KR20200129343A - Weather data processing device for energy management and energy management system - Google Patents

Abstract

The present invention relates to a weather data processing apparatus capable of providing weather data, which is used by an energy management apparatus to predict energy demands, on the day before a prediction target day while increasing its accuracy, and an energy management system using the same. According to one aspect of the present invention, the energy management system includes: a weather data processing apparatus creating a first weather data file by using predicted weather data (D day/D+1 day predicted weather data) of the D+1 day which is forecasted on the D day to predict energy consumption of the D+1 day, creating a second weather data file by modifying the first weather data file through machine learning, and providing the second weather data file to an energy management apparatus; and the energy management apparatus predicting energy consumption of the D+1 day in a district by using the second weather data file provided from the weather data processing apparatus, and managing energy of the district based on the predicted energy consumption.

Description

Weather data processing device and energy management system for energy management {WEATHER DATA PROCESSING DEVICE FOR ENERGY MANAGEMENT AND ENERGY MANAGEMENT SYSTEM}

The present invention relates to a meteorological data processing apparatus and energy management system for energy management.

Recently, attempts to use weather information to manage energy of buildings or areas are increasing. This is because the energy consumption of the area, etc. reacts sensitively to weather conditions. When an energy management device that manages energy such as a zone can accurately predict the amount of energy demand in a zone it is in charge of using weather information, the energy management device can efficiently manage the energy of the zone.

However, current attempts to predict energy demand using meteorological information are mostly focused on using weather data that is forecast within 3 hours of the target time point.

Predicted weather data forecast within 3 hours has the advantage of providing relatively accurate weather conditions, but the energy management device has a problem that the timing of forecasting energy demand through weather information is too late.

In recent energy management, the target area is divided into multiple levels (e.g., community-building-area, etc.) and energy is generated through demand-response interactions between the system, upper energy management device and lower energy management device. While trading, attempts are being made to maximize their own profits. Therefore, since an energy management device that intends to participate in the energy trading market through a demand response needs to predict energy demand at least one day in advance, there is a problem that the weather data provided within 3 hours is too late in time and thus its utilization is low.

In addition, when energy demand changes rapidly due to a sudden change in weather conditions (eg, a sudden rise or fall in temperature), it may be difficult to efficiently manage energy unless the energy management device predicts it in advance.

As described above, in order to efficiently manage energy using meteorological data, it is necessary to predict energy demand using meteorological data at an earlier time.

The present invention can provide a meteorological data processing device and an energy management system using the same, which can increase accuracy while providing meteorological data used by an energy management device to predict energy demand on the day before a target date.

One aspect of the present invention for achieving the above-described object, in order to predict the energy consumption of the D + 1 day, the predicted weather data of the D + 1 day (D day / D + 1 day predicted weather data) Weather data processing by generating a first weather data file using machine learning, generating a second weather data file modified from the first weather data file using machine learning, and providing the second weather data file to an energy management device Device; And an energy management device for predicting an energy consumption amount of D+1 in the area using the second weather data file provided from the meteorological data processing device, and managing energy in the area based on the estimated energy consumption amount. It is an energy management system that includes.

In the energy management system, the second weather data file may be at least one of temperature, humidity, wind speed, wind direction, and solar radiation in the first weather data file modified by the machine learning.

In the energy management system, the meteorological data processing device comprises: D-1 day/D day predicted weather data and D day actual weather data (D-day actual weather data ), and the machine learning corrects the predicted weather data of day D/D+1 by using the predicted weather data of day D-1/day D and the actual weather data of day D, +1 day corrected predicted weather data may be generated, and the weather data processing apparatus may generate the second weather data file using the D/D+1 corrected predicted weather data generated by the machine learning. .

In the energy management system, the machine learning is performed by inputting D-1 day/D day predicted weather data, D day actual weather data, and D day/D+1 day predicted weather data as inputs. One corrected predicted weather data may be output, but may be learned to reduce an error between the D/D+1 corrected predicted weather data and the D+1 actual weather data based on past weather data.

In the energy management system, the machine learning is performed by inputting an error between the predicted weather data of day D-1/day D and the actual weather data of day D, and the predicted weather data of day D/D+1 as inputs. D+1 day corrected predicted weather data may be output, but it may be learned to reduce an error between the D day/D+1 day corrected predicted weather data and D+1 day actual weather data based on past weather data.

In the energy management system, the machine learning may be learned to reduce errors in temperature, humidity, wind speed, wind direction, and insolation of the corrected predicted weather data on day D/D+1.

In the energy management system, the machine learning may be learned so that a weight is assigned to each of the temperature, humidity, wind speed, wind direction, and insolation, and the sum of errors considering the weight is minimized.

In the energy management system, the first weather data file updates the original weather data file (WDF) by extracting temperature, humidity, wind direction, and wind speed from the predicted weather data on day D/D+1 and calculating solar radiation. Can be created by

Another aspect of the present invention is a meteorological data processing device that provides a meteorological data file for use in predicting the energy consumption of D+1 day in the area by the energy management device for managing energy in the area, D-1 Predicted weather data for day D (D-1 day/D day predicted weather data), actual weather data for day D, and predicted weather data for day D+1 (day D/D+1 forecast) A weather information collection unit that collects weather data); Using the error information of the D-1/D day predicted weather data and the D day predicted weather data, D-day/D+1-day corrected predicted weather data are modified. Machine learning to generate; And a meteorological data file generation unit that generates the meteorological data file to be provided to the energy management device by using the D-day/D+1-day modified predicted weather data generated by the machine learning.

In the meteorological data processing apparatus, the machine learning is performed by inputting the predicted weather data of the D-1/D day, the actual weather data of the D, and the predicted weather data of the D/D+1 as inputs. /D+1 day corrected predicted weather data is output, but it may be learned to reduce an error between the D day/D+1 day corrected predicted weather data and D+1 day actual weather data based on past weather data.

In the meteorological data processing apparatus, the machine learning inputs an error between the predicted weather data of the D-1/D day and the actual weather data of the D, and the predicted weather data of the D/D+1 as inputs. It outputs the corrected predicted weather data for D/D+1, but learned to reduce the error between the corrected predicted weather data for D/D+1 and the actual weather data for D+1 based on past weather data. I can.

In the meteorological data processing apparatus, the first weather data file is updated by extracting temperature, humidity, wind direction, and wind speed from the predicted weather data on day D/D+1, calculating solar radiation, and updating the original weather data file (WDF). Is generated, and the weather data file is modified at least one of temperature, humidity, wind speed, wind direction, and solar radiation in the first weather data file based on the D-day/D+1-day modified predicted weather data output by the machine learning. Can be created.

Another aspect of the present invention is a meteorological data performed by a meteorological data processing device that provides a meteorological data file for use by an energy management device that manages energy in an area to predict the energy consumption of D+1 in the area. In the processing method, the meteorological data processing device is the forecasted weather data of the D day (D-1 day/D day predicted weather data), the actual weather data of the D day, and the predicted day D+1 of the day D. Collecting predicted weather data (D day/D+1 day predicted weather data); The weather data processing device causes machine learning to modify the D-day/D+1-day predicted weather data by using the error information of the D-1 day/D day predicted weather data and the D day's actual weather data. Instructing to generate corrected predicted weather data for +1 days; And generating, by the meteorological data processing device, the meteorological data file to be provided to the energy management device by using the D-day/D+1-day modified predicted weather data.

In the meteorological data processing method, the first weather data file is updated by extracting temperature, humidity, wind direction, and wind speed from the predicted weather data on day D/D+1, calculating solar radiation, and updating the original weather data file (WDF). Is generated, and the weather data file is modified at least one of temperature, humidity, wind speed, wind direction, and solar radiation in the first weather data file based on the D-day/D+1-day modified predicted weather data output by the machine learning. Can be created.

In the meteorological data processing method, the extracting of temperature, humidity, wind direction and wind speed from the D/D+1 predicted weather data comprises: converting the D/D+1 predicted weather data to XML; Extracting weather data at 3-hour intervals from the XML data; Generating weather data at 1 hour intervals from the weather data at 3 hour intervals using an interpolation method; And extracting temperature, humidity, wind direction, and wind speed from the weather data at an hourly interval.

In the meteorological data processing method, the meteorological data file may modify at least one of temperature, humidity, wind speed, wind direction, and solar radiation from the first weather data file by the machine learning.

In the meteorological data processing method, the machine learning is performed by inputting D-1 day/D day predicted weather data, D day actual weather data, and D day/D+1 day predicted weather data as inputs. The corrected predicted weather data for the first day is output, but may be learned to reduce an error between the corrected predicted weather data for the D/D+1 and the actual weather data for the D+1 based on past weather data.

According to the present invention, energy management using meteorological data can be performed more efficiently by increasing the accuracy while providing meteorological data used by the energy management device to predict energy demand on the day before the target date. .

1 illustrates an energy management system and a peripheral configuration thereof according to an embodiment.
2 and 3 are diagrams illustrating a method of processing weather data according to an exemplary embodiment over time.
4 is a diagram illustrating a process of processing weather data and predicting energy demand according to an embodiment.
5 is a diagram illustrating a procedure for generating a first weather data file according to an embodiment.
6 is a diagram illustrating a meteorological data processing apparatus according to an embodiment.
7 and 8 are diagrams illustrating machine learning according to an embodiment.
9 and 10 are diagrams illustrating machine learning according to an embodiment.
11 illustrates an energy management system and a peripheral configuration thereof according to an embodiment.
12 illustrates a weather data processing apparatus that can be used in the embodiment of FIG. 11.

Hereinafter, some embodiments of the present invention will be described in detail through exemplary drawings. In adding reference numerals to elements of each drawing, it should be noted that the same elements are assigned the same numerals as possible even if they are indicated on different drawings. In addition, in describing the present invention, if it is determined that a detailed description of a related known configuration or function may obscure the subject matter of the present invention, a detailed description thereof will be omitted.

In addition, in describing the constituent elements of the present invention, terms such as first, second, A, B, (a), (b) may be used. These terms are only used to distinguish the component from other components, and the nature, order, or order of the component is not limited by the term. When a component is described as being "connected", "coupled" or "connected" to another component, the component may be directly connected or connected to that other component, but another component between each component It should be understood that elements may be “connected”, “coupled” or “connected”.

1 illustrates an energy management system and a peripheral configuration thereof according to an embodiment.

The energy management system collects weather data forecasted by the weather forecaster 10 through the network 30, uses the collected weather data to predict future energy demand, and based on the predicted energy demand, the building or area You can manage the energy of the lights (hereinafter simply referred to as'zones').

To this end, the energy management system may include a meteorological data processing device 100, machine learning 110, and an energy management device 20 (EMS).

For convenience of explanation, in this specification, the day to predict energy consumption is referred to as'D+1 day', and the day before the day to predict energy consumption is referred to as'D day', and to predict energy consumption The day before the day will be referred to as'D-1 day' (see Fig. 2). In addition, the weather data of D+1 forecasted on D day will be referred to as'D day/D+1 predicted weather data'.

The meteorological data processing apparatus 100 may generate a meteorological data file for predicting energy consumption of D+1 using the predicted weather data of D/D+1. In order to increase the accuracy of the prediction of energy consumption for D+1, the meteorological data processing device 100 provides corrected predicted weather data for D/D+1, supplemented with D/D+1 predicted weather data from machine learning. After receiving, modifying the meteorological data file using this, it can be provided to the energy management device 20.

Exemplarily, the meteorological data processing apparatus 100 generates a first weather data file using D-day/D+1-day predicted weather data to predict energy consumption of D+1, and uses machine learning. A second weather data file obtained by modifying the first weather data file may be generated, and the second weather data file may be provided to an energy management device.

Machine learning 110 supplements the D-day/D+1-day predicted weather data to reduce errors that may occur when predicting the energy consumption of D+1 days using the D-day/D+1-day predicted weather data. One D-day/D+1-day corrected forecast weather data can be generated and provided to the meteorological data processing device 100.

To this end, by way of example, the machine learning 110 uses predicted weather data on day D-1/day D (weather data on day D predicted on day D-1), actual weather data on day D, and day D/D+1. It is possible to output the corrected predicted weather data for day D/D+1 by inputting the daily predicted weather data. In this case, the machine learning 110 may be in a state in which it is learned to reduce an error between the D/D+1 corrected predicted weather data and the D+1 actual weather data based on past weather data.

Exemplarily, the machine learning 110 inputs the error between the predicted weather data on day D-1/day D and the actual weather data on day D, and the predicted weather data on day D/D+1 as inputs. The corrected predicted weather data for the first day is output, but may be in a state that has been learned to reduce an error between the corrected predicted weather data for the D/D+1 and the actual weather data for the D+1 based on past weather data.

The energy management device 20 uses the meteorological data file provided from the meteorological data processing device 100 to predict the energy consumption of D+1 days in the area it manages, and based on the predicted energy consumption, the energy of the area Can manage. The energy management device 20 and the meteorological data processing device 100 may exchange information with each other through the network 30. For example, the energy management device 20 may operate in a cloud environment. Exemplarily, a plurality of energy management devices 20 can predict energy consumption in an area they manage and perform energy management using a weather data file provided by a single meteorological data processing device 100 in a cloud environment. I can. For example, the energy management device 20 may be a building energy management device (BEMS) that manages a building.

The meteorological data processing apparatus 100, the machine learning 110, and the energy management apparatus 20 may be implemented through conventional computing devices such as CPU, MCU, and microprocessor, respectively.

In addition, although the meteorological data processing device 100, machine learning 110, and energy management device 20 are illustrated in separate configurations in FIG. 1, the meteorological data processing device 100, machine learning 110, and energy At least two or more of the management devices 20 may be implemented through the same computing device.

In addition, although it is illustrated in FIG. 1 that the meteorological data processing apparatus 100 and the machine learning 110 are connected using a separate communication line from the network 30, the meteorological data processing apparatus 100 and the machine learning 110 May communicate via the network 30.

2 and 3 are diagrams illustrating a method of processing weather data according to an exemplary embodiment over time.

In this embodiment, the energy demand on the D+1 day is more accurately predicted before 0 o'clock on the D+1 day, thereby effectively utilizing the meteorological data for energy management. Therefore, in this embodiment, instead of using weather data for the day forecast on D+1 (e.g., weather data forecast within 3 hours), as illustrated in FIG. 2, the day before D+1 (day D) You can use the forecasted weather data. For example, the weather data of the D+1 (predicted weather data of the D/D+1) that is forecast at 17 o'clock on the D day may be used. However, if the D-day/D+1-day forecast weather data predicted by the weather forecaster is used as it is, there is a high possibility that an error may occur in the energy demand prediction of the D+1 day. In this embodiment, machine learning is used to D+1 day D/D+1 corrected predicted weather data supplemented with D+1 day predicted weather data is generated, and based on this, D+1 day before 0 o'clock (e.g., D day 23:00) Energy demand can be predicted (it should be understood that the meteorological data supplemented using machine learning in the present embodiment is limited to meteorological elements related to energy demand prediction, so it is not intended to predict the entire meteorological data). For example, if the bidding time of the demand response trading market on the D+1 day is 0 o'clock on the D+1 day, machine learning will predict the correction on the D/D+1 day at 23:00 on the D day before 0 o'clock on the D+1 day. Weather data is generated, and the energy management device can use it to predict energy demand for D+1 days, and then participate in the demand response trading market.

Machine learning collects D-1 day/D day predicted weather data and D day actual weather data to increase the accuracy of D-day/D+1-day modified predicted weather data, and based on their information (e.g., error). It is possible to generate corrected predicted weather data for day D/D+1, supplemented with predicted weather data for day D/D+1. In other words, machine learning can more accurately predict tomorrow's weather (meteorological factors related to energy consumption) by collecting and utilizing today's weather data and today's actual weather data predicted yesterday. These features of machine learning will be described in more detail below.

FIG. 2 shows only matters related to energy consumption prediction on day D+1, for convenience of explanation. Actually, as shown in FIG. 3, corrected predicted weather data on day D-1, day D, and day D+1 Will continue to be created. Machine learning can improve the accuracy of energy demand forecasting by continuously collecting predicted weather data and actual weather data from the day before the day to predict energy demand and analyzing the errors.

4 is a diagram illustrating a process of processing weather data and predicting energy demand according to an embodiment.

4 is a meteorological data processing and energy demand prediction procedure performed by the meteorological data processing device 100, machine learning 110, and energy management device 20, focusing on predicting the energy consumption of D+1 days. It is illustrated accordingly.

First, the meteorological data processing apparatus 100 may collect D-1 day/D day predicted weather data, D day actual weather data, and D day/D+1 day predicted weather data (steps S401, S403, and S405). . As described above, the D-1 day/D day predicted weather data is the weather data of the D day forecast on the D-1 day, the D day actual weather data is the D day actual weather data, and the D day/D+1 The daily predicted weather data may be understood as the weather data of the D+1 forecasted on the D day.

As step S407, the meteorological data processing apparatus 100 is based on at least one of the collected D-1 day/D day predicted weather data, D day actual weather data, and D day/D+1 day predicted weather data. Data files can be created. Exemplarily, the first weather data file is generated by extracting at least one of temperature, humidity, wind direction, and wind speed from the predicted weather data of D/D+1, calculating solar radiation, and updating the original weather data file (WDF). Can be.

Here, the original meteorological data file contains meteorological data for evaluating the energy demand of the air conditioner, and from meteorological observation records for a long period (at least 10 years), meteorological elements (temperature, humidity, This is the data extracted from wind direction, wind speed, solar radiation, etc.).

Exemplarily, the first weather data file is a meteorological element (temperature, humidity, which is highly related to energy consumption) extracted from the D/D+1 predicted weather data from the original weather data file, which is meteorological data based on long-term records , Wind direction, wind speed and insolation, etc.).

As step S409, the machine learning 110 is based on the D-1 day/D day predicted weather data, D day actual weather data, and D day/D+1 day predicted weather data provided from the meteorological data processing device 100. It is possible to generate corrected forecast weather data on D/D+1.

Exemplarily, the machine learning 110 corrects the predicted weather data on day D/D+1 using the predicted weather data on day D-1/day D and the actual weather data on day D, You can generate corrected forecast weather data. To this end, the machine learning 110 modifies the D-1/D+1 day by inputting the D-1 day/D day predicted weather data, the D day's actual weather data, and the D day/D+1 day predicted weather data as inputs. The predicted weather data may be output, but may be in a state learned to reduce an error between the D/D+1 corrected predicted weather data and the D+1 actual weather data based on past weather data.

Exemplarily, the machine learning 110 uses the error information between the predicted weather data on the D-1/D day and the actual weather data on the D day to modify the D/D+1 predicted weather data. +1 day corrected forecast weather data can be generated. To this end, by way of example, the machine learning 110 inputs the error between the predicted weather data of day D-1/day D and the actual weather data of day D, and the predicted weather data of day D/D+1 as inputs. Outputs the corrected forecast weather data for day/D+1, but may be in a state trained to reduce the error between the corrected predicted weather data for day D/D+1 and the actual weather data for day D+1 based on past weather data. .

For example, the machine learning 110 may be trained to reduce errors in temperature, humidity, wind speed, wind direction, and solar radiation of the corrected predicted weather data on day D/D+1. For example, the machine learning 110 may be trained to give a weight to each of temperature, humidity, wind speed, wind direction, and insolation, and to minimize the sum of errors considering the weight.

As step S411, the meteorological data processing apparatus 100 generates a second weather data file using the D-day/D+1-day corrected predicted weather data provided from the machine learning 110, and converts the second weather data file into energy. It can be provided to the management device (20).

Exemplarily, the second weather data file is at least one of temperature, humidity, wind speed, wind direction, and insolation in the first weather data file based on the D/D+1 corrected predicted weather data output by the machine learning 110. Can be modified and created.

In step S413, the energy management device 20 may predict the energy demand for D+1 days based on the second weather data file provided from the meteorological data processing device 100. Here, the prediction of the energy demand of the D+1 day by the energy management device 20 may be performed before 0 o'clock on the D+1 day. To this end, the generation of D/D+1 corrected forecast weather data by machine learning 110 and generation of the second weather data file by the weather data processing device 100 are also performed before 0 o'clock on D+1. Can be. In this way, the energy management device 20 can predict the energy demand of the D+1 day before 0 o'clock on the D+1 day, compared to the case of using weather data within 3 hours forecast on the day D+1. Energy operation of D+1 days can be performed efficiently.

5 is a diagram illustrating a procedure for generating a first weather data file according to an embodiment. The process of generating the first weather data file illustrated in FIG. 5 may be performed by the aforementioned weather data processing apparatus.

In step S501, the meteorological data processing apparatus may collect predicted weather data. The predicted weather data may be weather forecast data provided by a weather forecaster. The predicted weather data may include D-day/D+1-day predicted weather data. For example, D/D+1 forecast weather data is weather data for each region predicted by the Meteorological Agency, and includes temperature, humidity, precipitation probability, precipitation type, wind direction, wind speed, significant wave height, sky conditions, precipitation, snowfall, minimum temperature. , May optionally include information such as the maximum temperature. As an example, the predicted weather data for day D/D+1 may be a weather forecast at 3 hour intervals.

In step S503, the meteorological data processing apparatus may convert the collected D-day/D+1-day predicted weather data into XML.

In step S505, the meteorological data processing apparatus may extract numeric data from the XML converted file. As an example, the extracted numeric data may be data corresponding to a weather forecast every 3 hours.

In step S507, the meteorological data processing apparatus may generate predicted weather data at 1-hour intervals using an interpolation method on the predicted weather data at 3-hour intervals. To this end, as an example, a Lagrange interpolation method may be used.

In step S509, the meteorological data processing apparatus may extract meteorological elements that are important for calculating the energy consumption amount from the predicted weather data at 1-hour intervals. Meteorological factors that are important to the calculation of energy consumption may include at least one of temperature, humidity, wind speed, and wind direction.

As step S511, the meteorological data processing apparatus may calculate the amount of insolation. Insolation may not be included in the predicted weather data for D/D+1 provided by the weather forecaster. Since the amount of insolation is an important factor in predicting the amount of energy consumption, the meteorological data processing apparatus can calculate the amount of insolation using the collected meteorological data. As an example, the meteorological data processing apparatus may use a total solar radiation calculation model (Seo model) for calculating the solar radiation.

In step S513, the meteorological data processing apparatus may generate a first weather data file. The first weather data file may be generated by updating the original weather data file WDF. For example, the first weather data file may be generated by updating the temperature, humidity, wind direction, and wind speed information and calculated solar radiation information extracted through step S509 to the original weather data file. As described above, the original meteorological data file is data obtained by extracting meteorological factors that are important to the consumption of heating and cooling energy from weather observation records for a long period of time (at least 10 years or more). In this way, more accurate energy demand prediction can be made by updating the meteorological factors extracted/calculated from the D-day/D+1-day predicted weather data in the original weather data file.

Thereafter, as described with reference to FIG. 4, the first weather data file is supplemented using the D/D+1 corrected predicted weather data secured through machine learning to generate a second weather data file, and the energy management device Can predict the energy demand for D+1 days using the second weather data file.

6 is a diagram illustrating a meteorological data processing apparatus 100 according to an embodiment.

The meteorological data processing apparatus 100 may include a meteorological information collecting unit 120, a meteorological data file generating unit 130, and a communication unit 140 to perform the functions described with reference to FIGS. 1 to 5.

The meteorological information collection unit 120 may collect meteorological data announced by the weather forecaster. For example, the meteorological information collection unit 120 may collect weather data including predicted weather data on day D-1/day D, actual weather data on day D, and predicted weather data on day D/D+1.

As illustrated through FIG. 5, the meteorological data file generation unit 130 may generate a first weather data file based on the predicted weather data of day D/D+1. In addition, the meteorological data file generation unit 130 may generate a second weather data file to be provided to the energy management device by using the D-day/D+1-day corrected predicted weather data generated by machine learning.

The communication unit 140 may perform a communication function so that the meteorological data processing apparatus 100 can transmit and receive data to and from the outside. The communication unit 140 may include conventional wired or wireless communication means.

7 and 8 are diagrams illustrating machine learning 110 according to an embodiment.

The machine learning 110 uses D-day/D+1-day predicted weather data to reduce errors that may occur when predicting energy demand for D+1 days, and D-day/D+1-day predicted weather data It is possible to generate corrected predicted weather data for day D/D+1 by modifying. To this end, the machine learning 110 inputs D-1 day/D day predicted weather data, D day's actual weather data, and D day/D+1 day predicted weather data as inputs and predicts D/D+1 day correction. Weather data can be output. In this case, the machine learning 110 may be in a state in which it is learned to reduce an error between the D/D+1 corrected predicted weather data and the D+1 actual weather data based on past weather data.

8 is a diagram exemplarily illustrating a concept in which the machine learning 110 performs learning based on past weather data. As illustrated in FIG. 7, the machine learning 110 inputs D-1 day/D day predicted weather data, D day's actual weather data, and D day/D+1 day predicted weather data as inputs, In a state configured to output D+1 day corrected predicted weather data, past weather data (X-1 day/X day predicted weather data, X day actual weather data and X day/X+1 day predicted weather data, where X day is a direction to reduce the error (error of X+1 day) by comparing the corrected predicted weather data outputted when X day/X+1 day is output when inputting the date before day D) with actual weather data of day X+1. It may be in a learned state. The machine learning 110 can improve the accuracy of the D-day/D+1-day corrected predicted weather data by performing learning based on past weather data in this manner.

For example, the machine learning 110 may be trained to reduce an error of at least one of temperature, humidity, wind speed, wind direction, and solar radiation of the corrected predicted weather data on day D/D+1. For example, the machine learning 110 may be trained to give a weight to each of temperature, humidity, wind speed, wind direction, and insolation, and to minimize the sum of errors considering the weight. That is, the machine learning 110 may improve the accuracy of energy consumption prediction by focusing on meteorological elements that are highly related to energy consumption among meteorological elements included in the meteorological data and performing learning through past meteorological data.

9 and 10 are diagrams illustrating machine learning according to an embodiment.

The machine learning 110 uses D-day/D+1-day predicted weather data to reduce errors that may occur when predicting energy demand for D+1 days, and D-day/D+1-day predicted weather data It is possible to generate corrected predicted weather data for day D/D+1 by modifying. To this end, the machine learning 110 inputs the error of the D-1 day/D day predicted weather data and the D day's actual weather data, and the D day/D+1 day predicted weather data as inputs, and Corrected forecast weather data can be output. In this case, the machine learning 110 may be in a state in which it is learned to reduce an error between the D/D+1 corrected predicted weather data and the D+1 actual weather data based on past weather data.

10 is a diagram exemplarily illustrating a concept in which the machine learning 110 performs learning based on past weather data. As illustrated in FIG. 9, the machine learning 110 inputs the error of the predicted weather data on day D-1/day D and the actual weather data on day D, and the predicted weather data on day D/D+1 as inputs. In a state configured to output the corrected predicted weather data for the day/D+1, past weather data (error between the predicted weather data of the X-1/X day and the actual weather data of the X day and the prediction of the X day/X+1 day Weather data, where X day is the date before D day, and the error (error of X+1 day) by comparing the corrected predicted weather data output on day X/X+1 with actual weather data on day X+1. ) May be learned in the direction of decreasing. The machine learning 110 can improve the accuracy of the D-day/D+1-day corrected predicted weather data by performing learning based on past weather data in this manner.

For example, the machine learning 110 may be trained to reduce an error of at least one of temperature, humidity, wind speed, wind direction, and solar radiation of the corrected predicted weather data on day D/D+1. For example, the machine learning 110 may be trained to give a weight to each of temperature, humidity, wind speed, wind direction, and insolation, and to minimize the sum of errors considering the weight. That is, the machine learning 110 may improve the accuracy of energy consumption prediction by focusing on meteorological elements that are highly related to energy consumption among meteorological elements included in the meteorological data and performing learning through past meteorological data.

As described above, in this embodiment, by using past weather data, the machine learning 110 learns the effect of the error between the predicted weather data of the day and the actual weather data of the day on the weather prediction of the next day by learning the weather data of the next day ( By increasing the accuracy of forecasting of weather factors that are highly related to energy demand), the next day's energy consumption can be more accurately predicted.

11 illustrates an energy management system and its surrounding configuration according to an embodiment, and FIG. 12 illustrates a weather data processing apparatus that can be used in the embodiment of FIG. 11.

11 and 12, the meteorological data processing apparatus 1100 is different from the above-described meteorological data processing apparatus 110 in that it includes machine learning 110.

As such, the machine learning 110 is included in the meteorological data processing device 1100 and supplements the D-day/D+1-day predicted weather data under the control of the meteorological data processing device 1100 to improve the accuracy of energy demand prediction. Can be used to raise.

As described above, according to the present embodiment, energy management using meteorological data is more efficient by increasing the accuracy while providing meteorological data used by the energy management device to predict energy demand on the day before the target day. Can be done with

Terms such as "include", "consist of", or "have" described above, unless otherwise stated, mean that the corresponding component may be included, and thus other components are not excluded. It should be interpreted as being able to further include other components. All terms, including technical or scientific terms, unless otherwise defined, have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs. Terms generally used, such as terms defined in the dictionary, should be interpreted as being consistent with the meaning in the context of the related technology, and are not interpreted as ideal or excessively formal meanings unless explicitly defined in the present invention.

The above description is merely illustrative of the technical idea of the present invention, and those of ordinary skill in the art to which the present invention pertains will be able to make various modifications and variations without departing from the essential characteristics of the present invention. Accordingly, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain the technical idea, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be interpreted as being included in the scope of the present invention.

Claims (17)

In order to predict the energy consumption of D+1, a first weather data file was created using the predicted weather data of D+1 (predicted weather data of D/D+1) predicted on D day, and machine learning was performed. A meteorological data processing device that generates a second weather data file obtained by modifying the first weather data file by using, and provides the second weather data file to an energy management device; And
An energy management device that predicts the energy consumption of D+1 days of the area using the second weather data file provided from the weather data processing device, and manages energy of the area based on the estimated energy consumption;
Energy management system comprising a. The method according to claim 1,
The second weather data file is an energy management system, characterized in that at least one or more of temperature, humidity, wind speed, wind direction, and solar radiation from the first weather data file are modified by the machine learning. The method according to claim 2,
The meteorological data processing device,
Collecting D-1 day/D day predicted weather data and D day's actual weather data (D day actual weather data), which is the weather data of day D forecast on day D-1,
The machine learning,
The D-1/D day predicted weather data and the D day predicted weather data are modified using the D-day/D+1-day predicted weather data to generate the D-day/D+1-day corrected predicted weather data, and ,
The meteorological data processing device,
And generating the second weather data file using the modified predicted weather data on day D/D+1 generated by the machine learning. The method of claim 3,
The machine learning,
D-1/D-day predicted weather data, D-day actual weather data, and D-day/D+1-day predicted weather data are input as inputs, and corrected predicted weather data for D/D+1 are output, but past weather data The energy management system, characterized in that learning to reduce an error between the D/D+1 corrected predicted weather data and D+1 actual weather data based on. The method of claim 3,
The machine learning,
By inputting the error between the predicted weather data of D-1/D and the actual weather data of D, and the predicted weather data of D/D+1 as inputs, the corrected predicted weather data for D/D+1 is output. An energy management system, characterized in that learning to reduce an error between the D/D+1 corrected predicted weather data and D+1 actual weather data based on past weather data. The method according to claim 4 or 5,
The machine learning is an energy management system, characterized in that it is learned to reduce errors in temperature, humidity, wind speed, wind direction, and solar radiation of the corrected predicted weather data on day D/D+1. The method of claim 6,
The machine learning is an energy management system, characterized in that a weight is assigned to each of the temperature, humidity, wind speed, wind direction, and insolation, and the total amount of errors considering the weight is minimized. The method according to claim 1,
The first weather data file is generated by updating the original weather data file (WDF) by extracting temperature, humidity, wind direction, and wind speed from the predicted weather data on day D/D+1 and calculating solar radiation. Energy management system. In the meteorological data processing apparatus for providing a weather data file for use by an energy management device for managing energy in an area to predict the energy consumption of D+1 day in the area,
Forecasted weather data for day D (D-1/D day predicted weather data), predicted weather data for day D, and forecasted weather data for day D + predicted weather data for day D (day D/D+) A weather information collection unit that collects daily predicted weather data);
Using the error information of the D-1/D day predicted weather data and the D day predicted weather data, D-day/D+1-day corrected predicted weather data are modified. Machine learning to generate; And
A meteorological data file generator for generating the meteorological data file to be provided to the energy management device by using the D-day/D+1-day corrected predicted weather data generated by the machine learning;
Weather data processing device comprising a. The method of claim 9,
The machine learning,
The D-1/D day predicted weather data, the D day actual weather data, and the D day/D+1 day predicted weather data are inputted to output the D/D+1 day corrected predicted weather data, Meteorological data processing apparatus, characterized in that learning to reduce an error between the D/D+1 corrected predicted weather data and D+1 actual weather data based on past weather data. The method of claim 9,
The machine learning,
By inputting the error between the predicted weather data of the D-1/D day and the actual weather data of the D day, and the predicted weather data of the D/D+1 as inputs, the corrected predicted weather data of the D/D+1 However, the weather data processing apparatus, characterized in that learning to reduce an error between the D-day/D+1-day corrected predicted weather data and D+1-day actual weather data based on past weather data. The method of claim 9,
The first weather data file is generated by extracting temperature, humidity, wind direction, and wind speed from the predicted weather data for day D/D+1 and updating the original weather data file (WDF) by calculating solar radiation,
The weather data file is generated by modifying at least one of temperature, humidity, wind speed, wind direction, and solar radiation in the first weather data file based on the D-day/D+1-day modified predicted weather data output by the machine learning. Weather data processing device, characterized in that. In the meteorological data processing method performed by a meteorological data processing device that provides a meteorological data file for use by an energy management device for managing energy in a zone to predict energy consumption of the zone for D+1 days,
The meteorological data processing unit uses the forecasted weather data of D-1 (predicted weather data of D-1/D), actual weather data of D-day, and predicted weather data of D+1 predicted on D-1 ( Collecting predicted weather data on day D/D+1;
The weather data processing device causes machine learning to modify the D-day/D+1-day predicted weather data by using the error information of the D-1 day/D day predicted weather data and the D day's actual weather data. Instructing to generate corrected predicted weather data for +1 days; And
Generating, by a weather data processing device, the weather data file to be provided to the energy management device by using the D/D+1 corrected forecast weather data;
Meteorological data processing method comprising a. The method of claim 13,
The first weather data file is generated by extracting temperature, humidity, wind direction, and wind speed from the predicted weather data for day D/D+1 and updating the original weather data file (WDF) by calculating solar radiation,
The weather data file is generated by modifying at least one of temperature, humidity, wind speed, wind direction, and solar radiation in the first weather data file based on the D-day/D+1-day modified predicted weather data output by the machine learning. Weather data processing method, characterized in that. The method of claim 14,
Extracting temperature, humidity, wind direction, and wind speed from the predicted weather data on day D/D+1,
Converting the predicted weather data of the D/D+1 to XML;
Extracting weather data at 3-hour intervals from the XML data;
Generating weather data at 1 hour intervals from the weather data at 3 hour intervals using an interpolation method; And
And extracting temperature, humidity, wind direction, and wind speed from the meteorological data at an hourly interval. The method of claim 14,
The meteorological data file is a meteorological data processing method, characterized in that at least one of the temperature, humidity, wind speed, wind direction, and solar radiation from the first weather data file is modified by the machine learning. The method of claim 13,
The machine learning,
D-1/D-day predicted weather data, D-day actual weather data, and D-day/D+1-day predicted weather data are input as inputs, and corrected predicted weather data for D/D+1 are output, but past weather data Meteorological data processing method, characterized in that learning to reduce an error between the D/D+1 corrected predicted weather data and D+1 actual weather data.

Images