KR102612781B1 - Load power analysis apparatus and load power analysis method by clustering technique - Google Patents

Abstract

Disclosed is a load power analysis device and load power analysis method using a clustering technique. The load power analysis device using the clustering technique of the present invention is installed at the power inlet point of a consumer where at least one load device is connected and measures the power usage of the consumer, and the power usage measured by the smart power meter unit. It includes a load decomposition unit that analyzes the power usage for each load device and monitors the power usage for each load device. The load decomposition unit preprocesses the power usage data measured by the smart power meter unit to obtain active power and apparent power by time period, and the active power and Apparent power are used to classify the corresponding load devices according to load power characteristics.

Description

Load power analysis apparatus and load power analysis method by clustering technique}

The present invention relates to a load power analysis device and a load power analysis method, and more specifically, to a load power analysis device and a load power analysis method using a clustering technique based on the power characteristics of the load.

Due to environmental problems, policy efforts are continuing to lower the rate of power generation by fossil energy and increase the rate of power generation by new and renewable energy. In addition, a power management policy is being implemented to accurately predict power demand and manage supply and demand in a balanced manner to maintain supply at a certain level.

In order to realize this energy management policy, an intelligent remote metering device (Advanced Metering Infrastructure: AMI) is installed and operated to accurately predict demand-side power usage.

AMI metering data not only serves as a standard for calculating electricity rates, but is also used for various intelligent services in the fields of distribution/sales/customer service, such as placebo, challenge detection, electricity quality (voltage, current) monitoring, and real-time information provision.

However, in order to support planned/rational power consumption and voluntary power saving for customers with existing AMIs installed, AMI data must not only provide real-time information but also be utilized as more detailed and useful information.

To achieve this, it must be possible to analyze the load installed at the customer (by type of electronic product). In order to analyze the load by electronic product, a separate device must be installed at the customer or meaningful information can only be obtained through separate analysis equipment. there was.

However, there is a need for the development of a new load power analysis device and load power analysis method that can determine the power usage by load by analyzing the characteristics of the customer's load without installing additional devices other than the already installed AMI.

Patent Document 1: Registered Patent No. 10-1970186 (Public Notice Date: April 18, 2019) Patent Document 2: Registered Patent No. 10-1708709 (Public Notice Date: February 21, 2017) Patent Document 3: Publication Patent No. 10-2016-0143456 (publication date: December 14, 2016)

The present invention, created in response to the above-described need, provides a load power analysis device and a load power analysis method that can analyze the power usage and usage patterns of individual loads by identifying the power characteristics of each load of the customer and classifying them using a clustering technique. The purpose is to

In order to achieve the above object, a load power analysis device using a clustering technique according to an embodiment of the present invention is a smart power system installed at the power inlet point of a consumer to which at least one load device is connected and measures the power usage of the consumer. meter section; And a load decomposition unit that monitors the power usage for each load device by analyzing the power usage measured by the smart power meter unit, wherein the load decomposition unit analyzes the power usage data measured by the smart power meter unit. Through preprocessing, active power and apparent power are obtained for each time period, and the active power and apparent power are used to classify the load device according to load power characteristics.

In this case, the load decomposition unit clusters the measured power usage data into a plurality of clusters on a daily or monthly basis.

In this case, the load decomposition unit calculates the power factor using the active power and the apparent power, and calculates the measured power usage data based on the active power, the power factor, and the meter reading time for reading the active power. Cluster into multiple clusters.

In this case, the load decomposition unit classifies clusters showing similar change patterns into homogeneous load devices using the median active power, median power factor, and median reading time for each of the plurality of clusters.

In this case, the load decomposition unit calculates the total daily operation time by adding the daily average generation time of the clusters classified into the same load devices, and uses the active power and the total daily operating time to load the corresponding load. Classifies devices into one of several types of power changes.

In this case, the load decomposition unit divides the plurality of power change types into a first type having two power change states, a second type having a plurality of power change states, and a third type having a continuous power change state in a short period of time. and a fourth type having a repetitive power change state over a long period of time, and is classified into any one of the first to fourth types.

Meanwhile, the load decomposition unit receives user feedback through an application pre-installed on a user terminal pre-registered as a user of the smart power meter unit.

Meanwhile, the load decomposition unit receives identification information about the load device acquired by a smart outlet pre-installed in a consumer where the smart power meter unit is installed.

The load power analysis method according to another embodiment of the present invention includes a smart power meter unit installed at the power inlet point of a consumer to which at least one load device is connected and measuring the power usage of the consumer, and measurement by the smart power meter unit. In the load power analysis method using a load power analysis device using a clustering technique that includes a load decomposition unit that monitors the power usage of each load device by analyzing the measured power usage, the load decomposition unit provides information on the measured power usage. Preprocessing and calculating a power factor using active power, apparent power, and the active power and apparent power for each predetermined time period; The load decomposition unit clusters the measured power usage information into a plurality of clusters on a daily or monthly basis using the active power, the power factor, and the meter reading time of the active power. ; The load decomposition unit classifies clusters showing similar change patterns into homogeneous load devices using the median value of active power, median power factor, and median value of meter reading time for each of the plurality of clusters; And the load decomposition unit calculates the total daily operation time by adding the daily average generation time of the clusters classified into the same load devices, and uses the active power and the total daily operating time to determine the corresponding load device. Classifying the plurality of power change types into one of a plurality of power change types, wherein the load decomposition unit classifies the plurality of power change types into a first type having two power change states, a second type having a plurality of power change states, and a short period of time. It is classified into either a third type with a continuous power change state or a fourth type with a repetitive power change state over a long period of time.

In this case, the load decomposition unit receives user feedback through an application pre-installed on a user terminal pre-registered as a user of the smart power meter unit or a load acquired by a smart outlet pre-installed in a consumer where the smart power meter unit is installed. Receiving identifying information about the device; and updating classification information for the load device using the user's feedback or identification information about the load device.

According to various embodiments of the present invention, the power characteristics of the consumer's load can be analyzed and classified by load using a clustering technique, so that the consumer's power usage can be accurately predicted,

By analyzing power usage patterns by load and breaking down AMI power usage, customers can reduce costs by providing them with information on the use of unnecessary home appliances and progressive tax notifications.

By using the power consumption patterns of each home appliance, you can detect the breakdown of the home appliance and prevent problems caused by the breakdown of the home appliance in advance.

By analyzing the usage patterns of users in each household, it predicts emergency situations that may occur in households with elderly people living alone, thereby demonstrating the effectiveness of accident prevention.

1 is a block diagram illustrating a load power analysis device using a clustering technique according to an embodiment of the present invention;
2 is a graph of four types of power patterns illustrating the power characteristics of load devices;
Figure 3 is a three-dimensional simulation diagram according to active power, power factor, and meter reading time to explain the clustering technique of the present invention;
Figure 4 is a data distribution graph for each power factor group of six clusters according to an embodiment of the present invention;
Figure 5 is a data distribution graph of public holidays and weekdays of six clusters according to an embodiment of the present invention;
Figure 6 is a data distribution graph for each day of the week in six clusters according to an embodiment of the present invention;
Figure 7 is a data distribution graph for each power factor group of four clusters according to another embodiment of the present invention;
Figure 8 is a data distribution graph of public holidays and weekdays of four clusters according to another embodiment of the present invention;
Figure 9 is a data distribution graph for each day of the week in four clusters according to another embodiment of the present invention;
10 is a diagram illustrating an exemplary user interface for receiving feedback from a user terminal according to another embodiment of the present invention, and
11 is a flowchart illustrating a load power analysis method according to another embodiment of the present invention.

Preferred embodiments of the present invention will be described below with reference to the drawings and tables. The embodiment described below is a preferred example, and various design changes may be made within the scope of the technical idea of the present invention.

Figure 1 is a block diagram illustrating a load power analysis device using a clustering technique according to an embodiment of the present invention. Referring to FIG. 1, the load power analysis device 100 using the clustering technique includes a smart power meter unit 110 and a load decomposition unit 130.

The smart power meter unit 110 is installed at the power inlet point of a consumer to which at least one load device is connected and measures the power usage of the consumer. The smart power meter unit 110 measures the power usage of a plurality of load devices (L1 to L5) connected through the smart outlet 300.

According to one embodiment of the present invention, clustering can be done based on the difference between active power and reactive power. In the provided data, the active power, reactive ground, reactive real power, and apparent power of one or more home appliances are accumulated for 15 minutes. Therefore, there is a need to perform preprocessing to suit the structure of the data.

The load decomposition unit 130 preprocesses the power usage data measured by the smart power meter unit 110. Table 1 below shows an example of power usage data preprocessed by the load decomposition unit 130.

Referring to Table 1 above, METR_NO is the identification number of the smart power meter unit, MR_YMD is the meter reading date, MR_HHMI is the meter reading time, WHM_NDL_LP is the active power, ONEA_VAR_LP is the reactive ground power, CDNPH_VARNDL_LP is the reactive ground power, and PH_WHM_NDL_LP is the apparent power. .

At this time, the power factor refers to a value that indicates how efficiently the actually applied voltage and current work. In theory, when efficiency is 100%, it can be obtained as the ratio of active power to the total power amount (apparent power). Power factor = defined as active power/apparent power. Add a derived variable that defines the power factor based on real power and apparent power. Table 2 below is a table adding the power factor calculated in this way.

In Table 2 above, the derived variable that defines the calculated power factor is defined as factor_check. Additionally, the P, Q1, Q2, S, factor check, Time, holiday, and Day items are defined as shown in Table 3 below.

The load decomposition unit 130 obtains active power and apparent power for each time period from the preprocessed power usage data. The load decomposition unit 140 calculates the power factor using the active power and the apparent power.

The load decomposition unit 130 can classify the power factor value values into four criteria and generate factor group variable items. The factor group items created in this way can be created in the manner shown in Table 4 below.

As shown in Table 5 below, the power factor values with a frequency of 95% or higher are [0.94, 0.93, 1.0, 0.96, 0.99, 0.98, 0.92], and the types of power factor values with a frequency of 95% or higher are 8 in descending order. The load decomposition unit 130 extracts the power factor value by assuming that the daily product usage frequency is around 95% of the total.

The load decomposition unit 130 preprocesses the measured power usage data by dividing it into daily or monthly units. Table 6 below shows an example of a data set preprocessed on a daily or monthly basis.

The load decomposition unit 130 clusters the measured power usage data into a plurality of clusters based on active power, power factor, and the meter reading time for reading the active power. At this time, the load decomposition unit 130 adjusts the number of clusters according to changes in the size (distance) of the clustered groups. Preferably, the number of clusters is determined at a critical point where the size (distance) of the clustered groups rapidly decreases as the number of clusters increases. To determine the number of clusters, the logic shown in Table 7 below can be used.

In more detail, as shown in Table 8 below, we look at the change in size of individual clustered groups formed while increasing the number of clusters, and select the number of clusters at the point where the number of clusters rapidly decreases as the number for clustering.

When the number of clusters is determined in the above-described manner, the load decomposition unit 130 clusters the measured power usage data into daily or monthly units according to the determined number of clusters. Table 9 below shows an example of clustering into 6 groups.

Table 9 above shows meter reading time, active power, power factor value, and generation time for each six clusters. At this time, the meter reading time, active power, and power factor value of each cluster refer to the median value among the power usage data constituting the cluster, and the occurrence time refers to the daily average occurrence time.

Table 10 above classifies clusters 1, 2, and 4 into homogeneous clusters, and the criterion is the similarity of active power and power factor values. The combined daily average generation time of clusters 1, 2, and 4 accounts for 13.2 hours, and the load device showing this power usage time pattern can be inferred to be a refrigerator.

That is, the load decomposition unit 130 uses the median value of active power, median power factor, and median reading time for each of the plurality of clusters to classify clusters showing similar change patterns into homogeneous load devices.

When analyzing a clustered data set like this, if it shows a similar pattern, it can be inferred that it is a homogeneous load device. Referring to Table 11, the load decomposition unit 130 can be classified as a homogeneous load device because the active power and power factor values of clusters 8 and 10 are similar.

The load decomposition unit 130 can use the change pattern of active power and power factor value to classify the cluster into four types that match the power change pattern.

The load decomposition unit 130 calculates the total operating time per day by adding the daily average generation time of clusters classified into the same load devices. The load decomposition unit 130 classifies the corresponding load device into one of a plurality of power change types using active power and total operating time per day.

The load decomposition unit 130 divides the plurality of power change types into a first type having two power change states, a second type having a plurality of power change states, a third type having a short-term continuous power change state, and a long-term power change state. It is classified into a fourth type having a repetitive power change state, and is classified into one of the first to fourth types. Referring to Table 12 below, the four types can be distinguished by power change patterns, and in addition to the four representative change patterns, additional power change patterns can also be interpreted.

The load decomposition unit 130 receives user feedback through an application pre-installed on the user terminal 200, which is pre-registered as a user of the smart power meter unit 110. The load decomposition unit 130 receives identification information about the load device acquired by the smart outlet 300 pre-installed at the consumer where the smart power meter unit 110 is installed.

Figure 2 is a graph of four types of power patterns explaining the power characteristics of load devices. Referring to Figure 2, the four types are divided into types 1 to 4.

The first type is a type that has two power change patterns consisting of a section where the power is turned off and a section where the power is turned on (see Figure 2(a)).

The second type is a type that has various power change patterns in the operation section (see Figure 2(b)), for example, a first section in which the power is turned off, a second section that operates at the first state level value, and a second section in which the power is turned off. There are various types of power change patterns, including a third section that operates with a state level value and a fourth section that operates with a third state level value.

The third type is a type that shows a continuous power change pattern in a short-term operation section (see Figure 2(c)). For example, this is a type in which the power change pattern occurs in a short period of time, with a section where the power is turned off, a section where the power rises to the maximum, and a section where the power drops to a lower level than the maximum power section. Here, short term means tens of minutes or hours.

The fourth type is a type that has a repetitive power change pattern in a long-term operation section (see Figure 2(d)). For example, it exhibits a continuous power usage pattern, showing a repeating pattern at two power levels: operating power and standby power. This fourth type shows a pattern that lasts for tens to hundreds of hours.

Figure 3 is a three-dimensional simulation diagram according to active power, power factor, and meter reading time to explain the clustering technique of the present invention. Referring to FIG. 3, the load decomposition unit 130 clusters the measured power usage data into 10 clusters.

At this time, the standard elements for clustering are active power, meter reading time, and power factor value. When clusters with similar characteristics of active power, power factor value, and meter reading time of 10 clusters are grouped together, the cluster is divided into a power use section by homogeneous load devices. It can be interpreted.

Figure 4 is a data distribution graph for each power factor group of six clusters according to an embodiment of the present invention. Referring to Figure 4, the data distribution of six groups for four power factor groups can be confirmed.

In the case of the first cluster, the third power factor group contains about 600 power usage data densely. In the case of the second cluster, the first power factor group and the second power factor group are distributed with 980 and 1100 pieces of data, respectively. In the case of the third cluster, data of the third power factor group and the fourth power factor group are distributed. In the case of the fourth cluster, data of the first power factor group and the second power factor group are distributed. In the case of the fifth cluster, data of the first power factor group, the second power factor group, and the third power factor group are distributed. In the case of the 6th cluster, data of the 3rd power factor group and the 4th power factor group are distributed.

As shown in FIG. 4, the purpose is to accurately analyze the power usage of load devices even when various load devices are included in the same cluster through data distribution analysis for each power factor group for a plurality of clusters.

Figure 5 is a data distribution graph of holidays and weekdays of six clusters according to an embodiment of the present invention. Referring to Figure 5, you can check the distribution of power usage data on public holidays and weekdays.

In the case of clusters 1 to 6, the distribution of power usage data on public holidays and weekdays can be distinguished, so it is possible to distinguish what type of load device is used by analyzing this data distribution pattern.

For example, in the case of the sixth cluster shown in Figure 5, the distribution of power usage data on weekdays is much larger than the distribution of power usage data on holidays, so it is estimated to be a load device that is widely used on weekdays, and the accuracy of this estimation must be increased. In order to do this, if information about the characteristics of each load device is received from the user as feedback information and reflected, the corresponding load device can be inferred more precisely. If you analyze the power usage data distribution by day of the week more precisely than analyzing the power usage data distribution on weekdays or holidays, the classification accuracy of load devices can be further improved.

Figure 6 is a data distribution graph for each day of the week in six clusters according to an embodiment of the present invention. Referring to Figure 6, in the case of the second cluster, the distribution of power usage data is concentrated on Saturdays, so it is inferred that load devices are used frequently on Saturdays. In order to increase the inference accuracy of these load devices, it is necessary to receive feedback information from users or use smart devices. The load device can be classified by receiving identification information of the operating load device from equipment that can identify the type of load device, such as an outlet.

The power usage data pattern shown in FIG. 6 is an example, but when analyzing the power usage data pattern by day of the week of the second cluster, the power due to the use of load devices (e.g., washing machines) that are used evenly on weekdays and are frequently used on Saturdays Indicates a pattern of change.

Figure 7 is a data distribution graph for each power factor group of four clusters according to another embodiment of the present invention. Referring to FIG. 7, distribution characteristics for each power factor group for the first to fourth clusters can be confirmed. In the case of the third cluster, it can be seen that the power usage data distribution of the first power factor group appears, and the power usage data distribution of the second power factor group appears in the remaining clusters.

Figure 8 is a data distribution graph of public holidays and weekdays of four clusters according to another embodiment of the present invention. Referring to FIG. 8, through the power usage distribution patterns of the first to fourth clusters on holidays and weekdays, a specific cluster can be classified into a cluster that represents the operating characteristics of a specific load device.

For example, the fourth cluster can be classified as a load device characterized by a high frequency of use on weekdays, and the second cluster can be classified as a load device characterized by a high frequency of use on public holidays. If feedback information is received from the user to identify the characteristics of the load device, the type of major load device used in the cluster can be identified more accurately by using the cluster's power usage data distribution and feedback information.

Figure 9 is a data distribution graph for each day of the week in four clusters according to another embodiment of the present invention. Referring to Figure 9, data distribution characteristics can be confirmed by day of the week for the four clusters. In the case of the 10th cluster, it tends to be used frequently on Wednesdays, showing the characteristics of a load device that is frequently used on Wednesdays. It can be analyzed as a cluster.

Figure 10 is a diagram illustrating a user interface for receiving feedback from a user terminal according to another embodiment of the present invention. Referring to FIG. 10, if feedback information on usage cycle, usage time, usage day, etc. for each load device is received from the user terminal and referred to when analyzing data, the power change pattern in each cluster is analyzed after clustering to classify load devices by type. It can be disassembled to analyze power usage.

Figure 11 is a flowchart illustrating a load power analysis method according to another embodiment of the present invention. Referring to FIG. 11, the load power analysis method preprocesses the measured power usage information and calculates the active power, apparent power, and power factor for each predetermined time period (S1101), using the active power and the apparent power. Clustering the measured power usage information into a plurality of clusters on a daily or monthly basis using the meter reading time for power, power factor, and corresponding active power (S1103), valid for each cluster Classifying clusters showing similar change patterns into homogeneous load devices using the median value of power, median power factor, and median reading time (S1105), average daily occurrence time of clusters classified as homogeneous load devices It includes a step (S1107) of calculating the total operating time per day by adding them together and classifying the corresponding load device into one of a plurality of power change types using the active power and total operating time per day.

At this time, before and after executing step S1107, the user's feedback is received through an application pre-installed on the user terminal 200 that is pre-registered as a user of the smart power meter unit 110, or the smart power meter unit 110 is installed. A step (S1108) of receiving identification information about the load device acquired by the smart outlet 300 pre-installed at the consumer may be further performed. In addition, after step S1108, a step S1109 may be further performed to update or modify classification information for the load device using the user's feedback or identification information about the load device.

100: load power analysis device
110: Smart power meter unit
130: load disassembly unit
200: user terminal
300: Smart outlet

Claims (10)

In the load power analysis device using the clustering technique,
A smart power meter unit installed at the power inlet point of a consumer to which at least one load device is connected and measures the power usage of the consumer; and
Analyze the measured power usage data, calculate the power factor using active power and apparent power, and calculate the measured power usage data based on the active power, the power factor, and the meter reading time for reading the active power. Clusters are clustered into a plurality of clusters on a daily or monthly basis, and clusters showing similar change patterns are homogeneous using the median active power, median power factor, and median reading time for each of the plurality of clusters. It is classified as a load device, and the total daily operation time is calculated by adding the daily average generation time of the clusters classified as the same type of load device, and the active power and total daily operation time are used to determine the corresponding load device. It is classified into one of a plurality of power change types, and the plurality of power change types are a first type having two power change states, a second type having a plurality of power change states, and a third type having a continuous power change state in a short period of time. A load decomposition unit that classifies each load device into one of the first to fourth types and monitors the power usage for each load device; containing,
Load power analysis device using clustering technique.
delete delete delete delete delete According to claim 1,
The load decomposition unit is a load power analysis device using a clustering technique, characterized in that it receives user feedback through an application pre-installed on a user terminal pre-registered as a user of the smart power meter unit.
According to claim 1,
The load decomposition unit is a load power analysis device using a clustering technique, characterized in that it receives identification information about the load device acquired by a smart outlet pre-installed in the consumer where the smart power meter unit is installed.
delete delete