって予測日の入力データ（水需要量に対する関連条件データ：天候、最高気温、曜日、直近水需要時系列）に類似する（距離の近い）入力データを持つ過去日のデータほど予測値に対する影響が大きくなり、予測日の入力データに類似する（距離の近い）入力データを持つ過去日のデータを優先的に利用した予測が行われる。 Yt: Water demand at time t (time increments are in 1-hour units),
Z1: Binary variable (1: Yes, 0: No) that indicates whether or not the predicted date is sunny,
Z2: Binary variable indicating the presence or absence of rain on the predicted date,
Z3: Binary variable indicating the presence or absence of snow on the predicted date,
Z4: Maximum temperature on the predicted day,
Wi: Binary variable representing the day of the week of the predicted day (1 variable of W1 only when classifying by weekday / holiday, 8 variables of W1 to W8 corresponding to Monday to Sunday / holiday when classifying by Monday to Sunday / holiday)
Then, the input data (vector) consisting of the related condition data (weather, maximum temperature, day, latest water demand time series) for the water demand is set to Xt = (Yt, Yt-1, ..., Yt-m + 1, Z1, Z2, Z3, Z4, Wi)
Then, the water demand amount Yt + n at the time t + n (n = 1, ..., 24) is set to

Here, N: Number of days of learning data
αi: Regression parameters (i = 1, ..., N)
(Xt (i), Yt + n (i)): Predict using the model learning data of the i-day in the last N days (input / output pair data of the past day corresponding to the same time as the input / output data of the predicted day). .. Here, k (Xt, Xt (i)) is a Gaussian kernel function that represents the closeness between the input data Xt of the predicted date and the learning input data Xt (i) of the past i-day.

Here, || Xt-Xt (i) ||: Distance of input data (vector) Xt, Xt (i) (L2 norm, the square root of the residual sum of squares of each component of each vector)
β: Hyperparameter (β> 0)
Defined in. The Gaussian kernel function takes a continuous value of 0 to 1, and the closer the distance between the input data Xt on the predicted date and the learning input data Xt (i) on the past i-day, the closer the value to 1, and the farther the distance Take a value close to 0. Therefore, the data of the past day with input data similar (closer) to the input data of the forecast date (related condition data for water demand: weather, maximum temperature, day, latest water demand time series) has an influence on the forecast value. Forecasting is performed by preferentially using the data of the past day, which becomes larger and has the input data similar (closer to the distance) to the input data of the forecast date.

In the above explanation, the input data vector corresponding to the predicted date and the past day (learning date) is composed of the weather, the maximum temperature, the day, and the latest water demand time series, but here we focus only on the water demand time series. Then, as shown in the example of FIG. 7, it is assumed that there is water demand data corresponding to the predicted date and water demand data corresponding to the two learning days 1 and 2.

At this time, it is assumed that the other conditions (weather, maximum temperature, day of the week) on each day are the same. It is assumed that the learning day 1 data (most recent water demand) always exceeds the predicted date data by a predetermined amount, and the learning day 2 data fluctuates up and down by the predetermined amount from the predicted date data. At this time, in general, if the above other conditions are the same, it is expected that the above relationship tends to continue to some extent even in the water demand data after the prediction start time of the prediction date and the learning days 1 and 2. As described above, the distance (base maintenance) between the two input data vectors is calculated by the L2 norm, that is, the sum of squares of the errors of each component, so the distance between the predicted date data and the two learning days 1 and 2 data. As shown in Fig. 7, the actual water demand in the forecast time zone of the learning day data 1 and 2 is likely to be reflected in the water demand forecast on the day of the forecast to the same extent (Fig. 7 is an image diagram). And this is not always the case).

Here, if the difference value between the time before each time in the latest water demand time series is added as a component of the input data vector, the difference value in the latest water demand time series of the forecast date data and the learning day 1 data is the same. Since the value of the degree is taken, the distance (similarity) from the predicted day data is closer to the learning day 1 data than to the learning day 2 data, and as shown in FIG. 8, the water in the predicted time zone of the learning day 1 data. It is easy to obtain a forecast result in which the actual demand is largely reflected in the water demand forecast on the day of the forecast, that is, a forecast result with a waveform similar to the actual value that exceeds the actual value on the forecast date by a predetermined amount. However, in this case, since the predicted value always exceeds the actual value, the integration error of the water demand forecast will increase. Since the prediction error of water demand appears as an error between the actual and prediction of the distribution reservoir water level, an increase in the integration error may lead to deviation from the upper and lower limit water levels of the distribution reservoir water level. Therefore, for water operation, it may be desirable that the predicted value fluctuates up and down near the actual value so that the integration error does not become large, rather than the predicted result of water demand always exceeding the actual value.

Next, if the total value of the latest water demand time series is added as a component of the input data vector, the total value of the latest water demand time series of the forecast date data and the learning day 2 data will be the same value, so the forecast The distance (similarity) from the day data is closer to the learning day 2 data than to the learning day 1 data, and as shown in Fig. 9, the actual water demand in the predicted time zone of the learning day 2 data is the water on the predicted day. It tends to be a forecast result that is largely reflected in the demand forecast, that is, a forecast result that fluctuates up and down by a predetermined amount from the actual value on the forecast date. In this case, since the actual value and the total water demand value tend to be close to each other, it is possible to expect a reduction in the integration error of the water demand forecast.

Therefore, by newly adding the difference value or total value of the latest water demand time series to the input data vector corresponding to the forecast date and the past day (learning date), the forecast result whose waveform is similar to the actual water demand can be obtained. You will be able to select prediction results with a small integration error. The addition of the difference value or the total value to the input data vector is the addition of the difference data when the latest water demand amount is selected as the related condition data for the water demand described in the prediction condition setting process (2), and the total. The amount can be added, and the administrator of the computer system 101 may switch the above selection according to the purpose of the water demand forecast.

Here, the regression parameter αi (i = 1, ..., N) has an error of 2 for the prediction result when all the training data of the past days are input to (Equation 2) each time the prediction calculation is performed. It is determined to minimize the evaluation function R consisting of the sum of multiplications and the regularization term for preventing overfitting of the training data.

上記評価関数Ｒは回帰パラメータαｉの関数であり、Ｒを最小化する回帰パラメータαｉは解析的に算出可能である（数式は省略）。よって予測計算を行う毎に、時刻ｔまでに得られた予測日の入力データＸｔ、過去日（学習データ日）の入出力データＸｔ（ｉ）、Ｙｔ＋ｎ（ｉ）を用いて、回帰パラメータαｉが算出でき、（式２）より予測対象時間ｔ＋ｎにおける水需要量予測値Ｙｔ＋ｎが算出できる。ｎをｎ＝１，・・・，２４まで変化させて上記計算を繰り返すことにより（パラメータαｉもｎ毎に算出）、予測対象時間帯（時刻ｔ＋１から時刻ｔ＋２４までの２４時間分）の水需要量予測値が算出できる。

The evaluation function R is a function of the regression parameter αi, and the regression parameter αi that minimizes R can be calculated analytically (the mathematical formula is omitted). Therefore, every time the prediction calculation is performed, the regression parameter αi is set by using the input data Xt of the prediction date obtained by the time t, the input / output data Xt (i) of the past day (learning data day), and Yt + n (i). It can be calculated, and the predicted water demand value Yt + n at the prediction target time t + n can be calculated from (Equation 2). By repeating the above calculation by changing n to n = 1, ..., 24 (parameter αi is also calculated for each n), the water demand in the prediction target time zone (24 hours from time t + 1 to time t + 24). The quantity predicted value can be calculated.

The regularization term includes a parameter λ (not described) that defines the degree of prevention of overfitting to the training data. Β included in the kernel function and λ included in the regularization term are hyperparameters (parameters that control the behavior of the algorithm), and were calculated using the above calculation using the data of the predetermined past period. Determine in advance by the round-robin method or the like so as to minimize the sum of squares of the error between the predicted value and the actual value of water demand.

Therefore, the relevant condition data for the water demand on the forecast date (weather, maximum temperature, day of day, latest water demand time series, etc.), the above-mentioned related condition data for each day of the most recent predetermined period, and each of the above. Water demand for the most recent predetermined period (s) with the above-mentioned related condition data similar to the forecast date based on the kernel ridge regression model using the actual water demand data corresponding to the forecast target time zone of the day. It is possible to predict and calculate the amount of water demand on the day of the forecast by preferentially using the actual quantity value.

The water demand forecast processing (3) is basically performed once a day when the latest water demand measurement value up to the forecast start time is acquired, but as will be described later, there is a water demand forecast error. When the administrator of the computer system 101 determines that the expansion has occurred, the water demand forecast processing (re-prediction) can be performed according to the administrator's instructions.

From the forecast condition management table 133, the water demand forecasting unit 125 of the computer system 101 determines the forecast start time and the conditions related to the water demand (weather, maximum temperature, day of day, latest water demand) required for the water demand forecast. (Series, etc.), get the number of days in the last past day (number of days in training data).

Then, the water demand forecasting unit 125 uses the water demand data management table 131 and the day / weather data management table 132 to determine the latest water demand actual value up to the forecast start time on the forecast date and the predetermined past day (learning date). , Weather, maximum temperature, daytime data, and actual water demand data for the predicted time zone on the specified past day (learning day).

At this time, the water demand forecasting unit 125 changes the classification of the day information according to the setting contents made in the above process (2) (weekdays / holidays / days / holidays / days / holidays + 3 consecutive holidays or more). (Change the specified day of the day) and add water demand time series data related information to the above input data vector (water demand time series data alone / water demand time series data + difference data / water demand time series data + total Quantitative data). Then, the water demand prediction unit 125 acquires the hyperparameters β and λ of the kernel ridge regression model from the model parameter management table 134.

The water demand prediction unit 125 calculates the water demand prediction value in the prediction time zone of the prediction date by using the prediction calculation based on the kernel ridge regression model described above using the acquired data, and the prediction result. Is registered in the water demand forecast data management table 135 and distributed to the monitoring control system 103 via the network 111. As a result, the monitoring and control system 103 can formulate an operation plan for the pump 105 and a planned water level for the distribution reservoir 106 based on the delivered water demand forecast result. FIG. 10 shows an example of the water demand forecast data management table 135 registered by the water demand forecast unit 125.

As described above, the water demand prediction unit 125 performs the water demand prediction process.

Next, the display processing (4) of the water demand forecast result will be described. This process (4) is executed every time the latest water demand prediction result by the above process (3) is obtained, and every time the latest measured water demand value (actual value) by the above process (1) is obtained. Will be done. The forecast result display unit 126 of the computer system 101 acquires the latest forecast result from the water demand forecast data management table 135, acquires the latest water demand actual value from the water demand data management table 131, and predicts the start time. The integration error of water demand is calculated by taking the sum of the prediction errors (= predicted value-actual value) from (1 o'clock) to each time.

Then, the prediction result display unit 126 divides the calculated integration error by the cross-sectional area of the distribution reservoir 106 to calculate the water level conversion value of the distribution reservoir 106. Then, the prediction result display unit 126 registers the predicted date, the actual value, the integration error, and the distribution reservoir water level conversion value of the water demand integration error acquired and calculated in the prediction result display data management table 136. , A prediction result display screen in which the above data is displayed as a graph is created and displayed on the display unit 122.

FIG. 11 shows an example of the prediction result display data management table 136 registered by the prediction result display unit 126. Further, FIG. 12 shows an example of the prediction result display screen created by the prediction result display unit 126.

As described above, the prediction result display unit 126 performs the display processing of the water demand amount prediction result.

The administrator of the computer system 101 browses the forecast result display screen, determines whether or not the water demand forecast error is expanding, and if it is determined that the water demand forecast error is expanding, the administrator of the water demand using the latest data is used. Reprediction can be made. When performing reprediction, the administrator inputs a new prediction start time (current time) from the input unit 121 and instructs reprediction from the prediction start time as in the above process (2).

Then, the water demand prediction unit 125 uses the latest actual water demand value, weather data, etc. to predict and calculate the water demand from the new prediction start time to 24 hours ahead, as in the above process (3). Is performed, the forecast result is registered in the water demand forecast data management table 135, and the forecast result is distributed to the monitoring control system 103. This makes it possible to re-forecast water demand using the latest data even if the forecast error increases.

Next, the model parameter determination process (5) will be described. This process (5) may be performed periodically, for example, once a year or once every six months. The model parameter determination unit 127 sets the initial values (sufficiently large values) of the parameters β and λ of the kernel ridge regression model, and sets each day of the latest predetermined period (for example, 3 months) from the present to the predicted date. , The water demand forecasting unit 125 is called to execute the water demand forecasting process (3) in the forecasting time zone of each forecasting day. At this time, the conditions used for the prediction are those registered in the prediction condition management table 133.

Then, the model parameter determination unit 127 acquires the actual water demand value of each forecast day from the water demand data management table 131, and 2 of the prediction error between the predicted value and the actual value in all the predicted time zones and the predicted days. Calculate the sum of products.

The model parameter determination unit 127 repeats the prediction process while reducing the values of the parameters β and λ by a predetermined value, and optimizes the values of the parameters β and λ that minimize the sum of squares of the calculated prediction errors. It is registered in the model parameter management table 134 as a parameter value. FIG. 13 shows an example of the model parameter management table 134.

As described above, the model parameter determination unit 127 performs the model parameter determination process.

In the above-described embodiment, when the correction of the model parameters is pending and the learning of the kernel ridge regression model progresses, the computer system 101 calculates the water demand based only on the relevant conditions at the time of predicting the water demand. Can be predicted.

As described above, according to the embodiment of the present invention, the relevant condition data (weather, maximum temperature, day of day, latest water demand time series, etc.) for the water demand on the forecast date, and each of the latest predetermined period. Using the above-mentioned related condition data for the daily water demand and the water demand actual data corresponding to the forecast target time zone of each day, the day of the most recent predetermined period having the above-mentioned related condition data similar to the forecast date ( Water demand is forecasted on the day of the forecast by preferentially using the actual value of water demand (s). As a result, the forecast is made with reference to the water demand in the past day, which is close to the change in the water demand, so that the water demand can be predicted with high accuracy according to the change in the water demand.

100 ... Water distribution equipment 101 ... Computer system 102 ... Water supply system 103 ... Monitoring and control system 104 ... Water purification plant 105 ... Pump 106 ... Distribution reservoir 107 ... Water level gauge 108 ... Flow meter 109 ... Water distribution pipe network 110 ... Customer 111 ... Network 121 ... Input unit 122 ... Display unit 123 ... Data acquisition unit 124 ... Prediction condition setting unit 125 ... Water demand amount prediction unit 126 ... Prediction result display unit 127 ... Model parameter determination unit 131 ... Water demand amount data management table 132 ... Day / weather data Management table 133 ... Prediction condition management table,
134 ... Model parameter management table 135 ... Water demand forecast data management table 136 ... Forecast result display data management table

Claims (11)

It is a water demand forecasting system that predicts the amount of water demand for each reference period for a predetermined distribution area where water is distributed from the water distribution facility via the distribution pipe network.
With memory
With the controller
With
What the controller does based on the water demand forecasting program
The actual value of the water demand in the predetermined water distribution zone, which is continuously measured by the water distribution facility, is acquired from the monitoring and control system of the water distribution facility and recorded in the memory together with the weather information and the day of the week information. When,
Setting related conditions to be used for forecasting the water demand, the related conditions are weather information, day of the week information, and the above-mentioned for a predetermined period before the reference period. Including that it is the actual value of water demand
Regarding the actual value of the water demand for the past predetermined period recorded in the memory, the related condition is obtained for each reference period, and the related condition and the related condition for the reference period in which the water demand is predicted are used. Calculating similarity and
Based on the actual value of at least one water demand amount in the plurality of reference periods in the past predetermined period, the prediction calculation of the water demand amount is performed with reference to the similarity.
Sending the result of the prediction calculation to the monitoring and control system
including,
Water demand forecasting system. Performing the prediction calculation of the water demand is prioritized in the prediction calculation of the water demand as the actual value of the water demand in the reference period having a high degree of similarity in the reference period of the past predetermined period. Including using for
The water demand forecasting system according to claim 1. To perform the prediction calculation of the water demand amount, the related condition in the reference period for which the water demand amount is to be predicted is used as input data, and the related condition and the actual water demand amount in each reference period of the past predetermined period are used. Using a kernel ridge regression model that uses values as training data, water demand is predicted based on the similarity.
The water demand forecasting system according to claim 2. Setting the related condition used for the prediction of the water demand includes setting the time difference value of the water demand for a predetermined period before the reference period as the related condition.
The water demand forecasting system according to claim 1. The water demand according to claim 1, wherein setting the related condition used for forecasting the water demand includes adding the total value of the water demand for a predetermined period before the reference period to the related condition. Prediction system. To set the relevant conditions used to forecast the water demand, the information on the day of the week is regarded as Friday the day before the consecutive holidays of three days or more, the first day of the consecutive holidays of three days or more is regarded as Saturday, and three days or more. Includes at least one of the last days of the holidays being Sunday,
The water demand forecasting system according to claim 1. The controller includes the predicted water demand and the actual value of the water demand, the integration error between the predicted water demand and the actual value of the water demand from a predetermined time, and the abolished equipment of the integration error. Display at least one of the water level conversion values of the distribution reservoir of
The water demand forecasting system according to claim 1. The administrator of the water demand forecasting system can input a predetermined period before the reference period and the predetermined period in the past into the system via an input device.
The water demand forecasting system according to claim 1. The administrator of the water demand forecasting system can input the information of the day of the week as the related condition into the system via the input device, and the information of the day of the week has a plurality of modes, and the administrator has a plurality of modes. A predetermined mode can be selected from the modes of the above.
The water demand forecasting system according to claim 1. The administrator of the water demand forecasting system can input the mode of the actual value of the water demand as the related condition via the input device, and there are a plurality of modes of the actual value of the water demand. Allowed the administrator to select a predetermined aspect from a plurality of aspects,
The water demand forecasting system according to claim 1. This is a method in which a computer predicts the amount of water demand for each reference period for a predetermined distribution area where water is distributed from the distribution facility via the distribution pipeline network.
The computer
Acquire the water demand amount of the predetermined distribution area, the weather information of the distribution area, and the day of the week information.
As related conditions for the water demand amount in the reference period, the weather information, the day of the week information, and the water demand amount in the predetermined range before the reference period are set.
The degree of similarity between the related conditions in each reference period of the past predetermined period and the related conditions in the reference period for which the water demand is to be predicted is calculated, and the actual water demand in the reference period having the conditions with high similarity is calculated. Predicting the water demand in the reference period for which the water demand is to be predicted, using the value preferentially.
The method.

Images