JP2006227852A - Use plan value calculation method and computer system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To facilitate countermeasures to such a case that the result quantity of the object of consignment to be supplied from a consignment client to the consignment network of a consignment undertaker. <P>SOLUTION: This method may be applied when a penalty is imposed on a consignment client based on an error between the use result quantity of the object of consignment (power, gas or the like) by a customer and the use plan value of the customer. This computer system is configured to acquire the use result quantity of the object of consignment in every first unit time for the customer from the consignment undertaker computer, and to calculate the prediction plan value of the object of consignment of the customer in every first unit time according to a specific calculation method, and to calculate a correction value for the prediction plan value in every first unit time for reducing a penalty based on a cumulative error between the use result quantity and the use plan value in the second unit time, and for minimizing the penalty based on the error between the use result quantity and the use plan value in the first unit time, and to calculate the use plan value in every first unit time from the correction value and the prediction plan value. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an information processing technique executed when electric power, gas, or other energy or resources are supplied to a customer's facility via a consignment network of a consignment company.

  For example, if the consignment requester is A, the consignment carrier is B, and the object of consignment is gas, as shown in Fig. 1, the gas from A is the flow control valve, receiving meter, B network, and dispensing meter. To the customer's facility of the consignment client. When carrying out consignment, the flow control valve on the side A is controlled in order to perform simultaneous and same amount control so that the gas supply amount of the customer A and the gas consumption of the customer are the same. If the difference between the gas supply amount of the customer and the gas consumption amount of the customer exceeds a predetermined threshold value, the customer must pay a penalty fee to the customer.

  Japanese Patent Application Laid-Open No. 2004-178478 is known as a prior art based on such a concept. More specifically, when managing the contractor gas request amount in units of one hour, the comparison unit compares the signal of the one-hour integrated amount of the contractor gas request amount and the signal of the one-hour integrated amount of the city conduit discharge gas amount. Let's compare. If there is a difference in the amount of gas to be discharged or too much with respect to the contractor gas request amount, the correction calculation unit obtains the instantaneous amount of gas necessary to eliminate the difference in the remaining time of 1 hour, and The correction signal is added to the above-mentioned contractor gas request amount signal at the adding unit, and this is compared with the currently discharged gas amount signal as the set value by the dispensing control unit, and the flow control valve of the commercial conduit is selected according to the difference. The amount of discharged gas is adjusted by adjusting.

Japanese Patent Application Laid-Open No. 2002-230640 discloses a technique for reducing the cost for a connection supplier and making it easy to use a utility (combustible gas, power, water) supply network. . Specifically, the received amount measured by the receiving-side flow meter provided in the receiving facility and the paid-out amount measured by the dispensing-side flow meter provided in the paying facility are collected and stored in the data collection storage means, and these are stored. Based on the received amount and the paid-out amount, the consignment company calculates the necessary charges for each of the connected supplier and the customer. Then, the connected supplier is notified of the received amount and the paid-out amount and the charges of the connected supplier and the customer via the communication line by the notification means for the connected supplier. The notification means notifies the payout amount and the customer's fee via a communication line.
JP 2004-178478 A Japanese Patent Laid-Open No. 2002-230640

  However, the consignment is not always performed in the situation shown in FIG. That is, as shown in FIG. 2, when the gas of A and B is mixed, since the receiving meter measures the flow of A and B together, the supplied amount from the A may not be actually measured. In such a case, it is impossible to control such that the customer's gas consumption measured by the payout meter is the same as the gas supply amount of the former.

  Accordingly, an object of the present invention is to provide a computer system that is required when the actual amount of the object of consignment supplied to the consignment network of the consignor is not specified by the consignor requester.

  Another object of the present invention is to appropriately set the customer's use plan value, which is data required when the actual amount of the target of consignment supplied to the consignment network of the consignment carrier is not specified by the consignment requester. It is to provide a technique for calculating.

  The computer system according to the first aspect of the present invention provides a consignment target (power) by a consignment requester's customer without specifying the actual amount of the consignment target supplied to the consignment carrier's consignment network by the consignment requester. , Gas, other energy and resources), a computer for consignees used when a penalty is imposed based on the error between the actual usage amount of the customer and the planned usage value of the customer (planned value in the embodiment) System. The computer system obtains data on the actual usage amount to be entrusted for each first unit time (for example, 1 hour) from the computer of the consignment carrier and stores it in the actual amount data storage unit. Means for calculating a predicted plan value of a customer's consignment target for each first unit time using the data stored in the quantity data storage unit according to a specific algorithm, and storing it in the predicted plan value data storage unit; Using the data stored in the data storage unit, while reducing the penalty based on the accumulated error between the actual usage amount and the usage plan value in the second unit time (for example, three months) longer than the first unit time, A correction value for the predicted plan value for minimizing a penalty based on an error between the actual usage amount and the usage plan value in the first unit time is calculated and corrected every first unit time. Using the data stored in the data storage unit, the predicted plan value data storage unit, and the data stored in the correction value data storage unit, the usage plan value for each first unit time is calculated, and the usage plan value data is stored. Means for storing in the unit.

  If the utilization plan value is calculated in this way, the penalty paid by the consignment requester to the consignment company can be minimized. As mentioned above, since the correction value is calculated after taking into account long-term and short-term penalties, for example, unlike the simple demand forecast of the customer, for example, from the forecast for the first unit time, In some cases, the usage plan value is set so as to reduce the penalty for the second unit time even if there is a slight deviation.

  In addition, the utilization plan value calculation method according to the second aspect of the present invention includes the customer's utilization plan value for the consignment target supplied to the customer's facility of the consignment requester connected to the consignment company's consignment network, This is a method of calculation by a computer having a processing unit and a data storage unit. In this method, the processing unit uses the data stored in the actual amount data storage unit that stores the usage amount of the customer's consignment target every first unit time (for example, one hour), and uses the data stored in the actual amount data storage unit. The predicted plan value is calculated for each first unit time according to a specific algorithm, stored in the predicted plan value data storage unit, and the processing unit uses the data stored in the actual amount data storage unit, This is to reduce the accumulated error between the actual usage amount and the usage plan value in a second unit time (for example, three months) longer than one unit time, and is longer than the first unit time and shorter than the second unit time. A first correction value calculation step for calculating a first correction value in a third unit time (for example, a day) and storing the first correction value in the first correction value data storage unit; Quantity data storage and Using the stored data, the first correction value in the third unit time is satisfied, and the penalty based on the error between the actual usage amount and the usage plan value in the first unit time is minimized. The second correction value is calculated every first unit time within the unit time of 3 and stored in the second correction value data storage unit, and the predicted planned value data storage unit and the second correction value are processed by the processing unit. Using the data stored in the data storage unit, a usage plan value that is the sum of the predicted plan value and the second correction value is calculated for each first unit time within the third unit time, and the usage plan is calculated. Storing in the value data storage unit.

  In this way, after calculating the first correction value in the third unit time in consideration of the long-term penalty, the first correction value is satisfied so as to minimize the short-term penalty while satisfying the first correction value. Since the second correction value is calculated for each unit time, the predicted plan value is appropriately corrected.

  Note that the first correction value calculation step described above includes the step of specifying a candidate value of the first correction value by the processing unit, and the actual usage amount and the usage plan value in the second unit time by the processing unit. The data stored in the penalty function data storage unit that stores the parameter data that is the basis of the penalty function for the cumulative error, the actual amount data storage unit, and the predicted plan value data storage unit that also stores past predicted plan values And corresponding to the candidate value of the first correction value, the data of the probability density function for the cumulative prediction error between the actual usage amount and the predicted plan value in the third unit time, and the second unit time A first step of calculating penalty function data; a second step of calculating an expected value of penalty from the probability density function data and penalty function data by the processing unit; The parts for different candidate values of the first correction value by performing the first and second step, may include the steps of expected value of penalty identifying a first correction value becomes minimum. A more appropriate first correction value can be calculated.

  In addition, the first correction value calculation step described above uses the data stored by the processing unit in the actual amount data storage unit and the usage plan value data storage unit that also stores past usage plan values. , A step of calculating a cumulative error between the actual usage amount generated so far in the second unit time and the planned use value, and canceling the cumulative error with the remaining third unit time in the second unit time Calculating a correction value to be used as the first correction value. For example, the first correction value may be calculated in this way when it is far from the end of the second unit time.

  Furthermore, the second correction value calculating step described above includes a step of specifying a correction candidate value in a specific unit time for the first unit time by the processing unit, and a step in the first unit time by the processing unit. Penalty function data storage that stores the parameter data that is the basis of the penalty function for the error between the actual usage amount and the usage plan value, the actual amount data storage unit, and the predicted plan value data storage that also stores past forecast plan values Data corresponding to the correction candidate value, the probability density function data on the prediction error between the actual usage amount and the predicted plan value in a specific unit time, and the specific unit time. The first step of calculating the penalty function data and the processing unit calculate the expected value of the penalty from the probability density function data and the penalty function data. The second step and the processing unit execute the first and second steps on the plurality of correction candidate values for all the first unit times within the third unit time, and for the third unit time. A step of specifying a second correction value that minimizes the expected value of the penalty for each first unit time within the third unit time while satisfying the first correction value may be included. In this way, a more appropriate second correction value can be calculated.

  In addition, it may further include a step of executing a process of changing the parameter data in accordance with at least one of the cost to be consigned and the selling unit price, and storing the parameter data in the penalty function data storage unit. This makes it possible to estimate the penalty according to the situation.

  The computer system according to the third aspect of the present invention uses the consignment target by the consignee client without specifying the actual amount of consignment target supplied to the consignment network of the consignor by the consignment requester. This is a computer system for a consignment requester that is used when a penalty is imposed based on an error between the actual amount and the usage plan value of the customer. And this computer system acquires the data of the use amount of the usage object of the consignment object for every unit time about the customer from the computer of the consignment carrier, and stores it in the actual amount data storage unit and the actual amount data storage unit. Means for calculating the predicted plan value of the customer's consignment target for each unit time according to a specific algorithm and storing it in the predicted plan value data storage unit for storing the past predicted plan value, and the actual amount data Using the data stored in the storage unit and the predicted plan value data storage unit, for each unit time, calculate the statistic of the prediction error between the actual usage amount and the predicted plan value in the unit time, and the statistic Based on the above, the correction value for the predicted plan value is calculated every unit time, and stored in the correction value data storage unit, the predicted plan value data storage unit, and the correction value data storage unit Using chromatography data, calculates a utilization plan value per unit time, and means for storing a utilization plan value data storage unit. As described above, an appropriate usage plan value can be obtained even if a correction value is calculated using statistical data instead of simple prediction as described above and the prediction is corrected. .

  It is also possible to create a program for causing a computer to execute the method according to the present invention. The program is, for example, a storage medium or a storage device such as a flexible disk, a CD-ROM, a magneto-optical disk, a semiconductor memory, and a hard disk. Stored in In some cases, digital signals are distributed over a network. Note that data being processed is temporarily stored in a storage device such as a computer memory.

  According to the present invention, it is possible to provide a computer system that is required when the actual volume of the object of consignment supplied to the consignment network of the consignment carrier is not specified by the consignment requester.

  Further, as another aspect of the present invention, the customer's use plan value, which is data necessary when the actual amount of the target of consignment supplied to the consignment carrier's consignment network is not specified by the consignment requester, is appropriately calculated. can do.

  FIG. 3 shows a system outline diagram according to an embodiment of the present invention. For example, a network 1 that is the Internet is connected to a consignment company computer 7, a customer computer 9, a weather data distribution server 5, and a consignment planning system 3 that performs main processing in the present embodiment. The consignment planning system 3 includes an actual value data acquisition unit 301, a customer planned value data acquisition unit 302, a weather data acquisition unit 303, an actual value data storage unit 304, a customer planned value data storage unit 305, and a weather data storage. Unit 306, predicted consumption value calculation unit 307, predicted consumption value data storage unit 309, planned value data storage unit 310, correction value calculation unit 312, correction value data storage unit 313, planned value calculation unit 314, And a plan value data transmission unit 315.

  The actual value data acquisition unit 301 acquires the customer's consumption actual value data (for example, hourly data) of the previous day from the consignment company computer 7 via the network 1 and stores the acquired data in the actual value data storage unit 304. The customer plan value data acquisition unit 302 acquires the next day's customer plan value data (for example, hourly data) from the customer computer 9 via the network 1 and stores the data in the customer plan value data storage unit 305. The meteorological data acquisition unit 303 acquires meteorological data from the meteorological data distribution server 5 via the network 1 and stores it in the meteorological data storage unit 306. The consumption predicted value calculation unit 307 uses the data stored in the actual value data storage unit 304, the customer plan value data storage unit 305, and the weather data storage unit 306 to specify the type specified according to the prediction method specification 308. The consumption prediction value is calculated by the prediction method and stored in the consumption prediction value data storage unit 309. The correction value calculation unit 312 uses the data stored in the consumption predicted value data storage unit 309, the plan value data storage unit 310, and the actual value data storage unit 304, and the type specified according to the correction method specification 311. The correction value is calculated by the correction method and stored in the correction value data storage unit 313. The plan value calculation unit 314 calculates a plan value using data stored in the correction value data storage unit 313 and the consumption prediction value data storage unit 309 and stores the plan value in the plan value data storage unit 310. To do. The plan value data transmission unit 315 reads the next day's plan value data from the plan value data storage unit 310 and transmits it to the consignment company computer 7 via the network 1.

  The consignment planning system 3 may be a single personal computer. In that case, for example, a display processing unit that displays data stored in each data storage unit on a display device according to a user instruction is provided. In addition, the consignment planning system 3 may be a server. In that case, the consignment planning system 3 is stored in, for example, each data storage unit in response to a request from a client terminal connected via the in-house network of the consignor requester. For example, HTML data generated from the received data is transmitted to the requesting client terminal, and displayed on the display device of the client terminal by a client application program (for example, a Web browser).

  FIG. 4 shows a detailed functional block diagram of the correction value calculation unit 312. The correction value calculation unit 312 includes a parameter data storage unit 3124, a penalty function generation unit 3123, a penalty function data storage unit 3122, a daily correction amount calculation unit 3121, a daily correction amount data storage unit 3125, and a simultaneous same amount correction. A value calculation unit 3126 and a prediction error distribution data storage unit 3127 are included. The parameter data storage unit 3124 is parameter data of data for determining a penalty function for calculating a penalty fee that the consignment requester will pay to the consignment company (for example, sales unit price and cost on the consignment requester side). Is stored. The penalty function data storage unit 3122 stores the penalty function basic data defined in the contractor's contract and the like and the penalty function data to be currently used. The penalty function generation unit 3123 generates penalty curve data to be used from the data stored in the parameter data storage unit 3124 and the penalty function data storage unit 3122 and stores the penalty curve data in the penalty function data storage unit 3122. The daily correction amount calculation unit 3121 uses the data stored in the penalty function data storage unit 3122, the actual value data storage unit 304, the predicted consumption data storage unit 309, and the planned value data storage unit 310 to specify a correction method specification 311. Accordingly, the daily correction amount is calculated by the specified type of correction method and stored in the daily correction amount data storage unit 3125. At this time, the distribution data (probability density function data) of the daily prediction error (= consumption prediction value−actual value) is stored in the prediction error distribution data storage unit 3127. The simultaneous same amount correction value calculation unit 3126 uses the data stored in the actual value data storage unit 304, the predicted consumption data storage unit 309, the daily correction amount data storage unit 3125, and the penalty function data storage unit 3122 every hour. Are calculated at the same time, and are stored in the correction value data storage unit 313. At this time, prediction error distribution data storage unit 3127 stores prediction error distribution data for each hour.

  Next, the processing flow of the system shown in FIG. 3 will be described with reference to FIGS. For example, the following processing flow is executed by the consignment planning system 3 until the next day's plan value data is transmitted to the consignment company computer 7. First, the actual value data acquisition unit 301 acquires the consumption data amount of the customer of the previous day (data for every hour) from the consignment company computer 7 via the network 1 and cumulatively stores the actual value data storage unit 304. (Step S1). Note that the user of the consignment planning system 3 may input data on the actual consumption amount of the customer to the actual value data acquisition unit 301. For example, a format example of data stored in the actual value data storage unit 304 is shown in FIG. In the example of FIG. 6, the customer ID, date, time, and actual consumption amount are registered.

  Further, the customer plan value data acquisition unit 302 acquires the customer plan value data of the next day from the customer computer 9 via the network 1, and stores it in the customer plan value data storage unit 305 cumulatively (step S3). Note that the user of the consignment planning system 3 may input the customer plan value data of the customer of the next day to the customer plan value data acquisition unit 302. A format example of data stored in the customer plan value data storage unit 305 is shown in FIG. In the example of FIG. 7, the customer ID, date, time, customer plan value, and singular date flag are registered. In addition, a special day is a day with special circumstances that cannot refer to past consumption results.

  Further, the weather data acquisition unit 303 acquires the next day's weather forecast data and the previous day's actual weather data from the weather data distribution server 5 via the network 1, and stores them in the weather data storage unit 306 (step S5). An example format of data stored in the weather data storage unit 306 is shown in FIGS. The example in FIG. 8 shows actual weather data, and a location name or code, date, time, and temperature are registered. Actual weather data is stored cumulatively. Further, the example of FIG. 9 shows weather forecast data, and a location name or code, a date, a maximum temperature, and a minimum temperature are registered.

  Next, the predicted consumption value calculation unit 307 specifies a consumption amount prediction method from the prediction method designation 308 or the like (step S7). For example, one method (model) is (1) autoregressive model, (2) similar day prediction model, (3) regression model, (4) customer plan value correction model, (5) others, and the consignment planning system 3 users are designated. Alternatively, an appropriate method (model) may be set in advance, and the setting may be followed.

  Then, the consumption predicted value calculation unit 307 uses the data stored in the actual value data storage unit 304, the customer plan value data storage unit 305, and the weather data storage unit 306, and uses the consumption prediction method specified in step S7. The consumption predicted value (consumption predicted value for every hour) is calculated and stored in the consumption predicted value data storage unit 309 (step S9). Each method is outlined below.

(1) Autoregressive model (AR (Autoregressive Model) model)
Consumption predicted value W (t) = a0 + a1 * Y (t-1) + ... + an * Y (tn)
A predicted consumption value is calculated according to this equation. A0, a1, a2,. . . an is a model constant, Y (t) is a consumption actual value, and t is a time (unit time such as one hour or one day). About a model constant, n is defined, for example, and it calculates from the consumption performance value for the past predetermined number of days. Further, the optimal n and model coefficient are calculated from the actual consumption values for the past predetermined number of days on the basis of AIC (Akaike information amount, one of information amounts used for determining the optimal order of the time series model).

(2) Similarity Date Prediction Model A consumption prediction value is generated by proportionally correcting or adding a deviation between the predicted date and the actual consumption value on the date assumed to be a similar date.
For example, predicted consumption value W (t) = α × actual value A (t) on similar days (proportional correction)
Also, for example, consumption forecast value W (t) = β + actual value A (t) on similar days (deviation addition)
α and β are calculated statistically or empirically according to customer consumption trends, climate conditions, and the like.

(3) Regression model A consumption forecast value is calculated using parameters related to consumption, such as past consumption record, temperature (past record, forecast), day of the week, and specific days.
For example, predicted consumption value W (t) = a0 + a1 × Z1 (t−1) +... + An × Zn (tn)
Zi is an explanatory variable (consumption actual value, temperature, day of the week, etc.). It is also possible to make predictions using a neural network or the like using the explanatory variables used in the regression model.

(4) Customer planned value correction model The relationship between the customer planned value and the actual value is statistically analyzed, and the consumption predicted value is calculated by correcting the customer planned value.
For example, consumption forecast value W (t) = customer plan value P (t) + f (μ, σ)
Here, f (μ, σ) is a function for reducing the error distribution θ (μ, σ) between the customer plan value and the actual value.
In addition, for example, consumption forecast value W (t) = α × actual value P (t) + β on a day with a similar customer plan value
For example, the similar date of the customer plan value is searched under such a condition that the square error of the day is the smallest. Α and β are calculated from the relationship between the actual value on the similar date and the customer plan value and other factors.

(5) Others For example, a method of using an autoregressive model for calculating α and β of a similar date model or a customer plan value correction model is included.

  Since the method (model) described above is a well-known technique, further explanation is omitted. In addition, for all methods (models), the one with the smallest standard deviation or sum of squares of prediction errors may be automatically selected as the optimum method (model). You may make it select the method with the best fit for every customer, every day, and every time.

  The consumption predicted value data calculated in step S9 is stored in the consumption predicted value data storage unit 309, for example, in the data format shown in FIG. That is, in the example of FIG. 10, the customer ID, date, time, and consumption predicted value are registered.

  Returning to the description of FIG. 5, the daily correction amount calculation unit 3121 uses the data stored in the actual value data storage unit 304, the predicted consumption data storage unit 309, the planned value data storage unit 310, and the penalty function data storage unit 3122. Then, according to the correction method specification 311, the daily correction value for the predicted plan value is calculated according to the specified type of method, and stored in the daily correction amount data storage unit 3125 (step S 11). The processing of this step will be described in detail below.

  First, a penalty that is a premise for correction will be described. In the present embodiment, first, for example, when the actual value and the planned value are compared every hour and the error exceeds the allowable value, a penalty determined in advance in the terms and conditions is paid. ing. A graph for explaining this penalty is shown in FIG. In FIG. 11, the horizontal axis represents time, and the vertical axis represents consumption or planned value. The solid line A represents the time change of the planned value, the thick line B represents the time change of the consumption amount, and the dotted line C represents a penalty occurrence threshold offset from the plan value by a predetermined allowable value (for example, + 10%). Represents. In the example of FIG. 11, the thick line B exceeds the dotted line C only in the portion D, and a penalty occurs for this portion.

  In the present embodiment, not only the short-term penalty described above but also a long-term penalty is defined. In the present embodiment, it is assumed that a penalty corresponding to the value of cumulative error = (cumulative plan value−cumulative actual value) is defined with a determination period of three months. FIG. 12 shows a graph representing the change in accumulated error over time. In FIG. 12, the horizontal axis represents time (from April to July), and the vertical axis represents cumulative error. Since an accumulated error occurs every hour, if accumulated, the accumulated error was initially greater than 0 as in Case 1, but then becomes a negative value, or the accumulated error is initially 0 as in Case 2. However, various patterns may occur, such as once becoming a negative value and then becoming a positive value. However, no matter what route is taken, a penalty corresponding to the absolute value E or F of the accumulated error occurs at the end of the determination period.

  In the present embodiment, a penalty that occurs with respect to an error h between the actual value and the planned value every hour is defined as a simultaneous equal amount penalty L (h). This simultaneous equal amount penalty L (h) is a value as shown in FIG. That is, when an error of −10% or more of the planned value occurs, the simultaneous equal amount penalty L (h) increases along the straight line having the first slope according to the amount of the error. Further, when an error of the planned value of + 10% or more occurs, the simultaneous equal amount penalty L (h) increases along a straight line having the second slope according to the amount of the error. In the example of FIG. 13, the second slope is larger than the first slope. That is, if the customer consumes less than the planned value, more penalty is generated than if the customer consumes more than the planned value.

  In this embodiment, a penalty that occurs with respect to the cumulative error u between the actual value and the planned value every three months is defined as a cumulative penalty R (u). This cumulative penalty R (u) has a value as shown in FIG. In other words, for a cumulative error up to + 3% of the cumulative planned value, a cumulative penalty R (u) occurs along the straight line having the first slope. For example, in the present embodiment, the first slope is the difference between the sales unit price of the consignment requester and the purchase unit price of the consignment company. For an accumulated error of + 3% or more of the accumulated planned value, an accumulated penalty R (u) occurs along a straight line having the second slope. For example, in the present embodiment, the second slope is the cost unit price of the consignment requester. On the other hand, for negative cumulative errors (when consumption exceeding the cumulative planned value is made and the consignment company compensates for the cumulative planned value), the cumulative penalty R (u) along the straight line having the third slope Occurs. For example, in the present embodiment, the third inclination is the difference between the sales unit price of the consignment carrier and the sales unit price of the consignment requester.

  If such a penalty is defined, the cumulative penalty R (u) due to the cumulative error is reduced and the simultaneous equal amount penalty L (h) due to the error every hour is not generated (so as to be reduced). Plan values can be determined. Such a planned value is obtained by appropriately determining a correction value described below.

  In the present embodiment, the correction value is calculated in two stages. That is, the correction value for each day (next day) and the correction value for each hour on the day. In step S11, the former correction value is calculated. This process will be described with reference to FIG. First, the penalty function generation unit 3123 uses the cost and sales unit price data stored in the parameter data storage unit 3124 and the data stored in the penalty function data storage unit 3122 to update the latest according to the change in the cost and the sales unit price. Is generated and stored in the penalty function data storage unit 3122 (step S21). As shown in FIG. 14, in the present embodiment, the accumulated penalty R (u) depends on the sales unit price and cost unit price of the consignment requester. . Therefore, the data stored in the parameter data storage unit 3124 is updated as appropriate, and the penalty function generation unit 3123 updates the slope data. In the example of FIG. 14, an example in which changes in the sales unit price and the cost unit price affect the inclination is shown, but when other parameters are affected, the parameters are updated. More specifically, in the example of FIG. 14, data about a range (for example, 0 to 3%, 3% or more, less than 0%, etc.) is acquired from the penalty function data storage unit 3122, and the cost is stored in the parameter data storage unit. For example, the first inclination from 0 to 3% and the second inclination at 3% or more are calculated.

  Next, the daily correction amount calculation unit 3121 uses the latest actual value data stored in the actual value data storage unit 304 and the corresponding predicted consumption value data in the predicted consumption value data storage unit 309 to predict daily errors. Distribution data is generated and stored in the prediction error distribution data storage unit 3127 (step S23). In this step, prediction error distribution data (probability density function) P (u) (u is a daily correction value (day correction amount)) is calculated for a daily prediction error (= consumption prediction value−actual value). For example, the prediction error distribution function may be specified by calculating the average and standard deviation of the prediction error and assuming a normal distribution according to the average and standard deviation of the prediction error. Moreover, the prediction error distribution actually obtained may be used as it is, or may be applied to another distribution instead of the normal distribution.

In order to make the processing after this step easier to understand, the following specific example is assumed. That is, the planned value for 31 days is calculated on December 30. Here, the predicted consumption values on the 30th and 31st are 1000 respectively, and there are actual values up to the 29th as of the 30th. Further, the cumulative prediction plan value up to 29th is 30000, the cumulative actual value is 30400, and the cumulative prediction error is 400. As for the prediction error distribution up to the 29th, a normal distribution with an average of 0 and a standard deviation of 70 is assumed. Since the days with uncertain prediction results are two days (30 days and 31 days), the standard deviation σ = 2 ^0.5 × 70 = 100 is used. In addition, since it is known that the standard deviation spreads by the square root multiple of the uncertain number of days in the case of the normal distribution in this way, this law is used. Then, a normal distribution with an average of 0 and a standard deviation of 100 is obtained, and a prediction error distribution as shown in FIG. 16 is obtained. In FIG. 16, the vertical axis represents the occurrence probability, and the horizontal axis represents the prediction error.

  Next, the daily correction amount calculation unit 3121 temporarily sets the daily correction amount u (step S25). For example, u = 0. The daily correction amount calculation unit 3121 reads data from the prediction error distribution data storage unit 3127, sets the prediction error distribution P (u) for the daily correction amount u from the daily correction amount u, and the prediction error distribution data storage unit. It stores in 3127 (step S27). In this step, the prediction error distribution generated in step S23 is moved by the daily correction amount u. Here, since u = 0, it is not different from FIG.

  Further, the daily correction amount calculation unit 3121 sets the cumulative penalty function R (u) from the daily correction amount u using the data stored in the penalty function data storage unit 3122, and stores it in the penalty function data storage unit 3122. (Step S29). For example, when the cumulative penalty function R (u) is defined by the ratio (%) of the cumulative error to the cumulative plan value as shown in FIG. 14, it is necessary to calculate the actual value of the cumulative error. In the example described above, the predicted plan value cumulative up to 24:00 on the 31st is 32400 (= 30000 + 1000 + (1000 + 400)) (= total predicted plan value up to the 29th + 30th and 31st predicted plan value−cumulative forecast error) It is. Here, in order to cancel the cumulative prediction error up to the 29th, correction is made to the predicted plan value. Thus, 3% of the cumulative penalty function R (u) as shown in FIG. 14 is 972 = (32400 × 0.03). Thus, in the case of FIG. 14, the cumulative penalty function R (u) is determined.

  Then, the daily correction amount calculation unit 3121 calculates the expected loss value and stores it in a storage device such as a main memory (step S31). The expected loss value E (u) is calculated by the following equation.

  The result of the product of the cumulative penalty function R (u) after step S29 and the prediction error distribution function P (u) shown in FIG. 16 is as shown in FIG. In FIG. 17, the vertical axis represents penalty × occurrence probability, and the horizontal axis represents prediction error. The expected loss value E (u) is a value obtained by adding penalty × occurrence probability for all cumulative prediction errors.

  The daily correction amount calculation unit 3121 determines whether the expected loss value calculated in step S31 is the minimum (step S33). If it is determined that it is not the minimum, the daily correction amount u is adjusted (step S35). Then, the process returns to step S27. Since these steps are the same as the processing for calculating the optimum solution such as Newton's method, detailed description thereof is omitted here.

  On the other hand, when it is determined that the expected loss value is the minimum, the daily correction amount calculation unit 3121 selects the daily correction amount u corresponding to the minimum expected loss value and the prediction error total (for canceling the prediction error total). The sum of the values (400 in the above example) is stored in the daily correction amount data storage unit 3125 as the final daily correction amount (step S37). For example, the daily correction amount data storage unit 3125 stores data in a data format as shown in FIG. That is, the customer ID, date, and date correction amount are stored. Then, the process returns to the original process.

  In the specific example described above, for example, the daily correction amount u is set to 200. Then, in step S27, the prediction error distribution P (u) becomes a distribution obtained by shifting the distribution of FIG. 16 to the right by 200, that is, a distribution as shown in FIG. Furthermore, the penalty function P (u) has a predicted plan total value of 31260 (= 32400 + 200), so 3% is 978. The final daily correction amount is 600 (= 400 + 200).

  Accordingly, the product of the penalty function R (u) and the prediction error distribution is as shown in FIG. Compared to FIG. 17, the height of the left mountain of the two peaks is significantly reduced, and the height of the right mountain is slightly increased. Therefore, it can be seen that the expected loss value E (u) is a value obtained by adding penalty × occurrence probability for all prediction errors, and thus is reduced as compared with the case of FIG.

  The relationship between the expected loss value E (u) and the correction value is summarized as shown in FIG. In the example of FIG. 21, the vertical axis represents the expected loss value, and the horizontal axis represents the correction value. Since the correction value is 0 in the case described first, it is plotted at the point G in the graph of FIG. On the other hand, when the correction value is 200, it is plotted at the point H in the graph of FIG. When calculation is performed for the other cases, a graph as shown in FIG. 21 is formed, and in the case of point H, the expected loss value is the lowest. Such a correction value that minimizes the expected loss value is obtained by the Newton method as described above. Note that the expected loss value may be calculated comprehensively and the lowest case may be identified.

  Note that the daily correction amount may be simply calculated without performing the processing shown in FIG. Whether the method as described above or the simple method is used depends on, for example, the correction method designation 311. For example, a cumulative error (= plan value cumulative-actual value cumulative) is calculated from past data stored in the plan value data storage unit 310 and the actual value data storage unit 304, and the cumulative error is calculated for the cumulative penalty determination period. The value divided by the remaining days is used as the daily correction amount. That is, the cumulative error is gradually eliminated in the same manner on the remaining days of the determination period.

  Returning to the description of the processing flow in FIG. 5, the simultaneous equal amount correction value calculation unit 3126 then includes the actual value data storage unit 304, the predicted consumption data storage unit 309, the penalty function data storage unit 3122, and the daily correction amount data. Using the data stored in the storage unit 3125, the correction value for each hour (simultaneous same amount correction value) is calculated according to the daily correction amount so as to minimize the simultaneous same amount penalty L (h), and the correction value The data is stored in the data storage unit 313 (step S13). This process is basically the same as step S11.

  That is, when the simultaneous equal amount penalty function L (hi) is affected by the cost unit price or the sales unit price, the penalty function generation unit 3123 stores the data stored in the penalty function data storage unit 3122 and the parameter data storage unit 3124. Is used to calculate the base data of the simultaneous equal amount penalty function L (hi). The simultaneous same amount correction value calculation unit 3126 generates prediction error distribution data for every hour of the next day using the latest actual value data and the corresponding predicted plan value data. The prediction error distribution data is stored in the prediction error distribution data storage unit 3127. Then, the simultaneous equal amount correction value hi is temporarily set for each time, and the prediction error distribution P (hi) and the simultaneous equal amount penalty function L (hi) corresponding to the simultaneous equal amount correction value hi are specifically set. Then, an expected loss value is calculated for each time according to the following formula.

  Unlike the processing shown in FIG. 15, (a) next day total correction amount Σhi = next day daily correction amount u, and (b) hourly consumption predicted value fi + correction value hi ≧ 0 (the planned value is (C) The correction value hi for each time at which the expected loss value is minimized is calculated according to the Newton method, for example, while satisfying the constraint condition (because it cannot be negative).

  Here, a specific example is shown for easy understanding. For example, the daily correction amount u is set to −300, the consumption predicted value for each hour from 0 o'clock to 8 o'clock is set to 30, the predicted consumption value for each hour from 8 o'clock to 22 o'clock is set to 50, 23 and 24 o'clock for each hour The predicted consumption value is 30. The daily consumption forecast value is 1000 in total.

  In such a case, prediction error distribution data is generated from the predicted consumption value stored in the predicted consumption value data storage unit 309 and the actual value stored in the actual value data storage unit 304 for each time. Here, for example, when the average is 0 and the standard deviation is calculated as 5, this is applied to the normal distribution.

  For example, when 1 o'clock is taken as an example, the predicted value is 30, and when the simultaneous equal amount correction value h1 for the time is set to 0, the 10% point of the simultaneous equal amount penalty (see FIG. 13) is 3, and the simultaneous same amount A penalty function L (hi) is determined. When the simultaneous same amount correction value h1 is set to another value, 10% is calculated for the consumption predicted value fi + the simultaneous same amount correction value hi.

  When the simultaneous equal amount correction value h1 is set to 0, as shown in FIG. 22, a curve K representing the prediction error distribution P (h) and a curve J representing the simultaneous equal amount penalty function L (hi) can be obtained. . The relationship between the penalty times the occurrence probability prediction error is as shown in FIG. The expected loss value is obtained by adding the penalty × occurrence probability for all prediction errors.

  As a result of such a calculation, for example, a calculation result is obtained in which the same amount correction value from 0:00 to 8:00 and 23:00 and 24:00 is -9, and the same amount correction value from 8:00 to 22:00 is -15. If it is, the total is −300 (= day correction amount).

  As described above, the simultaneous same amount correction value hi that satisfies the constraints (a), (b), and (c) is calculated for each time, and stored in the correction value data storage unit 313. The correction value data storage unit 313 stores data in a data format as shown in FIGS. 24 and 25, for example. The example of FIG. 24 is a format for the data calculated this time, and the customer ID, date, time, and plan value (consignment plan value) are registered. Further, the plan value data storage unit 310 stores past plan value data in a data format as shown in FIG. 25, for example. That is, the customer ID, date, time, plan value (consignment plan value), sending history, and sending time are registered. For example, numerical values are assigned to the sending history in order from the oldest.

In some cases, a method in which steps S11 and S13 are simplified together is employed. For example, a statistical value (average and standard deviation σ) of prediction errors may be calculated every hour, and a correction value may be calculated as, for example, prediction error average × (1−σ / σ _max ). Note that σ _max is the maximum value of the standard deviation in each time zone. In this way, the correction of the prediction error average is made smaller in the time zone where the standard deviation is larger.

  Returning to the description of FIG. 5, the plan value calculation unit 314 reads data from the correction value data storage unit 313 and the consumption prediction value data storage unit 309, and calculates a plan value (= consumption prediction value−correction value) for each time. The calculated value is stored in the planned value data storage unit 310 (step S15).

  In the example described above, the plan value for each hour from 1 o'clock to 8 o'clock and 23 and 24 o'clock is 21 (= predicted value + correction value = 30−9), and the plan from 8 o'clock to 22 o'clock A value of 35 (= predicted value + correction value = 50-15) is obtained.

  At this stage, the display as shown in FIG. 26 may be performed on the display device of the consignment planning system 3 or the display device of the client terminal connected to the consignment planning system 3 to confirm the user.

  In the screen example of FIG. 26, a scheduled delivery date (next day), a scheduled delivery value column, a customer planned value column from a customer, a consumption forecast value column, and a planned correction value column are provided. Here, a prediction method selection field for a consumption prediction value is provided, and the user can select a specific prediction method. Also, a correction method selection field for the planned correction value is provided, for example, “No correction” “Cumulative error correction 1 (perform processing for reducing penalty due to cumulative error according to the processing flow of FIG. 15)” “ It is possible to select from cumulative error correction 2 (implementing a process for reducing a penalty due to a simple cumulative error), “simultaneous correction of the same amount (a process in which steps S11 and S13 are simplified together)”, and the like.

  The user may change the prediction method to another method on this screen, or change the correction method to another method. Also, the user may directly correct or set specific numerical values on the screen as shown in FIG.

  After step S15, the plan value data transmission unit 315 transmits the plan value data for each time for the next day stored in the plan value data storage unit 310 to the consigner computer 7 via the network 1 ( Step S17). At this time, the transmission time and the transmission history in the format shown in FIG.

  In this way, the planned value is notified to the daily consignment company. After that, since the planned final value of the next day is notified from the consignment carrier computer 7, the consignment planning system 3 receives the notification. If the notification indicates an abnormality that satisfies a predetermined condition, an alarm mail or the like is transmitted to the person in charge. The data stored in the data storage unit in the consignment planning system 3 is read and output in response to access from an authorized user.

  Although one embodiment of the present invention has been described above, the present invention is not limited to this. For example, the functional blocks of the consignment planning system 3 shown in FIGS. 3 and 4 do not necessarily match the program module configuration. Further, the consignment planning system 3 may be configured by a plurality of computers.

  Also, the prediction method of the predicted consumption value is not limited to the method described above, and various methods can be employed. Further, the correction value is not limited to the method described above, and other methods may be used. However, as described above, the correction value for each time is calculated so as to minimize the cumulative penalty and the simultaneous equal amount penalty.

  Further, the consignment planning system 3, the consignment carrier computer 7, the customer computer 9, and the weather data distribution server 5 are computer devices. As shown in FIG. 27, the computer device includes a memory 2501 (storage unit) and a CPU 2503. (Processing unit), hard disk drive (HDD) 2505, display control unit 2507 connected to display device 2509, drive device 2513 for removable disk 2511, input device 2515, and communication control unit 2517 for connecting to the network Are connected by a bus 2519. Application programs including an operating system (OS) and a Web browser are stored in the HDD 2505, and are read from the HDD 2505 to the memory 2501 when executed by the CPU 2503. If necessary, the CPU 2503 controls the display control unit 2507, the communication control unit 2517, and the drive device 2513 to perform necessary operations. Further, data in the middle of processing is stored in the memory 2501 and stored in the HDD 2505 if necessary. Such a computer realizes various functions as described above by organically cooperating hardware such as the CPU 2503 and the memory 2501 described above with the OS and necessary application programs.

It is a schematic diagram which shows the outline | summary of the conventional consignment. It is a schematic diagram which shows the outline | summary of the consignment in this invention. It is a functional block diagram in an embodiment of the present invention. It is a functional block diagram of a correction value calculation unit. It is a figure which shows the main process flow in this Embodiment. It is a figure which shows the example of a format of the data stored in a performance value data storage part. It is a figure which shows the example of a format of the data stored in a customer plan value data storage part. It is a figure which shows the example of a format of the data stored in a weather data storage part. It is a figure which shows the example of a format of the data stored in a weather data storage part. It is a figure which shows the example of a format of the data stored in a consumption estimated value data storage part. It is a figure for demonstrating the outline | summary of simultaneous equal amount penalty. It is a figure for demonstrating the outline | summary of a cumulative penalty. It is a figure which shows the outline | summary of a simultaneous equal amount penalty function. It is a figure which shows the outline | summary of a cumulative penalty function. It is a figure which shows the processing flow of the calculation process of a daily correction amount. It is a figure which shows an example of the prediction error distribution of a day unit. It is a figure showing the relationship between a penalty x occurrence probability and a prediction error. It is a figure which shows an example of the data stored in a daily correction amount data storage part. It is a figure which shows the example at the time of correcting the prediction error distribution of a day unit according to a correction value. FIG. 20 is a diagram illustrating a relationship between penalty × occurrence probability and prediction error in the case of FIG. 19. It is a figure showing the relationship between a correction value and a loss expected value. It is a figure showing the simultaneous amount penalty function and the prediction error distribution of specific time. It is a figure showing the relationship between prediction error and penalty x occurrence probability. It is a figure which shows the example of a format of the data stored in a correction value data storage part. It is a figure which shows the example of a format of the data stored in a plan value data storage part. It is a figure which shows the example of a display screen. It is a functional block diagram of a computer.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Network 3 Consignment planning system 5 Weather data delivery server 7 Consignment company computer 9 Customer computer 301 Actual value data acquisition part 302 Customer planned value data acquisition part 303 Weather data acquisition part 304 Actual value data storage part 305 Customer planned value data storage part 306 Weather data storage unit 307 Consumption predicted value calculation unit 308 Prediction method designation 309 Consumption prediction value data storage unit 310 Plan value data storage unit 311 Correction method designation 312 Correction value calculation unit 313 Correction value data storage unit 314 Plan value calculation unit 315 Plan Value data transmission unit 3121 Daily correction amount calculation unit 3122 Penalty function data storage unit 3123 Penalty function generation unit 3124 Parameter data storage unit 3125 Daily correction amount data storage unit 3126 Simultaneous same amount correction value calculation unit 3127 Prediction error distribution data storage unit

Claims (9)

Without specifying the actual amount of the consignment target supplied to the consignment carrier's consignment network by the consignment requester, the error between the actual use amount of the consignment target by the customer of the consignment requester and the planned use value of the customer A computer system for consignees used when penalties are imposed
Means for obtaining data of the actual usage amount of the object of consignment for each first unit time for the customer from the computer of the consignment carrier, and storing it in the actual amount data storage unit;
Using the data stored in the actual amount data storage unit, the predicted plan value of the customer to be consigned is calculated for each first unit time according to a specific algorithm, and stored in the predicted plan value data storage unit Means,
Using data stored in the actual amount data storage unit, while reducing a penalty based on a cumulative error between the actual usage amount and the usage plan value in a second unit time longer than the first unit time. Calculating a correction value for the predicted plan value for each first unit time in order to minimize a penalty based on an error between the actual usage amount and the use plan value in the first unit time; Means for storing in the correction value data storage unit;
Means for calculating a use plan value for each first unit time using data stored in the predicted plan value data storage unit and the correction value data storage unit and storing the use plan value in the use plan value data storage unit; ,
A computer system. A method for calculating a customer's use plan value for a consignment target to be supplied to a customer facility of a consignment requester connected to a consignment network of a consignment carrier by a computer having a processing unit and a data storage unit,
Using the data stored in the actual amount data storage unit that stores the actual usage amount of the customer to be entrusted by the processing unit for each first unit time, the predicted planned value of the customer to be entrusted by the processor is obtained. Calculating for each of the first unit time according to a specific algorithm, and storing in the predicted plan value data storage unit;
Using the data stored in the actual amount data storage unit, the processing unit reduces an accumulated error between the actual usage amount and the usage plan value in a second unit time longer than the first unit time. A first correction value calculated in a third unit time longer than the first unit time and shorter than the second unit time, and stored in a first correction value data storage unit. A value calculation step;
Using the data stored in the actual amount data storage unit and the first correction amount data storage unit by the processing unit, while satisfying the first correction value in the third unit time, the first unit A second correction value is calculated for each first unit time within the third unit time so as to minimize a penalty based on an error between the actual usage amount and the planned use value in a unit time. Storing in the second correction value data storage unit;
Using the data stored in the predicted plan value data storage unit and the second correction value data storage unit by the processing unit, the prediction plan for each first unit time within the third unit time. Calculating a use plan value that is the sum of the value and the second correction value, and storing the use plan value in a use plan value data storage unit;
Usage plan value calculation method including The first correction value calculating step includes:
Identifying a candidate value for the first correction value by the processing unit;
A penalty function data storage unit for storing parameter data as a basis of a penalty function for a cumulative error between the actual usage amount and the usage plan value in the second unit time by the processing unit, and the actual amount data storage unit And the predicted plan value data storage unit that also stores the past predicted plan value, and corresponding to the candidate value of the first correction value, in the third unit time, A first step of calculating probability density function data on a cumulative prediction error between the actual usage amount and the predicted planned value, and the penalty function data in the second unit time;
A second step of calculating an expected value of a penalty from the probability density function data and the penalty function data by the processing unit;
Executing the first and second steps for candidate values different in the first correction value by the processing unit to identify the first correction value that minimizes the expected value of the penalty;
The utilization plan value calculation method according to claim 2, including: The first correction value calculating step includes:
The processing unit uses the data stored in the actual amount data storage unit and the usage plan value data storage unit that also stores past usage plan values to the present in the second unit time. Calculating a cumulative error between the generated usage amount and the planned usage value;
Calculating a correction value for canceling the accumulated error by the remaining third unit time number in the second unit time as the first correction value;
The utilization plan value calculation method according to claim 2, including: The second correction value calculating step includes:
Identifying a correction candidate value in a specific unit time for the first unit time by the processing unit;
A penalty function data storage unit that stores parameter data that is a basis of a penalty function for an error between the actual usage amount and the usage plan value in the first unit time, and the actual amount data storage unit. Using the data stored in the predicted plan value data storage unit that also stores past predicted plan values, corresponding to the correction candidate values, the actual usage amount in the specific unit time and the A first step of calculating data on a probability density function for a prediction error from a predicted plan value and data on the penalty function in the specific unit time;
A second step of calculating an expected value of a penalty from the probability density function data and the penalty function data by the processing unit;
The processing unit executes the first and second steps for a plurality of correction candidate values for all the first unit times within the third unit time, and the processing unit performs the first unit time for the third unit time. Identifying a second correction value that minimizes the expected value of the penalty for each of the first unit times within the third unit time while satisfying a first correction value;
The utilization plan value calculation method according to claim 2, including: The utilization plan value according to claim 3 or 5, further comprising a step of performing a process of changing the parameter data in accordance with a change in at least one of a consignment cost and a sales unit price, and storing the parameter data in the penalty function data storage unit. Calculation method.   The program for making a computer perform the utilization plan value calculation method of any one of Claims 2 thru | or 6. Without specifying the actual amount of the consignment target supplied to the consignment carrier's consignment network by the consignor requester, the error between the actual use amount of the consignment target by the customer of the consignment requester and the planned use value of the customer A computer system for consignees used when penalties are imposed
Means for obtaining data on the actual usage amount of the target of consignment for each unit time for the customer from the computer of the consignment carrier, and storing it in the actual amount data storage unit;
Using the data stored in the actual amount data storage unit, the forecast plan value of the customer to be entrusted is calculated for each unit time according to a specific algorithm, and the forecast plan value is also stored in the past forecast plan value Means for storing in the data storage;
Using the data stored in the actual amount data storage unit and the predicted plan value data storage unit, for each unit time, statistics of the prediction error between the actual use amount and the predicted plan value in the unit time Means for calculating an amount, calculating a correction value for the predicted plan value for each unit time based on the statistic, and storing the correction value in a correction value data storage unit;
Means for calculating a use plan value for each unit time using data stored in the predicted plan value data storage unit and the correction value data storage unit, and storing in the use plan value data storage unit;
A computer system. A computer system that calculates a customer's usage plan value for a consignment target that is connected to a consignor's customer facility connected to a consignment carrier's consignment network,
Using the data stored in the actual amount data storage unit that stores the actual usage amount of the customer to be entrusted by the processing unit for each first unit time, the predicted planned value of the customer to be entrusted by the processor is obtained. Means for calculating the first unit time according to a specific algorithm and storing it in the predicted plan value data storage unit;
Using the data stored in the actual amount data storage unit, the processing unit reduces an accumulated error between the actual usage amount and the usage plan value in a second unit time longer than the first unit time. A first correction value calculated in a third unit time longer than the first unit time and shorter than the second unit time, and stored in a first correction value data storage unit. A value calculating means;
Using the data stored in the actual amount data storage unit and the first correction amount data storage unit by the processing unit, while satisfying the first correction value in the third unit time, the first unit A second correction value is calculated for each first unit time within the third unit time so as to minimize a penalty based on an error between the actual usage amount and the planned use value in a unit time. Means for storing in the second correction value data storage unit;
Using the data stored in the predicted plan value data storage unit and the second correction value data storage unit by the processing unit, the prediction plan for each first unit time within the third unit time. Means for calculating a use plan value that is the sum of the value and the second correction value, and storing the use plan value in a use plan value data storage unit;
A computer system.

Images