JP2011214794A - Air conditioning system control device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain an air conditioning system control device capable of automatically preparing a model applicable to each building by learning a physical parameter of a versatile building model.SOLUTION: The air conditioning system control device includes: an air conditioner operation data acquisition part 5 acquiring air conditioner operation data from outside; a weather data acquisition part 4 acquiring weather data; a parameter learning part 10 having the versatile building model 10a based on a heat conduction equation and determining the physical parameter in the building model 10a by learning; a thermal load prediction part 13 predicting a thermal load based on the physical parameter and the building model 10a; a schedule preparation part 14 determining an operation schedule 15 of each air conditioner 2 based on the predicted thermal load; and an operation schedule output part 6 transmitting the operation schedule 15 to each air conditioner 2.

Description

  The present invention has an air conditioning load prediction function for accurately predicting the load of an air conditioner installed in a building such as a building so that the heat output of a heat source device can be effectively used corresponding to the air conditioning load. The present invention relates to an air conditioning system control apparatus that plans and controls an operation schedule of an air conditioner.

In recent years, demands for energy saving of various air conditioners constituting an air conditioning system such as a building have increased, and in order to satisfy these demands, many air conditioning system control devices with reduced power of the air conditioners have been proposed.
Conventionally, in many air conditioning system control devices, a control method for changing the operating state of an air conditioning system in accordance with an air conditioning load has been proposed as follows.

  First, as a first prior art, an air conditioner to be controlled and past performance data about the surrounding environment are used as input information, and an air conditioner load of the air conditioner is used as an output layer, and an air conditioning load model is constructed by a neural network. A method of performing air conditioning load prediction and heat source operation control for energy saving has been proposed (see, for example, Patent Documents 1 and 2).

  In addition, as a second conventional technique, a detailed building model (hereinafter simply referred to as “building model”) based on the building construction specifications (wall members, thickness, physical property values, etc.), or the amount of solar radiation and the outside temperature Calculate the air conditioning load by a simple formula that multiplies the measurement data by a coefficient, and plan the start time of the multi air conditioner based on whether the day before the operation day is a holiday or the number of days off. A preheat dispersion control system has also been proposed (see, for example, Patent Document 3).

However, in the control technology based on the simplified formulas of Patent Documents 1, 2, and 3, there is a physical phenomenon that cannot be reproduced because the physical formula is not considered.
For example, the effects of solar radiation on air conditioning load include components that directly enter the room through windows or the like and become air conditioning loads, and components that are once absorbed by the outer wall and become air conditioning loads as through-flow heat. It cannot be reproduced without a physical formula.
On the other hand, in the control technique based on the detailed building model based on the building specification without using the neural network model or the simple formula described in the above-mentioned Patent Document 3, a huge amount of setting parameters necessary for air conditioning load calculation are obtained for each building. Difficult to do.

Japanese Patent No. 2983159 JP-A-6-147598 Japanese Patent Application No. 2006-29694

  Conventional air-conditioning system control devices use a neural network model based on input / output data described in Patent Documents 1 and 2, and a simple calculation formula described in Patent Document 3, both of which take physical formulas into consideration. Since there is a physical phenomenon that cannot be reproduced, there is a problem that the prediction accuracy of the heat load deteriorates.

  On the other hand, it is conceivable to solve the above problem using a detailed building model that conforms to the physical formula without using the above neural network model or simple formula, but enormous setting parameters are required for each target building. However, it is practically difficult to obtain the required setting parameters, and there is a problem that it cannot be put into practical use.

  The present invention has been made to solve the above-described problems, and realizes a building model that predicts an air conditioning load according to a physical formula without setting a huge amount of parameters necessary for the building model. The air conditioning load predicted by the building model is used as an input variable to determine the operating state of the air conditioning system that minimizes the total required power of the air conditioning system that constitutes the air conditioning system, and each air conditioner is determined according to the determined target value. The purpose is to obtain an air-conditioning system control device that controls and efficiently air-conditions multiple air-conditioning spaces and realizes energy saving.

  An air conditioning system control device according to the present invention is an air conditioning system control device that controls an air conditioner installed in a building, and a weather data acquisition unit that performs data communication with a weather data distribution company via a communication means; Air conditioner operation data acquisition unit that performs data communication with air conditioners via communication means, data storage unit that stores data acquired by the weather data acquisition unit and air conditioner operation data acquisition unit, and building specifications Owns a building model based on it, obtains predetermined learning input data from the data storage unit, obtains a thermal characteristic parameter by learning based on the building model, and acquires predetermined prediction input data from the data storage unit A thermal load prediction unit for predicting the thermal load of the building based on the thermal characteristic parameters, the building model, and predetermined input data for prediction; Based on the heat load obtained by the prediction unit and predetermined input data for prediction, a schedule creation unit for planning an air conditioner operation schedule for energy saving, and an operation schedule output for transmitting the operation schedule obtained by the schedule creation unit to the air conditioner The data storage unit stores the thermal characteristic parameters and the operation schedule.

  According to the present invention, since the general building model and the parameter learning unit that learns and calculates the physical parameters can be calculated, the thermal characteristic parameters of the individual buildings can be calculated. An operation schedule can be made and an energy saving effect can be realized in each building.

It is a block diagram which shows the air conditioning system control apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows the calculation process of the air conditioning system control apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows typically the building model of the air-conditioning system control apparatus which concerns on Embodiment 1 of this invention. It is a block diagram which shows the air conditioning system control apparatus which concerns on Embodiment 2 of this invention. It is explanatory drawing which shows the relationship between the learning period and prediction accuracy in the air conditioning system control apparatus which concerns on Embodiment 2 of this invention. It is explanatory drawing which shows the relationship between the learning period in the air-conditioning system control apparatus which concerns on Embodiment 2 of this invention, and the estimated thermal characteristic parameter. It is a block diagram which shows the air conditioning system control apparatus which concerns on Embodiment 3 of this invention.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing an air conditioning system control apparatus according to Embodiment 1 of the present invention.
In FIG. 1, the air conditioning system control device 3 is configured to be capable of data communication with a weather data distribution company 1 and an air conditioner 2 via a communication means 100 such as the Internet. Based on the input information from the air conditioner 2 and the air conditioner characteristic data 8d (for example, input by the user), the air conditioner 2 installed in a building (not shown) to be controlled is controlled.

  The air conditioning system control device 3 includes a weather data acquisition unit 4 connected to the weather data distribution company 1, an air conditioner operation data acquisition unit 5 and an operation schedule output unit 6 connected to the air conditioner 2, and a data storage unit 8. The parameter learning unit 10 having the building model 10a, the thermal load prediction unit 13 that predicts the thermal load of the building, and the schedule creation unit 14 that plans the operation schedule 15 of the air conditioner 2 are provided.

  The data storage unit 8 stores the data acquired through the weather data acquisition unit 4 and the air conditioner operation data acquisition unit 5, inputs the learning input data 9 to the parameter learning unit 10, and stores the prediction input data 12. Input to the thermal load prediction unit 13.

The data storage unit 8 stores the thermal characteristic parameters 11 calculated by the parameter learning unit 10 and the operation schedule 15 calculated by the schedule creation unit 14 as calculation results.
The operation schedule output unit 6 transmits the operation schedule 15 stored in the data storage unit 8 to the air conditioner 2 via the communication unit 100.

  The parameter learning unit 10 acquires predetermined learning input data 9 from the data storage unit 8, and based on a general-purpose building model 10a based on the heat conduction equation, the learning characteristic calculates the thermal characteristic parameter 11 (physical characteristics in the building model 10a). Parameter).

  The thermal load prediction unit 13 is based on the building model 10 a in the parameter learning unit 10, the thermal characteristic parameter 11 calculated by the parameter learning unit 10, and predetermined prediction input data 12 stored in the data storage unit 8. Predict building heat load.

The schedule creation unit 14 devises an operation schedule 15 of the air conditioner 2 that saves energy in consideration of the thermal characteristics, based on the thermal load obtained by the thermal load prediction unit 13 and the prediction input data 12, and a data storage unit 8 is stored.
The operation schedule output unit 6 reads the operation schedule 15 obtained by the schedule creation unit 14 from the data storage unit 8 and transmits it to the air conditioner 2.

  In addition, in FIG. 1, although the one air conditioner 2 is shown typically, the several air conditioner 2 is accept | permitted. The air conditioner 2 may be a large heat source machine such as a building multi-air conditioner, a packaged air conditioner, a room air conditioner, or an absorption chiller composed of an outdoor unit and an indoor unit.

Next, the operation according to the first embodiment of the present invention shown in FIG. 1 will be described with reference to FIGS.
FIG. 2 is a flowchart showing a specific calculation process by the air conditioning system control device 3 of FIG.
Moreover, FIG. 3 is explanatory drawing which shows the building model 10a which the air-conditioning system control apparatus 3 has typically, and has shown the example of each factor considered with the building model 10a.

In FIG. 2, the air conditioning system control device 3 first acquires and stores each data through the weather data acquisition unit 4 and the air conditioner operation data acquisition unit 5 (step S21).
That is, the air conditioning system control device 3 acquires meteorological data at a predetermined point from the weather data distribution company 1 through the weather data acquisition unit 4, and the air conditioning system 2 of the air conditioner 2 through the air conditioner operation data acquisition unit 5. The operation data is acquired and each data is stored in the data storage unit 8.

  The weather data at the predetermined point includes at least the amount of solar radiation and the outside temperature, and includes not only past data but also future forecast values. In addition, when the amount of solar radiation cannot be obtained with a future forecast value, it may be weather forecast information such as sunny or cloudy, and typical solar radiation amount data may be corrected based on the weather forecast.

The operation data of the air conditioner 2 includes at least the amount of heat supplied (or removed) by each air conditioner 2 to the room and the room temperature of the room where the air conditioner 2 is installed. If the amount of heat supplied from the air conditioner 2 cannot be obtained, other parameters necessary for calculating the amount of heat (for example, data that can be acquired by existing sensors such as the operating frequency, evaporation temperature, and condensation temperature of each outdoor unit) It may be. Moreover, you may substitute room temperature with the suction temperature of an indoor unit.
Furthermore, the data acquisition process (step S21) may be executed regularly and automatically.

Subsequently, air conditioner characteristic data 8d representing the performance of the air conditioner 2 is registered in the data storage unit 8 (step S22).
The air conditioner characteristic data 8d may be registered in advance in the data storage unit by a user operation, or may be obtained by estimation from the operation data of the air conditioner 2.

  The air conditioner characteristic data 8d includes at least the relationship between the power consumption of each air conditioner 2 and the supply (removal) heat amount. Alternatively, when the amount of heat supplied (removed) by the air conditioner 2 cannot be obtained in the air conditioner operation data acquisition unit 5 and it is necessary to calculate the amount of heat from the operating frequency of the outdoor unit, the air conditioner characteristic data 8d is It includes the relationship between the operating frequency, evaporation temperature, condensation temperature, supply heat quantity and power consumption of each outdoor unit.

Next, calculation conditions such as the number of the building to be controlled and the learning period are set by a user operation (step S23).
Subsequently, the parameter learning unit 10 extracts and acquires the necessary learning input data 9 from the data storage unit 8 (step S24), and estimates a thermal characteristic parameter by a learning process based on the building model 10a (step S25). ).
The learning input data 9 includes at least past weather data of a predetermined point, past operation data of each air conditioner 2 installed in the target building, and air conditioner characteristic data 8d.

Here, the parameter learning unit 10 having the building model 10a shown in FIG. 3 will be described more specifically.
In the building model 10a shown in FIG. 3, as an influence factor of the heat load, the outside air temperature (T _O ) 41, the amount of solar radiation (Q _S ) 42, the adjacent room temperature (T _OZ ) 43, and the room temperature (T _Z ) 44, air conditioning removal heat quantity (Q _HVAC ) 45, and indoor generated heat quantity (Q _OCC + Q _EQP ) (human body + OA equipment + lighting) 46 are considered.

  When the relationship of the influence factors of the heat load is expressed by a theoretical formula (heat conduction equation), the following formulas (1) to (3) are derived.

In formulas (1) to (3), _{Q S} is the amount of solar radiation _{^{[kW / m 2], Q}} OCC human body heating value _{[kW], Q EQP} is OA equipment heating value _{[kW], Q HVAC} the air conditioner 2 The amount of heat removed (supplied) [kW].
T _O is the outside air temperature [° C.], T ₁ is the outside wall room outer surface temperature [° C.], T ₂ is the outside wall room inner surface temperature [° C.], _TZ is the room temperature [° C.], and T _OZ is the adjacent room temperature [ ° C].

R ₁ is the outdoor heat transfer coefficient [W / (m ² · K)], R ₂ is the outer wall heat conductivity [W / (m · K)], and Rz is the indoor heat transfer coefficient [W / (m ² · K)], R _OZ is the inner wall thermal conductivity [W / (m · K)], and R _WIN is the window heat transfer coefficient [W / (m ² · K)].

C ₁ is the outer wall outdoor heat capacity [J / K], C ₂ is the outer wall indoor heat capacity [J / K], and Cz is the indoor heat capacity [J / K].
Furthermore, α is the coefficient of solar radiation that penetrates into the room (−), β is the coefficient of solar radiation to irradiate the outer wall (−), γ is the coefficient of heat generation of OA equipment (−), and δ is the amount of heat removed from air conditioning (supply). The coefficient (−) and ρ are coefficients (−) of the amount of heat generated by the human body.

Here, regarding the calorific value Q _OCC , 30% (0.3 × Q _OCC ) affects the inner wall surface temperature T ₂ by radiation, and the remaining 70% (0.7 × Q _OCC ) is due to convection. It is assumed that the rise in the room temperature (Tz) 44 is affected. This ratio can be changed arbitrarily.
Also, when only 1 zone one floor is present, since the adjoining room temperature _(T OZ) 43 is not present, the next room temperature _(T OZ) 43 and an inner wall thermal conductivity _{R OZ} are neglected.

  When Expressions (1) to (3) are replaced with a state space model, they are expressed as Expressions (4) and (5) below.

In Equations (4) and (5), θ is a vector of unknown thermal characteristic parameters, and is determined so that the error between the actual room temperature and the calculated room temperature is minimized.
Here, although the prediction target is room temperature, it is also possible to predict the heat load (= air conditioner removal heat quantity Q _HVAC ) by giving the room temperature as the target room temperature.
In Expression (4), A (θ) is represented by the following parameter matrix (state transition matrix).

  B (θ) is represented by the following parameter matrix (state transition matrix).

  Furthermore, θ and C are each represented by the following parameter matrix (observation value matrix).

The human body heat generation amount Q _OCC and the OA device heat generation amount Q _EQP are difficult to measure, and may be estimated from the power consumption (power consumption data) of the target floor.
For example, when power consumption data of an air conditioner, lighting, and OA equipment is available, the OA equipment heat generation amount Q _EQP may be obtained by summing the lighting power consumption and the OA equipment power consumption Q.

Further, the human body calorific value Q _OCC may be obtained by calculating the following equation (6) by multiplying the “number of people in the room” by the “calorific value per person”.

Q _OCC = “Heat generation amount per person” * “Number of people in the room” (6)

  Further, the number of people in the room can be determined by installing sensors on the floor and counting the number of people, or by using the probability density function P (x) and the “maximum occupancy rate” for the OA equipment power consumption Q as in the following equation (7). The “number of people in the room” may be given, and the calculation may be performed from an actual measurement value (or predicted value) of the OA device power consumption Q.

  Number of people in the room = “Maximum number of people in the room” * P (x) (7)

If the OA device power consumption Q cannot be obtained, representative data of the human body heat generation amount Q _OCC and the OA device heat generation amount Q _EQP may be used in accordance with the application of the target building.

On the other hand, if the air conditioner removal heat quantity Q _HVAC can be acquired by the air conditioner operation data acquisition unit 5, the value may be used as it is. In addition, when the air conditioner removal heat quantity Q _HVAC cannot be obtained, the heat quantity can be calculated using the air conditioner characteristic data 8d.
For example, as the air conditioner characteristic data 8d, a relational expression among the compressor frequency f of the outdoor unit, the evaporation temperature ET, the condensation temperature CT, and the supply heat amount is prepared as follows.

In the above relational expression, the coefficients a, b, c, and d are given different values depending on the type of the compressor. Each coefficient a, b, c, d may be obtained from an actual measurement value.
Further, the relationship among the compressor frequency f, the evaporation temperature ET, the condensation temperature CT, and the supply heat amount of the outdoor unit may be calculated using an air conditioner model that models the refrigerant circuit of the outdoor unit.

The parameter matrices A (θ), B (θ), and C can be identified based on the above data and the above-described equations (4) and (5).
The identification procedure of the parameter matrices A (θ), B (θ) and C is as shown in steps A1 to A7 below.

  First, an initial value vector θ and a state vector X of learning parameters are determined (step A1), and state transition matrices A (θ) and B (θ) and an observation value matrix C (θ) are defined by the initial value vector θ. (Step A2).

  Subsequently, using the state transition matrices A (θ) and B (θ), the observed value vector u (k) at t = k, and the estimated state value X (k−1) at t = k−1, A state vector X (k) is generated (step A3).

Further, the system response vector Y ′ (k + 1) is estimated from the observation value matrix C (θ) and the state vector X (k) (step A4).
Next, the difference e (θ, k) between the observed value (actually measured value) and the predicted value is calculated by the following equation (step A5).

  e (θ, k) = Y (k) −Y ′ (k−1)

  Thereafter, the above steps A2 to A5 are executed for all the observed values, and a parameter matrix E (θ) is obtained as in the following equation (8) (step A6).

E (θ) = [e (θ, 1), e (θ, 2), e (θ, 3),..., E (θ, n)]
... (8)

In equation (8), n is the number of observations.
Finally, a parameter vector θ * that minimizes the norm of the parameter matrix E (θ) is obtained as shown in (9) below using the nonlinear least square method (step A7).

  θ * = argmin (E (θ) * E (θ)) (9)

In steps A3 to A7, a maximum likelihood method may be used in addition to the nonlinear least square method.
Further, a subspace method for directly obtaining the state transition matrices A and B and the observation value matrix C by matrix decomposition of the set u of the observation value vectors u (k) may be used.
As described above, the parameter learning unit 10 obtains the thermal characteristic parameter 11 by learning.

Returning to FIG. 2, the air conditioning system control device 3 stores the thermal characteristic parameter 11 obtained by the parameter learning unit 10 in the data storage unit 8 (step S <b> 26).
Next, in response to a user request, it is determined whether or not the operation schedule 15 of the air conditioner 2 is to be continuously planned (step S27). If it is determined not to plan (that is, NO), the processing routine of FIG. finish.

  On the other hand, if it is determined in step S27 that the operation schedule of the air conditioner 2 will be continuously planned according to the user request (that is, YES), calculation conditions such as the target building number and the learning period are set by the user operation ( Step S28).

Subsequently, the thermal load prediction unit 13 acquires the thermal property parameter 11 of the target building, the building model 10a, and the prediction input data 12 necessary for prediction (step S29), and predicts the thermal load of the target building. (Step S30).
The prediction input data 12 basically includes the same items as the learning input data 9, but future predicted values may be included in the data.

  Next, the schedule creation unit 14 can minimize the power consumption of the air conditioner 2 and remove the necessary heat load based on the air conditioner characteristic data 8d with respect to the heat load predicted by the heat load prediction unit 13. The operation schedule 15 of the air conditioner 2 is drafted (step S31).

  In addition, although the operation schedule 15 of the air conditioner 2 includes at least the operation stop time of the air conditioner 2, the supply (removal) heat amount of each air conditioner 2 may be included. In particular, in the case of a building multi-air conditioner, a packaged air conditioner, or a room air conditioner, the operating frequency of the outdoor unit or the set temperature of the indoor unit may be included.

Here, the calculation example of the driving schedule 15 by the schedule creation part 14 is demonstrated concretely.
The operation schedule 15 uses the total power consumption of the air conditioner 2 as an objective function, and solves an optimization problem using the supply heat amount and operation stop of each outdoor unit as control variables, thereby obtaining the following equations (10) to (14): )

In Expressions (10) to (14), Pi is the power consumption of the outdoor unit i, Qi is the amount of heat supplied to the outdoor unit i, ui is the shutdown of the outdoor unit i [0/1], Q _total is the indoor thermal load, Tz is the room temperature, maxTz is the upper limit value of the room temperature, minTz is the lower limit value of the room temperature, maxQi is the upper limit value of the supplied heat amount of the outdoor unit i, and minQi is the lower limit value of the supplied heat amount of the outdoor unit i.

The schedule creation unit 14 obtains the operation stop of the outdoor unit, which is a discrete value, by combination optimization, and obtains the optimum outdoor unit supply heat amount in each combination by nonlinear optimization.
In addition, the combination is changed so as to approach the optimal solution efficiently from a huge number of combinations, and the solution having the minimum objective function is output as the optimal operation schedule 15 from the limited candidates.
As an optimization method, dynamic programming or problem space search may be used for combinatorial optimization, or quadratic programming may be used for nonlinear optimization.

Returning to FIG. 2, the operation schedule 15 created by the schedule creation unit 14 is stored in the data storage unit 8 (step S32).
Finally, the operation schedule 15 is transmitted from the operation schedule output unit 6 to the air conditioner 2 via the communication unit 100 (step S33), and the processing routine of FIG.

Thereby, the air conditioner 2 is operated based on the operation schedule 15.
The above steps S21 to S33 can all be automatically executed by registering the calculation conditions in the data storage unit 8 in advance.

  As described above, according to the first embodiment (FIGS. 1 to 3) of the present invention, in the air conditioning system control device 3 that controls the air conditioner 2 installed in the building, the weather data distribution company 1 The meteorological data acquisition unit 4 that performs data communication via the communication unit 100, the air conditioner operation data acquisition unit 5 that performs data communication with the air conditioner 2 via the communication unit 100, the weather data acquisition unit 4 and The data storage unit 8 stores the data acquired by the air conditioner operation data acquisition unit 5 and the building model 10a based on the building specifications, and acquires predetermined learning input data 9 from the data storage unit 8, and the building And a parameter learning unit 10 that obtains the thermal characteristic parameter 11 by learning based on the model 10a.

Further, the air conditioning system control device 3 acquires predetermined prediction input data 12 from the data storage unit 8 and predicts the thermal load of the building based on the thermal characteristic parameter 11, the building model 10a, and the predetermined prediction input data 12. Based on the thermal load obtained by the thermal load predicting unit 13 and the predetermined input data 12 for prediction, the schedule creating unit 14 for planning the operation schedule 15 of the air conditioner 2 to be energy-saving, and schedule creation And an operation schedule output unit 6 for transmitting the operation schedule 15 obtained by the unit 14 to the air conditioner 2.
The data storage unit 8 stores the thermal characteristic parameter 11 and the operation schedule 15.

  Furthermore, as physical parameters in the building model 10a, at least the heat resistance value (reciprocal of heat transfer coefficient) and heat capacity of the outer wall and the window, which represent heat inflow from the outside, and the inner wall representing heat inflow from the adjacent zone. It includes thermal resistance value (reciprocal of heat transfer coefficient), indoor heat capacity, absorption rate of solar radiation to the outer and inner walls, and absorption rate of internal heat generation to the inner wall.

  With the above configuration, the parameter learning unit 10 can automatically learn the thermal characteristic parameters 11 of individual buildings, so that the thermal characteristic parameters 11 of each building, position information (for example, orientation), and detailed buildings Even without obtaining specifications (for example, window area and roof area), it is possible to predict a heat load with high accuracy based on a physical formula.

Further, by periodically performing the learning calculation of the thermal characteristic parameter 11, seasonal fluctuations (for example, fluctuations in the amount of solar radiation Qs incident from the window) and changes in the indoor environment (for example, blinds are installed in the windows). Etc.) can be modeled.
Accordingly, the accuracy of the thermal load prediction in each individual building is improved, and the schedule creation unit 14 can obtain the operation schedule 15 of the air conditioner 2 according to the thermal characteristics of the building, so that an energy saving effect can be obtained in each building. .

Embodiment 2. FIG.
In the first embodiment (FIG. 1), the learning period in the parameter learning unit 10 is not specifically mentioned. However, as shown in FIG. 4, a learning period determining unit 10b is provided in the parameter learning unit 10A. Also good.

FIG. 4 is a block diagram showing an air conditioning system control device 3A according to Embodiment 2 of the present invention. Components similar to those described above (see FIG. 1) are denoted by the same reference numerals as those described above or after the reference numerals. A detailed description is omitted with “A”.
In FIG. 4, the difference from the above (FIG. 1) is only that a learning period determination unit 10b is added in the parameter learning unit 10A.

Normally, the more the learning input data 9 in the parameter learning unit 10A is more abundant, the closer to the actual value of the thermal characteristic parameter 11 can be estimated. However, when the learning input data 9 is long-term, As a result, the calculation load on the parameter learning unit 10A increases, and a large amount of time is required for the parameter estimation calculation.
If the accuracy of the building model 10a itself is low, the accuracy of parameter learning will not be improved even if the learning input data 9 is acquired over a long period of time.

  In the second embodiment of the present invention, as shown in FIG. 4, by adding a learning period determination unit 10b to the parameter learning unit 10A, an optimum learning period (learning input) including the improvement of the accuracy of the building model 10a. The data period acquired as data 9 can be determined.

Hereinafter, the learning period determination unit 10b according to the second embodiment of the present invention will be described with reference to FIGS.
FIG. 5 is an explanatory diagram showing the relationship between the learning period and the prediction accuracy in the air conditioning system control device 3A of FIG. 4, and the accuracy change when the learning period is changed from 1 day to 10 days for a certain target building. Is represented by a mean square error (RMSE: Root Mean Square Error) between the predicted value and the actual measurement value of the prediction target day.
FIG. 6 is an explanatory diagram showing the relationship between the learning period of the air conditioning system control device 3A of FIG. 4 and the estimated thermal characteristic parameter 11 (wall heat capacity [MJ / K]).

In FIG. 5, when the learning period is 4 days or more, the error average value RMSE is less than 1 ° C., and it seems that the parameter estimation is successful.
On the other hand, in FIG. 6, when the learning period is 5 days or longer, the fluctuation of the wall heat capacity [MJ / K] is small.

Among the heat characteristic parameters 11, the wall heat capacity is basically an invariable physical property value and should not change due to the difference in the learning period. Therefore, as is clear from FIG. 6, the learning period is less than 5 days. However, it cannot be said that a value close to true can be estimated.
Therefore, it is necessary not only to select a learning period in which the error average value in FIG. 5 is less than a certain value, but basically to determine a learning period in which fluctuations in invariable physical property values (such as wall heat capacity) are less than a certain value. There is.

  For example, when the learning period determination mode is selected at the time of calculation condition setting (step S23 in FIG. 2), the learning period determination unit 10b includes about 10 days to several tens of days including the learning input data 9 Within this range, extra data is acquired, and an appropriate learning period is determined by simulating the error between the predicted value and the actual measurement value of the predetermined thermal characteristic parameter 11 and the fluctuation of the predetermined thermal characteristic parameter 11. To do.

Here, the appropriate learning period is, for example, the shortest learning period in which the error between the predicted value and the actual measurement value of the thermal characteristic parameter 11 is equal to or less than the first threshold, and the variance of the thermal characteristic parameter 11 is equal to or less than the second threshold Of the shortest learning periods, the longer value may be selected.
The learning period determined by the learning period determination unit 10 b is stored in the data storage unit 8 together with the thermal characteristic parameter 11.

  In addition, you may verify the change of a building member by monitoring the thermal characteristic parameter 11 learned regularly on a long-term basis. For example, when the learning results of four times a year are stored for 10 years, a change in the building member (for example, deterioration of the outer wall reflecting paint) can be detected by comparing 40 learning results.

  As described above, the parameter learning unit 10A of the air conditioning system control device 3A according to Embodiment 2 (FIGS. 4 to 6) of the present invention has the learning period determination unit 10b, and the learning period determination unit 10b The shortest learning period in which the error between the predicted value of the characteristic parameter 11 (power consumption in the building and the like) and the actual measurement value is equal to or smaller than the first threshold value and the variation of the thermal characteristic parameter is equal to or smaller than the second threshold value is calculated.

Thereby, since an appropriate learning period can be determined for each building, calculation time can be shortened and prediction accuracy can be improved.
In addition, since the prediction accuracy of the thermal characteristic parameter 11 is improved, modeling close to an actual building is possible.
Furthermore, since the physical property value close to the true value can be monitored by improving the prediction accuracy of the thermal characteristic parameter 11, it is possible to evaluate the performance of the building (e.g., the evaluation of heat insulation performance).

Embodiment 3 FIG.
In the first embodiment (FIG. 1), the correction of the building model 10a is not mentioned, but the building model correction unit 10c may be provided in the parameter learning unit 10B as shown in FIG.

FIG. 7 is a block diagram showing an air conditioning system control device 3B according to Embodiment 3 of the present invention. Components similar to those described above (see FIG. 1) are denoted by the same reference numerals as those described above or after the reference numerals. Detailed description is omitted with “B”.
In FIG. 7, the difference from the above (FIG. 1) is only that a building model correcting unit 10c is added in the parameter learning unit 10B.

  In FIG. 7, the more detailed the building model 10a is, the more the thermal characteristic parameter 11 can be estimated. However, the building model 10a is a general-purpose that can be applied to many buildings. Since it must be a model, if it is detailed, it becomes a model in which prediction accuracy differs greatly between buildings.

  Further, as is clear from FIG. 5 described above, even if the learning period is increased from 4 days to 10 days, the prediction accuracy is not dramatically improved. That is, FIG. 5 shows the limit of the building model 10a, and shows that the prediction accuracy will not be improved even if the learning period is extended unless the building model 10a itself is changed.

Therefore, as shown in FIG. 7, a building model 10a most suitable for the target building can be created by adding a building model correcting unit 10c in the parameter learning unit 10B.
For example, in the case of room temperature prediction, the building model correction unit 10c uses a building model for obtaining a representative temperature of one zone when building the building model 10a, or obtains a representative temperature of each zone by dividing the zone into two zones east and west. It is automatically determined whether it is a building model or a building model that is divided into smaller zones and obtains the representative temperature of each zone.

  For example, if n outdoor unit data is included in the air conditioner operation data, the minimum unit of zone division is determined to be n, and the accuracy at the time of n zone division is compared with the accuracy without zone division. To do.

  In addition, when dividing n zones, the degree of influence between zones can be calculated by comparing the inner wall heat transfer coefficient between the zones. The optimum number of zone divisions is obtained by comparing the accuracy of division. Alternatively, zone division may be defined in advance by a user operation.

Moreover, you may create the building model 10a which added not only zone division but the new thermal characteristic parameter.
For example, when the accuracy required by the building model 10a cannot be obtained, some options may be prepared and the detailed building model 10a satisfying the required accuracy of the following conditions may be constructed.

That is, as an optional condition,
(A) Increase the heat capacity setting point of the wall with respect to the building model 10a.
(B) Add the heat capacity of the window,
(C) Dividing the outer wall into a ceiling and a floor,
Etc.

  As described above, the parameter learning unit 10B of the air-conditioning system control device 3B according to Embodiment 3 (FIG. 7) of the present invention updates the building model 10a so as to approach the target building. Therefore, the building model 10a corresponding to the target building can be automatically created, and the prediction accuracy of the thermal characteristic parameter 11 is improved.

That is, modeling including spatial differences (for example, azimuth) in the room and changes in zone division (for example, movement of partitions) can be performed.
Also, by learning the physical parameters of a general (basic) building model, a model that can be applied to each building can be created automatically, and non-standard buildings (for example, extremely insulating) It is possible to automatically create a building model 10a that can cope with a high building).

  1 weather data distribution company, 2 air conditioners, 3, 3A, 3B air conditioning system control device, 4 weather data acquisition unit, 5 air conditioner operation data acquisition unit, 6 operation schedule output unit, 8 data storage unit, 8d air conditioner characteristic data 9 Input data for learning 10, 10A, 10B Parameter learning unit, 10b Learning period determination unit, 10a Building model, 10c Building model correction unit, 11 Thermal characteristic parameter, 12 Prediction input data, 13 Thermal load prediction unit, 14 Schedule creation unit, 15 operation schedule, 100 communication means, S21 data acquisition / storage step, S22 air conditioner characteristic data registration step, S23 calculation condition setting step, S24 learning input data acquisition step, S25 parameter learning step, S26 parameter storage step , S28 calculation condition setting Step, S29 prediction input data acquisition step, S30 heat load calculation step, S31 operation schedule creation step, S32 operation schedule storage step, S33 operation schedule output step, 41 outside air temperature, 42 solar radiation amount, 43 adjacent room zone temperature, 44 indoors Temperature, 45 Air-conditioner supply heat, 46 Indoor heat generation.

Claims (3)

In an air conditioning system controller that controls an air conditioner installed in a building,
A weather data acquisition unit that performs data communication with a weather data distribution company via a communication means;
An air conditioner operation data acquisition unit for performing data communication with the air conditioner via a communication means;
A data storage unit for storing data acquired by the weather data acquisition unit and the air conditioner operation data acquisition unit;
A parameter learning unit that has a building model based on the specifications of the building, obtains predetermined learning input data from the data storage unit, and obtains a thermal characteristic parameter by learning based on the building model;
Obtaining a predetermined prediction input data from the data storage unit, and based on the thermal characteristic parameter, the building model and the predetermined prediction input data, a thermal load prediction unit for predicting a thermal load of the building;
Based on the thermal load obtained by the thermal load prediction unit and the predetermined input data for prediction, a schedule creation unit that formulates an operation schedule of the air conditioner to be energy-saving,
An operation schedule output unit that transmits the operation schedule obtained by the schedule creation unit to the air conditioner, and
The data storage unit stores the thermal characteristic parameter and the operation schedule. The parameter learning unit includes a learning period determination unit,
The learning period determination unit calculates a shortest learning period in which an error between the predicted value and the actual measurement value of the thermal characteristic parameter is equal to or less than a first threshold value, and a variation in the thermal characteristic parameter is equal to or less than a second threshold value. The air-conditioning system control apparatus according to claim 1.   The air conditioning system control device according to claim 1, wherein the parameter learning unit includes a building model correction unit that updates the building model so as to approach the building.

Images