JPH0678459A - Demand control unit - Google Patents

Abstract

PURPOSE:To improve a flow-up characteristic and stability in demand control, by carrying out fuzzy inference with regard to a demand characteristic parameter for indicating a historical changing trend of working power or a stationary degree, and calculating a single predictive value from temporary predictive values calculated by a plurality of temporary prediction calculating means. CONSTITUTION:In a fuzzy characteristic operation unit 7, fuzzy inference is carried out with regard to a demand characteristic parameter generated from a demand characteristic parameter extracting unit 2 for indicating a historical changing trend of service power or a stationary degree so that weighting values for each predictive value from predicting units a3, b4, and c5 are determined on the basis of a given fuzzy rule by referring a membership function in a demand characteristic membership function unit 6. Then, a weight adding step is carried out by a weight adding unit 8 so that the determined weight values correspond to each predictive value from the predicting units 3 to 5. A center-of- gravity operation unit 9 generates a final power demand predictive value from a given formula on the basis of the predictive value from the predicting units 3 to 5 together with each corresponding weighting value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an improvement of a demand control device for controlling power consumption during demand time so as not to exceed contracted power.

[0002]

2. Description of the Related Art Conventionally, in a demand control device, a calculation formula such as, for example, the following formula (1) is used as a predictive calculation formula of power consumption, and the value calculated by this formula is used at the end of the demand time period. It was used as the predicted value of power consumption.

R = P + (Δp / Δt) × T _n (1) where R: predicted value of power consumption at the end of demand time period P: current value of power consumption at demand time period Δp: sampling time Power consumption (increase in power consumption) Δt: Sampling time T _n : Demand time remaining time. Practically, Δp is a discrete quantity because it is input in the form of a pulse having a constant power value conversion weight.

As is clear from the above formula (1), when the remaining time T _n of the demand time period is large, even if the sampling time Δt is the same, the power consumption Δp during that period is
A slight change in the value will appear as a large change in the predicted value R. This fluctuation range becomes significant when the remaining time T _n is large or when the sampling time Δt is short.

From the viewpoint of the followability of the demand control, for example, when the load suddenly becomes heavy, this is promptly reflected in the predicted value, it is desirable that the sampling time Δt is short. Stability, e.g. avoiding demand interruption due to short-term fluctuations in the load or inconsistent power fluctuations due to load inrush current,
From this point of view, it is desirable that the sampling time Δt is long.

As one of the conventional prediction methods, for example, in the demand time period, in the range where the remaining time T _n can be considered to be large, the sampling time Δt is set to be longer than the range where the remaining time T _n is considered to be small, and Some tried to satisfy the conditions of. That is, by increasing the sampling time Δt, the sampling time Δt
This is an attempt to suppress fluctuations in the predicted value by averaging power fluctuations having a duration shorter than t.

[0007]

For the purpose of suppressing the fluctuation of the above-mentioned predicted value, the sampling time Δ
The longer t is, the more effective it is. However, if the averaging section is long, for example, when a large-capacity load rises, the predicted value will actually be affected, but there will be a time delay. Depending on the size of the, the timing of demand control (load cutoff / return) may be lost.

For example, FIG. 11 shows a case where pulses (power pulses) are input at substantially equal intervals within a relatively long sampling time Δt, and FIG. 12 remarkably occurs within a relatively long sampling time Δt. This is the case where there is a bias in the arrival intervals of the pulses. It is assumed that the former is in a steady operation with a constant load, and the latter is a rise from a light load to a heavy load.

From the above, in the conventional prediction calculation formula (1), even if the remaining time T _n is large, if the sampling time Δt is simply lengthened, both of them (states of FIGS. 11 and 12) The distinction was difficult, and there was a risk of delaying demand control as described above.

In view of the above points, an object of the present invention is to provide a demand control device capable of improving the followability and stability of demand control.

[0011]

SUMMARY OF THE INVENTION The present invention provides a plurality of provisional prediction calculation means for tentatively predicting power consumption at the end of a demand time period, and a change tendency of power consumption up to the present from sampling values at past sampling times. A demand feature parameter calculating means for calculating a demand feature parameter having a steady degree, and a membership function for the demand feature parameter calculated by the means, which corresponds to a specific demand pattern in advance. , A weighting amount for each of the temporary prediction values from the plurality of temporary prediction calculation means is determined based on a predetermined fuzzy rule, and thereby the temporary prediction values are weighted to obtain a final single value. And a fuzzy calculation means for calculating a predicted value of

[0012]

The fuzzy calculating means refers to a membership function corresponding to a specific demand pattern in advance for a demand feature parameter indicating a change tendency or a steady state of the power consumption up to the present, and determines the predetermined value. Based on the fuzzy rule, a weighting amount for each of the temporary prediction values from the plurality of temporary prediction calculation means is determined, and each temporary prediction value is weighted by each weighting amount, that is, fuzzy inference is performed, and finally A single predictive value is calculated.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below based on the illustrated embodiments.

FIG. 1 is a block diagram of the demand control device 11. The demand control device 11 measures electric power used by a load (not shown) and outputs electric power with a transmitter for transmitting a pulse proportional to the measured value. Receive pulses from meter 1. This pulse is used by the predictor 3, which independently performs prediction calculation.
4, 5 (hereinafter referred to as predictors a, b, c) and a demand feature parameter extractor 2 respectively.

The predictors a to c are distinguished from each other by the characteristics such as different sampling times Δt in the above-mentioned prediction calculation formula (1) or using a moving average power value.

In short, the demand feature parameter extractor 2 is for quantitatively determining whether the power consumption is increasing or decreasing. Specifically, for Δt _{1 to} Δt ₃ shown in FIG. 2, for example, the following calculation is performed and the determination is performed.

The time t ₀ is the elapsed time from the start of the demand time period to the present time, and the electric power used during this elapsed time t ₀ is P
_When it is set to ₀ , the average power consumption P _{A0 up to} t ₀ is represented by P _A0 = (P ₀ / t ₀ ). Furthermore, section A, section B, and section C in FIG.
Letting P ₁ , P ₂ , and P ₃ be the respective power consumptions, the average power consumptions P _A1 , P _A2 ,
P _A3 can be expressed as P _A1 = (P ₁ / Δt ₁ ) P _A2 = (P ₂ / Δt ₂ ) P _A3 = (P ₃ / Δt ₃ ).

At this time, the relative variation amount δP ₁ of the electric power used at the demand time t ₁ is defined by the following equation.

ΔP ₁ = (P _A1 −P _A0 ) / P _A0 Similarly, the relative fluctuation amounts δP ₂ and δP ₃ of the used power at the demand times t ₂ and t ₃ are defined by the following equations.

ΔP ₂ = (P _A2 −P _A0 ) / P _A0 δP ₃ = (P _A3 −P _A0 ) / P _A0 Now, the demand feature parameter extraction interval ΔT (determined regardless of the sampling time Δt , I.e., load fluctuations between the demand times t ₀ and t ₃ , typical load fluctuations are as follows.

ΔP ₁ <δP ₂ <δP ₃ ... The load is gradually increasing.

ΔP ₁ = δP ₂ = δP ₃ ... The load is almost steady.

ΔP ₁ > δP ₂ > δP ₃ ... The load is gradually decreasing.

FIG. 3 shows a representative line of the relative variation amount δP in these cases.

For example, if the relative variation amount δP is always above the representative line A, the load rises fairly rapidly, and if it is near the representative line B, the load is almost steady, and the representative line C. If it is below, the load is likely to fall fairly quickly.

Considering a general case, the straight lines as shown by the representative lines A to C are not formed, and therefore handling in that case is necessary.

In this embodiment, when the fuzzy feature calculator 7 of FIG. 1 refers to the membership function held in the demand feature membership function unit 6, the relative variation amount δP is not linear. Take action.

The demand feature membership function section 6 holds a membership function as shown in, for example, FIGS. 4 (a) to 4 (c). FIG. 4 (a) shows the load at the demand time t ₁ . This is a membership function for determining increase / decrease, and FIG.
(b) and (c) are increase / decrease determination membership functions for demand times t ₂ and t ₃ , respectively.

The degree of coincidence with the representative line at the demand time t ₁ is, for example, as shown in FIG. 5, the intersection with each membership function when the relative variation amount δP ₁ is taken on the horizontal axis is read. It can be calculated by Similar calculations are performed for the relative fluctuation amounts δP ₂ and δP ₃ . Thus, for example degree of coincidence of increasing or decreasing trend of the average power with respect to the representative line A are obtained respectively as _{Cf At1, Cf At2, Cf At3} . This average value, that is, (Cf _At1 + Cf _At2 + Cf _At3 ) / 3 is taken as the average degree of coincidence with the representative line A. In addition, each of the matching degree _{Cf Bt1 ~Cf Bt3, Cf Ct1 ~Cf} Ct3
The root mean square of the deviation from the average degree of coincidence, that is, the variance is calculated.

The fuzzy feature calculator 7 uses the membership function as shown in FIG. 6 from the calculated average degree of coincidence and variance, and, for example, a rule as shown in FIG. ..) is used to determine the weighting amount of the prediction value for each of the predictors a to c.

A specific calculation procedure will be described assuming that the rule of FIG. 7 corresponds to the representative line A of FIG. 3, for example. The representative lines A, B, and C are not limited to straight lines, and may be three or more.

In this case, the average degree of coincidence in the rule is
It is the average degree of coincidence with the representative line A, and the same applies to the variance.

"Rule 1" shown in FIG. 7 will be described.

“If (“ the average degree of coincidence is high ”and“ variance is small ”) then (“ the weight of the predictor c is “1” ”and“ the weight of the predictor a is “0” ””] For the above-mentioned “high average degree of coincidence”, as shown in FIG. 8, by calculating the current average degree of coincidence and the value of the intersection of the membership functions corresponding to “high”, the degree of membership r For “small variance”, the membership level s is obtained by calculating the intersection of the current variance and the membership function corresponding to “0”, as shown in FIG. .

An and operation is performed within the "rule 1" if. That is, for example, of the belonging degree r and the belonging degree s,
The calculation (Min calculation) is performed to take the smaller value. (The smaller value is used when determining the weight of the predictor to be described later, and when the vertical fluctuation of the load is large, the weight of the predictor having a short sampling time is used. To make it smaller). Let this value be w.

Below "the rule 1" is "the weight of the predictor c is" 1 "", that is, the weight output value for the predictor c in the "rule 1" is "1
Xw = w _C1 ”. Further, since “the weight of the predictor a is“ 0 ””, “0 × w = w _a1 ” is similarly calculated for the predictor a, and this is set as the weight output value for the predictor a. . The predictor c is calculated using the function f _C (t) that changes with the passage of time within the demand time limit.
The weighted output value with respect to may be calculated as “f _C (t) × w = w _C1 ”.

Generally, since there are a plurality of rules, each rule is
By performing the above-mentioned calculation over all the parameters, for example, as weights for the predictor c, w _C1 , w _C2 , w _C3 , ... W
get _cn Of these, the largest value is determined as the weight wc for the predictor c (Max operation). Similarly, one weight corresponding to each of the other predictors a and b is determined.

The weight adder 8 of FIG. 1 associates the predicted values output from the predictors a to c with the weights determined as a result of the above calculation.

The center-of-gravity computing unit 9 outputs the respective predicted values (hereinafter, referred to as Ra, Rb, Rc) output from the predictors a to c and the respective weight values (hereinafter, wa and w).
b), and a function of outputting a single predicted value R using the following equation, for example.

R = (Ra.wa + Rb.wb + Rc.wc) / (wa + wb + wc) The above operation is performed using appropriate timing issued by the demand time counter 10.

The membership function used for fuzzy inference may change with the passage of demand time,
It may change depending on the absolute value of the average power consumption. Also, the membership function need not be triangular or trapezoidal, and need not be symmetric.

Further, the demand feature parameter extraction interval ΔT does not need to be a constant value, and as shown in FIG. 10, it moves with the passage of time with respect to the demand period (ΔT ₁ → Δ
There is no problem even if T ₂ → ΔT ₃ ). Further, it is not necessary for the lengths of the characteristic data sampling sections to form an integer ratio with each other, and the number of sections is not limited to three. Furthermore, it is not necessary to start sampling in each section at the same time. Furthermore, the sampling operations of the predictors a to c and the sampling operation of the demand feature parameter extractor 2 do not have to be synchronized.

Further, although the variance is used as the demand feature parameter in this embodiment, it may be the square root of the variance, that is, the standard deviation, or the so-called fluctuation obtained by dividing the standard deviation by the average degree of coincidence. It may be a coefficient or an average of deviations instead of variance.

Here, the meaning of rules will be described.

The predictor a has a sampling time Δt of "7 minutes" in the conventional prediction calculation formula (1), and the predictor b has a sampling time of "3 minutes", for example.
The predictor c has, for example, a sampling time of “1 minute”. Therefore, for example, the predictor a is suitable when the load is steady, and the predictor c is suitable when the load shows a sharp change.

The fact that the average degree of coincidence with the representative line A is high indicates that the average power is on the rise, and the fact that the variance is small means that
This is considered to indicate that there is little variation in the degree of increase, that is, it is not a temporary change in load but that the load is increasing reliably.

In such a situation, the sampling time Δt
It is common sense that the short prediction method is suitable, that is, the weight of the predictor c having a short sampling time Δt can be set to be close to "1" (see "Rule 1"). .

On the contrary, when the load fluctuates in a short time,
Alternatively, when it is determined that the load is stationary within the time of about the demand feature parameter extraction interval ΔT, for example, by making the weight of the predictor a having a long sampling time Δt close to “1”. , The stability of the predicted value can be improved.

When such an operation is performed by a simple binary logical judgment, the prediction value naturally differs for each prediction method of the predictor a and the predictor c, so that even if the load fluctuation is continuous, Although the predicted value does not become continuous, the present embodiment adopts a configuration in which the weighting of the predictor is continuously changed by fuzzy inference, so that the predicted value can be continuously changed. It is possible.

As described above, in the present embodiment, the short predicted value of the sampling time Δt is output under the objective and highly reliable condition that the average power continuously increases. This indicates that the prediction method according to the present embodiment is more advantageous than the conventional prediction method in determining the load cutoff / recovery of the demand control.

According to the present embodiment, when predicting the power consumption, the point that the sampling time is simply changed by the remaining time of the demand time limit by the conventional prediction calculation formula is amended, and the sampling time is set to be long. Even if it is, the fluctuations in the power value are estimated to be “sufficiently steep” in the light of past experience values. Since it is immediately reflected in the predicted value, the followability of demand control can be improved, and if the power consumption is estimated to be "almost constant," the sampling time can be changed. Since the long prediction calculation formula is used, it is possible to suppress the fluctuation of the prediction value due to the fluctuation of the pulse arrival interval and improve the stability of the demand control.

(Modification) In the present embodiment, the demand time is started within the demand time, the sampling value which is the electric power used in the sampling time obtained from the remaining time of the demand time, the number of pulses from the watt hour meter with a transmitter, and the demand time. From the current value of the power used from the time to the current time, the power used at the end of the demand time period is provisionally predicted, but the present invention is not limited to this.For example, when the freezer is a load, the temperature and It can be tentatively predicted by the capacity put inside. In addition, the current value of power consumption is assumed to be the value from the start of the demand time period to the current time, but if there is a power outage within the demand time period, the power used in the previous demand time period may be changed as necessary. You may make it use a value suitably.

[0053]

As described above, according to the present invention,
Regarding the demand feature parameters that show the tendency of change in power consumption up to the present and the degree of steady state, a plurality of them are based on a predetermined fuzzy rule by referring to the membership function that corresponds to a specific demand pattern in advance. Determines a weighting amount for each temporary prediction value from the temporary prediction calculation means, weights each temporary prediction value by each weighting amount, that is, performs fuzzy inference, and calculates a final single predicted value. is doing.

Therefore, it is possible to improve the followability and stability of demand control.

[Brief description of drawings]

FIG. 1 is a demand controller according to an embodiment of the present invention.
FIG. 3 is a block diagram showing a mobile device.

FIG. 2 is a diagram showing an example of sampling intervals for extracting demand feature parameters in an embodiment of the present invention.

FIG. 3 is a diagram showing an example of a representative line of load fluctuation for extracting demand feature parameters in one embodiment of the present invention.

FIG. 4 is a diagram showing an example of a membership function corresponding to a representative line of load fluctuations in FIG.

FIG. 5 is a diagram showing an example of the degree of coincidence with the representative line when the relative fluctuation amount of the average power is specifically obtained in the embodiment of the present invention.

FIG. 6 is a diagram showing an example of a case where the average degree of coincidence and variance are membership functions and a case where the weighting function is not a constant in one embodiment of the present invention.

FIG. 7 is a diagram showing an example of a fuzzy reasoning rule according to an embodiment of the present invention.

FIG. 8 is a diagram showing an example for calculating a specific degree of belonging from a membership function of the average degree of coincidence in one embodiment of the present invention.

FIG. 9 is a diagram showing an example for calculating a specific degree of membership from a membership function of distribution according to an embodiment of the present invention.

10 is a diagram showing an example of a case where the demand feature parameter extraction interval of FIG. 2 moves with the passage of demand time.

FIG. 11 is a diagram showing arrival intervals of power pulses within a relatively long sampling time when it is expected that there is almost no load fluctuation.

FIG. 12 is a diagram showing arrival intervals of power pulses within a relatively long sampling time when it is expected that load fluctuations are large.

[Explanation of symbols]

 1 Energy Meter with Transmitter 2 Demand Feature Parameter Extractor 3 Predictor (a) 4 Predictor (b) 5 Predictor (c) 6 Demand Feature Membership Function Part 7 Fuzzy Feature Calculator 8 Weighting Unit 9 Center of gravity calculator 10 Demand time coefficient device 11 Demand control device

Claims (2)

[Claims] 1. A remaining time counting means for instructing a remaining time of a demand time period, a plurality of temporary prediction calculation means for temporarily predicting electric power used at the end time of the demand time period, and a sampling value from a past sampling time to a present time. A demand feature parameter calculating means for calculating a change tendency of power consumption and a demand feature parameter of a steady degree, and a demand feature parameter calculated by the demand feature parameter calculating means are specified in advance. The membership function corresponding to the demand pattern is referred to, and the weighting amount for each of the temporary prediction values from the plurality of temporary prediction calculation means is determined based on a predetermined fuzzy rule. A demand control device comprising: a fuzzy calculation means for weighting the prediction values to calculate a final single prediction value. 2. The plurality of provisional prediction calculation means are a remaining time instructed by the remaining time counting means, a sampling value which is electric power used at a sampling time obtained from the number of pulses from a watt hour meter with a transmitter, and a demand time limit. 2. The demand control device according to claim 1, wherein the demand control device is means for tentatively predicting the power consumption at the end of the demand time period from the current value of the power consumption from the start of the demand time period to the current time.