JP6833660B2 - Reliability monitoring system, reliability evaluation method, and program - Google Patents

Description

Embodiments of the present invention relate to reliability monitoring systems, reliability evaluation methods, and programs.

In a power system including a power facility that generates power, a power system monitoring system is known that monitors the power system and estimates the power supply corresponding to an event that occurs in the future in order to stably supply power. ..

In the electric power system (hereinafter, appropriately referred to as "system"), the degree of ability to deliver electric power to consumers is called supply reliability. When looking at supply reliability from the standpoint of a grid operator, there are two concepts: adequacy and security. Adecacy is the static supply reliability, and is a concept regarding whether or not the equipment configuration meets the power demand in consideration of operational restrictions such as planned shutdown of grid equipment. On the other hand, security is dynamic supply reliability, which is related to whether the stability of the system can be maintained and the frequency and voltage can be kept normal against unexpected disturbances such as sudden failures. It is a concept.

At the stage where the grid operator operates the power system in real time, as a standard for ensuring the reliability of these supplies, even if one of the facilities that make up the grid fails, the remaining facilities supply power. The "N-1 reliability standard" that enables the continuation of the above is widely adopted. In addition, for important equipment, the "N-2 reliability standard" that enables continuous power supply even if two equipments fail may be adopted.

Among the controls for maintaining the supply reliability, there is an idea of preventive control in which measures are taken in advance before the occurrence of an assumed accident in order to avoid the interruption of power supply caused by an assumed accident. In preventive control, the current system information is acquired and the state is estimated, and the reliability when an assumed accident occurs with respect to the estimated system state is examined. If the power supply is interrupted in the grid, proactive measures such as generator output adjustment and phase adjustment control will be considered. However, in normal preventive control, the system is controlled so that they cooperate with each other in consideration of both the demerit of reliability due to supply interruption and the demerit of economy due to proactive measures.

In recent years, there is concern that uncertainty will increase in future systems due to power system reforms and increased renewable energy. In the conventional power system, analysis by physical model-based detailed simulation such as continuous tidal current calculation and stability calculation has been performed for a specific tidal current case assumed in advance from the rule of thumb of the grid operator. However, in recent systems, uncertainty in the power system has increased due to fluctuations in renewable energy power sources and factors that the system operator cannot completely manage and control, such as separation of power transmission and distribution. As described above, in recent systems, it is required to determine the stability of the system for innumerable tidal current cases that can be taken in the future, and the calculation time may become enormous in the conventional physical model-based analysis method.

Therefore, a learning model using statistical or machine learning based on a combination of measurable information (input) such as voltage, phase, load, generator output, and power flow, and multiple pieces of information on system stability (output). A method of instantly estimating the output from the input can be applied to control the system instead of the physical model-based analysis method.

When operating a power system, it is preferable to select a statistical learning model that makes it easy to explain the relationship between inputs and outputs. There is a regression model as a model for clarifying the relationship between input and output. However, in a power system including renewable energy, if the combination of input and output that becomes enormous and complicated due to fluctuations in renewable energy is modeled by a single regression model, the estimation error may become large.

Japanese Unexamined Patent Publication No. 2013-208042

The problem to be solved by the present invention is a reliability monitoring system, reliability that can estimate supply reliability with high accuracy by using a suitable regression model applied to the system state assumed in the power system. To provide a degree evaluation method and a program.

The reliability monitoring system of the embodiment has an offline analysis unit, a system state estimation unit, an estimation model creation unit, and a reliability estimation unit. The offline analysis unit calculates the supply reliability of the power system in advance by using the system state of the power system categorized in advance in the power system as input information. The system state estimation unit estimates the future system state based on the measurement information acquired from the power system and the information on the power demand. The estimation model creation unit creates a plurality of regression models based on the teacher data that combines the input information input to the offline analysis unit and the supply reliability output by the offline analysis unit. The reliability estimation unit selects the first regression model whose regression error with the teacher data is less than the threshold value from the plurality of regression models created by the estimation model creation unit, and selects the first regression model and the selected regression model. The future supply reliability of the power system is estimated based on the future system state estimated by the system state estimation unit.

The figure which shows an example of the structure of the reliability monitoring system 1 of 1st Embodiment. The figure which shows an example of the content of the assumed accident aspect information 13C of 1st Embodiment. The figure which shows an example of the probability distribution of the renewable energy power source output prediction information 13D of 1st Embodiment. The figure which shows an example of the renewable energy power source output prediction information 13D of 1st Embodiment. It is an example of the estimated systematic state of the first embodiment. The figure which shows an example of the branch data in the systematic state information of 1st Embodiment. The flowchart which shows an example of the flow of processing executed by the estimation model creation unit 16 of 1st Embodiment. The figure which shows an example of the addition method of the regression model of 1st Embodiment. FIG. 5 is a flowchart showing an example of processing executed by the reliability estimation unit 17 in step S106 of the first embodiment. The figure which shows an example of the 1st neighborhood data extracted by the reliability estimation part 17 of 1st Embodiment. The figure which shows an example of the regression model to which the first neighborhood data of 1st Embodiment belongs. The figure which shows the state which the 1st neighborhood data is included in 3 regression models of 1st Embodiment. FIG. 5 is a flowchart showing an example of a flow of processing executed by the estimation model creation unit 16 of the second embodiment. FIG. 5 is a flowchart showing an example of a flow of clustering processing executed by the estimation model creation unit 16 in step S206 of the second embodiment.

Hereinafter, the reliability monitoring system, the reliability evaluation method, and the program of the embodiment will be described with reference to the drawings.

(First Embodiment)
FIG. 1 is a diagram showing an example of the configuration of the reliability monitoring system 1 of the first embodiment. The reliability monitoring system 1 evaluates the supply reliability of the power supplied by the power system 21 when it is assumed that a pre-categorized accident has occurred in the power system 21 to which the renewable energy power source 23 is connected. It is a system.

The reliability monitoring system 1 estimates the future system operation state (for example, 30 to 60 minutes after the present time) from the current system operation state of the power system 21, and an accident assumed in the future system operation state may occur. Evaluate supply reliability when it occurs. Then, the reliability monitoring system 1 calculates and outputs the preventive control contents if a trouble occurs in the power supply. Specific evaluation items of supply reliability include, for example, overload, transient stability, voltage stability, frequency stability, and the like. If these evaluation items exceed a predetermined control value (threshold value), equipment failure / dropout may occur, causing a power outage. Therefore, the reliability monitoring system 1 evaluates the supply reliability based on the evaluation items and suppresses the power failure.

The reliability monitoring system 1 controls, for example, a system information collection unit 12, a data storage unit 13, a system state estimation unit 14, an offline analysis unit 15, an estimation model creation unit 16, and a reliability estimation unit 17. A content calculation unit 18 and a display unit 19 are provided. A power system 21 is connected to the reliability monitoring system 1. The power system 21 includes, for example, a generator 22 (22a, 22b), a renewable energy power source 23 (23a, 23b, 23c, 23d), and a consumer 24 (24a).

First, the power system 21 controlled by the reliability monitoring system 1 will be described. The generator 22 is, for example, a large-scale generator that performs thermal power generation, hydroelectric power generation, nuclear power generation, and the like. The renewable energy power source 23 is, for example, a power source that generates electricity using renewable energy such as solar power and wind power. Renewable energy sources can be created on a variety of scales. The individual renewable energy power sources 23 and consumers 24 may be grouped by region, for example, or may be treated as a regional set 31 (31a, 31b, 31c).

The power system 21 includes, for example, a plurality of measuring devices 25 (25a, 25b, ..., 25n) and a plurality of control terminal devices 26 (26a, 26b, ..., 25n) as devices for linking with the reliability monitoring system 1.・ ・, 26n). The measuring device 25 measures system information such as voltage, phase, and power flow of various parts of the power system 21, and the state of various power system facilities such as a transformer, a generator 22, and a renewable energy power source 23, and the measured system information. Is transmitted to the reliability monitoring system 1.

The control terminal device 26 receives a control command from the reliability monitoring system 1 and controls the system control device and the generator 22 (output amount, etc.) and suppresses the output of the renewable energy power source 23. Here, the system control device includes, for example, a phase-advancing capacitor, a shunt reactor, an SVC (Static Var Compensator), and the like. The tidal current means the flow of electricity, and is an index represented by the magnitude of, for example, active power, ineffective power, and the like.

Next, the configuration of the reliability monitoring system 1 will be described. In the reliability monitoring system 1, for example, a system information collection unit 12, a data storage unit 13, a system state estimation unit 14, an offline analysis unit 15, an estimation model creation unit 16, a reliability estimation unit 17, and a control content calculation unit 18. A part or all of each functional unit is realized by executing a program (software) by a processor such as a CPU (Central Processing Unit). In addition, some or all of these functional units are driven by hardware such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), and GPU (Graphics Processing Unit). It may be realized, or it may be realized by the collaboration of software and hardware.

The system information collecting unit 12 acquires various information about the current power system from the measuring device 25 installed in the power system 21. By inputting information to the data storage unit 13, the grid operator relates to power demand such as demand forecast information 13A, operation plan information 13B, assumed accident aspect information 13C, and renewable energy power output forecast information 13D for each region. Information is stored. The data storage unit 13 is realized by, for example, an HDD (Hard Disc Drive), a flash memory, an EEPROM (Electrically Erasable Programmable Read Only Memory), a ROM (Read Only Memory), a RAM (Random Access Memory), or the like.

The demand forecast information 13A predicts the magnitude of demand in each time zone of the day, and is predicted with high accuracy from the empirical rule of the system operator. The operation plan information 13B is information on the operation plan of the power system 21 determined in advance, such as the introduction / disconnection of the phase adjustment equipment, the start / stop of the generator, and the increase / decrease of the generator output.

The assumed accident aspect information 13C is information in which the assumed accidents in the power system 21 are categorized in advance. Typical pre-categorized accidents include, for example, transmission line failure, bus failure, generator failure (generator 22 failure), transformer failure, and the like.

FIG. 2 is a diagram showing an example of the contents of the assumed accident aspect information 13C of the first embodiment. As shown in FIG. 2, the assumed accident aspect information 13C is, for example, No. 1 indicating an accident. On the other hand, it is information in which the accident target location (accident point) and the accident aspect are associated with each other. The accident target location is a location where an accident is expected to occur in the power system 21. The accident aspect is a type of accident that occurs in the power system 21. Information such as "3Φ3LG" in the figure is a code representing the accident aspect.

Returning to FIG. 1, the renewable energy power supply output prediction information 13D is information regarding the output prediction of the renewable energy power supply 23 in the future (for example, 30 minutes or 60 minutes after the present time). The renewable energy power source 23 has uncertainty that the output fluctuates due to the influence of the weather and the like. The renewable energy power source output prediction information 13D is represented by, for example, a probability distribution in consideration of the uncertainty of the renewable energy power source 23.

FIG. 3 is a diagram showing an example of the probability distribution of the renewable energy power supply output prediction information 13D of the first embodiment. In the graph of FIG. 3, the vertical axis represents the probability of occurrence and the horizontal axis represents the renewable energy output. As shown in FIG. 3, the renewable energy output is defined as having some probability distribution because the output fluctuates. Renewable energy output uncertainty is expressed, for example, by defining the output range to be within ± σ (σ: standard deviation) of the predicted value (68.27%).

Note that this range can be arbitrarily determined. For example, the range of renewable energy output may be defined as being inside ± 2σ or ± 3σ of the predicted value, or an index other than the standard deviation, for example, It may be defined as inside ± 10% of the predicted value by using the capacity ratio of the renewable energy power source 23.

The renewable energy power source output prediction information 13D including such uncertainty is stored in the data storage unit 13 for each renewable energy power source 23 or the regional set 31.

FIG. 4 is a diagram showing an example of the renewable energy power supply output prediction information 13D of the first embodiment. As shown in 4, in the renewable energy power source output prediction information 13D, a value corresponding to ± σ in the renewable energy output prediction value and the probability distribution is given for each area (regional set 31). By using such renewable energy power output prediction information 13D, the reliability monitoring system 1 can be used for future power systems for any combination pattern of system states, including renewable energy outputs that fluctuate due to uncertainty. It is possible to evaluate whether or not the reliability of 21 can be maintained.

Returning to FIG. 1, in the system state estimation unit 14, the current system information acquired by the system information collection unit 12, the demand forecast information 13A given to the data storage unit 13, the operation plan information 13B, and the renewable energy power supply output prediction. Based on the information 13D, at least the future system state of the power system 21 is estimated.

FIG. 5 is a diagram showing an example of the estimated system state of the first embodiment. FIG. 6 is a diagram showing an example of branch data in the systematic state information of the first embodiment. In the node data of FIG. 5, information on voltage, phase, generator output (active power output, reactive power output), load (active power load, reactive power load), and phase adjustment equipment is associated with each node name. Has been done. Further, in the branch data of FIG. 6, information on the active power flow, the ineffective power flow, the active power loss, and the ineffective power loss is associated with each branch name.

Returning to FIG. 1, the offline analysis unit 15 inputs the demand forecast information 13A, the operation plan information 13B, the renewable energy power supply output prediction information 13D, and the assumed accident case input by the system operator and stored in the data storage unit 13. As information, the supply reliability of a plurality of assumed patterns is calculated in advance.

For example, the offline analysis unit 15 performs a power flow calculation based on the information stored in the data storage unit 13 as a preliminary step for calculating the supply reliability, and performs the same type of information (systematic state) as the state estimation result (see FIG. 5). ) Is used as the input information (X) to calculate the supply reliability (Y) by offline analysis. The offline analysis unit 15 outputs the combination of the input information (X) and the supply reliability (Y) to the estimation model creation unit 16 as teacher data.

The estimation model creation unit 16 divides a space area of the teacher data based on the teacher data output by the offline analysis unit 15, for example, and a plurality of regressions for estimating the actual supply reliability in the power system 21. Create a model. The estimation model creation unit 16 first creates the first regression model based on the teacher data. The estimation model creation unit 16 calculates the regression error based on the teacher data in the created first regression model.

The estimation model creation unit 16 repeatedly creates another regression model from the first regression model so that the calculated regression error is less than a predetermined threshold value, and creates a plurality of regression models. For example, in the process of creating another regression model, the estimation model creation unit 16 is based on the distance between the teacher data and the future systematic state, and the regression error from the teacher data exceeds the threshold on the positive side. And, of the teacher data, the negative side deviation teacher data whose regression error exceeds the threshold on the negative side is extracted.

The estimation model creation unit 16 creates a plurality of regression models by creating regression models based on the positive deviation teacher data and the negative deviation teacher data.

The reliability estimation unit 17 is based on the future system state estimated by the system state estimation unit 14 and the plurality of regression models created by the estimation model creation unit 16, and the future system state estimation unit 14 estimates the future system state. Estimate the supply reliability of the system state.

For example, the reliability estimation unit 17 extracts and extracts the minimum necessary regression model from the plurality of regression models created by the estimation model creation unit 16 so as to reduce the regression error of the plurality of regression models. From each regression model, the first regression model for estimating the supply reliability of the future power system is selected. Then, the reliability estimation unit 17 estimates the supply reliability of the future power system 21 based on the selected first regression model and the future system state estimated by the system state estimation unit 14.

The reliability estimation unit 17 calculates the spatial distance between the teacher data and the future systematic state in, for example, a plurality of regression models created by the estimation model creation unit 16. The reliability estimation unit 17 divides a plurality of regression models based on the calculated distance, and selects the first regression model having the minimum distance from each of the divided regression models.

The reliability estimation unit 17 determines, for example, in a plurality of regression models, a regression error with respect to the system state of the future power system from the teacher data based on the spatial distance between the teacher data and the future system state. The first neighborhood data existing in a predetermined range is extracted.

When the first neighborhood data belongs to a single regression model, the reliability estimation unit 17 selects the single regression model as the first regression model. When the first neighborhood data belongs to a plurality of regression models, the reliability estimation unit 17 determines the supply reliability based on the combination of the data included in the first neighborhood data belonging to each of the plurality of regression models. Analyze the sensitivity to the input information. The sensitivity will be described in detail later. The reliability estimation unit 17 extracts a variable whose sensitivity exceeds a predetermined threshold value from the input information.

The reliability estimation unit 17 extracts highly sensitive input information as a division factor variable, and calculates the distance between the teacher data and the future system state as the division factor distance using only the division factor variable. The reliability estimation unit 17 extracts the teacher data whose division factor distance is less than the threshold value as the second neighborhood data, and selects the regression model to which the second neighborhood data belongs most as the first regression model.

The reliability estimation unit 17 sets the change in the supply reliability with respect to the change in the input information as the first sensitivity, and sets the input information having the first sensitivity of a predetermined value or more as the division factor variable. The reliability estimation unit 17 excludes the input information whose first sensitivity is less than a predetermined value, and then calculates the change in the input information with respect to the change in the supply reliability as the second sensitivity. The reliability estimation unit 17 uses input information having a second sensitivity of a predetermined value or more as a division factor variable, and calculates the distance based on the division factor variable. The specific procedure for estimating the supply reliability by the reliability estimation unit 17 will be described later.

The control content calculation unit 18 assumes that an accident categorized in advance in the future system state has occurred based on the supply reliability of the future system state estimated by the reliability estimation unit 17, and the power system 21 Determine if there is a supply disruption. When the control content calculation unit 18 determines that a supply failure occurs in the power system 21, the control content calculation unit 18 calculates the control content of various control devices and generators necessary for suppressing the supply failure.

The control content calculation unit 18 estimates the system state after control for suppressing the supply trouble based on the calculated control content, and outputs the estimation result to the reliability estimation unit 17. The reliability estimation unit 17 evaluates the supply reliability after control for suppressing the supply trouble based on the estimation result output by the control content calculation unit 18.

The display unit 19 displays the evaluation result evaluated by the reliability estimation unit 17 and the calculation result calculated by the control content calculation unit 18. The display unit 19 is, for example, a display device such as an LCD (Liquid Crystal Display).

Next, an example of the processing executed by the estimation model creation unit 16 will be described. FIG. 7 is a flowchart showing an example of the flow of processing executed by the estimation model creation unit 16.

The estimation model creation unit 16 determines an initial regression variable for creating a regression model based on the teacher data (step S101). For example, the estimation model creation unit 16 extracts a variable that has a correlation with the supply reliability (Y) acquired from the offline analysis unit 15 from the input information (X) based on the height of the correlation coefficient. To do. The estimation model creation unit 16 creates the first regression model based on the variables extracted from the input information (X) (step S102). The regression model created by the estimation model creation unit 16 is, for example, a linear regression model. Hereinafter, a first-order linear regression model will be described as an example, but the estimation model creation unit 16 may use an estimation model other than the first-order linear regression model as the regression model.

The estimation model creation unit 16 determines the regression variables and the regression equation of the created regression model, and obtains the coefficients of the regression equation. The estimation model creation unit 16 calculates the regression error in, for example, calculating the coefficient of the regression equation (step S103). The regression error is, for example, the difference between the estimated value of the regression model for each teacher data and the true value of the supply reliability (Y). The true value is, for example, the supply reliability (Y) calculated by offline analysis. The estimation model creation unit 16 calculates the coefficient so that the sum of squares error of the regression error is minimized by, for example, the least squares method. The estimation model creation unit 16 applies the calculated coefficient to the regression equation to create the current regression model.

The estimation model creation unit 16 compares the calculated regression error of the current regression model with a predetermined threshold value (step S104). The estimation model creation unit 16 evaluates the regression error based on, for example, the average of a plurality of regression errors calculated for each teacher data, the maximum value of the regression error, and the like. In the evaluation of the regression error, the estimation model creation unit 16 determines whether or not the regression error calculated for all the teacher data is less than a predetermined threshold value (step S105).

When the regression error of all the teacher data becomes less than the threshold value in step S105 and a positive judgment is obtained, the estimation model creation unit 16 determines to use the current regression model, and the processing of the flowchart ends. If at least one of all the teacher data has a regression error equal to or greater than the threshold value in step S105 and a negative determination is obtained, the estimation model creation unit 16 adds a regression model (step S106). The process returns to step S104. The details of the process of step S106 will be described later with reference to FIG.

FIG. 8 is a diagram showing an example of a method of adding a regression model. The estimation model creation unit 16 compares, for example, the regression error of the teacher data with the threshold, and extracts the teacher data whose regression error exceeds the threshold on the positive side with respect to the regression line as the positive deviation teacher data. Of the teacher data, the data whose regression error exceeds the threshold on the negative side with respect to the regression line is extracted as negative deviation teacher data. The estimation model creation unit 16 creates a regression model based on each of the extracted positive deviation teacher data and negative deviation teacher data, and adds the regression model. The estimation model creation unit 16 creates a plurality of regression models by such processing.

Next, the process executed by the reliability estimation unit 17 will be described. The reliability estimation unit 17 extracts, for example, data in the vicinity of the teacher data representing the future systematic state based on the distance from the teacher data. The reliability estimation unit 17 extracts a variable whose sensitivity exceeds a predetermined threshold value from the input information based on the sensitivity derived from the relationship between the change in the input information and the change in the supply reliability from the extracted data. To do. The reliability estimation unit 17 selects a first regression model from a plurality of regression models based on the extracted variables. The reliability estimation unit 17 estimates the supply reliability of the future power system based on the selected first regression model and the future system state estimated by the system state estimation unit 14.

FIG. 9 is a flowchart showing an example of the process executed by the reliability estimation unit 17 in step S106 in the flowchart of FIG. 7. In the following description, the "future system state estimated by the system state estimation unit 14" is referred to as a "future state". The reliability estimation unit 17 calculates the distance between the future state and the teacher data. The "distance" is, for example, a distance in space, and is represented by, for example, an Euclidean distance, a distance obtained by a distance function based on a correlation coefficient, a cosine distance, a Manhattan distance, or the like.

The reliability estimation unit 17 calculates the distance between the future state and the teacher data based on, for example, the future state and the input information (X) of the teacher data. The reliability estimation unit 17 searches for the teacher data having the minimum distance with respect to the future state as the minimum distance teacher data by the distance calculation (step S121). Next, the reliability estimation unit 17 selects the minimum distance teacher data as a result of the search, and extracts the teacher data within a predetermined distance range from the selected minimum distance teacher data as the first neighborhood data (step S122).

FIG. 10 is a diagram showing an example of the first neighborhood data extracted by the reliability estimation unit 17. In the example of FIG. 10, X is treated as a one-dimensional variable for the sake of simplicity. The first neighborhood data is data existing within a predetermined range with reference to the selected minimum distance teacher data. The predetermined range is, for example, a range in which the standard deviation width (± σ) is given with reference to the minimum distance teacher data.

Here, the predetermined range may be arbitrarily determined by the user, or may be determined based on a statistical amount of information. The statistical amount of information is, for example, the standard deviation of each vector of X in the teacher data. In the example of FIG. 10, the teacher data in the range within ± σ in the X direction with respect to the selected minimum distance data is extracted as the first neighborhood data.

FIG. 11 is a diagram showing an example of a regression model to which the first neighborhood data belongs. As shown in FIG. 11 (a), the first neighborhood data may belong to a single regression model. The first neighborhood data in FIG. 11A belongs to the regression model B and does not belong to the regression model A. Further, as shown in FIG. 11B, the first neighborhood data may belong to a plurality of regression models. The first neighborhood data in FIG. 11B belongs to the regression model A and the regression model B.

Returning to FIG. 9, the reliability estimation unit 17 determines whether or not the extracted first neighborhood data belongs to a plurality of regression models (step S123). The reliability estimation unit 17 determines that the extracted first neighborhood data belongs to a single regression model, and if a negative determination is obtained in step S123, the single regression model is used as the first regression model for the future state. Is selected as (step S124).

The reliability estimation unit 17 determines that the extracted first neighborhood data belongs to a plurality of regression models, and if a positive judgment is obtained in step S123, extracts the division factor variable from the first neighborhood data ( Step S125). The division factor variable is used to divide a plurality of regression models to which the extracted first neighborhood data belong and select the first regression model, as described later.

Hereinafter, selection of the division factor variable by the reliability estimation unit 17 will be described. FIG. 12 is a diagram showing a state in which the first neighborhood data is included in the three regression models. As shown in the figure, a regression model A, b regression model B, and c regression model C belong to the first neighborhood data.

1. 1. The reliability estimation unit 17 calculates the change of X between a plurality of regression models in order to evaluate the regression error. For example, the change in X between a plurality of regression models is defined as follows.

In the above, (A⇔B) indicates the change between the regression model A and the regression model B. Assuming that the number of teacher data (first neighborhood data) belonging to the regression model A is a and the number of teacher data (first neighborhood data) belonging to the regression model B is b, there are a × b combinations. Also, p is the number of dimensions of X.

2. 2. The reliability estimation unit 17 defines the change in Y between a plurality of regression models as follows.

3. 3. The reliability estimation unit 17 calculates the first sensitivity between the plurality of regression models. The first sensitivity is an index showing the magnitude of the change in Y with respect to the change in X. The reliability estimation unit 17 defines the first sensitivity as follows, for example. The first sensitivity is used, for example, as an index for excluding X having a small effect on Y, that is, having a low sensitivity, as a variable and extracting X having a high sensitivity to Y. High first sensitivity means, for example, equal to or higher than a predetermined threshold value, and low first sensitivity means lower than a predetermined threshold value.

Using the above definition, the reliability estimation unit 17 excludes the calculated X having the low first sensitivity. Here, each first sensitivity S is represented by, for example, a value of the ratio of the amount of change in X to the amount of change in Y, as shown in the equation (1).

4. The reliability estimation unit 17 calculates the first sensitivity of X. The reliability estimation unit 17 is 3. Calculate the first sensitivity between multiple regression models calculated in. As the method for calculating the first sensitivity, the average of each first sensitivity calculated in 3 may be used, or the calculation may be performed based on the maximum value of the first sensitivity. In the following example, as shown in equation (2), a method using the average of the first sensitivities among a plurality of regression models will be described.

5. The reliability estimation unit 17 compares the first sensitivity and the threshold value, and excludes X having a low first sensitivity. The reliability estimation unit 17 excludes the calculated variable with low first sensitivity from X as the number of dimensions after reducing q, for example, based on the equation (2). In the process of exclusion, the reliability estimation unit 17 may compare the first sensitivity and the threshold value as described above, and specify the number of dimensions q after exclusion in advance.

6. The reliability estimation unit 17 defines the second sensitivity between the plurality of regression models. The second sensitivity is an index showing the change of X with respect to the change of Y. For example, the second sensitivity is defined as follows. The second sensitivity is, for example, an index for extracting X whose amount of change is larger than the amount of change in Y, among X excluding variables having low sensitivity to Y based on the first sensitivity. is there. From X excluding variables with low first sensitivity, variables with high second sensitivity (division factor variables) are further extracted. High second sensitivity means that the second sensitivity is equal to or higher than a predetermined threshold value.

ここで、各第二感度Ｓ´は、例えば、式（３）に示されるように、Ｘの変化量に対するＹの変化量の比の値で示される。

Here, each second sensitivity S'is represented by, for example, a value of the ratio of the amount of change in Y to the amount of change in X, as shown in the equation (3).

The reliability estimation unit 17 calculates the distance between the teacher data and the future state using the variable X extracted based on the second sensitivity, thereby causing a difference in Y among a plurality of regression models and of Y. Distance calculations can be performed by variables that change the difference more significantly. By executing such a process by the reliability estimation unit 17, the accuracy of assigning the future state to the regression model can be improved.

7. The reliability estimation unit 17 calculates the second sensitivity of X. The reliability estimation unit 17 calculates the second sensitivity of X from the second sensitivity between the regression models calculated in 6. The second sensitivity of X may be calculated from the average of the first sensitivity calculated in 6, or may be calculated from the maximum value of the first sensitivity calculated in 6. In the following description, the second sensitivity of X uses the average of the first sensitivities between the regression models, as shown in equation (4).

8. The reliability estimation unit 17 extracts variables X having high second sensitivity calculated by the equation (4), and uses the extracted r variables X as division factor variables. The dividing factor variable may be extracted by comparing the second sensitivity with the threshold value as described above, or the number r of the dividing factor variable may be specified in advance and extracted. The reliability estimation unit 17 extracts r division factor variables by performing the above processes 1 to 8.

Here, there are a plurality of teacher data that are close to the future state. The partitioning factor variable is used from the neighboring area to select a regression model suitable for indicating the future state when there are multiple regression models in the neighboring area to which multiple teacher data close to the future state belong. It is a variable used to divide the variable used in the first regression model.

Next, the reliability estimation unit 17 calculates the division factor distance, which is the distance from the future state, using the extracted division factor variable (step S126). Next, the reliability estimation unit 17 extracts the teacher data that minimizes the division factor distance from the division factor variables extracted for the target future state.

The reliability estimation unit 17 extracts the teacher data within a predetermined distance range from the teacher data having the minimum division factor distance as the second neighborhood data (step S127). The predetermined distance range may be arbitrarily determined by the user or may be determined based on a statistical amount of information. The statistical amount of information is, for example, the standard deviation of each vector of X in the teacher data.

The reliability estimation unit 17 selects a first regression model to which a target future state is assigned from a plurality of regression models to which the extracted second neighborhood data belongs (step S128). The reliability estimation unit 17 selects, for example, a regression model to which the second neighborhood data belongs most. That is, the second neighborhood data is the variable used in the first regression model divided from the space of the teacher data. The reliability estimation unit 17 inputs the target future state into the selected first regression model and calculates the estimated value of Y (step S129).

According to the embodiment described above, the reliability monitoring system 1 can estimate the supply reliability with high accuracy by using a suitable regression model applied to the system state assumed in the power system 21. .. Then, the reliability monitoring system 1 can easily explain the relationship between the estimation result and the physical model by using the regression model as the estimation model for estimating the supply reliability Y. Further, the reliability monitoring system 1 keeps a plurality of regression models created by each process of the estimation model creation unit 16 and the reliability estimation unit 17 within a predetermined error range, and minimizes the division process. Can be done.

Further, the reliability monitoring system 1 contributes to the change of Y between the regression models and causes the change of Y even when the teacher data having a short distance to X belongs to a plurality of regression models. By calculating the distance only with variables with a large value, it is possible to improve the accuracy of allocation of future states to the regression model.

(Second Embodiment)
In the first embodiment, the estimation model creation unit 16 divides a plurality of regression models based on the distance. In the second embodiment, the estimation model creation unit 16 differs in the process of dividing the plurality of regression models by clustering. In the following description, the same name and reference numeral will be used for the same configuration as that of the first embodiment, and duplicate description will be omitted as appropriate.

FIG. 13 is a flowchart showing an example of the flow of processing executed by the estimation model creation unit 16. The processes from step S201 to step S205 are executed in the same manner as the processes from steps S101 to S105 of the first embodiment.

FIG. 14 is a flowchart showing an example of the flow of the clustering process executed by the estimation model creation unit 16 in step S206. When at least one of all the teacher data whose regression error is equal to or higher than the threshold value exists in step S205 and a negative judgment is obtained, the estimation model creation unit 16 uses a predetermined clustering method in the teacher data. The number of clusters is added, and clustering is performed again (step S206).

In clustering, the estimation model creation unit 16 uses, for example, a clustering method that is roughly classified into hierarchical clustering and non-hierarchical clustering such as the k-means method. For example, when hierarchical clustering is used, the estimation model creation unit 16 has a predetermined number of clusters based on a dendrogram (tree diagram) formed based on the distance between two points of two data selected in the teacher data. To divide.

The estimation model creation unit 16 performs clustering that increases the number of clusters by one in the teacher data (step S221). The estimation model creation unit 16 creates a regression model in each clustered cluster (step S222). The estimation model creation unit 16 calculates, for example, the distance between all the teacher data and all the created regression models (step S223). Here, the estimation model creation unit 16 also calculates the distance from the regression model created in the cluster other than the cluster to which the teacher data belongs.

Further, as the distance to the regression model, the distance of the perpendicular line from the coordinates in space to the regression line, the absolute value of the regression error, etc. can be used, but from the viewpoint of minimizing the regression error, the absolute value of the regression error is used. Is preferable. Therefore, the estimation model creation unit 16 uses the absolute value of the regression error in the distance calculation with the regression model.

The estimation model creation unit 16 reassigns each teacher data to the cluster of the regression model that minimizes the calculated distance for each teacher data, and updates the cluster to which each teacher data belongs (step S224). The estimation model creation unit 16 updates the teacher data belonging to each cluster, and then starts updating the regression model again for each cluster (step S225).

The estimation model creation unit 16 determines whether or not the correction amount of the regression coefficient is less than the threshold value in updating each regression model (step S226). If the estimation model creation unit 16 obtains a negative determination in step S226, the estimation model creation unit 16 returns to the process of step S223. If the estimation model creation unit 16 obtains a positive determination in step S226, the estimation model creation unit 16 redetermines the regression variables of the regression model of each cluster (step S227). Here, when redetermining the regression variable, the estimation model creation unit 16 extracts, for example, a variable having a high correlation coefficient with respect to the supply reliability (Y) from the input information (X).

After the processing of step S227, the estimation model creation unit 16 determines whether or not the regression variable of the regression model has been changed (step S228). If a negative opinion is obtained in step S228, the processing of this flowchart ends, and the processing proceeds to step S204 (see FIG. 13). If a positive opinion is obtained in step S228, the process proceeds to step S223. By iterative processing from step S223 to step S227, it is possible to allocate variables to the clusters having the smallest regression error in the current number of clusters.

In this embodiment, by applying clustering to the division of the regression model, it is possible to obtain the minimum required number of regression models within a predetermined error while maintaining the difference in features between the regression models as much as possible.

According to at least one embodiment described above, the reliability monitoring system calculates the supply reliability of the power system in advance by using the system state of the power system categorized in advance in the power system as input information. The analysis unit, the system state estimation unit that estimates the future system state based on the measurement information acquired from the power system and the information on the power demand, the input information input to the offline analysis unit, and the offline analysis. Of the estimation model creation unit that creates a plurality of regression models based on the teacher data combined with the supply reliability output by the unit and the plurality of regression models created by the estimation model creation unit, the teacher Select a first regression model whose regression error with the data is less than the threshold, and based on the selected first regression model and the future system state estimated by the system state estimation unit, the future supply reliability of the power system. By having a reliability estimation unit for estimating the degree, it is possible to estimate the supply reliability with high accuracy by using a suitable regression model applied to the system state assumed in the power system.

Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, as well as in the scope of the invention described in the claims and the equivalent scope thereof.

1 ... Reliability monitoring system, 12 ... System information collection unit, 13 ... Data storage unit, 13A ... Demand forecast information, 13B ... Operation plan information, 13C ... Assumed accident aspect information, 13D ... Renewable energy power source output forecast information, 14 ... system state estimation unit, 15 ... offline analysis unit, 16 ... estimation model creation unit, 17 ... reliability estimation unit, 18 ... control content calculation unit, 19 ... display unit, 21 ... power system, 22 ... generator, 22a ... Generator, 22b ... Generator, 23 ... Renewable energy power supply, 23a ... Renewable energy power supply, 23b ... Renewable energy power supply, 23c ... Renewable energy power supply, 23d ... Renewable energy power supply, 24 ... Consumer, 24a ... Consumer, 25 ... Measuring device, 26 ... Control terminal device, 31 ... Regional set, 31a ... Regional set, 31b ... Regional set, 31c ... Regional set

Claims (13)

An offline analysis unit that calculates the supply reliability of the power system in advance by using the system state of the power system categorized in advance in the power system as input information.
A system state estimation unit that estimates the future system state based on the measurement information acquired from the power system and information on the power demand,
An estimation model creation unit that creates a plurality of regression models based on teacher data that combines the input information input to the offline analysis unit and the supply reliability output by the offline analysis unit.
Among the plurality of regression models created by the estimation model creation unit, the first regression model in which the regression error with the teacher data is less than the threshold is selected, and the selected first regression model and the systematic state estimation unit are selected. A reliability estimation unit that estimates the supply reliability of the power system in the future based on the future system state estimated by
Reliability monitoring system. The estimation model creating unit calculates the regression error based on the teacher data in the created regression model, and from the one regression model so that the calculated regression error is less than a predetermined threshold value. By repeating the process of creating the regression model of the above, the plurality of regression models are created.
The reliability estimation unit calculates the spatial distance between the teacher data and the future system state in the plurality of regression models created by the estimation model creation unit, and the calculation is based on the calculated distance. The first regression model that minimizes the distance is selected from a plurality of regression models.
The reliability monitoring system according to claim 1. In the process of creating the other regression model, the estimation model creation unit positively exceeds the threshold value on the positive side of the regression error from the teacher data based on the distance between the teacher data and the future system state. The side deviation teacher data and the negative side deviation teacher data whose regression error exceeds the threshold on the negative side of the teacher data are extracted, respectively, and regression is performed based on the positive side deviation teacher data and the negative side deviation teacher data. By creating each model, the plurality of regression models are created.
The reliability monitoring system according to claim 2. When creating the plurality of regression models, the estimation model creation unit clusters the teacher data based on a predetermined clustering method, and creates a regression model for each clustered cluster.
The reliability monitoring system according to claim 1. Based on the distance between the teacher data and the plurality of regression models, the estimation model creation unit reassigns the teacher data to the cluster to which the calculated regression model with the minimum distance belongs.
The reliability monitoring system according to claim 4. An offline analysis unit that calculates the supply reliability of the power system in advance by using the system state of the power system categorized in advance in the power system as input information.
A system state estimation unit that estimates the future system state based on the measurement information acquired from the power system and information on the power demand,
An estimation model creation unit that creates a plurality of regression models based on teacher data that combines the input information input to the offline analysis unit and the supply reliability output by the offline analysis unit.
Data in the vicinity of the teacher data representing the future system state is extracted based on the distance from the teacher data, and from the extracted data, from the relationship between the change in the input information and the change in the supply reliability. Based on the derived sensitivity, a variable whose sensitivity exceeds a predetermined threshold is extracted from the input information, and a first regression model is selected from the plurality of regression models based on the extracted variable and selected. It includes the first regression model and a reliability estimation unit that estimates the supply reliability of the power system in the future based on the future system state estimated by the system state estimation unit.
Reliability monitoring system. When selecting the first regression model, the reliability estimation unit determines the teacher data based on the spatial distance between the teacher data and the plurality of future system states in the plurality of regression models. When the first neighborhood data in which the regression error for the system state of the power system in the future exists in a predetermined range is extracted from the data, and the first neighborhood data belongs to a plurality of regression models, the plurality of regressions are performed. Based on the combination of data included in the first neighborhood data belonging to each of the models, the sensitivity derived from the relationship between the change in the input information and the change in the supply reliability is analyzed, and the sensitivity derived from the input information is analyzed. A variable whose sensitivity exceeds a predetermined threshold is extracted, and the first regression model is selected based on the extracted variable.
The reliability monitoring system according to any one of claims 2 to 6. The reliability estimation unit calculates the distance between the teacher data and the future system state as the division factor distance using the variable, and extracts the teacher data in which the division factor distance is less than the threshold as the second neighborhood data. Then, the regression model to which the second neighborhood data belongs most is selected as the first regression model.
The reliability monitoring system according to claim 7. The reliability estimation unit uses the change in supply reliability with respect to the change in input information as the first sensitivity, the input information having the first sensitivity of a predetermined value or more as a division factor variable, and the first sensitivity as a predetermined value. After excluding the input information that is less than, the change of the input information with respect to the change of the supply reliability is calculated as the second sensitivity, and the input information whose second sensitivity is equal to or more than a predetermined value is used as a division factor variable. The distance is calculated based on the division factor variable.
The reliability monitoring system according to claim 8. The computer
The supply reliability of the power system is calculated in advance by using the system state of the power system categorized in advance in the power system as input information.
The future system state is estimated based on the measurement information acquired from the power system and the information on the power demand.
A plurality of regression models are created based on the teacher data that combines the input information and the calculated supply reliability.
From the plurality of created regression models, the first regression model whose regression error with the teacher data is less than the threshold value is selected.
The supply reliability of the future power system is estimated based on the selected future system state estimated as the first regression model.
Reliability evaluation method. On the computer
Using the system state of the power system categorized in advance in the power system as input information, the supply reliability of the power system is calculated in advance.
The future system state is estimated based on the measurement information acquired from the power system and the information on the power demand.
A plurality of regression models are created based on the teacher data that combines the input information and the calculated supply reliability.
From the plurality of the regression models created, the first regression model whose regression error with the teacher data is less than the threshold value is selected.
The supply reliability of the future power system is estimated based on the future system state estimated as the first regression model selected.
program. The computer
The supply reliability of the power system is calculated in advance by using the system state of the power system categorized in advance in the power system as input information.
The future system state is estimated based on the measurement information acquired from the power system and the information on the power demand.
A plurality of regression models are created based on the teacher data that combines the input information and the calculated supply reliability.
Data in the vicinity of the teacher data representing the future systematic state is extracted based on the distance from the teacher data.
From the extracted data, a variable whose sensitivity exceeds a predetermined threshold is extracted from the input information based on the sensitivity derived from the relationship between the change in the input information and the change in the supply reliability.
The first regression model is selected from the plurality of regression models based on the extracted variables.
The supply reliability of the future power system is estimated based on the selected future system state estimated as the first regression model.
Reliability evaluation method. On the computer
Using the system state of the power system categorized in advance in the power system as input information, the supply reliability of the power system is calculated in advance.
The future system state is estimated based on the measurement information acquired from the power system and the information on the power demand.
A plurality of regression models are created based on the teacher data that combines the input information and the calculated supply reliability.
Data in the vicinity of the teacher data representing the future systematic state is extracted based on the distance from the teacher data.
From the extracted data, a variable whose sensitivity exceeds a predetermined threshold is extracted from the input information based on the sensitivity derived from the relationship between the change in the input information and the change in the supply reliability.
The first regression model is selected from the plurality of regression models based on the extracted variables.
The supply reliability of the future power system is estimated based on the future system state estimated as the first regression model selected.
program.

Images