JP2018173837A - Prediction device, prediction system, prediction method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a prediction device, a prediction system, a prediction method and a program capable of performing demand prediction at low cost and high speed.SOLUTION: A prediction device 100 comprises: a waveform acquisition part 110 acquiring waveforms representing in a time-series values of objective variables measured in the past; a classification processing part 131 classifying similar waveforms among the plurality of waveforms into one pattern and setting an explanatory variable indicating a characteristic common to the waveforms included in the one pattern; a probability calculation part 132 calculating probability distribution of values which the objective variable may take in a prediction target period by selecting a pattern matching the explanatory variable in the future prediction target period and performing statistical processing on the waveform included in the pattern; and a prediction part 133 predicting a value of the objective variable in the prediction target period on the basis of the probability distribution.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a prediction device, a prediction system, a prediction method, and a program.

In order to efficiently use energy and resources, the consumer needs to create a plan for predicting the demand for energy and resources and securing supply according to the demand.
For example, when using electric power, a consumer predicts the demand per unit time (for example, 30 minutes) and concludes a contract for purchasing necessary electric power from an electric power company. At this time, a consumer sets as the electric power purchased from an electric power provider, ie, contract electric power, based on the largest demand amount among the demand amounts according to unit time.
In addition, the consumer pays the electric power company a basic fee according to the contracted power and a pay-as-you-go fee according to the amount of power actually used. The customer can reduce the basic charge as the contracted power is reduced, but if the customer uses (purchases) the power exceeding the contracted power, he / she has to pay an excess penalty to the power company. . For this reason, a consumer needs to perform highly accurate demand prediction in order to reduce a basic charge and an excess penalty.

As a method for predicting power demand, for example, Patent Literature 1 discloses a method for predicting power demand for each electrical load by inputting conditions such as weather into a prediction model prepared for each electrical load. .
In addition, a physical model that imitates equipment and the like possessed by a customer is created, and a large number of simulations in which a plurality of conditions (explanatory variables) are stochastically changed in the physical model are executed to obtain the power demand (objective variable). A prediction method using Monte Carlo simulation is known.

JP, 2015-201972, A

However, in the conventional prediction method, it is necessary to create a detailed physical model that covers a large number of facilities that a customer has, and it takes a lot of time to create such a physical model.
Moreover, in the prediction method using the Monte Carlo simulation, it is necessary to execute many simulations in order to improve the prediction accuracy, and it takes a long time to obtain the prediction result. For this reason, for example, when predicting the power demand amount after 30 minutes as the prediction target period, it is not possible to complete all the simulations by the prediction target period, and the prediction result can be obtained before actually procuring power. There is no possibility.

  The present invention has been made in view of such problems, and provides a prediction device, a prediction system, a prediction method, and a program capable of performing demand prediction at high speed.

In order to solve the above problems, the present invention employs the following means.
According to the first aspect of the present invention, the prediction device (100) includes a waveform acquisition unit (110) that acquires a waveform representing a value of a target variable measured in the past in time series, and a plurality of the waveforms Categorizing similar waveforms into one pattern and setting an explanatory variable indicating characteristics common to the waveforms classified into the one pattern, and the explanation in a future prediction target period A probability calculator (132) that calculates a probability distribution of values that can be taken by the objective variable in the prediction target period by selecting a pattern that matches the variable and performing statistical processing on the waveform classified into the pattern; A prediction unit (133) for obtaining a value of the objective variable in the prediction target period based on the probability distribution.
As described above, the prediction device calculates and predicts the probability distribution for the waveform classified into the pattern that matches the explanatory variable in the prediction target period, and calculates the value of the objective variable based on the probability distribution. Can be executed quickly. For this reason, for example, the time required for calculation and prediction can be shortened as compared with an existing prediction method using Monte Carlo simulation.

According to the second aspect of the present invention, in the prediction device according to the above aspect, the prediction unit determines that the objective variable is within a range of confidence intervals determined in advance in the prediction target period based on the probability distribution. Among the possible values, an upper limit value and a lower limit value are obtained.
By doing in this way, the prediction apparatus can obtain the upper limit value and the lower limit value with high accuracy after eliminating the value with low certainty from the values that the objective variable can take in the prediction target period.

According to the third aspect of the present invention, the prediction device according to any one of the above aspects is based on the value of the objective variable obtained by the prediction unit, and the supply and demand plan for the object related to the objective variable. Is further provided with a plan creation unit (134).
Thus, since the prediction apparatus creates a supply and demand plan in the prediction target period based on the value of the objective variable predicted by the prediction unit, it is possible to optimally operate the equipment in a planned manner in the prediction target period.

According to the fourth aspect of the present invention, the prediction system includes a measuring device (10) that measures an actual value of an objective variable in a plant, and values that the objective variable can take in a future prediction target period of the plant. A prediction device (100) to be obtained.
The prediction device includes a waveform acquisition unit (110) that acquires a waveform representing the value of the objective variable measured by the measurement device in time series, and a plurality of the waveforms that are similar to each other in a single pattern. A classification processing unit (131) that classifies and sets an explanatory variable indicating characteristics common to the waveforms classified into the one pattern, and a value that the objective variable can take in the prediction target period that matches the explanatory variable A probability calculation unit (132) that calculates the probability distribution based on statistical processing, and a prediction unit (133) that calculates the value of the objective variable in the prediction target period based on the probability distribution.

  According to the fifth aspect of the present invention, the prediction method includes a waveform acquisition step of acquiring a waveform representing the value of the objective variable measured in the past in time series, and a similar waveform among the plurality of waveforms. A classification process step for classifying into one pattern and setting an explanatory variable indicating characteristics common to the waveforms classified into the one pattern; and the target variable is taken in a future prediction target period that matches the explanatory variable. A probability calculating step of calculating a probability distribution of values to be obtained based on statistical processing; and a predicting step of obtaining a value of the objective variable in the prediction target period based on the probability distribution.

  According to the sixth aspect of the present invention, the program uses a computer of the prediction device, a waveform acquisition unit that acquires a waveform representing the value of the objective variable measured in the past in time series, and a similarity among the plurality of waveforms. A processing unit for classifying waveforms to be classified into one pattern and setting an explanatory variable indicating characteristics common to the waveforms classified into the one pattern; and the purpose in a future prediction target period that matches the explanatory variable It functions as a probability calculation unit that calculates a probability distribution of values that a variable can take based on statistical processing, and a prediction unit that calculates the value of the target variable in the prediction target period based on the probability distribution.

  According to the prediction device, the prediction system, the prediction method, and the program according to the present invention, it is possible to perform demand prediction at high speed.

It is a figure which shows the function structure of the prediction apparatus which concerns on one Embodiment of this invention. It is the 1st figure showing the processing flow of the classification device concerning one embodiment of the present invention. It is a figure for demonstrating the classification | category process which concerns on one Embodiment of this invention. It is a 2nd figure which shows the processing flow of the classification device based on one Embodiment of this invention. It is a 1st figure for demonstrating the prediction process which concerns on one Embodiment of this invention. It is a 2nd figure for demonstrating the prediction process which concerns on one Embodiment of this invention. It is a figure which shows the hardware constitutions of the prediction apparatus which concerns on one Embodiment of this invention.

(Functional configuration of prediction system)
Hereinafter, a prediction system according to an embodiment of the present invention will be described with reference to FIGS.
In the present embodiment, a plant that consumes electric power will be described as an example of a consumer. The plant is a facility such as a power plant, a factory, a warehouse, or a commercial facility. Moreover, in this embodiment, the prediction system demonstrates the example which estimates the electric power consumed in a plant, ie, the demand amount of electric power.
In the following description, the demand amount predicted by the prediction system is also referred to as “object variable”. An objective variable is a variable to be predicted. The objective variable is also called “dependent variable” or “external criterion” and can be regarded as the result of things.

FIG. 1 is a diagram illustrating a functional configuration of a prediction device according to an embodiment of the present invention.
As shown in FIG. 1, the prediction system 1 includes a measurement device 10, a server 20, and a prediction device 100.
The measuring device 10 is provided in the plant A, and measures the amount of power used by the plant A (power demand amount) and the state of the plant A every unit time (for example, 30 minutes).
For example, when the plant A is a factory, the state information of the plant A includes information such as the production amount of the plant A and the operating state of the equipment. The operation state of the production facility is an operation rate for each unit time of facilities that consume power, such as air conditioning, lighting, and production facilities.
The measuring device 10 includes a wattmeter 11 for measuring the actual power demand in the plant A and a sensor unit 12 for measuring the actual production amount and the operating state of the plant A.
In addition, the measuring device 10 transmits “demand information” that associates the measured power demand amount and measurement date and time “state information” that associates the measured state and measurement date and time to the prediction device 100 as a measurement result.
The server 20 is communicably connected to the prediction device 100 via the network NW.
The server 20 collects “weather information” for each region and each time and transmits it to the prediction device 100. The weather information includes information such as weather, maximum temperature, minimum temperature, humidity, rainfall, and wind speed.

The prediction device 100 predicts the power demand amount in the future prediction target period of the plant A for each unit time based on the measurement result received from the measurement device 10 and the weather information acquired from the server 20. The prediction target period is, for example, 0:00 to 24:00 on the next day.
As illustrated in FIG. 1, the prediction device 100 includes a waveform acquisition unit 110, a variable acquisition unit 120, a processing unit 130, and a storage unit 140.

The waveform acquisition unit 110 acquires a waveform representing the power demand amount of the plant A measured in the past in time series.
Specifically, the waveform acquisition unit 110 divides a plurality of demand information received from the measuring device 10 for each predetermined period, and arranges the divided demand information in time series, thereby changing the power demand amount over time. Acquires the waveform represented in the series. Then, the waveform acquisition unit 110 stores and accumulates the waveform in the storage unit 140.
In the present embodiment, the waveform acquisition unit 110 acquires a daily waveform, for example, by setting a predetermined period to one day (each day from 0:00 to 24:00).

The variable acquisition unit 120 acquires explanatory variables. An explanatory variable is a variable that describes an objective variable. Explanatory variables are also called “independent variables” and can be taken as the cause of things. In the present embodiment, the explanatory variable indicates a factor (characteristic) that affects the power demand.
Specifically, the variable acquisition unit 120 acquires “facility information” including information such as the location of the plant A, holidays, start hours, and end hours stored in advance in the storage unit 140 as an explanatory variable. The storage unit 140 stores a calendar, and the variable acquisition unit 120 may acquire the day of the week corresponding to the date from the calendar.
Further, “state information” is acquired from the measuring device 10 and “weather information” of the area including the location of the plant A is acquired from the server 20 as an explanatory variable.
Furthermore, the variable acquisition unit 120 acquires “planned information” in the prediction target period of the plant A as an explanatory variable. For example, when the plant A is a factory, the schedule information includes the planned production amount of the plant A in the prediction target period, the planned operating state of the equipment, and the like. The schedule information may be stored in advance in the storage unit 140 or may be input by the operator via the server 20.

The processing unit 130 includes a classification processing unit 131, a probability calculation unit 132, a prediction unit 133, and a plan creation unit 134.
The classification processing unit 131 classifies similar waveforms among a plurality of waveforms accumulated in the storage unit 140 into one pattern (as one pattern / one type), and the waveforms classified into the one pattern. Set an explanatory variable indicating the characteristics common to. The classification processing unit 131 performs a machine learning using a plurality of waveforms, thereby creating a decision tree indicating the relationship of what waveform tendency (pattern) appears to the characteristics.
The characteristic indicates a value of at least one of information included in state information, weather information, and facility information. For example, when many waveforms similar to holidays of the plant A appear, the characteristic common to these waveforms is “holiday”. That is, the characteristics common to the waveforms are factors that change the waveforms by affecting the power demand.
Note that “common” is not limited only to the value matching (for example, the acquired date is “holiday”), but the value is within a predetermined range (for example, the maximum temperature is “10 ° C. or more and less than 15 ° C.”). )) And a value exceeding a predetermined threshold (for example, the maximum temperature is “30 ° C. or higher”, “less than 0 ° C.”)

The probability calculation unit 132 selects a pattern that matches the explanatory variable in the future prediction target period, and performs statistical processing on the waveform classified into the pattern, thereby probing the probability that the power demand can take in the prediction target period. Calculate the distribution. For example, in the present embodiment, the probability calculation unit 132 describes an example of calculating a probability distribution using a normal distribution method as statistical processing, but the present invention is not limited to this. In another implementation, the probability calculation unit 132 may calculate the probability distribution using a method such as maximum likelihood estimation, Poisson distribution, t distribution, binomial distribution, or the like as statistical processing.
In the present embodiment, the probability calculation unit 132 further divides the prediction target period by unit time and calculates a probability distribution for each unit time.

  The prediction unit 133 predicts the power demand amount in the prediction target period based on the probability distribution calculated by the probability calculation unit 132.

  The plan creation unit 134 creates a power procurement plan for the prediction target period of the plant A based on the power demand amount predicted by the prediction unit 133.

  The storage unit 140 sequentially stores the waveform acquired by the waveform acquisition unit 110 and the explanatory variable acquired by the variable acquisition unit 120. The storage unit 140 stores facility information in advance.

(Prediction system processing flow)
FIG. 2 is a first diagram illustrating a processing flow of the classification device according to the embodiment of the present invention.
FIG. 3 is a diagram for explaining classification processing according to an embodiment of the present invention.
The prediction device 100 accumulates the waveform of the plant A and performs processing for classifying the waveform before predicting the power demand. Hereinafter, the details of the classification process will be described with reference to FIGS.

As illustrated in FIG. 2, the waveform acquisition unit 110 acquires a waveform that represents a change in power demand in time series based on a plurality of pieces of demand information received from the measurement device 10 (step S100).
In the present embodiment, the waveform acquisition unit 110 acquires a waveform obtained by arranging a plurality of demand information in time series. Furthermore, the waveform acquisition unit 110 acquires a daily waveform (each day from 0:00 to 24:00) by dividing the waveform every day. In addition, the waveform acquisition unit 110 stores and accumulates daily waveforms in the storage unit 140.
At this time, the variable acquisition unit 120 acquires the state information of the plant A and the weather information measured during the same period as each waveform from each of the measurement device 10 and the server 20.

Next, the classification processing unit 131 performs machine learning using a plurality of waveforms stored in the storage unit 140, and creates a decision tree (step S101).
Based on the characteristics (value of at least one of the information included in the state information, weather information, and facility information) and the waveform, the classification processing unit 131 uses a waveform to resemble a waveform that is similar to any characteristic. Classify whether it appears a lot. Then, as shown in FIG. 3, the classification processing unit 131 classifies similar waveforms into one pattern, and sets a decision tree in which explanatory variables indicating characteristics common to the waveforms classified into the one pattern are set. create.
For example, as shown in FIG. 3, it is assumed that similar waveforms are classified into pattern 1, pattern 2, pattern 3,. At this time, when many waveforms of pattern 1 appear when “maximum temperature is 20 ° C. or higher”, a characteristic that “maximum temperature is 20 ° C. or higher” is set as an explanatory variable of pattern 1.

  The prediction apparatus 100 updates the decision tree by repeating the above process every time a new waveform is acquired or at a predetermined timing (for example, every other month).

FIG. 4 is a second diagram illustrating a processing flow of the classification device according to the embodiment of the present invention.
FIG. 5 is a first diagram for explaining a prediction process according to an embodiment of the present invention.
FIG. 6 is a second diagram for explaining the prediction processing according to the embodiment of the present invention.
Hereinafter, with reference to FIG. 4 to FIG. 6, the details of processing in which the prediction device 100 performs prediction of power demand and creation of a power procurement plan will be described.
As shown in FIG. 4, first, the variable acquisition unit 120 acquires explanatory variables in the future prediction target period of the plant A (step S200).
For example, when predicting the power demand for the next day, the variable acquisition unit 120 acquires various information associated with the date of the next day from the storage unit 140 and the server 20. For example, the variable acquisition unit 120 acquires, from the storage unit 140 or the server 20, schedule information including the scheduled production amount of the plant A on the next day and the planned operation state of the equipment. Further, the variable acquisition unit 120 acquires the weather information of the area including the location of the plant A on the next day from the server 20.

  Next, the probability calculation unit 132 estimates and selects a waveform pattern in the prediction target period based on the explanatory variable acquired by the variable acquisition unit 120 and the decision tree created by the classification processing unit 131 (step S201). ).

Next, the probability calculation unit 132 divides the prediction target period by unit time by performing statistical processing on a plurality of waveforms included in the selected pattern, and the probability of the value that the power demand can take by unit time Distribution is calculated (step S202).
For example, when it is assumed that the possible value of the power demand amount follows a normal distribution, the probability calculation unit 132 represents the probability distribution by a normal distribution as shown in FIG.

Next, the prediction unit 133 predicts the lower limit value and the upper limit value of the power demand amount included in the predetermined confidence interval based on the normal distribution for each unit time (step S203). For example, when a confidence interval with a confidence level of 97% is set, if the average power demand is μ and the standard deviation from the average is σ, the lower limit (MIN in FIG. 5) is approximately indicated by “μ-2σ”. The upper limit value (MAX in FIG. 5) is approximately “μ + 2σ”.
Then, as shown in FIG. 6, the prediction unit 133 arranges the lower limit value and the upper limit value of the power demand calculated for each unit time in time series, and shows a waveform representing the transition of the lower limit value in the prediction target period (in FIG. 6). MIN) and a waveform (MAX in FIG. 6) representing the transition of the upper limit value are predicted.

Next, the plan creation unit 134 creates a power procurement plan for the prediction target period of the plant A based on the power demand amount predicted by the prediction unit 133 (step S204).
For example, the plan creation unit 134 creates a procurement plan in which contract power is set so that power greater than or equal to the largest value of the power demand amount predicted by the probability calculation unit 132 can be procured. In addition, the plan preparation part 134 uses the power generation apparatus and cogeneration system which the plant A has, and when the electric power procured (electricity purchase) from an electric power provider can be reduced, the procurement plan which considered the said reduction | decrease is considered. You may make it create.

  The prediction device 100 repeats the above-described process at every predetermined timing (for example, each day), predicts the power demand amount in the future prediction target period, and creates a power procurement plan.

(Hardware configuration of prediction device)
FIG. 7 is a diagram illustrating a hardware configuration of a prediction device according to an embodiment of the present invention.
Hereinafter, the hardware configuration of the prediction apparatus 100 according to the present embodiment will be described with reference to FIG.

The computer 900 includes a CPU 901, a main storage device 902, an auxiliary storage device 903, an input / output interface 904, and a communication interface 905.
The prediction apparatus 100 described above is implemented in a computer 900. And the operation | movement of each part of the prediction apparatus 100 mentioned above is memorize | stored in the auxiliary storage device 903 which each computer 900 has in the format of the program. The CPU 901 (processing unit 130) reads the program from the auxiliary storage device 903, expands it in the main storage device 902, and executes the above processing according to the program. In addition, the CPU 901 ensures a storage area corresponding to the storage unit 140 in the main storage device 902 according to the program. In addition, the CPU 901 ensures a storage area for storing data being processed in the auxiliary storage device 903 according to the program.
The computer 900 may be connected to the external storage device 910 via the input / output interface 904, and the storage unit 140 may be secured in the external storage device 910. The computer 900 may be connected to the external storage device 920 via the communication interface 905, and the storage unit 140 may be secured in the external storage device 920.

  In at least one embodiment, the auxiliary storage device 903 is an example of a tangible medium that is not temporary. Other examples of the tangible medium that is not temporary include a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, and a semiconductor memory connected via the input / output interface 904. When this program is distributed to the computer 900 via a communication line, the computer 900 that has received the distribution may develop the program in the main storage device 902 and execute the above processing.

  The program may be for realizing a part of the functions described above. Further, the program may be a so-called difference file (difference program) that realizes the above-described function in combination with another program already stored in the auxiliary storage device 903.

(Function and effect)
As described above, the prediction device 100 according to the present embodiment includes the waveform acquisition unit 110 that acquires a waveform representing the value of the power demand (objective variable) measured in the past in time series, and among the plurality of waveforms, A classification processing unit 131 that classifies similar waveforms into one pattern and sets an explanatory variable indicating characteristics common to the waveforms classified into the one pattern, and a pattern that matches an explanatory variable in a future prediction target period And performing a statistical process on the waveform classified into the pattern, a probability calculation unit 132 that calculates a probability distribution of values that the power demand can take in the prediction target period, and a prediction target based on the probability distribution A prediction unit 133 that predicts a power demand amount in a period (determines an objective variable).
As described above, the prediction device 100 calculates the probability distribution only for the waveform classified into the pattern that matches the explanatory variable in the prediction target period, and calculates the value of the objective variable based on the probability distribution. And the prediction can be performed quickly. For this reason, for example, the time required for calculation and prediction can be shortened as compared with an existing prediction method using Monte Carlo simulation.
Moreover, in the above-described embodiment, the prediction device 100 performs the classification process in advance based on a plurality of past waveforms before performing the power demand amount prediction process. For this reason, when performing the prediction process, the prediction device 100 can easily estimate and select a pattern estimated in a future prediction target period based on a decision tree created in advance as a result of the classification process. it can. As a result, the calculation processing in the probability calculation unit 132 can be further speeded up.
Furthermore, since the prediction apparatus 100 only needs to be provided with the power demand amount and explanatory variables (state information of the plant A, facility information, weather information) measured in the past, a detailed physical model that covers the facilities of the plant A Can be omitted. Thereby, the prediction device 100 can reduce the cost for the demand prediction.

Further, the prediction unit 133 obtains an upper limit value and a lower limit value among values that can be taken by the power demand in the range of the confidence interval determined in advance in the prediction target period based on the probability distribution.
By doing in this way, the prediction apparatus 100 can obtain | require a highly accurate upper limit value and lower limit value, after removing the value with low certainty among the values which the electric power demand amount can take in the prediction target period. .

The prediction device 100 further includes a plan creation unit 134 that creates a power procurement plan based on the power demand predicted by the prediction unit 133.
Thus, since the prediction apparatus 100 creates a procurement plan in the prediction target period based on the predicted power demand, it is possible to optimally operate the equipment in a planned manner in the prediction target period. For example, by making a contract based on the plan created by the plan creation unit 134, setting contract power that greatly exceeds the predicted power demand amount in the forecast period, or setting contract power to be small and excess violation The possibility of generating gold can be reduced. As a result, the prediction device 100 can suppress an increase in the cost of purchasing power.

As mentioned above, although embodiment of this invention was described in detail, unless it deviates from the technical idea of this invention, it is not limited to these, A some design change etc. are possible.
For example, in the above-described embodiment, the example in which the prediction system 1 predicts the power demand has been described, but the present invention is not limited to this.
In another embodiment, for example, the plant A may be equipment such as a gas turbine, an air conditioner, and a refrigerator provided in the facility. Moreover, the prediction system 1 may predict the demand amount of energy (heat, etc.) and resources (water, steam, fuel, etc.) in the plant A, or may predict the demand amount and sales of products to be sold.
In this case, the measuring device 10 includes a measuring instrument for measuring the amount of energy, resources, and commodities instead of the wattmeter 11.
In addition, the plan creation unit 134 of the prediction device 100 creates a procurement plan for energy, resources, and products (inventory). The plan creation unit 134 may create a supply plan for energy, resources, and commodities supplied to the plant A based on the demand amount and the sales forecast.

  Moreover, although the above-mentioned embodiment demonstrated the aspect in which the one prediction apparatus 100 performs the demand prediction in the one plant A, it is not restricted to this. In another embodiment, one prediction device 100 may be connected to the measurement devices 10 of a plurality of plants to perform demand prediction of each of the plurality of plants.

1 Prediction System 10 Measuring Device 11 Wattmeter 12 Sensor Unit 100 Prediction Device 110 Waveform Acquisition Unit 120 Variable Acquisition Unit 130 Processing Unit 131 Classification Processing Unit 132 Probability Calculation Unit 133 Prediction Unit 134 Plan Creation Unit 140 Storage Unit 20 Server

Claims (6)

A waveform acquisition unit that acquires a waveform representing the value of the objective variable measured in the past in time series, and
A classification processing unit that classifies similar waveforms among a plurality of the waveforms into one pattern, and sets an explanatory variable indicating characteristics common to the waveforms classified into the one pattern;
A probability distribution of values that can be taken by the objective variable in the prediction target period is calculated by selecting a pattern that matches the explanatory variable in the future prediction target period and performing statistical processing on the waveform classified into the pattern A probability calculator,
Based on the probability distribution, a prediction unit for obtaining a value of the objective variable in the prediction target period;
A prediction device comprising: The prediction unit obtains an upper limit value and a lower limit value among values that can be taken by the objective variable within a range of confidence intervals determined in advance in the prediction target period based on the probability distribution.
The prediction device according to claim 1. Based on the value of the objective variable obtained by the prediction unit, further comprising a plan creation unit that creates a supply and demand plan for the object related to the objective variable.
The prediction device according to claim 1 or 2. A measuring device that measures the actual value of the objective variable in the plant;
A prediction device for obtaining a possible value of the objective variable in a future prediction target period of the plant;
With
The prediction device is
A waveform acquisition unit for acquiring a waveform representing the value of the objective variable measured by the measurement device in time series;
A classification processing unit that classifies similar waveforms among a plurality of the waveforms into one pattern, and sets an explanatory variable indicating characteristics common to the waveforms classified into the one pattern;
A probability calculator that calculates a probability distribution of values that the objective variable can take in the prediction target period that matches the explanatory variable based on statistical processing;
Based on the probability distribution, a prediction unit for obtaining a value of the objective variable in the prediction target period;
A prediction system. A waveform acquisition step for acquiring a waveform representing the value of the objective variable measured in the past in time series;
A classification processing step of classifying similar waveforms among a plurality of the waveforms into one pattern, and setting an explanatory variable indicating characteristics common to the waveforms classified into the one pattern;
A probability calculating step of calculating a probability distribution of values that the objective variable can take in a future prediction target period that matches the explanatory variable based on statistical processing;
A prediction step for obtaining a value of the objective variable in the prediction target period based on the probability distribution;
A prediction method comprising: Predictor computer
A waveform acquisition unit that acquires a waveform representing the value of the objective variable measured in the past in time series,
A classification processing unit that classifies similar waveforms among a plurality of the waveforms into one pattern, and sets an explanatory variable indicating characteristics common to the waveforms classified into the one pattern,
A probability calculator that calculates a probability distribution of values that the objective variable can take in a future prediction target period that matches the explanatory variable based on statistical processing;
A prediction unit for obtaining a value of the objective variable in the prediction target period based on the probability distribution;
Program to function as.

Images