KR101838393B1 - Apparatus and method for 24 hour electrical load forecasting - Google Patents

Abstract

The present invention relates to a 24-hour power demand forecasting apparatus and a power demand forecasting method using hourly temperature, and a power demand forecasting method according to an embodiment of the present invention is a method for predicting a demand for power demand, A predicted temperature reception step of receiving temperature data; A similarity extracting step of comparing the predicted temperature data with a preset temperature history database to extract a similarity corresponding to the temperature change pattern of the predicted temperature data; And a predicted power demand pattern generation unit that extracts a measured power demand pattern measured by the power demand forecasting apparatus on the similar day from a predetermined power demand database and generates a predicted power demand pattern for the target day using the measured power demand pattern, Step < / RTI >

Description

TECHNICAL FIELD [0001] The present invention relates to a 24-hour power demand forecasting apparatus and a power demand forecasting method using time-

The present invention relates to a power demand forecasting apparatus and a power demand forecasting method for a next 24 hours of the next day. The present invention searches for past power demand having a similar temperature change pattern using the predicted hourly temperature forecasts for the next day, To a power demand forecasting apparatus and a power demand predicting method capable of predicting power demand using the power demand forecasting apparatus and the power demand forecasting apparatus.

Power demand forecasting is an important factor for stable and smooth power system operation and power supply and demand planning. Because power demand forecasts are used to determine power prices or to operate power systems, errors in forecasting power demand can interfere with stable power system operation and cause massive economic losses. Therefore, various power demand forecasting techniques such as time series analysis, regression analysis, artificial neural network, and knowledge based expert system have been suggested to reduce the power demand forecast error.

However, the conventional power demand forecasting techniques are predicated only using the maximum and minimum daily temperatures, and it is difficult to sufficiently reflect the characteristics of the power demand that is sensitive to temperature. For example, when an exceptional pattern of temperature change occurs, the demand pattern of electric power can be changed according to the temperature change. However, it is difficult to accurately predict the electric power demand based on the temperature change only by the conventional prediction method. Therefore, conventionally, there has been a problem that a power demand forecast error occurs at a time when an exceptional temperature change occurs.

The present application aims to provide a power demand forecasting apparatus and a power demand forecasting method capable of predicting power demand using time-series temperature change patterns.

The power demand forecasting method according to an embodiment of the present invention includes: a predicted temperature reception step of receiving forecasted temperature data corresponding to a temperature forecast of a target day; A similarity extracting step of comparing the predicted temperature data with a preset temperature history database to extract a similarity corresponding to the temperature change pattern of the predicted temperature data; And a predicted power demand pattern generation unit that extracts a measured power demand pattern measured by the power demand forecasting apparatus on the similar day from a predetermined power demand database and generates a predicted power demand pattern for the target day using the measured power demand pattern, Step < / RTI >

The similarity extracting step may include calculating a degree of similarity between the temperature data of each measurement day and the predicted temperature data stored in the temperature history database using a predetermined similarity extraction model, The measurement date of the set number can be extracted by the similar day.

Here, the similarity extraction model

, And the degree of similarity can be measured by calculating the Euclidean distance using the similarity extraction model. ^{_{Here, ΔT max = T f max -}} T p max, ΔT min = T f min - T and ^p _min, T ^f _max is the maximum temperature, T ^p _max is the maximum temperature of the measurement days, T ^f _min of the target date is targets The minimum temperature of day, T ^p _min , may be the lowest temperature of the day of measurement. In addition, TC = 1 - R and R can be the correlation coefficient between the target day and the temperature of the day of measurement, and Δslope _m = | slope ^f _m - slope ^p _m |, Δslope _a | | slope ^f _a - slope ^p _a | Slope ^f _m is the morning temperature slope of the target day, slope ^p _m is the morning temperature slope of the measurement day, slope ^f _a is the afternoon temperature slope of the target day, and slope ^p _a is the afternoon temperature slope of the measurement day. ^{_{Also, ΔH max = H f max -}} H p max, ΔH min = H f min - H and ^p _min, H ^f _max is the maximum temperature time of occurrence of the target date, H ^p _max is a maximum temperature time of occurrence of the measurement work, H ^f _min is the minimum temperature of the target day, H ^p _min is the minimum temperature of the measurement day, T ^f _h is the hourly temperature of the target day, T ^p _h is the hourly temperature of the day of measurement, and 0 ≤ h ≤ 24, and h can be an integer. On the other _{hand, ω 11, ω 12, ω} 13, ω 14, ω 15, ω 21, ω 22, ω 23, ω 24, ω 25, ω 26, ω 27, ω 31, ω 32, ω 33, ω 34, ? ₃₅ is a weight value, and the weights may be calculated by a least squares method using a regression equation.

Wherein the predictive power demand pattern generation step comprises: an exponential smoothing process of generating an exponential smoothing demand pattern by applying an exponential smoothing method to the extracted plurality of measured power demand patterns; A normalization process of normalizing the exponential smoothing demand pattern to generate a regular demand pattern; And generating a predicted power demand pattern by receiving the maximum power demand forecast value and the minimum power demand forecast value for the target day and applying the maximum power demand forecast value and the minimum power demand forecast value to the regular demand pattern, .

Here, the exponential smoothing process

SDL ^p _h is the exponential smoothing demand value at the time _h of the target day, and SDL ⁿ _h is the power demand value at the time _h of the similarity with the nth degree of similarity, , and? can be an exponential smoothing weight.

Here, the normalization process

를 이용하여 상기 정규수요패턴을 생성하는 것으로서, PU_SDL^p _h는 목표일의 h시간에서의 정규화된 정규수요값이고, SDL^p _h는 목표일의 h시간에서의 지수평활수요값, SDL^p _max는 목표일에서의 지수평활수요값의 최대값, SDL^p _min은 목표일에서의 지수평활수요값의 최소값일 수 있다.

As with the generating the normal demand pattern, PU_SDL ^p _h is the normalized normal demand value at the h time of the target date, SDL ^p _h is exponential smoothing demand value at the h time of the target date, SDL ^p _max is The maximum value of the exponential smoothing demand value at the target date, SDL ^p _min , may be the minimum value of the exponential smoothing demand value at the target date.

Here, the pattern generation process

를 이용하여 상기 예측전력수요패턴을 생성하는 것으로,

는 목표일의 h시간에서의 예측전력수요값이고, PU_SDL^p _h는 상기 목표일의 h시간에서의 정규화된 정규수요값,

는 상기 최대전력수요예측값,

는 상기 최소전력수요예측값일 수 있다.

To generate the predicted power demand pattern,

Is a predicted power demand value at time _h of the target day, PU_SDL ^p _h is a normalized normal demand value at time _h of the target day,

The maximum power demand prediction value,

May be the minimum power demand predicted value.

An apparatus for predicting electric power demand according to an embodiment of the present invention includes a data receiving unit for receiving predicted temperature data corresponding to a target day's temperature forecast; A similarity search unit for comparing the predicted temperature data with a predetermined temperature history database to extract a similarity corresponding to the temperature change pattern of the predicted temperature data; And a demand pattern generation unit for extracting a measured power demand pattern measured on the similar day from a predetermined power demand database and generating the predicted power demand pattern of the target day using the measured power demand pattern.

In addition, the means for solving the above-mentioned problems are not all enumerating the features of the present invention. The various features of the present invention and the advantages and effects thereof will be more fully understood by reference to the following specific embodiments.

According to the electric power demand forecasting apparatus and the electric power demand forecasting method according to an embodiment of the present invention, it is possible to accurately predict the electric power demand in consideration of the temperature change. It is thus possible to perform accurate power demand forecasting even on exceptional temperature changes.

1 is a block diagram showing a power demand predicting apparatus according to an embodiment of the present invention.
FIG. 2 is a graph showing a general temperature change pattern of a meteorological day according to an embodiment of the present invention.
FIG. 3 is a graph illustrating an exceptional temperature change pattern of a meteorological day according to an embodiment of the present invention.
4 is a graph showing a power demand pattern according to an embodiment of the present invention.
5 is a graph illustrating a predicted power demand pattern according to an embodiment of the present invention.
6 to 8 are flowcharts illustrating a power demand predicting method according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, in order that those skilled in the art can easily carry out the present invention. In the following detailed description of the preferred embodiments of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. In the drawings, like reference numerals are used throughout the drawings.

In addition, in the entire specification, when a part is referred to as being 'connected' to another part, it may be referred to as 'indirectly connected' not only with 'directly connected' . Also, to "include" an element means that it may include other elements, rather than excluding other elements, unless specifically stated otherwise. Also, the terms "part,""module," and the like in the specification mean units for processing at least one function or operation, which may be implemented by hardware or software or by a combination of hardware and software.

Generally, as shown in Fig. 2, the 24-hour temperature change pattern gradually rises with sunrise after reaching the minimum temperature at dawn time, and falls again after reaching the maximum temperature around 15:00 to 16:00 do. That is, in general weather days, there is a temperature change pattern in which the lowest temperature occurs at dawn time and the highest temperature occurs in the afternoon regardless of the season.

However, there are cases where the maximum temperature occurs before or after sunrise. For example, when the highest temperature occurs at the early morning time, as shown in FIG. 3 (a) on December 16, 2014, the temperature change pattern may be continuously lowered. In addition, when the maximum temperature is generated after sunset, such as a change in the temperature of the time zone on December 14, 2014 in Fig. 3 (b), the temperature pattern may continuously rise.

On the other hand, FIG. 4 shows the 24-hour power demand and temperature on December 16, 2014, when the downward temperature pattern occurred and the surrounding days. On Dec. 16, 2014, there was a frostbite in the whole country, resulting in mostly cloudy or snowy areas. On the other hand, on December 14, 2014, there was a southerly low pressure and the weather was warmer than other days. Both measurement days belong to the early winter, but occur when unusual climates occur and the temperature is continuously falling and rising, and this change in temperature has a great impact on the hourly power demand pattern.

Specifically, referring to FIG. 4, on December 16, 2014, when the downward temperature pattern occurred, the peak power demand occurred at 18:00. However, with the exception of this date, the peak power demand for December 15, 2014, December 17, 2014 through December 19, 2014, occurred at 10:00 and 11:00. In other words, the electricity demand pattern on December 16, 2014 was the increase in afternoon power demand due to the continuous decrease in temperature. In addition, after December 16, 2014, there is a general pattern of the temperature again, and it can be seen that the electric power demand pattern returns to the form in which the maximum electric power demand occurs in the morning. Therefore, if the electricity demand on December 16, 2014 is predicted in the same manner as before, the power demand due to the exceptional temperature pattern can not be accurately predicted, so that the prediction error may increase greatly. In order to cope with such an exceptional temperature change, a power demand prediction method considering a time-varying temperature change pattern is required.

1 is a block diagram showing a power demand predicting apparatus according to an embodiment of the present invention.

Referring to FIG. 1, the power demand predicting apparatus 100 according to an embodiment of the present invention may include a data receiving unit 110, a similarity searching unit 120, and a demand pattern generating unit 130.

Hereinafter, a power demand predicting apparatus according to an embodiment of the present invention will be described with reference to FIG.

The data receiving unit 110 may receive the predicted temperature data corresponding to the target temperature for forecasting the power demand. Here, the predicted temperature data may be a 24-hour temperature forecast of the target day received from the meteorological office. Depending on the embodiment, the 3-hour unit temperature forecast received from the Korea Meteorological Administration may be processed by 24-hour temperature forecasting using an interpolation method or the like have. The data receiving unit 110 may be connected to the meteorological office through wired or wireless data communication, and may receive the predicted temperature data using an Open API provided by the meteorological office.

The similarity search unit 120 may compare the predicted temperature data with the predetermined temperature history data base 200. [ Here, in the temperature history history database 200, the result of measuring the 24-hour temperature change of the measurement date measured in the past can be stored. For example, the 24-hour temperature graph of the past Seoul area shown in FIG. 2 to FIG. 4 may be included in the temperature history history database 200. Accordingly, the similarity search unit 120 can extract a measurement date having a temperature change pattern similar to the temperature change pattern of the predicted temperature data through comparison with the temperature history database 200, As shown in Fig.

Specifically, the similarity search unit 20 can extract a similarity having a similar temperature change pattern using a predetermined similarity extraction model. That is, the degree of similarity between the respective temperature data of each measurement day stored in the temperature history database 200 and the predicted temperature data can be calculated, and a predetermined number of measurement days can be extracted in a similar sequence according to the order of high degree of similarity. At this time, to compute the similarity of the temperature change patterns, eight temperature elements of the following table can be utilized.

Temperature element Contents

Highest temperature difference between target and measurement day

Minimum temperature difference between target and measurement day

1, the value obtained by subtracting the correlation coefficient (R) between the target day and the temperature of the measurement day

The morning temperature slope of the target day and the morning temperature slope difference of the measurement day

The afternoon temperature slope of the target day and the afternoon temperature slope difference of the measurement day

Temperature deviation by target time and measurement day

Difference in peak time between target and measurement day

The difference between the target temperature and the measurement temperature

Where T ^f _max is the maximum temperature of the target day, T ^p _max is the maximum temperature of the measurement day, T ^f _min is the minimum temperature of the target day, T ^p _min is the minimum temperature of the measurement day, R is the target temperature, . The slope ^f _m is the slope of the morning temperature at the target day, and the slope ^p _m is the morning temperature of the day of the measurement (in milliseconds) Slope ^f _a is the afternoon temperature slope of the target day, and slope ^p _a is the afternoon temperature slope of the measurement day. T ^f _h is the hourly temperature of the target day, T ^p _h is the hourly temperature of the day of measurement, h is an integer satisfying 0 ≤ h ≤ 24, H ^f _max is the maximum temperature occurrence time of the target day, H ^p _max is the maximum temperature occurrence time of the measurement day, H ^f _min is the minimum temperature occurrence time of the target day, and H ^p _min is the minimum temperature occurrence time of the measurement day.

The similarity search unit 120 can calculate the similarity degree using the similarity extraction model, and in some embodiments, the Euclidean distance model can be used as a similarity extraction model. At this time, the similarity extraction model can be generated using the eight temperature elements, and various kinds of models can be generated according to the temperature element to be used.

Specifically, the similarity search unit 120 can extract a similarity using any one of the following four similarity extraction models.

Here, EDM1 to EDM3 are models to which a weighted Euclidean distance is applied, and EDM4 is a model to which a general Euclidean model is applied. That is, the result value calculated in each equation is the Euclidean distance, and the smaller the Euclidean distance, the higher the degree of similarity. Here, ω _ij is the weight of the jth temperature element of the i th model.

On the other hand, since power demand is sensitive to changes in temperature, similarity searches can be performed based on the same season. Therefore, the last one month (30 days) based on the target date can be set to the same season and climatic condition as the target date. It is also possible to extend the population size for similarity searches by including ± 30 days on the same date as the past one year and ± 30 days as the same date on the same two years as the past one year. For example, if you set Jan. 21, 2010 as the target date, you will be required to sign a contract between January 19, 2010 and December 22, 2009, from February 20, 2009 to December 22, 2008, And the similarity search period may be set as the similarity search interval from the 20th of the month to the December 22nd of 2007.

In addition, the weighting factor ω _ij of the temperature element can be calculated by applying the least squares method after constructing a regression equation for each model using historical data. Here, the regression formula for weight calculation and the application of the least squares method are well-known contents, so a detailed description of the application method is omitted.

The similarity extracting unit 120 may calculate the Euclidean distance between the temperature change pattern of the measurement day included in the predetermined similar day search period and the predicted temperature data of the target day by applying any one of the similarity extraction models . Thereafter, in accordance with the order in which the Euclidean distance between the target date and the target date is closest to the measurement date, a predetermined number of extraction days can be extracted in a similar manner. According to the embodiment, it is also possible to extract three measurement days having the shortest Euclidean distance as the similar days.

The demand pattern generation unit 130 can extract the measured power demand pattern measured on the similar day from the power demand database 300 and generate the predicted power demand pattern of the target day using the measured power demand pattern have. Here, the power demand database 300 may store time-specific power demand patterns measured in the past measurement date, and the power demand database 300 may be provided from the Korea Power Exchange.

First, the demand pattern generation unit 130 may generate an exponential smoothing demand pattern by applying an exponential smoothing method to a plurality of extracted measured power demand patterns. The following equation represents the exponential smoothing demand pattern when three similar jobs are extracted.

Where SDL ^p _h is the power demand at time _h of the target day, SDL ⁿ _h is the power demand at time h with the nth highest degree of similarity, and α means the exponential smoothing weight.

Thereafter, the demand pattern generation unit 130 generates a normal demand pattern by normalizing the exponential smoothing demand pattern, and generates a predicted power demand pattern by applying the maximum power demand forecast value and the minimum power demand forecast value to the regular demand pattern .

Specifically, the demand pattern generation unit 130

를 이용하여 정규수요패턴을 생성할 수 있으며, 여기서 PU_SDL^p _h는 목표일의 h시간에서의 정규화된 정규수요값, SDL^p _max는 목표일에서의 지수평활수요값의 최대값, SDL^p _min은 목표일에서의 지수평활수요값의 최소값이다.

, Where PU_SDL ^p _h is the normalized normal demand value at time _h of the target day, SDL ^p _max is the maximum value of the exponential smoothing demand value at the target day, and SDL ^p _min is the maximum demand value at the target day, Is the minimum value of the exponential smoothing demand value at the target date.

를 이용하여 최종적으로 예측전력수요패턴을 생성할 수 있다. 여기서,

는 목표일의 h시간에서의 예측전력수요값이고,

는 최대전력수요예측값,

는 최소전력수요예측값이다. 여기서, 최대전력수요예측값 및 최소전력수요예측값은 종래의 방식에 의하여 계산하는 것으로서, 상기 최대전력수요예측값 및 최소전력수요예측값은 전력수요예측장치가 직접 계산하여 구하거나, 또는 외부로부터 제공받을 수 있다.
Thereafter, the demand pattern generation unit 130

The predicted power demand pattern can be finally generated. here,

Is the predicted power demand value at time h of the target day,

Is the maximum power demand forecast value,

Is the minimum power demand forecast. Here, the maximum power demand forecast value and the minimum power demand forecast value are calculated by a conventional method, and the maximum demand forecast value and the minimum power demand forecast value can be directly calculated or obtained from the power demand forecasting apparatus or can be provided from the outside .

5 is a graph showing a result of performing power demand prediction using the electric power demand forecasting apparatus according to an embodiment of the present invention. Specifically, in the case of FIG. 5 (a), the electric power demand prediction result in the descending temperature pattern in which the temperature gradually falls, and FIG. 5 (b) shows the electric power demand prediction result in the rising temperature pattern in which the temperature gradually increases. As shown in FIG. 5, when the prediction technique of the present invention is utilized, it can be seen that the prediction error can be improved as compared with the conventional technique.

Also, referring to FIG. 5, it can be seen that for each temperature pattern, the similarity extraction model in which the maximum error occurs the smallest is different from the similarity extraction model in which the average error occurs the least. Therefore, according to the embodiment, it is also possible to minimize the prediction error by selectively applying the similarity extraction model according to each air temperature pattern, or depending on the type of error to be improved.

6 and 7 are flowcharts illustrating a power demand forecasting method according to an embodiment of the present invention.

Referring to FIGS. 6 and 7, the method of predicting power demand according to an embodiment of the present invention includes a predicted temperature receiving step S100, a similarity extracting step S200, and a predicted power demand pattern generating step S300 And the predicted power demand pattern generation step S300 may include an exponential peace process S310, a normalization process S320, and a pattern generation process S330.

Hereinafter, a power demand forecasting method according to an embodiment of the present invention will be described with reference to FIGS. 6 and 7. FIG.

In the predicted temperature receiving step (S100), the power demand forecasting apparatus can receive the predicted temperature data corresponding to the target day's temperature forecast. Here, the predicted temperature data may be a 24-hour temperature forecast of the target day received from the meteorological office. Depending on the embodiment, the 3-hour unit temperature forecast received from the Korea Meteorological Administration may be processed by 24-hour temperature forecasting using an interpolation method or the like have. The power demand forecasting device can be connected to the Korea Meteorological Agency through wired or wireless data communication, and can receive the predicted temperature data by using the Open API provided by the Korea Meteorological Administration.

In the similarity extraction step S200, the power demand prediction apparatus can compare the predicted temperature data with a predetermined temperature history data base. Here, the temperature history database may be the result of measuring the 24-hour temperature change of the measurement date measured in the past. Accordingly, in the similarity extraction step (S200), the similarity can be extracted from the measurement date corresponding to the temperature change pattern of the predicted temperature data through comparison with the temperature history database.

Specifically, in the similarity extraction step (S200), a similarity having a similar temperature change pattern can be extracted using a predetermined similarity extraction model. That is, the degree of similarity between the temperature data of each measurement day stored in the temperature history database and the predicted temperature data can be calculated, and a predetermined number of measurement days can be extracted in a similar sequence according to the order of high degree of similarity. At this time, a plurality of temperature elements can be utilized to calculate the degree of similarity of the temperature change patterns. Since the air temperature element has been described above, detailed description is omitted here.

According to an embodiment, a similarity extraction model for calculating the similarity using the Euclidean distance model in the similarity extraction step (S200) can be applied, and various similarity extraction models can be applied according to the temperature element to be used. Specifically, in the similarity extraction step (S200), the similarity can be extracted using the following four similarity extraction models.

Here, EDM1 to EDM3 are models to which a weighted Euclidean distance is applied, and EDM4 is a model to which a general Euclidean model is applied. That is, the result value calculated in each equation is the Euclidean distance, and the smaller the Euclidean distance, the higher the degree of similarity. Here, ω _ij represents the weight of the jth temperature element of the i th model.

The similarity search can be performed based on the same season, and the last one month (30 days) based on the target date can be set to the same season and climatic conditions as the target date. Here, the population size for the similarity search can be extended by including ± 30 days on the same date as the past one year, and ± 30 days as the similar date search interval on the same day as the past two years . The weighting factor ω _ij of the temperature element can be calculated by applying the least squares method after constructing a regression equation for each model using historical data.

On the other hand, in the similarity extraction step (S200), one of the similarity extraction models may be applied to calculate the Euclidean distance between the temperature change pattern of the measurement day included in the predetermined similar day search interval and the predicted temperature data of the target day have. Thereafter, in accordance with the order in which the Euclidean distance between the target date and the target date is closest to the measurement date, a predetermined number of extraction days can be extracted in a similar manner. According to the embodiment, it is possible to extract three measurement dates which are the closest to the calculated Euclidean distance by the similar days.

In the predicted power demand pattern generation step S300, a measured power demand pattern measured on a similar day is extracted from a predetermined power demand database, and a predicted power demand pattern of a target day is generated using the measured power demand pattern .

7, the predicted power demand pattern generation step S300 may include an exponential smoothing process S310, a normalization process S320, and a pattern generation process S330.

를 이용하여 상기 지수평활수요패턴을 생성할 수 있다. 여기서, SDL^p _h는 목표일의 h시간에서의 지수평활수요값이고, SDLⁿ _h은 유사도가 n번째로 높은 유사일의 h시간에서의 전력수요값, α는 지수평활 가중치일 수 있다. In the exponential smoothing process (S310), an exponential smoothing demand pattern can be generated by applying exponential smoothing to the extracted plurality of measured power demand patterns. At this time,

Can be used to generate the exponential smoothing demand pattern. Here, SDL ^p _h is the exponential smoothing demand value at the time _h of the target day, SDL ⁿ _h is the power demand value at the time _h of the similarity with the nth highest similarity degree, and?

를 이용하여 상기 정규수요패턴을 생성할 수 있다. 여기서, PU_SDL^p _h는 목표일의 h시간에서의 정규화된 정규수요값이고, SDL^p _h는 목표일의 h시간에서의 지수평활수요값, SDL^p _max는 목표일에서의 지수평활수요값의 최대값, SDL^p _min은 목표일에서의 지수평활수요값의 최소값일 수 있다. In the normalization process (S320), the exponential smoothing demand pattern is normalized to generate a normal demand pattern. In the normalization process

The normal demand pattern can be generated. Here, PU_SDL ^p _h is the normalized normal demand value at the h time of the target date, SDL ^p _h is exponential smoothing demand value at the h time of the target date, SDL ^p _max is the maximum of exponential smoothing demand value at the target day Value, SDL ^p _min may be the minimum value of the exponential smoothing demand value at the target date.

Also, in the pattern generation process (S330), after receiving the maximum power demand forecast value and the minimum power demand forecast value for the target day, the maximum power demand forecast value and the minimum power demand forecast value are applied to the regular demand pattern, A power demand pattern can be generated.

를 이용하여 생성할 수 있다. 여기서,

는 목표일의 h시간에서의 예측전력수요값이고, PU_SDL^p _h는 상기 목표일의 h시간에서의 정규화된 정규수요값,

는 상기 최대전력수요예측값,

는 상기 최소전력수요예측값일 수 있다. 최대전력수요예측값과 최소전력수요예측값은 종래의 방식에 의하여 계산하는 것으로서, 상기 최대전력수요예측값과 최소전력수요예측값은 전력수요예측장치가 직접 계산하여 구하거나, 외부로부터 제공받을 수 있다.
Specifically, the predicted power demand pattern

. ≪ / RTI > here,

Is a predicted power demand value at time _h of the target day, PU_SDL ^p _h is a normalized normal demand value at time _h of the target day,

The maximum power demand prediction value,

May be the minimum power demand predicted value. The maximum power demand forecast value and the minimum power demand forecast value are calculated by a conventional method. The maximum demand forecast value and the minimum power demand forecast value can be directly calculated or obtained from the power demand prediction apparatus, or can be provided from outside.

Meanwhile, the power demand forecasting method according to an embodiment of the present invention can be performed according to the flowchart of FIG.

Referring to FIG. 8, the user can extract power demand patterns corresponding to the 24-hour temperature change of the days measured in the past (S1). That is, information on the daily temperature change patterns and power demand patterns shown in FIG. 4 can be extracted from a predetermined database, a meteorological office, a power exchange, and the like.

Thereafter, the user can input a target date for forecasting electric power demand, and the temperature forecast for the target day can be collected (S2). Here, the temperature forecast is information that predicts the 24-hour temperature change of the target day, and can be collected in such a manner as receiving from the weather station.

The user can set a similarity extraction model to extract a similarity having a temperature change pattern similar to the temperature forecast of the target day (S3). Specifically, any one of EDM1, EDM2, EDM3, and EDM4 described above can be set as the similarity extraction model. However, depending on the embodiment, it is possible to set in advance such that different similarity extraction models are selected according to various situations or conditions. For example, EDM3 may be used when the temperature forecast of the target day includes the rising pattern, EDM4 may be used when the falling pattern is included, or a predetermined similarity may be set depending on the season You can set the work extraction model to be selected.

After the target date is determined, the search range of the past date can be set to extract the date having the temperature change pattern similar to the target date (S4). For example, based on the target date, the last month (30 days) can be regarded as the same season and climatic condition as the target date, and the search range can be set. However, in this case, it may be difficult to extract a date sufficiently close to the target date because the search range is narrow. Therefore, you can extend the population size for similarity searches by including ± 30 days on the same day in the past year and ± 30 days on the same day in the past two years in the similarity search interval. .

When the similarity extraction model and the search range are determined, it is possible to compare the similarity of the temperature change pattern of past days included in the search range with the temperature forecast of the target day. Specifically, the Euclidean distance for the temperature change pattern of the target date and the past date can be calculated using the similarity extraction model (S5). The closer the Euclidean distance is, the higher the degree of similarity. Therefore, it is possible to compare the Euclidean distances, and to measure the temperature change of the target day and the degree of similarity to the temperature change of each past day.

Thereafter, as a result of the calculation of the Euclidean distance, the upper three dates near the distance can be extracted as similar days (S6). That is, three similar days having a temperature change pattern most similar to the target day's temperature forecast among a plurality of past days included in the search range can be extracted. Here, extraction of the top three similar works is exemplified, but it is obvious that the number of similar works extracted according to the embodiment can be varied.

When three similar jobs are extracted, the 24-hour power demand pattern of the target day can be calculated using the power demand pattern according to the 24-hour temperature change of the extracted similar jobs (S7). Since demand for power is sensitive to changes in temperature, the 24-hour power demand pattern of similar days with a pattern of temperature changes similar to the target day is very similar to the 24-hour power demand pattern of the actual target day. Therefore, a 24-hour power demand pattern with high accuracy can be obtained by calculating the 24-hour power demand pattern of the target day by applying the exponential smoothing method to the 24-hour power demand pattern of three similar days. However, according to the embodiment, various techniques other than the exponential smoothing method may be utilized.

On the other hand, after the 24-hour power demand pattern is calculated for the target day, it can be used to predict the 24-hour power demand of the target day, and the prediction result can be utilized in various aspects (S8).

The present invention is not limited to the above-described embodiments and the accompanying drawings. It will be apparent to those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

100: Power demand prediction device 110: Data reception unit
120: similarity searching unit 130: demand pattern generating unit
200: temperature history database 300: electric power demand database
S100: Estimated temperature reception step S200: Pseudo-day extraction step
S300: Predicted power demand pattern generation step S310: Exponential smoothing process
S320: Normalization process S330: Pattern generation process

Claims (11)

delete delete A predicted temperature receiving step of receiving the predicted temperature data corresponding to a target temperature of the target day;
The power demand predicting device may include a similarity extracting step of comparing the predicted temperature data with a predetermined temperature history database to extract a similarity corresponding to a temperature change pattern of the predicted temperature data; And
Wherein the power demand predicting device is configured to calculate a predicted power demand pattern that extracts a measured power demand pattern measured on the similar day from a predetermined power demand database and generates the predicted power demand pattern of the target day using the measured power demand pattern Generating step;
Lt; / RTI >
Wherein the similarity extracting step comprises:
Calculating a degree of similarity between each of the temperature data of each measurement date stored in the temperature history database and the predicted temperature data by using a predetermined similarity extraction model, Extract,
In the similarity extraction model,

To calculate the Euclidean distance and to measure the similarity,
^{_{ΔT max = T f max - T}} p max, ΔT min = T f min - T and ^p _min, T ^f _max is the maximum temperature of the target date, T ^p _max is a maximum temperature, T ^f _min of the measurement day is the target day the lowest temperature, T ^p _min is a minimum temperature, TC = 1 of the measurement day - and R, R is the correlation coefficient, Δslope _m = between one target routine measurement temperature ^{_{| slope f m - slope p m}} |, Δslope a = | slope ^f _a - slope ^p _a | a, slope ^f _m am a temperature gradient of the target date, slope ^p _m am a temperature gradient of the measurement date, slope ^f _a is a measure afternoon temperatures slope, slope ^p _a target one Wherein the afternoon temperature slope, ω ₁₁ , ω ₁₂ , ω ₁₃ , ω ₁₄ , and ω ₁₅ are weights, and the weights are calculated by a least squares method using a regression equation. A predicted temperature receiving step of receiving the predicted temperature data corresponding to a target temperature of the target day;
The power demand predicting device may include a similarity extracting step of comparing the predicted temperature data with a predetermined temperature history database to extract a similarity corresponding to a temperature change pattern of the predicted temperature data; And
Wherein the power demand predicting device is configured to calculate a predicted power demand pattern that extracts a measured power demand pattern measured on the similar day from a predetermined power demand database and generates the predicted power demand pattern of the target day using the measured power demand pattern Generating step;
Lt; / RTI >
Wherein the similarity extracting step comprises:
Calculating a degree of similarity between each of the temperature data of each measurement date stored in the temperature history database and the predicted temperature data by using a predetermined similarity extraction model, Extract,
In the similarity extraction model,

To calculate the Euclidean distance and to measure the similarity,
^{_{ΔT max = T f max - T}} p max, ΔT min = T f min - T and ^p _min, T ^f _max is the maximum temperature of the target date, T ^p _max is a maximum temperature, T ^f _min of the measurement day is the target day the lowest temperature, T ^p _min is a minimum temperature, TC = 1 of the measurement day - and R, R is the correlation coefficient, Δslope _m = between one target routine measurement temperature ^{_{| slope f m - slope p m}} |, Δslope a = | slope ^f _a - slope ^p _a | a, slope ^f _m am a temperature gradient of the target date, slope ^p _m am a temperature gradient of the measurement date, slope ^f _a is a measure afternoon temperatures slope, slope ^p _a target one PM temperature ^{_{gradient, ΔH max = H f max -}} H p max, ΔH min = H f min - and H ^p _min, H ^f _max is the maximum temperature time of occurrence of the target date, H ^p _max is generated the maximum temperature of the measurement work time , a H ^f _min is the minimum temperature time of occurrence of the target date, H ^p _min is generated with minimum temperature of the measured one _{hour, ω 21, ω 22, ω} 23, ω 24, ω 25, ω 26, ω 27 is the weight, The weights are calculated by applying a regression equation The electricity demand forecasting method to calculate a multiplication. A predicted temperature receiving step of receiving the predicted temperature data corresponding to a target temperature of the target day;
The power demand predicting device may include a similarity extracting step of comparing the predicted temperature data with a predetermined temperature history database to extract a similarity corresponding to a temperature change pattern of the predicted temperature data; And
Wherein the power demand predicting device is configured to calculate a predicted power demand pattern that extracts a measured power demand pattern measured on the similar day from a predetermined power demand database and generates the predicted power demand pattern of the target day using the measured power demand pattern Generating step;
Lt; / RTI >
Wherein the similarity extracting step comprises:
Calculating a degree of similarity between each of the temperature data of each measurement date stored in the temperature history database and the predicted temperature data by using a predetermined similarity extraction model, Extract,
In the similarity extraction model,

To calculate the Euclidean distance and to measure the similarity,
^{_{ΔT max = T f max - T}} p max, ΔT min = T f min - T and ^p _min, T ^f _max is the maximum temperature of the target date, T ^p _max is a maximum temperature, T ^f _min of the measurement day is the target day the lowest temperature, T ^p _min is the measurement date, minimum temperature, TC = 1 - R, R is the correlation coefficient between the work target work and the measurement temperature, and ΔH _max = H ^f _max - H ^p _max, ΔH _min = H ^f _min - and H ^p _min, H ^f _max is the maximum temperature time of occurrence of the target date, H ^p _max is the maximum temperature of the measurement work time of occurrence, H ^f _min is the lowest temperature time of occurrence of the target date, H ^p _min is the minimum of the measurement days the temperature time _{occurrence, ω 31, ω 32, ω} 33, ω 34, ω 35 is a weight, the weight has a power demand prediction method for calculating the least squares method by applying the respective regression equation. A predicted temperature receiving step of receiving the predicted temperature data corresponding to a target temperature of the target day;
The power demand predicting device may include a similarity extracting step of comparing the predicted temperature data with a predetermined temperature history database to extract a similarity corresponding to a temperature change pattern of the predicted temperature data; And
Wherein the power demand predicting device is configured to calculate a predicted power demand pattern that extracts a measured power demand pattern measured on the similar day from a predetermined power demand database and generates the predicted power demand pattern of the target day using the measured power demand pattern Generating step;
Lt; / RTI >
Wherein the similarity extracting step comprises:
Calculating a degree of similarity between each of the temperature data of each measurement date stored in the temperature history database and the predicted temperature data by using a predetermined similarity extraction model, Extract,
In the similarity extraction model,

, Where T ^f _h is the hourly temperature of the target day, T ^p _h is the hourly temperature of the day of measurement, 0 ≤ h ≤ 24, h is an integer satisfying Of electric power demand forecasting method. delete A predicted temperature receiving step of receiving the predicted temperature data corresponding to a target temperature of the target day;
The power demand predicting device may include a similarity extracting step of comparing the predicted temperature data with a predetermined temperature history database to extract a similarity corresponding to a temperature change pattern of the predicted temperature data; And
Wherein the power demand predicting device is configured to calculate a predicted power demand pattern that extracts a measured power demand pattern measured on the similar day from a predetermined power demand database and generates the predicted power demand pattern of the target day using the measured power demand pattern Generating step;
Lt; / RTI >
Wherein the similarity extracting step comprises:
Calculating a degree of similarity between each of the temperature data of each measurement date stored in the temperature history database and the predicted temperature data by using a predetermined similarity extraction model, Extract,
Wherein the predicted power demand pattern generation step comprises:
An exponential smoothing process for generating an exponential smoothing demand pattern by applying exponential smoothing to the extracted plurality of measured power demand patterns;
A normalization process of normalizing the exponential smoothing demand pattern to generate a normal demand pattern; And
Generating a predicted power demand pattern by receiving the maximum power demand forecast value and the minimum power demand forecast value for the target day and applying the maximum power demand forecast value and the minimum power demand forecast value to the regular demand pattern;
/ RTI >
The exponential smoothing process includes:

SDL ^p _h is the exponential smoothing demand value at the time _h of the target day, and SDL ⁿ _h is the power demand value at the time _h of the similarity with the nth degree of similarity, , and α is exponential smoothing weight. 9. The method of claim 8, wherein the normalization process comprises:

As with the generating the normal demand pattern, PU_SDL ^p _h is the normalized normal demand value at the h time of the target date, SDL ^p _h is exponential smoothing demand value at the h time of the target date, SDL ^p _max is The maximum value of the exponential smoothing demand value at the target day, and SDL ^p _min is the minimum value of the exponential smoothing demand value at the target day. 10. The method of claim 9,

To generate the predicted power demand pattern,

Is a predicted power demand value at time _h of the target day, PU_SDL ^p _h is a normalized normal demand value at time _h of the target day,

The maximum power demand prediction value,

Is the minimum power demand forecast value.

delete

Images