JP5763955B2 - Air conditioning load prediction device and air conditioning load prediction method - Google Patents

Description

  The present invention relates to an air conditioning load prediction device and an air conditioning load prediction method for predicting an air conditioning load of an air conditioning device.

  Predicting the air conditioning load, which is the energy consumption of the air conditioning equipment, is extremely important for efficiently operating the air conditioning equipment and improving the energy efficiency, and in particular, air conditioning equipped with heat storage equipment that effectively uses nighttime power. The system is highly demanded in terms of reducing energy consumption. In addition, the operation of conventional heat source equipment has a case of excessive heat storage on the premise of maximum air conditioning load, which does not lead to reduction of equipment running cost and environmental load, and makes accurate air conditioning load prediction. A feed-forward operation method suitable for the load is required.

  Examples of predicting such an air conditioning load include those described in Patent Documents 1 to 3, for example.

  Patent Document 1 describes the use of external network condition data, indoor environment condition data, and air conditioning system operating condition data when predicting the air conditioning load in units of hours on the next day using a neural network model constructed on a computer. Input is performed, and the learning coefficient for each input data is optimized based on the past input data and the air conditioning load value in the time unit of the next day.

  The thing of patent document 2 inputs these each data with respect to the prediction model learned using the air-conditioning load performance data, the weather data, the calendar data, and the air-conditioning equipment operation schedule data, and predicts the air-conditioning load.

  Patent Document 3 describes a past actual measurement with respect to the actual measurement data input as a new case when an air conditioning load is predicted by an ARMA model expression to which a daily fluctuation pattern model expression is added. The highest load of the day is predicted by phase case-based modeling that performs prediction using a case base with data as a case, and the predicted load based on the ARMA model formula is corrected based on the highest load.

JP 2006-78009 A JP 2005-226845 A Japanese Patent No. 3168529

  As described above, predicting the air conditioning load of an air conditioner has been extremely important in order to improve the energy efficiency by efficiently operating the air conditioner.

  Therefore, the present invention does not require the conventional weather data and outside air data used when predicting the air conditioning load of the air conditioning equipment, uses the rising load on the prediction day, eliminates the complexity of parameter input, and is easy for the operator. It is intended to improve energy efficiency by operating air conditioning equipment efficiently.

The present invention relates to an hourly load distribution rate calculating means for calculating a load distribution rate for one day at each time by dividing an air conditioning load amount for each time of a plurality of days in a past fixed period by an air conditioning load amount for one day. From the air conditioning load amount at each time during the predetermined time when the air conditioning starts on the prediction day and the load distribution ratio at each time during the predetermined time when the air conditioning starts on the day in the past plural days calculated by the hourly load distribution rate calculating means. The daily load amount predicting means for predicting the air conditioning load amount for one day on the prediction day , and the air load load amount for one day predicted by the daily load amount predicting means, calculated by the hourly load distribution rate calculating means. is multiplied by the load distribution rate of the time have a, a load prediction means when predicting air conditioning load amount of each time, and sets the weighting factor to the air conditioning load of each time at a certain time the air conditioner rises in the predicted day , Weighting factor is characterized that you have been set closer small time during which the air-conditioning rises.

According to the present invention, the load distribution ratio for one day at each time is calculated by dividing the air-conditioning load amount for each time of a plurality of days in a past fixed period by the air-conditioning load amount for one day . From the air conditioning load amount at each time during the predetermined time when the air conditioning is started up and the calculated load distribution ratio at each time during the predetermined time when the air conditioning is started over the past several days, the air conditioning load amount for one day on the predicted day is calculated. Predict. Then, the air conditioning load amount at each time is predicted by multiplying the predicted air conditioning load amount for one day by the calculated load distribution rate at each time. Thus, by predicting the air conditioning load amount at each time in one day, the air conditioning equipment can be efficiently operated to improve the energy efficiency.
In addition, according to the present invention, the weighting coefficient set for the air conditioning load amount at each time is set to be smaller as the air conditioning rises closer to the time when the air conditioning starts. With such a setting, when past data (load amount) is used, the accuracy decreases as the data traces back to the past from the latest data, so that the air conditioning load prediction can be made more accurate.

It is a block diagram which shows the whole structure of the air-conditioning load prediction apparatus concerning the 1st Embodiment of this invention. It is a whole lineblock diagram of a heat source system provided with a thermal storage tank. It is a flowchart which shows the air-conditioning load prediction method by the air-conditioning load prediction apparatus of FIG. It is a graph which shows the cooling load allocation rate for every time from 1 am to midnight in 24 hours a day for a fixed period. 4A is a graph showing a state in which the load distribution ratio for one day at each time on a plurality of days excluding exception days among a plurality of days for a certain period is calculated, and FIG. It is a graph which shows the state which computed the average value of. A value that minimizes the sum of the squares of the difference between the actual load amount of the day and the predicted load amount indicated by a straight line with respect to the actual load amount as an index of the weighting coefficient for the air conditioning load at each time in the rising period It is a graph for calculating | requiring (the least squares method). It is a graph which shows the state where the prediction load amount and the actual load amount which is an actual air conditioner operation load amount are shown for each time, and the daily prediction load amount and the actual load amount decrease as time passes. It is a graph which shows setting the driving | running pattern of the refrigerator shown in FIG. 2 with respect to the estimated load amount shown in FIG. 7, and a heat exchanger. It is a graph which shows the operating state of the refrigerator shown in FIG. 2 with respect to the operation pattern setting shown in FIG. 8, and a heat exchanger.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  As shown in FIG. 1, the load prediction apparatus according to an embodiment of the present invention mainly includes a load data collection unit 1, an air conditioning load prediction unit 3, a heat source facility control unit 5, and a heat source facility device 7. Yes.

  The load data collection unit 1 collects and accumulates the air conditioning load amount at each time of a plurality of days in a past fixed period, and functions as a central monitoring device that also stores temperature data of a heat storage tank, which will be described later. . The air conditioning load prediction unit 3 predicts the air conditioning load in the heat source equipment 7 based on the data collected and accumulated by the load data collection unit 1 by regression analysis as described later. The heat source facility control unit 5 controls the heat source facility device 7 in consideration of the air conditioning load predicted by the air conditioning load prediction unit 3.

  As shown in FIG. 2, the heat source equipment 7 is an air conditioning system including a heat storage tank 9, and an indoor air conditioner 11 such as a fan coil that performs indoor air conditioning, and the heat storage tank 9 mainly use nighttime power. Refrigerators 13 and 15 for storing heat, a heat exchanger 17 for performing air conditioning by the indoor air conditioner 11 using a heat source of the heat storage tank 9, and a refrigerator 19 for performing indoor air conditioning ing.

  Reference numerals 21 and 23 are headers, and reference numerals 25 to 35 are pumps.

  Next, a load prediction method by the above-described load prediction device will be described based on the flowchart shown in FIG. First, the load data collection unit 1 collects and accumulates the air conditioning load amount at each time of a plurality of days in a past fixed period (step S1), and adopts it for air conditioning load prediction from the accumulated past data. The period to be set is set to either a certain period of the same period of the previous year (for example, one month of the same month of the previous year) or a certain period immediately before (for example, three weeks) (step S2). This setting operation is performed manually by the operator.

Then, by dividing the air conditioning load amount q _dn for each day of the plurality of days in the adopted period by the air conditioning load amount Q _d for one day, the load distribution ratio r _dn for one day at each time is obtained. Calculate (step S3). That is, r _dn = q _dn ÷ Q _d . Therefore, the air-conditioning load prediction unit 3 calculates the load distribution ratio for each day at each time by dividing the air-conditioning load amount for each time of a plurality of days in a past fixed period by the air-conditioning load amount for one day. An allocation rate calculation means is included.

FIG. 4 shows the cooling load distribution ratio r _dn at each time from 1:00 am to midnight in 24 hours a day for a certain period, and the cooling load distribution ratio r _{dn for} this certain period for a plurality of days. Is calculated (step S4).

Subsequently, for each day of the above-mentioned fixed period, calculate the square of the difference between the load distribution ratio at each time and the average value of the load distribution ratio at each time for a certain period of time, and the calculated numerical value In order of increasing size, an exceptional date is set to exclude the cooling load distribution ratio r _dn of about 5 days from the average value calculation in step S4 if it is a one month period (step S5).

FIG. 5A shows a state in which the load distribution ratio r _dn for each day of each time in a plurality of days excluding the exceptional days described above is calculated (step S6), and FIG. ) Shows a state where the average value is calculated with respect to FIG. 5A (step S7). In addition, let the load distribution rate with respect to 1st of each time in the several day except this exceptional day be an effective load distribution rate.

Here, in parallel with the calculation of the load distribution ratio r _{dn in} step S6, the operator manually set a time (for example, 5:00 to 8:00) at a certain time when the air conditioning on the day used for the load prediction calculation starts. After (step S8), a weighting coefficient K by exponential diminishing method is set to the load amount at each time of rising (step S9). This weight coefficient is maximized at the last time when the air conditioning (heat source) is operating, and is decreased as it goes back in the past. In other words, the weighting factor is made smaller as it is closer to the time when the air conditioning is started. This is due to the fact that when using past data (load amount), the accuracy worsens as it goes back to the past from the latest data.

  The above weighting factor K is calculated by the following equation.

K ₀ = (1−α) ¹
K ₋₁ = (1−α) ²
K _-2 = (1-α) ³
K _-3 = (1-α) ⁴
However, α is an index smaller than 1 (α <1), and the sum of the squares of the differences between the daily actual load amount shown in FIG. 6 and the predicted load amount indicated by a straight line with respect to the actual load amount Is the smallest (least square method) value, and this is set as the index α.

  In FIG. 6, the horizontal axis is the load amount of the entire rising period (5 o'clock to 8 o'clock), and the vertical axis is the load amount for one day (24 hours).

  Next, the daily load amount Q is predicted from the total numerical value obtained by multiplying the load amount at each time in the rising period by the weighting factor and the effective load distribution ratio calculated in step S7 (step S10). ). That is, the air-conditioning load predicting unit 3 predicts the air-conditioning load amount for one day from the air-conditioning load amount at each time in a certain time during which air-conditioning starts on one day, among the air-conditioning load amounts for each time of a plurality of days. Includes prediction means.

Q = {(Q _o × K ₀ ) + (Q ₋₁ × K ₋₁ ) + (Q ₋₂ × K ₋₂ ) + (Q ₋₃ × K ₋₃ )} ÷ {(R _o × K ₀ ) _{+ (R -1 × K -1)} + (R -2 × K -2) + (R -3 × K -3)}
However, Q _o : Load amount at the predicted time (actual value)
Q _-1 : Load amount one hour before the predicted time (actual value)
Q _-2 : Load amount 2 hours before the predicted time (actual value)
Q- ₃ : Load amount (actual value) 3 hours before the predicted time
R _o : Load distribution ratio at the predicted time R ₋₁ : Load distribution ratio at one hour before the predicted time R ₋₂ : Load distribution ratio at two hours before the predicted time R ₋₃ : Load distribution ratio at three hours before the predicted time K _o : Weighting coefficient for prediction time K _-1 : Weighting coefficient for 1 hour before prediction time K _-2 : Weighting coefficient for 2 hours before prediction time K _-3 : Weighting coefficient for 3 hours before prediction time Multiply the daily load amount (daily load amount) Q by the load distribution rate at each time to calculate the load amount (hour load amount) q at each time by the following equation (step S11), and calculate the calculated hourly load amount data. It outputs to the heat source equipment 7 of FIG. 1 (step S12). That is, the air conditioning load predicting unit 3 multiplies the air conditioning load amount for one day predicted by the daily load amount predicting unit by the load distribution rate at each time calculated by the hourly load distribution rate calculating unit, and the air conditioning load at each time. It includes hourly load amount predicting means for predicting the amount.

q ₁ = Q × R ₁
q ₂ = Q × R ₂
q ₃ = Q × R ₃
・ ・ ・ ・ ・ ・
q ₂₄ = Q × R ₂₄
However, q ₁ : Predicted load at 1 o'clock
q ₂ : Predictive load at 2 o'clock
q ₃ : Predictive load at 3 o'clock
・ ・ ・ ・ ・ ・
q ₂₄ : Predicted load amount at 24:00 FIG. 7 shows the predicted load amount (broken line A) calculated in this way and the actual load amount (solid line B) that is the actual air conditioner operating load amount for each time. It can be seen that the difference between these is extremely small, and the load amount is accurately predicted.

The numerical value corresponding to 1 o'clock of the broken line C corresponds to the calculated daily predicted load Q _d, and the predicted load for each time indicated by the broken line A is obtained at the previous time as time passes. The predicted load is subtracted sequentially. A solid line D indicates the actual load amount indicated by the solid line B by sequentially subtracting the actual load amount used at the previous time as time elapses. The numerical values of the broken line C and the solid line D correspond to the scale on the right side of FIG.

  FIG. 8 shows the predicted load amount indicated by the broken line A in FIG. 7 as an example, the hourly load prediction in which the rising load at 9 o'clock is predicted and the predicted processing load heat amount, and the refrigerator 13 shown in FIG. The simulation of the operation plan of the day of 15 and 19 and the heat exchanger 17 is shown. Since the data in FIG. 8 and the data in FIG. 7 are for different days, the numerical values are different from each other.

  In FIG. 8, the required indoor air-conditioning load amount from 9:00 to 17:00 is calculated by subtracting the processing load predicted heat amount (dashed line C) predicted at 9:00 from 9:00 (E part) and 17:00 (F part). Assuming that the refrigerator 19 (H section) is operated so as to consume the amount of heat stored at night (region below the two-dot chain line G), the fluctuation of the indoor air conditioning load at each time is assumed to be the heat exchanger 17 (I section). ) Set the operation pattern to be carried out. At this time, an operation pattern is assumed in which the refrigerator 13 (J section) is operated from 9 o'clock to 11 o'clock in order to compensate for the indoor air conditioning load that cannot be covered by the night heat storage amount. As a result, the amount of heat stored in the heat storage tank 9 is simulated so that the heat of the heat storage tank is used up at 17:00 at the end of the high load time of the indoor air conditioning load, and the optimum operation of the refrigerators 13, 15, 19 and heat exchanger 17 by feedforward is performed. The plan can be set easily.

  FIG. 9 shows the load prediction shown in FIG. 8 and the actual operation results and actual load results of the refrigerators 13, 15, 19 and the heat exchanger 17 shown in FIG.

  In FIG. 9, the heat storage tank 9 is heated by operating the refrigerators 13 and 15 (J, K part) using the nighttime power from 22:00 to 8:00, and the indoor air conditioning load is performed by the refrigerator 19 (H part). At 9 o'clock, the predicted load amount at each time in consideration of the rising load is calculated and shown as the predicted load heat amount (broken line A). From 9 o'clock to 11 o'clock when the air conditioning load increases, the heat source equipment operation plan simulated at the time of load prediction by the refrigerator 19 (H part) and the refrigerator 13 (J part) was performed, and the heat storage tank 9 was stored. The operation | movement made to follow indoor air-conditioning load with the heat exchanger 17 (I part) using calorie | heat amount is shown. From 12:00 to 17:00, the base operation of the refrigerator 19 (H part) and the amount of heat stored in the heat storage tank 9 are tracked by the heat exchanger 17 (I part) to the indoor air conditioning load. 13 (J section) is stopped (peak cut operation). Moreover, the operation | movement which followed the indoor air-conditioning load in the refrigerator 19 (H part) which is a base machine is shown from 18:00 to 22:00.

  Thus, from FIG. 8 and FIG. 9, in the conventional feedback control, the heat source device is chased with the required load heat amount (actual result) at that time, and residual heat is generated due to excessive heat storage in the heat storage tank. The operation pattern of the refrigerators 13, 15, 19 is determined from the required heat amount predicted for the load, the refrigerator 13 (J section) is systematically stopped at 11:00, and the heat stored at night is converted into the heat exchanger 17 (I section). ), It is possible to perform simulation in consideration of effective use, and in fact, the relationship between the predicted load heat quantity (broken line A) and the actual load heat quantity (solid line B) in FIG. The relationship with the processing burden heat amount (solid line D) has an accuracy within ± 5%, respectively, and the optimized operation of the heat source device and the efficient use of energy can be obtained.

As described above, according to the present embodiment, the air conditioning load amount (hour load amount) q _dn for each time of a plurality of days in the past fixed period is set to the air conditioning load amount (daily load amount) Q _d for one day. In addition to calculating the load distribution ratio r _dn for each day at each time, each time in a certain period (for example, 5:00 to 8 o'clock) during which air-conditioning starts on one day out of the air-conditioning load amount for each time on multiple days The air conditioning load amount Q _d for one day is predicted from the air conditioning load amount.

In this case, the air conditioning load amount Q _d for one day can be identified with a certain accuracy (for example, an accuracy of 95%) by a specific explanatory variable (a load amount during a certain period when the air conditioning is started up).

The air conditioning load amount q at each time is predicted by multiplying the predicted air conditioning load amount Q _d for one day by the calculated load distribution ratio r _dn at each time. Thus, by predicting the air conditioning load amount q at each time in one day, it is possible to efficiently operate the heat source equipment device 7 that is an air conditioning device and improve energy efficiency.

  At that time, in the present embodiment, the following effects are obtained.

  (1) Simple program structure and low cost.

  (2) A prediction result can be clearly obtained with a simple algorithm.

  (3) It is easy to update the setting parameter (a past fixed period adopted for the air conditioning load prediction) based on the accumulation of the actual value of the air conditioning load. That is, XX month XX day to YY month YY can be set, and can be arbitrarily specified by month, season, or the like.

  (4) It is possible to set a parameter (a load distribution ratio specific to a building based on the above-described regression analysis) suitable for the load characteristics specific to the building using the air conditioner. For example, for office buildings, the load distribution ratio can be set according to the working hours and peak hours of office building activities, and for commercial buildings, it is a busy time zone, such as weekday evenings and holiday afternoon peak hours. The load distribution ratio can be set accordingly. Since these buildings have a uniform pattern and regularity in the fluctuation pattern of the hourly load amount, the load distribution rate specific to the building can be calculated with high accuracy.

Moreover, the improvement of the air conditioning load prediction makes it easy to adopt the optimized operation control of the air conditioning system equipped with the heat storage device, thereby realizing an automated operation with a simple operation. This improves the coefficient of performance of the equipment, leading to maximum utilization of late-night power, power peak cut, peak shift, etc., and can contribute to energy saving and CO ₂ reduction.

  In addition, changes in equipment contents and operational contents can be made without changing the program structure, etc. without changing the parameters of the prediction formula (for the past fixed period used for air conditioning load prediction and for the fixed time when the air conditioning of the day starts up. This can be easily handled by changing the (time).

  Further, in this embodiment, a weighting factor is set for the air conditioning load amount at each time in a certain time when air conditioning rises in one day, and this weighting factor is set smaller as the time when air conditioning rises is closer to the past. ing. With such a setting, when past data (load amount) is used, the accuracy decreases as the data traces back to the past from the latest data, so that the air conditioning load prediction can be made more accurate.

  Moreover, in this embodiment, since the air conditioning load amount for each time of a plurality of days in a certain period is an average value at the same time on a plurality of days, a more accurate air conditioning load prediction can be performed.

  Further, in the present embodiment, each time except for the day when the difference between the load distribution ratio for one day at each time and the average value of the load distribution ratio at the same time for a plurality of days in the past for a certain period is the largest. Therefore, the air conditioning load prediction can be performed with higher accuracy.

  In the embodiment described above, the cooling load is used as the air conditioning load. However, the present invention can be similarly applied to a heating load.

  3 Air conditioning load prediction unit (hour load distribution rate calculation means, daily load amount prediction means, hour load amount prediction means)

Claims (4)

A load distribution rate calculating means for calculating a load distribution rate for one day at each time by dividing the air conditioning load amount for each time of a plurality of days in a past fixed period by the air conditioning load amount for one day;
From the air conditioning load amount at each time at a certain time when air conditioning starts on the prediction day and the load distribution ratio at each time at a certain time when air conditioning rises on the first day of the plurality of past days calculated by the hourly load distribution rate calculating means , A daily load amount predicting means for predicting an air conditioning load amount for one day on the prediction day ;
Time load amount prediction for predicting the air conditioning load amount at each time by multiplying the air load load amount for one day predicted by the daily load amount prediction unit by the load distribution rate at each time calculated by the hour load distribution rate calculating unit. and it means, possess,
The set weighting factors to the air conditioning load of each time at a certain time the air conditioner rises in the prediction day, the weight coefficient, the air-conditioning load prediction apparatus characterized that you have been reduced set closer to the time that the air conditioner rises . The air-conditioning load of a plurality of days for each of the time in a certain period of time, the air-conditioning load prediction apparatus according to claim 1, characterized in that an average value at the same time of the multi-day. The hourly load distribution ratio calculating means excludes a day in which the difference between the load distribution ratio for one day at each time and the average value of the load distribution ratio at the same time for a plurality of days in the past fixed period is at least the largest. The air-conditioning load prediction apparatus according to claim 1 or 2 , wherein a load distribution ratio for one day at each time is calculated. Divide the air conditioning load amount for each time of multiple days in the past fixed period by the air conditioning load amount for one day to calculate the load distribution ratio for one day at each time, and at the fixed time when air conditioning on the predicted day starts From the air conditioning load amount at each time and the calculated load distribution ratio at each time during a certain time when air conditioning starts on the day in the past, the air conditioning load amount for one day on the prediction day is predicted. Multiplying the air conditioning load amount for one day by the calculated load distribution rate at each time to predict the air conditioning load amount at each time ,
An air conditioning load prediction method , wherein a smaller weight coefficient is set to an air conditioning load amount at each time in a predetermined time when air conditioning rises on the prediction day, closer to the time when air conditioning rises .

Images