JP5980195B2 - Distribution system load prediction device and distribution system load prediction method - Google Patents

Description

  The present invention relates to a load prediction device for a distribution system and a load prediction method for a distribution system, and more particularly to a load prediction device for a distribution system formed by a plurality of distribution sections and a load prediction method for a distribution system formed by a plurality of distribution sections. It is.

  2. Description of the Related Art In recent years, a power load prediction apparatus that predicts a power load of a distribution system that supplies power to a load device such as a consumer electronics product has been proposed. For example, Patent Document 1 discloses a power load prediction device that predicts a power load of a distribution system in a predetermined distribution section based on past power load results and weather information. Patent Literature 2 discloses that the energy consumption amount in a new building is estimated from a measurement value database that stores the measurement value of the energy consumption amount of the building and the attribute information of the building in association with each other. Patent Document 3 discloses that a power load prediction is performed by obtaining a human behavior and a device operation pattern based on weather information and an occurrence probability thereof from past power load results and weather information. Patent Document 4 discloses that the reference data for estimating the power consumption is corrected for each of the devices included in the house according to the conditions regarding the house or the resident.

JP 2010-193605 A JP 2011-128930 A JP 2011-232903 A JP 2013-33401 A

  However, in the above-described power load prediction device, although the future load can be predicted for a distribution section or a customer who has measured past load results, the load is determined for any other distribution section or any other customer. I cannot make predictions. Even for power distribution sections and customers measured in the past, it is not possible to make a load prediction that takes into account the equipment penetration rate and the equipment usage time. Future changes in load due to the spread of equipment cannot be predicted.

  In addition, the number of devices and device availability in the distribution section depend on the attributes of the consumers included in the distribution section, but it is not possible to make a load prediction that takes into account the proportion or bias of the consumer attributes in the distribution section. .

  Furthermore, in order to predict the power load in the distribution section using the above-described power load prediction device, detailed information such as facilities and equipment held by each consumer included in the distribution section is provided for each individual consumer. Although it is necessary to obtain such information, it is difficult for an electric power company or the like that wants to predict the power load in the distribution section to obtain detailed information on such consumers. For this reason, it is difficult to predict the load of the entire distribution section by predicting and summing the loads of each customer belonging to the distribution section one by one.

  Specifically, in Patent Document 1, a load is not measured for each model as a past load record. Therefore, even if a change in load due to weather information can be predicted, a change in load due to a change in the penetration rate of each model and a change in the performance of each model cannot be predicted.

  In Patent Document 2, the power consumption of a newly constructed building is predicted from past results of similar building attributes. However, this is a prediction of the power consumption of the entire building, and the power consumption prediction for each device type is performed. I can't. In addition, the estimation process does not use a model that divides the device operation probability and power consumption for each device type, so long-term predictions such as future changes in operation probability and power consumption can be made. I can't. Furthermore, there is a description in which only factors that have a large effect on energy consumption are selected and compared as judgments on the similarity of building attributes, but it is not always obvious which factor has a large effect. If a factor having a small influence is selected by mistake, a similar past performance cannot be used, and thus the prediction accuracy decreases.

  In patent document 3, based on the past performance and weather information, an apparatus operation pattern and a human action pattern (present / absent) are analyzed for each customer, and the pattern and its probability are modeled. This pattern is peculiar to the consumer who measured the past performance, and has not been modeled to the general consumer. For this reason, it is impossible to estimate the pattern of an arbitrary consumer. Further, in Patent Document 3, when load prediction is performed for a customer other than the measurement customer, the facility information of the building (family structure, type of owned equipment, total floor area, etc.) is similar. A method is also described in which a past performance is searched and a load of another customer is predicted based on the searched past performance. When searching for a similar consumer, the details of the individual facility information of each consumer are required, as in Patent Document 2. However, it is difficult to obtain the details of the facility information of all the individual consumers in the distribution section where the load prediction is performed.

  The technology of Patent Document 4 is intended to predict the energy consumption of individual houses, and cannot predict the power load of a distribution system that supplies power to many houses. Absent.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a technology capable of accurately predicting a power load in an arbitrary distribution section of a distribution system.

  The load prediction device for a distribution system according to the present invention predicts a power load of a distribution system that has a plurality of distribution sections and supplies power to a load device. The load prediction device of the distribution system includes a ownership ratio model storage unit, an operation probability model storage unit, a power consumption model storage unit, a device number estimation unit, an operation number prediction unit, a power consumption estimation unit, and a load total unit And have. The holding ratio model storage unit stores a holding ratio model for each type of load device for each first housing information category for each distribution section. The operation probability model storage unit stores an operation probability model for each type of load device for each second housing information category and for each time. The power consumption model storage unit stores a power consumption model for each active load device for each model. The number-of-devices estimation unit is based on the ownership rate model stored for each first housing information category in the ownership rate model storage unit and the number-of-homes related information for each first housing information category for one distribution section. The number of models by type is estimated for each first housing information category for one distribution section. The number-of-operating-unit predicting unit includes the number of units for each model estimated for each first housing information section by the number-of-devices estimating unit, and the operating probability model stored for each second housing information section in the operating probability model storage unit. Used to predict the expected number of operating units by model and time. The power consumption estimation unit is based on the expected number of operating units by type and time predicted by the operating unit prediction unit, and the power consumption model per unit stored for each model in the power consumption model storage unit, Estimate the expected power consumption by model and time. The load summation unit calculates the power load by time by summing the expected power consumption value by model and time by time estimated by the power consumption estimation unit by time.

  The load prediction method for a power distribution system according to the present invention predicts a power load of a power distribution system that has a plurality of power distribution sections and supplies power to a load device, and includes the following steps. The ownership model for each type of load device stored for each distribution section in the holding ratio model storage unit for each distribution section, the number of houses related information for each first distribution section regarding one distribution section, Based on the above, the number-of-devices estimation unit estimates the number of models for each first housing information category for one distribution section. Based on the number of models estimated for each first housing information category by the number-of-devices estimation unit and the operation probability model stored for each second housing information category for each distribution section in the operation probability model storage unit, The operating unit predicting unit predicts the operating unit expected value for each model and time. Based on the expected number of operating units for each model and time predicted by the operating unit prediction unit, and the power consumption model for each operating load device stored for each model in the power consumption model storage unit The power consumption estimation unit estimates the expected power consumption value by model and time. By summing the expected power consumption values for each model and time estimated by the power consumption estimation unit for each time, the load summation unit calculates the power load for each time.

  The “first housing information category” and the “second housing information category” in the above may be common categories.

  According to the present invention, the total power consumption of the load equipment of the distribution system is predicted in consideration of the dependence on the model of the load equipment and the housing information category for any distribution section. Thereby, the power load in an arbitrary distribution section of the distribution system can be accurately predicted.

It is a block diagram which shows roughly the structure of the load prediction apparatus in Embodiment 1 of this invention. It is a figure which shows an example of the model memorize | stored in the possession rate model memory | storage part of FIG. It is a graph which shows an example of the model memorize | stored in the possession rate model memory | storage part of FIG. It is a graph which shows an example of the model memorize | stored in the operation probability model memory | storage part of FIG. FIG. 5 is a graph showing an example of an operation probability distribution at time t ₁ of the operation probability in FIG. 4. It is a graph which shows an example of the correction function applied to the model memorize | stored in the operation probability model memory | storage part of FIG. It is a graph which shows an example of the questionnaire survey result of the utilization factor of the heating utilization of an air conditioner. It is a graph which shows an example of the model memorize | stored in the operation probability model memory | storage part of FIG. It is a graph which shows an example of the model memorize | stored in the power consumption model memory | storage part of FIG. It is a graph which shows an example of the model using temperature as a parameter memorize | stored in the power consumption model memory | storage part of FIG. It is a graph which shows an example of the calculation result performed in the power consumption estimation part of FIG. It is a graph which shows an example of the calculation result performed in the load total part of FIG. It is a graph which shows an example of the calculation result performed in the load total part of FIG. It is a flowchart which shows schematically the structure of the load prediction method in Embodiment 1 of this invention. It is a block diagram which shows roughly the structure of the load prediction apparatus in Embodiment 2 of this invention. It is a flowchart which shows schematically the structure of the load prediction method in Embodiment 2 of this invention. It is a block diagram which shows roughly the structure of the load prediction apparatus in Embodiment 3 of this invention. It is a block diagram which shows roughly the structure of the load prediction apparatus in Embodiment 4 of this invention. It is a block diagram which shows roughly the structure of the load prediction system in the modification of Embodiment 4 of this invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following drawings, the same or corresponding parts are denoted by the same reference numerals, and description thereof will not be repeated.

<Embodiment 1>
(About overall structure)
FIG. 1 schematically shows a configuration of a load prediction apparatus 901 in the present embodiment. The load prediction device 901 predicts a power load of a distribution system that has a plurality of distribution sections and supplies power to a load device. As will be described later, the power load can be predicted for any distribution section in the distribution system. Load devices to which power is supplied include, for example, IH (Induction Heating) cooking heaters, dishwashers, water heaters, living outlets, and air conditioners. In the following, these models may be collectively referred to as “model n”, and may be represented by “n” in mathematical expressions. In addition, a load device that receives power supply from an arbitrary same distribution section may be referred to as “same distribution section device”.

  The load prediction device 901 includes an information storage unit 101, a control unit 201, an input unit 300, and an output unit 400. The information storage unit 101 includes an ownership model storage unit 110, an operation probability model storage unit 120, a power consumption model storage unit 130, and a house number related information storage unit 170. The control unit 201 includes a device number estimation unit 210, an operating number prediction unit 220, a power consumption estimation unit 230, and a load total unit 240.

  The information storage unit 101 is composed of, for example, at least one of a RAM, a ROM, and a hard disk. The control unit 201 is composed of, for example, a CPU. The input unit 300 is composed of, for example, a keyboard. The output unit 400 is composed of a display, for example. Such a hardware configuration is the same in other embodiments.

(Regarding the number of housing related information storage unit 170)
The number-of-housing-related information storage unit 170 is for storing the number-of-housing-related information for each housing information category in the distribution section to be predicted. This information may be stored in the information storage unit 101 when the load prediction device 901 is used, or may be stored in the information storage unit 101 in advance. In the former case, the information may be received by the input unit 300, for example, manually input by the user. In the latter case, information may be stored for a plurality of power distribution intervals. In this case, when the load prediction device 901 is used, the input unit 300 may accept designation of a power distribution section to be predicted.

  The number-of-homes-related information for each home information section in the power distribution section is typically information on the number of houses for each home information section, or information on the total number of houses and the ratio for each home information section. This information may be manually input by the user using the input unit 300 based on a questionnaire, a statistical survey such as a population ratio of the city and a nationwide consumption survey, or distribution sales information, a distribution distribution system, etc. It may be automatically input in cooperation with other systems. Further, instead of the current value or the predicted value, a future estimated value, an estimated value before the new town construction, or the like may be input. In this case, long-term simulation and prediction before the construction of a new town are possible. “Housing information classification” will be described later.

(About the ownership model storage unit 110)
Referring to FIG. 2, the ownership ratio model storage unit 110 is classified by load device type n for each housing information category k (first housing information category) for each distribution section (for example, each of the distribution sections A to C). It stores the ownership model. “Housing information classification” can classify the houses in the distribution section. A set of housing information sections (for example, housing information section k) has exactly one section corresponding to each house when all houses in a specific distribution section (for example, distribution section A) are classified thereby. Configured as follows.

  In the example shown in FIG. 2, the area, the detached house or the apartment house, the power contract type, and the building age are shown as the housing information category k. For example, the housing information category k = “detached” and “all electrified”, “detached” and “conventional light A”, “collection” and “all electrified”, and so on are possible. The “region” column may be further subdivided, and may be expressed in units of substations, for example. The column “electric power contract” indicates the type of contract between the electric power company and the house, and includes, for example, “conventional lamp A”, “by time zone”, and “all electrification”.

  Note that the ownership model is not limited to the table as shown in FIG. 2. For example, as shown in FIG. 3, a function (formula) in which the ownership and the housing information category (age in FIG. 3) are associated with each other. ) May be applied. In addition, the functions (formulas) and tables applied to the ownership rate model are manually input by the user based on a nationwide consumption survey or a questionnaire survey and stored in the above-described ownership rate model storage unit 110. Alternatively, it may be automatically generated by another system.

  The ownership rate model storage unit 110 can represent the ownership rate of the model n for each house information category k. Solar power generation, water heaters, heating equipment (air conditioners, heat storage heaters / floor heating), and cooking appliances, etc., by type n are cold, whether it is a detached house, an apartment house, or an all-electric house. There is a large correlation with the housing information category such as whether the area is warm or warm. Since the ownership model storage unit 110 stores the ownership model for each house information category k, it is possible to appropriately represent the ownership in the distribution section.

  As with the number-of-homes related information, a model assumed in the future may be used in place of the current model. In this case, for example, it is possible to perform a prediction simulation of a power load when a new device is widely used.

(About the device number estimation unit 210)
The number-of-devices estimation unit 210 includes the ownership rate model stored in the ownership rate model storage unit 110 for each model n and for each home information category k, and the housing information category k for one distribution section (for example, distribution section A in FIG. 2). Based on the information related to the number of houses for each, the number of models for each house information section k is estimated for one distribution section. Specifically, the number-of-devices estimation unit 210 (FIG. 1) is configured to store the house number related information stored in the house number related information storage unit 170 for each house information category k in the distribution section and the house information category k for each model n. Based on the ownership rate model stored in the ownership rate model storage unit 110 every time, the number N _{n, k} of devices for each model n and for each housing information category _k is calculated for the same distribution section device.

N _{n, k} = number of houses in the housing information category k in the distribution section A
× Ownership ratio of model n in housing information category k

For example, when k = “conventional light A” and n = “solar power generation device”, the number of devices N _{n, k} is as follows when the number-of-housing-related information includes information on the number of houses for each housing information category. It can be calculated by a formula.

    Number of houses with “conventional light A” contract in distribution section A × solar power generation ownership rate with “conventional light A” contract

  Moreover, when the number-of-homes related information includes information on the housing ratio for each housing information category, it can be calculated by the following equation.

    Total number of houses in distribution section A × Ratio of houses in “Conventional Light A” contract in distribution section A × Solar power generation ownership ratio in “Conventional Light A” contract

(About the operation probability model storage unit 120)
The operation probability model storage unit 120 (FIG. 1) stores an operation probability model for each type of load device and for each time for each house information section k (second house information section common to the first house information section) for each distribution section. It is something to remember. The operation probability model is a model indicating the operation probability (operation ratio) of the load device.

FIG. 4 shows an example of an operation probability model stored in the operation probability model storage unit 120. In this example, as an operation probability model, a function f _{n, k} (t) _, which is a probability formula indicating the probability (percentage) (%) that the model n is activated (operated) at time t, for each house information category k. Has been applied. For the function f _{n, k} (t), for example, an approximate expression such as an mth order expression is used.

  Note that the operation probability model storage unit 120 is not limited to the function (expression) shown in FIG. 4, but includes, for example, a table (not shown) in which operation probabilities (operation ratios) and times are associated with each other. May be applied. The function (formula) and table applied to the operation probability model storage unit 120 may be manually input by the user and stored in the above-described storage unit, or may be automatically generated by another system. There may be.

  With the operation probability model storage unit 120 as described above, it is possible to express the operation probability (operation ratio) of different load devices depending on the life pattern by type and time, such as the operation probability of the IH cooking heater and the dishwasher. Become. And if it is often not operated during the daytime on weekdays, lower the daytime operation probability of the operation probability model storage unit 120, or if the spread of rate plan setting and equipment with a timer (dishwasher) increases It is possible to perform a simulation that appropriately changes the operation probability for each model and time, such as shifting the operation probability peak of the operation probability model to midnight.

A graphical user interface (GUI) may be used for at least one of model change and input. For example, even if the function (expression) or table content that is currently input is displayed as a graph on the output unit 400 and the operation probability is partially changed by a user drag-and-drop operation received by the mouse of the input unit 300, etc. Good. At that time, when a specific one time is selected and the numerical value is moved, interpolation is performed so that the numerical values around the time are automatically adjusted so that the smoothness of the graph is maintained. Also good. The interpolation may be performed so that the graph smoothly changes in units of unevenness by using the convex and concave portions of the graph. For example, if the point at time t ₁ is selected and the value is slightly lowered, the peak of the mountain (t ₁ , f _{n, k} (t ₁ )) will be lowered, and the peak of the mountain will not be dented. no, f1 _n, as in _{k (t) (t 1,} f n, k (t 1)) the height of peaks whose apex falls entirely smooth the. Further, an example in which the peak time that is the peak of the mountain in the evening time zone is selected and shifted to a later time is shown in f2 _{n, k} (t). In addition to selecting and operating a vertex that is convex above the graph, the vertex of the convex portion may be selected and operated below the graph. You can also specify the whole graph or part and move it up or down in parallel, or increase or decrease x% from the original value at each time. It may be increased or decreased. At this time, correction is performed so that the deformation is less than 100% when the deformation exceeds 100%, and is 0% or more when the deformation is less than 0%. In the above, an example of changing the operation probability model by changing the convex or concave graph of the graph by mouse drag and drop operation has been shown, but instead of the mouse drag and drop operation, the portion of the graph to be changed is selected (selected) In this case, the graph may be automatically divided into areas with unevenness so that it can be selected), vertex addition (vertex time and value specified), peak value change (peak value specified), peak time Menus such as change (specify time), move up and down in parallel (specify parallel movement value), and increase / decrease x% from the original value at each time may be selected. With the GUI, the probability model can be easily and intuitively changed. In particular, the simulation considering the peak becomes easy by performing the deformation considering the convex or concave portion of the graph.

The operation probability model in FIG. 4 is an average value of the probability that the model n is operating at each time, but the operation probability model has a probability distribution instead of a single value at each time like the average value. It may be. FIG. 5 shows an example in which the probability distribution of the activation probability f _{n, k} (t ₁ ) at time t ₁ of the model n is a normal distribution with a standard deviation σ _n . When the operation probability model storage unit 120 has the operation probability model in such a probability distribution, for example, 2 × σ _n is estimated as an error (predetermined error), and the operation probability is calculated from the average value by the error (2 × σ _n ). It is possible to change only in the increasing direction (see the two-dot chain line graph f _{n, k} (t) + 2σn in FIG. 4). The predetermined error may be received from the user by the input unit 300.

In the above description, the error (predetermined error) in the operation probability is 2 × σ _n and is set for each model n, but the time is constant. However, the activation probability error (predetermined error) is not limited to this, and may be set for each time t and model n, for example, σ _n (t). Further, the error (predetermined error) in the activation probability of the model n may be set separately for an error in the increasing direction (for example, standard deviation σ _{n +} ) and an error in the decreasing direction (for example, standard deviation σ _n− ). Further, the error (predetermined error) in the activation probability of the model n may be a function (formula) format or a table format. The function (formula) or table applied to the error (predetermined error) may be manually input by the user and stored in the above-described storage unit, or may be automatically generated by another system. May be.

The operation probability model stored in the operation probability model storage unit 120 may store different models for each temperature and season. Specifically, the correction model may be further stored in the operation probability model storage unit 120. For example, the operation probability model is corrected by multiplying the operation probability f _{n, k} (t) of FIG. 4 by a correction function C1 _n (T) indicating a correction ratio using the temperature T as a parameter as shown in FIG. The That is, the operation probability f _{n, k} (t, T) is obtained as follows.

f _{n, k} (t, T) = f _{n, k} (t) × C1 _n (T)
However, when f _{n, k} (t, T)> 100%, it is considered that f _{n, k} (t, T) = 100%. The same applies to the following.

Here, the correction function C1 _n (T) for correcting the operation probability model (function f _{n, k} (t)) by multiplication has been described as an example of the correction function, but the correction function is not limited to this. . As shown in the following equation, a correction function C2 _n (T) that corrects the activation probability model (function f _{n, k} (t)) by addition may be applied.

f _{n, k} (t, T) = f _{n, k} (t) + C2 _n (T)
However, when f _{n, k} (t, T) <0%, it is considered that f _{n, k} (t, T) = 0%. The same applies to the following.

In the above, the configuration in which the function C1 _n (T) or C2 _n (T) with the temperature T as a parameter is applied as the correction function has been described, but the correction function is not limited to this. For example, a correction function C1 _n (t, T) or C2 _n (t, T) having a dependency on time t may be applied.

f _{n, k} (t, T) = f _{n, k} (t) × C1 _n (t, T)
f _{n, k} (t, T) = f _{n, k} (t) + C2 _n (t, T)

Further, the operation probability model and the correction function are not limited to those defined by the function (formula) as described above. For example, as an operation probability model, a table in which operation probabilities are associated with times may be applied, and a table in which correction values are associated with temperatures (or correction values, temperatures, and times are associated with each other). Table) may be adapted. Further, the operation probability model may be represented by an operation probability and a correction function, or the function and table of the activation probability and the correction function may be calculated and expanded to be represented in the form of f _{n, k} (t, T). .

  The function or table applied to the correction model may be manually input by the user and stored in the storage unit described above, or may be automatically generated by another system.

  As described above, since the operation probability model for each model n is stored for each house information category k, more accurate modeling of the operation probability is possible. Such an example will be described below.

  For example, the operating probability of heating equipment varies greatly depending on areas such as Tohoku and Kanto, whether or not all-electric homes are used. FIG. 7 shows the results of a questionnaire survey of the air-conditioner usage rate for heating. In cold areas, heat storage heaters and oil fan heaters are mainly used as heating equipment, so the operating probability of air conditioners is lower than in warm areas. Moreover, in the all electrified house, since the ratio of using an air conditioner as a heating device is higher than in a house that is not all electrified, the operation probability of the air conditioner tends to increase. Such a tendency can be appropriately handled by the operation probability model storage unit 120 storing an operation probability model for each region and each housing information category such as whether or not it is an all-electric housing.

  In addition, the family composition including senior generations (such as retired workers) has a higher day-to-day home rate on weekdays than the family structure does not include senior generations. In a house with a high home stay rate during the daytime on weekdays, the equipment operation rate during the daytime on weekdays tends to be high. Therefore, the function of the operation probability model of the IH cooking heater shown in FIG. 8 uses not only a parameter of the time t but also a weekday (Monday to Friday) or a holiday (Saturday, Sunday Congratulations) parameter.

(About the power consumption model storage unit 130)
The power consumption model storage unit 130 (FIG. 1) stores a power consumption model for each active load device for each model.

FIG. 9 shows an example of a device power characteristic model stored in the power consumption model storage unit 130. The device power characteristic model is a model that indicates the power consumption consumed by one load device during startup (in operation) for each model. In this example, as a power consumption model, an example is shown in which a probability distribution of power consumption consumed by one model n during startup (operation) is a normal distribution with an average P _n and a standard deviation σ _n . .

  The power consumption model is not limited to the normal distribution, but may be another probability distribution, for example, an approximate expression such as an mth order expression. Further, the functions (expressions) are not limited to those shown in FIG. 9, for example, average power consumption, maximum / minimum values, first quartile, third quartile, etc. (for example, arithmetic mean, center) A table (not shown) in which values and modes are associated with times may be applied. The function (formula) and table applied to the power consumption model may be manually input by the user and stored, or may be automatically generated by another system.

  With the appliance power characteristic model as described above, it is possible to express different power consumption characteristics depending on the model n such as an IH cooking heater or a dishwasher. In the future, when energy-saving equipment is developed and the power consumption of model n changes, the simulation of changing the average power consumption of each model as appropriate, such as changing the power consumption of model n of the equipment power characteristic model It becomes possible to carry out.

The device power characteristic model stored in the power consumption model storage unit 130 may be a model that varies depending on the temperature or season. FIG. 10 shows an example of a power consumption model in which the average power consumption P _n of the model n is represented by a function P _n (T) with the temperature T as a parameter. The power consumption model as shown in FIG. 10 is obtained, for example, by multiplying or adding the function (expression) shown in FIG.

  Note that the power consumption model is not limited to the function (expression) as shown in FIG. 10. For example, a table (not shown) in which the average power consumption of one model n and the temperature T is associated with each other. May be applied. In addition, functions (formulas) and tables applied to the power consumption model may be manually input by the user and stored in the storage unit described above, or may be automatically generated by another system. Good.

If the power consumption model stored in the power consumption model storage unit 130 is used, it is possible to obtain an average value P _{n of the} power consumption of the model n. In addition, for example, σ _n can be estimated as an error (predetermined error), and (P _n + σ _n ) can be used instead of the average power consumption P _n of the power consumption model. It is also possible to use a scale point.

In the above description, the power consumption error (predetermined error) is σ _n and is set for each model n, but is constant with respect to the temperature. However, the power consumption error (predetermined error) is not limited to this, and may be set for each temperature T and model n, for example, σ _n (T). In FIG. 10, the average power consumption P _n of the power consumption model is indicated by a solid line, and the average power consumption P _n changed by an error (σ _n (T)) is indicated by a broken line.

Further, the error (predetermined error) of the power consumption P _n of the model n may be set separately for an error in the increasing direction (for example, standard deviation σ _{n +} ) and an error in the decreasing direction (for example, standard deviation σ _n− ). . Further, the error (predetermined error) of the power consumption P _n of the model n may be in the form of a function (formula) or in the form of a table. Further, the function (formula) or table applied to the error (predetermined error) may be manually input by the user and stored in the storage unit described above, or the operation measured according to the model type and the temperature. The controller described above may be automatically generated based on power measurement data.

(About the operating unit prediction unit 220)
The number-of-operating-unit predicting unit 220 (FIG. 1) stores the number of units by type estimated by the number-of-devices estimating unit 210 for each home information category (first information category) and the housing information category (first) in the operation probability model storage unit 120. And the operation probability model stored for each information category and the second housing information category common to each other), the expected number of operating units by type and by time is predicted. Specifically, the number-of-operating-unit predicting unit 220 accumulates the products for each common house information category of the number of models and the operation probability model. This process will be described below.

The number-of-operating-unit prediction unit 220 adds the above-described number-of-devices estimation unit 210 (FIG. 1) to the operation probability model (f _{n, k} (t)) stored in the operation probability model storage unit 120 for each model n and for each house information category k. Multiply by the number of devices N _{n, k} for each housing information category k and model n of the same distribution section device estimated in), and further accumulate for the housing information category k, by type and time of the same distribution section device The expected value g _n (t) of the number of startups is calculated. For example, the operating number prediction unit 220 calculates the expected value g _n (t) of the number of activated units for each time t and model n of the same distribution section device by the following equation.

Note that unlike the second embodiment described later, in this embodiment, the housing number classification storage unit 110 uses the housing information classification and the operation probability model storage unit 120 stores the device number estimation unit 210 to estimate the number of devices. The house information category k is common to the house information category of the operation probability model to be used (for example, other than k⊇ all-electric house and all-electric house). Thus, when the housing information classification of the ownership ratio model storage unit 110 and the housing information classification of the operation probability model stored in the operation probability model storage unit 120 are common, g _n (t) is _{expressed by the} above formula. Can be calculated.

(About the power consumption estimation unit 230)
The power consumption estimation unit 230 (FIG. 1) stores the expected number of operating units g _n (t) for each model and time predicted by the operating unit prediction unit 220 and the model n in the power consumption model storage unit 130. Based on the power consumption model per unit, the expected power consumption value h _n (t) for each model n and time t of the same distribution section equipment is estimated. Specifically, the following equation is calculated.

h _n (t) = g _n (t) × power consumption of model n obtained from power consumption model FIG. 11 shows an expected value h _n (power consumption of model n calculated by the power consumption estimation unit 230 using the above equation. An example of t) is shown. In this example, n = expected value h _n (t) of the electric water heater is shown. That is, an expected value h _n (t) of power consumption consumed at time t by all electric water heaters that receive power supply from the distribution system in the same distribution section is shown.

(About load total unit 240)
The load summation unit 240 (FIG. 1) calculates the power load by time by summing the expected value of power consumption by model and time by type estimated by the power consumption estimation unit 230 for each model for each time. It is. That is, the load summing unit 240 sums up the expected power consumption value h _n (t) for each time t (here, summing up all of the models n), thereby obtaining all the loads that receive power supply from any one of the same distribution sections. The power consumption total E (t) for each time t of the device (all devices in the same distribution section) is calculated. For example, the load summation unit 240 calculates the total power consumption E (t) at the time t of the entire power distribution section device by the following equation.

  FIG. 12 shows an example of the total power consumption E (t) calculated by the load summation unit 240 using the above equation. FIG. 13 shows a simulation in which the ratio of all-electric homes in the housing information section in the distribution section included in the distribution section information is increased by 30% from the state where the total power consumption E (t) shown in FIG. 12 is obtained. A simulation result in the case where the ratio of newly constructed houses (less than 10 years old in FIG. 2) included in the distribution section information is 50% of the original is shown.

(Summary of load prediction methods)
As described above, while describing each of the functional blocks (FIG. 1), details of the load prediction method for predicting the power load of the distribution system that supplies power to the load device are also described. To summarize, with reference to FIG. 14, the load prediction method for the distribution system predicts the power load of the distribution system that has a plurality of distribution sections and supplies power to the load device, and has the following steps. .

In step S10, the ownership ratio model for each load device type n stored for each distribution section in the ownership ratio model storage unit 110 for each distribution section and the housing information section for one distribution section (for example, distribution section A). Based on the number-of-homes-related information for each k (first housing information category), the number-of-devices estimation unit 210 estimates the number N _{n, k} for each model n for each housing information category k for one distribution section.

In step S20, the number of devices N _{n, k} for each model n estimated by the number-of-devices estimation unit 210 for each house information category _k and the operation probability model storage unit 120 store the information for each house information category k for each distribution section. Based on the operating probability model (for example, f _{n, k} (t)), the operating unit prediction unit 220 predicts the operating unit expected value g _n (t) for each model n and for each time t.

In step S30, the expected number of operating units g _n (t) for each model n and for each time t predicted by the operating unit predicting unit 220, and the operating load stored for each model in the power consumption model storage unit 130 Based on the power consumption model per device, the power consumption estimation unit 230 estimates the expected power consumption value h _n (t) for each model n and for each time t.

In step S40, the load summation unit 240 sums the expected power consumption value h _n (t) for each model n and for each time t estimated by the power consumption estimation unit 230 for each time t. Calculate the load E (t).

(Function and effect)
According to the present embodiment, based on the number-of-homes-related information in an arbitrary distribution section and the model of the ownership model storage unit 110, the number N _n of devices in the same distribution section by time and by housing information category _{, k} is required. Based on this N _{n, k} and the model (for example, f _{n, k} (t)) of the operation probability model storage unit 120, the expected value h _n (t ) Is required. With this configuration, it is possible to predict the power load caused by the load device for any distribution section, taking into account the difference in the housing information category for each model and time. Therefore, it is possible to accurately predict the power load of the distribution system.

  In addition, since each model of the ownership rate model storage unit 110 and the operation probability model storage unit 120 is set for each home information category, accurate load devices having different ownership rates or operation probabilities for each home information category are also accurate. Prediction is possible.

  Due to changes in the electricity bill menu for each time zone or improvements in the power efficiency of devices, the device penetration rate by device type, the device operating time zone, and the power consumption of the load device are expected to change in the future. Changes in the device penetration rate can be accommodated by adjusting the model stored in the possession rate model storage unit 110. Changes in the device operating time zone can be dealt with by adjusting the model of the operation probability model storage unit 120. Changes in the power consumption of the load device can be dealt with by adjusting the model of the power consumption model storage unit 130. Therefore, according to the present embodiment, it is possible to accurately predict future changes in the power load. Furthermore, since the load of the distribution section is also related to the attributes of the houses included in the distribution section, the information stored in the number-of-housing-related information storage unit 170 is changed, or the house information classification is also changed. It is possible to perform future simulations such as forecasting future power distribution sections (such as New Town) and increasing the number of elderly households.

<Embodiment 2>
(About overall structure)
FIG. 15 schematically shows a configuration of a load prediction apparatus 902 in the present embodiment. Configurations that are the same as or similar to those of the load prediction device 901 according to the first embodiment are denoted by the same reference numerals, and different configurations will be mainly described. The load prediction device 902 includes an information storage unit 102 and a control unit 202. The information storage unit 102 includes a housing number related information storage unit 170V and an operation probability model storage unit 120V. The control unit 202 includes an operating number prediction unit 220V.

(About the ownership model storage unit 110)
As in the first embodiment, the ownership rate model storage unit 110 stores the ownership rate model for each type of load device for each housing information category for each distribution section. This house information section is referred to as a first information section k1 in the present embodiment.

(About the operation probability model storage unit 120V)
The operation probability model storage unit 120 stores an operation probability model for each type of load device and for each time for each second house information section k2 for each distribution section. The second house information section k2 is different from the first information section k1 used by the ownership model storage unit 110.

(Regarding the number of housing related information storage unit 170V)
The housing number related information storage unit 170V includes storage units 171 and 172 for each of the first and second housing information sections. Each of the storage units 171 and 172 for each of the first and second house information sections is for storing the number of houses related information for each house information section k1 and k2 in the distribution section to be predicted. is there. As in the first embodiment, this information may be stored in the information storage unit 102 when the load prediction device 902 is used, or may be stored in the information storage unit 102 in advance.

(About the device number estimation unit 210)
The number-of-devices estimation unit 210 is the same as that in the first embodiment. In this embodiment, the number-of-equipment estimation unit 210 stores one ownership ratio model stored in the ownership ratio model storage unit 110 for each first house information category k1. On the basis of the number-of-homes related information for each first housing information section k1 regarding the current distribution section, the number of models for each first housing information section k1 is estimated for one power distribution section. In addition, what is memorize | stored in the memory | storage part 171 for every 1st house information division can be used as the number-of-homes related information for every 1st house information division k1.

(About the operating unit prediction unit 220V)
The number-of-operations predicting unit 220V is the number of models for each model estimated by the number-of-devices estimation unit 210 for each first house information section k1, and the operation stored in the operation probability model storage unit 120 for each second house information section k2. Using the probability model and the number-of-houses-related information for each second house information section k2 regarding one power distribution section, the expected number of operating units by model and by time is predicted. In addition, what is memorize | stored in the memory | storage part 172 for every 2nd house information division can be used as the number-of-homes related information for every 2nd house information division k2.

  For example, the first housing information section k1 = {all electrified house, non-all electrified house}, and the second house information section k2 = {senior household, non-senior household}. The number-of-devices estimation unit 210 uses the number of houses information for each first information category k1 and the model of the ownership model storage unit 110, so that the number of dishwashers in all-electric houses is X, It is estimated that the number of dishwashers in an electrified house is Y units. The operating unit prediction unit 220V refers to, for example, information that senior households = 30% and non-senior households = 70% as the housing information for each second housing information category k2. And the operation number prediction part 220V performs the following calculations.

Number of appliances in all-electric homes and senior households = X units x 30% × Operation rate of dishwashers in senior households Number of appliances in non-all-electric homes and senior households = Y units × 30% × Senior Household dishwasher utilization rate Number of appliances in all-electric homes and non-senior households = X units x 70% x Non-senior household dishwasher availability Non-all-electric homes and non-senior household dishwashers Number of devices = Y units × 70% × Non-senior household dishwasher operation rate And the operation number prediction unit 220V predicts the number of dishwasher operations by summing all four combinations. That is, the total number of operating devices is integrated for all combinations of the house information categories k1 and k2.

  In other words, the above example can be explained as follows.

A house information section k12 having all combinations of the first and second house information sections k1 and k2 is created, and the ownership model storage unit 110 stores the model using the house information section k12. The number-of-devices estimation unit 210 obtains the number of devices for each house information category k12. At this time, the possession probability model f _{n, k2} (t) (k2 = m2) is used for the number of models n belonging to the housing information category k12 (k12 = m1 × m2, k1⊇m1 and k2⊇m2), Calculate with the following formula.

(Here, N _{n, k12} is the number of models n owned by a house belonging to the house information category k12 of the same distribution section device).

  As a method of creating the housing information section k12, there is a method of obtaining by multiplying the ratio of the first and second housing information sections k1 and k2. For example, category k1 = {all electrified housing, non-all electrified housing}, category k2 = {senior household, non-senior household}, respectively, all electrified housing 20%, non-all electrified housing 80%, senior household = 30%, Non-senior household = 70%. The ratio of all-electric housing and senior households is 20% x 30% = 6%. The device ownership rate for each model n is the same as that of all-electric homes, and the number of operating devices is calculated using the operating probability model for senior households.

  The percentage of senior households as information on the number of houses is not directly available to electric power companies, but is estimated using the population ratio (such as the percentage of people 65 years of age or older) in the area where the power distribution section exists. Can do.

(Summary)
As described above, each of the functional blocks (FIG. 15) has been particularly described with respect to differences from those of the first embodiment (FIG. 1). In summary, with reference to FIG. 16, the load prediction method for the distribution system predicts the power load of the distribution system that has a plurality of distribution sections and supplies power to the load device, and includes the following steps. .

In step S10V, the ownership ratio model for each load device type n stored in the ownership ratio model storage unit 110 for each first distribution section k1 for each distribution section, and one distribution section (for example, distribution section A) Based on the number-of-homes-related information for each first housing information category k1, the device number estimation unit 210 calculates the number of devices N _{n, k1} for each model n for each first housing information category k1 for one distribution section. presume.

In step S20V, the number of devices N _{n, k1} for each model n estimated by the number-of-devices estimation unit 210 for each first housing information category k1 and the distribution probability section (first housing) in the operation probability model storage unit 120V. Based on the operation probability model (for example, f _{n, k} (t)) stored for each second house information section k2 (which is different from the information section k1), the operation number predicting unit 220V is classified by model n and time t Another operation number expectation value g _n (t) is predicted.

  Thereafter, steps S30 and S40 are performed in the same manner as in the first embodiment.

  According to the present embodiment, the ownership rate model storage unit 110 and the operation probability model storage unit 120 can use different house information categories.

<Embodiment 3>
(overall structure)
FIG. 17 schematically shows the configuration of the load prediction device 903 in the present embodiment. Configurations that are the same as or similar to those of the load prediction device 901 according to the first embodiment are denoted by the same reference numerals, and different configurations will be mainly described. The load prediction device 903 includes an information storage unit 103 and a control unit 203. The information storage unit 103 includes a house information estimation model storage unit 150 and a distribution section information storage unit 180. The control unit 203 includes a housing information estimation unit 250.

(About distribution section information storage unit 180)
The power distribution section information storage unit 180 stores power distribution section information. The distribution section information may be stored in the information storage unit 103 when the load prediction device 903 is used, or may be stored in the information storage unit 103 in advance. The distribution section information includes, for example, information that can be known by the electric power company, such as the number of houses in the distribution section and electrical contract information, and statistical information such as the population ratio of an area (such as a municipality) where the distribution section exists.

(About the housing information estimation model storage unit 150)
The house information estimation model storage unit 150 stores a model for deriving the number of houses related information to be stored in the house number related information storage unit 170 from the distribution section information stored in the distribution section information storage unit 180. is there. Information related to the number of houses to be derived may be information that the electric power company cannot directly know, such as life patterns, family composition, number of families, floor area of houses, detailed building age, roof orientation, main lighting. This is for each category using the orientation of the surface. For example, based on the rate of all-electric housing and population ratio statistics in the distribution section area, all-electric homes and senior households = 5%, all-electric homes and non-senior households = 20%, non-all-electric homes and senior households = The number of housing-related information such as 20%, non-all-electric homes and non-senior households = 55% is estimated.

(About the housing information estimation unit 250)
The housing information estimation unit generates the number of houses related information for each housing information category based on the distribution section information stored in the distribution section information storage unit 180 and the derived model stored in the house information estimation model storage unit 150. To do.

(Modification)
In the present embodiment described above, the configurations of the house information estimation model storage unit 150, the power distribution section information storage unit 180, and the house information estimation unit 250 correspond to those applied to the load prediction device 901 of the first embodiment. The above configuration may be applied to the load prediction device 902 (FIG. 15) of the second embodiment instead of the load prediction device 901. In this case, the house information estimation model storage unit 150 stores a derivation model that guides the number of houses related information for each of the first and second house information sections k1 and k2 from the distribution section information. Further, the housing information estimation unit 250 performs first and second housing information classifications based on the distribution section information stored in the distribution section information storage unit 180 and the derived model stored in the house information estimation model storage unit 150. The number-of-homes related information for each of k1 and k2 is generated.

(Function and effect)
Like the method for creating the housing information section k12 from the first and second housing information sections k1 and k2 described in the second embodiment, the all-electric housing ratio and the population ratio statistical information existing in the distribution section area By simply combining the two, it is impossible to estimate the bias that non-all electrified houses have a lower percentage of senior households than all electrified houses. By providing the house information estimation model storage unit 150 as in the present embodiment, it is possible to estimate such a bias or information related to the number of houses that the power company cannot know. For example, it is possible to estimate the proportion of homes that are home during the daytime on weekdays and those that are absent during the daytime on weekdays.

<Embodiment 4>
(overall structure)
FIG. 18 schematically shows the configuration of the load prediction apparatus 904 in the present embodiment. Configurations that are the same as or similar to those of the load prediction device 901 according to the first embodiment are denoted by the same reference numerals, and different configurations will be mainly described. The load prediction device 904 includes an information storage unit 104 and a control unit 204. The information storage unit 104 has a measurement database 160. The control unit 203 has a model generation unit 260. Preferably, model generation unit 260 includes a correlation calculation unit 261.

(About the measurement database 160)
The measurement database 160 stores power consumption data classified by model and time measured in each measurement target house, and house information classification data indicating a house information classification of each measurement target house. The measurement target house is a house used as a sample for obtaining data. The housing information classification data stored in the measurement database 160 includes, for example, the power distribution section (A to C), region, detached house / collection, power contract, age of building, and the like illustrated in FIG.

(About the model generation unit 260)
The model generation unit 260 generates a retention rate model stored in the retention rate model storage unit 110 and an operation probability model stored in the operation probability model storage unit 120 based on the measurement database 160. As illustrated in FIG. 18, the model generation unit 260 may further generate a power consumption model stored in the power consumption model storage unit 130 based on the measurement database 160. Specifically, the following processing is performed.

  First, a method for generating the ownership model stored in the ownership model storage unit 110 will be described. The model generation unit 260 extracts from the measurement database 160 data of homes to be measured that belong to the same housing information category (for example, housing information category k = housing data for four family members in an all-electric housing in the Tohoku region). Such). Using the extracted measured power data for each home information category, the device ownership of the model n in the home information category k is calculated by the following formula.

    {Number of houses that hold data of model n among houses to be measured belonging to house information category k / total number of measured houses in house information category k} × 100

Secondly, a method for generating an operation probability model stored in the operation probability model storage unit 120 will be described. First, as described above, the model generation unit 260 extracts, from the measurement database 160, data of the measurement target houses that belong to the same house information category. Measurement data of the model n in the time section Δt (= [ta, tb)) is extracted from the measured power data for each extracted house information section. The operation probability f _{n, k} (t) is calculated by the following equation.

f _{n, k} {(t _a + t _b ) / 2}
= 100 × (the number of measurement data of the extracted measurement data that is greater than or equal to the operating power threshold P _Th )
/ (Total number of extracted measurement data)

Here, the operating power threshold value P _Th (W) in the numerator on the right side of the above formula is set, for example, a minimum power (W) at the time of operation that is different for each model n. Therefore, the number of measurement data shown in the molecule corresponds to the number of measurement data of the model n in operation. Therefore, for example, when the model n continues to operate for Δt (= [ta, tb)) (that is, the operating power continues to be greater than or equal to the operating power threshold P _{Th for} Δt), the numerator measurement data And the number of measurement data in the denominator are the same, and the activation probability of the model n in the time interval Δt (= [ta, tb)) is 100 (%).

  When the measurement data includes not only operating power but also an operating state indicating whether the model n is operating, the following equation can be used.

f _{n, k} {(t _a + t _b ) / 2}
= 100 x (number of measurement data in operating state among extracted measurement data)
/ (Total number of extracted measurement data)

In the above description, an example in which the operation probability model is obtained for each time interval Δt for each model n has been described. Further, measurement data of the model n in the temperature interval ΔT (= [T _a , T _b )) is extracted, and the above method is performed. In the same manner as described above, the function f _{n, k} (ΔT) indicating the operation probability may be obtained using the temperature T as a parameter. Further, the probability distribution of the operation probability may be obtained by repeating the operation probability calculation trial twice or more, for example, several times, and the average value or the standard deviation value σ _n may be obtained from the probability distribution. As a method of repeating the calculation trial, a certain number of measurement data used for calculating the operation probability is extracted from the database while changing the data sample to calculate the activation probability, and the standard deviation is calculated from the variation in the activation probability. There are a method and a method of calculating the activation probability every day and calculating the standard deviation from the variation of the activation probability.

Third, a method for generating a power consumption model stored in the power consumption model storage unit 130 will be described. From the measurement database 160, measurement data (for example, measurement data equal to or higher than the operating power threshold P _Th (W)) of the operating model n is extracted. Then, an average value of operating power indicated by the measurement data, an approximate expression of power distribution, and the like are obtained. A power consumption model may be obtained for each home information category.

(Correlation calculation unit 261)
The correlation calculation unit 261 calculates the degree of association between each home information category included in the home information category data stored in the measurement database 160 and the power consumption data. When the correlation calculation unit 261 can be provided, the model generation unit 260 determines the home information category to be used by the model based on the calculation result of the correlation calculation unit 261. And each model mentioned above is produced | generated using the housing information classification with high relevance.

  As a method for calculating the degree of correlation, the correlation coefficient between the measurement data of the model n and the house information, the average distance between the measurement data for each house information category, or the like can be used. The degree of association may be calculated using a predetermined house information category, and it may be determined whether to integrate or divide the house information category using a threshold value. Clustering methods may be used to group measurement data based on the distance between measurement data, etc., and automatically generate optimal housing information categories. The clustering threshold is obtained from the load prediction accuracy. For example, when it is desired to perform load prediction with an accuracy of 5%, the home information classification is divided when a difference of 5% occurs. Further, for example, measurement data may be projected onto a two-dimensional space by principal component analysis or factor analysis, and clustering between the data may be performed based on the proximity of the projected data. Further, the degree of association may be calculated together with the temperature category, the time category, the weekday / holiday category, and the optimum category may be automatically determined.

(Function and effect)
According to the present embodiment, by using the measurement database 160 and the model generation unit 260, the ownership rate model, the operation probability model, and the power consumption model can be automatically generated. Further, by updating the measurement database 160, it is possible to automatically generate each model reflecting the latest data.

  Further, when the correlation calculation unit 701 is provided, it is possible to automatically extract a housing information category having a large influence on each model. For this reason, it is not necessary to store an unnecessarily large number of models in each model storage unit. Therefore, the data capacity of the information storage unit 104 is reduced. Further, it is not necessary to input unnecessarily much information at the time of load prediction or simulation. Further, it is not necessary to perform an unnecessarily many pattern calculations in the control unit 204.

(Modification 1)
Based on the power consumption data and the house information classification data stored in the measurement database 160, the model generation unit 260 may further generate the house number related information stored in the house number related information storage unit 170. For example, the following processing is performed by the model generation unit 260.

  From the measurement database 160, for example, a category such as a housing information category k = “weekday absent home” is extracted. In addition, the record to the measurement database 160 of whether each measurement target house is a “weekday daytime absence house” may be made based on a survey of the measurement target house, or the measured power data of the measurement target house may be used. It may be made based on the guess used.

  And the number-of-homes related information is calculated as follows with reference to the measurement database 160.

{The number of houses that belong to the housing information category k among the houses to be measured / the total number of houses to be measured}
× 100

  Instead of the above calculation, a calculation using only the measurement target house that satisfies the specific condition X may be performed as follows.

{Number of houses belonging to housing information category k among measurement target houses that satisfy condition X / Number of measurement target houses that satisfy condition X} × 100

  The condition X may be included in one type of housing information category, such as “all-electric home”. In addition, a model may be generated after learning the relationship between the data in the measurement database 160 and the number of houses related information as correct data using machine learning or a neural network.

  In this modification, the correlation calculation unit 261 calculates the degree of association between the measurement database 160 and the number of houses related information. Based on the calculation result of the correlation calculation unit 261, the model generation unit 260 determines the home information category to be used by the home count related information. Then, the above-described housing number related information is generated using the housing information classification having a high degree of association.

  The model generation unit 260 further generates the number-of-housing-related information stored in the number-of-housing-related information storage unit 170V while applying the number-of-housing-related information storage unit 170V (FIG. 15) instead of the number-of-housing-related information storage unit 170. May be. In this case, the number-of-homes related information for each of the first and second housing information sections is generated.

  According to this modification, the number-of-homes related information can also be automatically generated. Further, by updating the measurement database 160, it becomes possible to automatically generate the number-of-housing related information reflecting the latest data.

(Modification 2)
Referring to FIG. 19, in this modification, a system having a load prediction device is used instead of the load prediction device alone. Specifically, the load prediction system 904S includes a load prediction device 904D and a database server 990 connected thereto via a network 999. The load prediction device 904D includes a communication unit 290 for connecting to the network 999. Unlike the load prediction device 904 (FIG. 18), the load prediction device 904D does not have the measurement database 160. Instead, the database server 990 has a measurement database 160. According to this modification, the load prediction apparatus itself does not need to have the measurement database 160 when performing load prediction.

  It should be noted that the present invention can be freely combined with each other within the scope of the invention, and each embodiment can be appropriately modified or omitted.

  101-104 Information storage unit, 110 Ownership model storage unit, 120, 120V Operation probability model storage unit, 130 Power consumption model storage unit, 150 Housing information estimation model storage unit, 160 Measurement database, 170, 170V Housing number related information storage , 171, 172 Storage unit for each first and second housing information section, 180 Power distribution section information storage unit, 201-204 Control unit, 210 Number of devices estimation unit, 220, 220V Operation number prediction unit, 230 Power consumption estimation Unit, 240 load total unit, 250 housing information estimation unit, 260 model generation unit, 261 correlation calculation unit, 290 communication unit, 300 input unit, 400 output unit, 701 correlation calculation unit, 901-904, 904D load prediction device, 904S Load prediction system, 990 database server, 999 network Click.

Claims (8)

A load prediction device for a distribution system that predicts the power load of a distribution system that has a plurality of distribution sections and supplies power to load equipment,
A holding ratio model storage unit that stores a holding ratio model for each type of the load device for each first housing information section for each distribution section;
An operation probability model storage unit for storing an operation probability model for each type of load device and for each time for each second housing information section for each distribution section;
A power consumption model storage unit that stores a power consumption model per unit of the load device in operation, for each model;
Based on the ownership model stored for each of the first housing information sections in the ownership model storage unit and the number-of-homes related information for each of the first housing information sections regarding one power distribution section, A number-of-equipment estimation unit that estimates the number of models for each of the first housing information categories for one distribution section;
The number of models by model estimated by the number-of-devices estimation unit for each first home information category, and the operation probability model stored for each second home information category in the operation probability model storage unit, Using the number of operating units prediction unit that predicts the expected number of operating units by model and time,
Based on the expected number of operating units for each model and time predicted by the operating unit prediction unit and the power consumption model per unit stored in the power consumption model storage unit for each model, And the power consumption estimation part which estimates the power consumption expectation value according to time,
A load prediction device for a distribution system, comprising: a load summation unit that calculates a power load for each time by summing the expected power consumption values for each model and time estimated by the power consumption estimation unit for each time. The first and second housing information sections are common housing information sections,
The operating number prediction unit integrates the product for each common house information category of the number of models by type and the operating probability model,
The load prediction apparatus for a power distribution system according to claim 1. The first and second housing information sections are different housing information sections,
The operating number prediction unit uses the number-of-housings related information for each of the second housing information sections related to the one distribution section,
The load prediction apparatus for a power distribution system according to claim 1. A housing information estimation model storage unit for storing a derivation model for deriving housing number related information for each of the first and second housing information sections from distribution section information;
A housing information estimation unit that generates the number-of-homes related information for each of the first and second housing information sections based on the distribution section information and the derived model stored in the housing information estimation model storage unit; The load prediction device for a distribution system according to any one of claims 1 to 3, further comprising: Based on a measurement database that stores power consumption data by type and time measured in each measurement target house, and housing information classification data indicating a housing information classification that each measurement target house has, the possession 5. The power distribution according to claim 1, further comprising a model generation unit configured to generate an ownership rate model stored in the rate model storage unit and an operation probability model stored in the operation probability model storage unit. System load prediction device. The model generation unit includes a correlation calculation unit that calculates the degree of association between each home information section and the power consumption data included in the home information section data stored in the measurement database.
The load prediction device for a power distribution system according to claim 5. The model generation unit generates the number-of-homes related information for each of the first and second home information sections based on the power consumption data and the home information section data stored in the measurement database.
The load prediction apparatus for a distribution system according to claim 5 or 6. A load prediction method for a distribution system that predicts the power load of a distribution system that has a plurality of distribution sections and supplies power to load equipment,
The ownership model for each type of load device stored for each distribution section in the ownership ratio model storage unit for each distribution section, and the number-of-homes related information for each first distribution section regarding one distribution section And, based on the above, a step in which the number-of-devices estimation unit estimates the number of models for each of the first housing information sections with respect to the one power distribution section;
The number of each model estimated for each of the first house information sections by the number-of-devices estimation section, and the operation probability model stored for each second house information section for each distribution section in the operation probability model storage section, Based on the step, the operation number prediction unit predicts the expected number of operation units by model and time, and
The expected number of operating units by type and time predicted by the operating unit prediction unit, and the power consumption model per unit of the operating load device stored by model in the power consumption model storage unit Based on the step, the power consumption estimation unit estimates the power consumption expected value by model and time,
A load summation unit calculating a power load for each time by summing the expected power consumption for each model and time estimated by the power consumption estimation unit for each time Method.

Images