Patent Application
20160077164
This application is based upon and claims the benefit of priority from Japan Patent Application NO. 2014-188575, filed on Sep. 17, 2014, the entire contents of which are incorporated herein by reference.
An embodiment of the present disclosure relates to a system for diagnosing a failure sign of an electrical power grid by obtaining measurement data from AMI and detecting a statistically deviated value using the data, and a method thereof.
In recent years, a wattmeter, called a smart meter, having a power measuring function and a bilateral communication function is becoming widespread. Thus, an environment in which a voltage and an electrical power amount at each customer spot can be monitored in an integrated manner is close to completion. Then, the AMI including the smart meter attracts attention as a sensor for monitoring the electrical power grid in view of its scale and the like.
The AMI is a main facility having the smart meter as its major configuration and comprises, in addition to the smart meter, a head end for collecting data from the smart meter and making settings of the smart meter and the like, a data management device of the smart meter, and moreover, a communication network for connecting these components and the like. Such AMI were initially contemplated for improving efficiency of charging operations according to the power consumption amount, but at present, it is considered as an infrastructure of a smart grid for controlling a power supply and a load, and is expected to be a monitoring sensor of the electrical power grid.
The electrical power grid is an infrastructure to be a contact point between an electrical power system and public customers, and in terms of its nature, not only high reliability but also high operating rate are required. On the other hand, the electrical power grids in many advanced countries were constructed before the 1970s, and an increase in failures caused by aging is concerned. There is great interest in maintenance of the electrical power grid under such circumstances.
The maintenance of the electrical power grid includes two policies roughly classified by replacement timing of devices comprised in the electrical power grid, that is, corrective maintenance and preventive maintenance. In the corrective maintenance, a failed device is replaced. However, since the device is replaced after a failure actually occurs in the corrective maintenance, it results in a lowered operating rate.
On the other hand, in the preventive maintenance, a device is forcedly replaced at a time when time or a load has reached a specific condition even before a failure occurs. Therefore, in the preventive maintenance, the operating rate of the electrical power grid can be improved. However, since the device with no concern of a failure is also replaced, the cost can become high.
Thus, a policy for improving the cost aspect while maintaining a stance of the preventive maintenance of replacement of a device before occurrence of a failure has been contemplated, and a system for statistically diagnosing a failure maintenance sign of an electrical power grid by detecting a statistically deviated value is proposed. In this diagnosis system, if a statistically deviated value is confirmed by monitoring an operating state of the electrical power grid, this is considered to be a sign of a failure, and a device in which occurrence of a failure is predicted is found and replaced. That is, since only a device which might be failed is replaced, a device with no concern of a failure is not replaced, whereby a cost can be suppressed.
In order to diagnose a failure sign of the electrical power grid with high accuracy, reinforcement of a monitoring system of the electrical power grid is important. Thus, it is expected to perform a failure sign diagnosis of the electrical power grid by using the AMI expected to be a monitoring sensor of the electrical power grid. However, the following problems are pointed out for a failure sign diagnosis technology of the electrical power grid using the AMI.
In general, in the electrical power grid, several thousands of customers are present under one substation for electrical power distribution. Thus, assuming that one smart meter is installed in one customer's home, several thousands of smart meters are included in the AMI monitoring one electrical power grid. That is, the number of dimensions of measurement data obtained from the AMI amounts to several thousand dimensions in correspondence to the number of installed smart meters.
At present, since there are only several tens of voltage-value monitoring sensors under one distribution substation, the number as several thousand dimensions is a great leap. As a result, it has been difficult to make a diagnosis of a failure sign for every single sensor value both from the viewpoints of calculation amounts and diagnosis accuracy. Thus, development of the failure sign diagnosis technology of the electrical power grid which can handle high-dimensional data obtained from the AMI has been required.
An embodiment of the present disclosure is proposed in order to solve the above-described problems and has an object to provide a failure sign diagnosis system of an electrical power grid capable of statistically diagnosing a failure sign of the electrical power grid by using high-dimensional measurement data obtained from the AMI, and a method thereof.
In order to achieve the above-described object, the embodiment of the present disclosure is a system for diagnosing a failure sign of the electrical power grid by detecting a statistically deviated value using measurement data obtained from the AMI and has the following components (a) to (g):
(a) A normal operation period calculation unit for calculating a period that can be considered to be at normal operation from a set of the measurement data;
(b) A load time slot calculation unit for calculating a high load time slot in which a load of the electrical power grid in a predetermined time slot is higher than a threshold value set in advance or a low load time slot in which the load is lower than the threshold value;
(c) A diagnosis section specifying unit for specifying a diagnosis section having a predetermined time width in the normal operation period corresponding to the time slot with a high load of the electrical power grid and the normal operation period corresponding to the time slot with a low load of the electrical power grid on the basis of calculation results of the normal operation period calculation unit and the load time slot calculation unit;
(d) A learning data calculation unit for calculating a representative value of each of the diagnosis sections for the measurement data corresponding to the diagnosis section as learning data;
(e) A normal subspace calculation unit for calculating a normal subspace that can be considered to be normal from a set of the learning data;
(f) A diagnosis data calculation unit for calculating a representative value of a diagnosis target period set in advance for the measurement data as diagnosis data; and
(g) A failure sign detection unit for detecting presence of a failure sign of the electrical power grid by calculating a difference between the diagnosis data and the normal subspace and comparing the difference with a threshold value set in advance.
Moreover, a failure sign diagnosis method of the electrical power grid is one aspect of the present disclosure.
Embodiments of the present disclosure will be described below with reference to the attached drawings.
A first embodiment of the present disclosure is a system for diagnosing a failure sign of an electrical power grid by detecting a statistically deviated value using measurement data obtained from a smart meter installed at a customer and uses AMI for obtaining the measurement data.
An AMI 1 is configured by connecting the smart meters S of each customer D connected to the identical electrical power grid C. The AMI 1 connects a communication network comprising a plurality of the smart meters S to the superordinate apparatus via a wide-area communication network. Methods of the communication network of the smart meter S include, for example, a wireless mesh method, a wide-area wireless method, a PCL method and the like.
The four calculation units are a diagnosis data calculation unit 6, a normal subspace calculation unit 7, a normal operation period calculation unit 8, and a high-load time slot calculation unit 18. Moreover, the failure sign diagnosis system 2 of the electrical power grid C has a detection unit 13 for detecting whether or not a failure sign has occurred on the basis of data calculated by each of the calculation units and an estimating unit 15 for estimating an occurrence spot of a failure and damage caused thereby on the basis of a detection result of the detection unit 13.
The six data storage units will be described.
The measurement data storage unit 5 is a time-series database for accumulating the measurement data transmitted from the AMI 1.
The maintenance information storage unit 9 stores maintenance information which is a failure occurrence history of the electrical power grid C and specifically, as illustrated in
The supply amount information storage unit 10 stores supply amount information of the electrical power grid C and specifically, as illustrated in
The low-load time normal subspace storage unit 11 and the high-load time normal subspace storage unit 12 are units for storing the normal subspace calculated in the normal subspace calculation unit 7 separately for the low-load time and the high-load time of the electrical power grid C, respectively.
The FMEA state storage unit 14 stores FMEA data associating a failure mode according to FMEA of the electrical power grid C with an ID of a device included in the electrical power grid C and a damage situation assumed by the failure. The FMEA is abbreviation of Failure Mode and Effect Analysis and means a potential failure mode and effect analysis. The FMEA data holds special associations, that is, the failure mode is associated with the related device ID and the assumed damage details.
For example, as illustrated in
Subsequently, the calculation units will be described. The normal operation period calculation unit 8 acquires a period in which none of the devices included in the electrical power grid C fails from a set of the measurement data obtained from the AMI 1 on the basis of the maintenance information stored in the maintenance information storage unit 9, as a normal operation period. An example of the normal operation period is illustrated in
In the example illustrated in
(1) 0:00 to 11:00
(2) 13:00 to 17:10
(3) 18:20 to 21:00
(4) 22:20 to 24:00
The high-load time slot calculation unit 18 is a unit for calculating a time slot in which a load of the electrical power grid C is at a threshold value set in advance or more in one day on the basis of the supply amount information stored in the supply amount information storage unit 10. In the first embodiment, when the high-load time slot calculation unit 18 acquires a high-load time slot in one day, the remaining time slots of the same day are assumed to be low-load time slots.
That is, when the high-load time slot calculation unit 18 acquires a high-load time slot, the low-load time slots are automatically determined. While there can be various threshold values for classifying the loads of the electrical power grid C, for example, 9 MW is set as a threshold value to classify the loads of the electrical power grid C.
Taking the supply amount information illustrated in
The normal subspace calculation unit 7 functions as a diagnosis section specifying unit and is a unit for calculating the normal subspace. Specification of a diagnosis section c (illustrated in
The diagnosis section c is a section specified as a unit of diagnosis and has a time width of approximately several seconds to 1 hour, for example. The diagnosis section c is set in advance as a constant in the system. Specifically, in the example illustrated in
(1) 0:00 to 11:00; and
(4) 22:20 to 24:00
are data obtained in the “normal and low-load section”.
Moreover, white circles indicated in the periods of:
(2) 13:00 to 17:10; and
(3) 18:20 to 21:00
are data obtained in the “normal and high-load section”. The normal subspace calculation unit 7 sends these two types of data indicated by the black circles and the white circles obtained in the diagnosis section c to the diagnosis data calculation unit 6.
In the example illustrated in
The normal subspace calculation unit 7 makes the data for each diagnosis section c specified in the diagnosis data calculation unit 6 as learning data d1 (illustrated in
As described above, in actuality, the data obtained from the smart meters S through the AMI 1 is in several thousand dimensions, but for convenience of description, it is assumed in
(Equation 1)
z=w
1
x
1
+ . . . +w
n
x
n (1)
Variance sz2 of the linear combination z is given by the following equation (2):
(Equation 2)
s
z
2
=w
1
2
s
1
2
+ . . . +w
n
2
s
n
2
+w
1
w
2
s
1
s
2
+ . . . w
n-1
w
n
s
n-1
2
s
n (2)
At this time, z at which the variance sz2 is maximized is considered to be a subspace expressing a feature of the data the best. Thus, w={w—1, . . . , w_n} which maximizes the variance sz2 becomes a conversion coefficient of the acquired subspace. Specifically, in order to determine w_i, the method of Lagrange multiplier is used. Since w_i is a conversion coefficient, an equation (3) which is a Norm normalization equation is introduced as a constraint equation.
(Equation 3)
w
1
2
+ . . . +w
n
2=1 (3)
From this equation (3) and the above-described equation (2), the following equation (4) is obtained:
(Equation 4)
L=s
z
2−λ(w12+ . . . +wn2−1) (4)
Moreover, each of the variables is subjected to partial differentiation, and n sets of simultaneous equations (5) together with 0 of a condition for extremal value can be obtained (only the example of i=1 will be described below):
Each of the above-described equations is organized for w={w—1, . . . , w_n}, and the following simultaneous linear equation (6) can be obtained:
Algebraically expressed, the above-described matrix is (S−λI) W=0, and λ is an eigenvalue of S, and W is a corresponding eigenvector. That is, λ can be acquired by solving an eigenvalue problem of S. S is a variance-covariance matrix of X, and the variance can be all specifically determined as:
and covariance can be all specifically determined as:
For a method for acquiring an eigenvalue from S, many known methods are proposed, and any arbitrary one can be used. Moreover, the eigenvalues are theoretically acquired in n pieces at the maximum, that is, in the same number as the original number of dimensions and the following configuration can be considered. For example, the preset number of eigenvalues are selected from the largest or a calculation formula called a degree of contribution selects some eigenvalues which constitute a combination exceeding a predetermined threshold. A set of w corresponding to λ selected as above becomes data corresponding to the normal subspace.
The normal subspace calculation unit 7 calculates a high-load time normal subspace Z_HIGH (illustrated in
The diagnosis data calculation unit 6 is a unit for calculating the learning data d1 for calculating the normal subspaces Z_HIGH and Z_LOW and diagnosis data d2 (illustrated in
The learning data d1 is data to be a unit of making a diagnosis of a failure sign and is a value represented by one actual value for each smart meter S in each diagnosis section c. The diagnosis data calculation unit 6 receives data corresponding to the diagnosis section c from the normal subspace calculation unit 7, divides the data for each diagnosis section c and calls the measurement data corresponding to each diagnosis section c from the measurement data storage unit 5.
The diagnosis data calculation unit 6 acquires the learning data d1 by calculating a representative value of each diagnosis section c from the measurement data called from the measurement data storage unit 5. That is, the learning data d1 is data to be a unit of making a diagnosis of a failure sign and is a value represented by one actual value for each smart meter S in each diagnosis section c.
How to acquire the representative value is not particularly specified in this embodiment, but any one of an average value, a maximum value, and a minimum value of the diagnosis section c or a first observation value and a last observation value can be used. An example of a certain diagnosis section c for voltage data of a certain smart meter S is illustrated in
Such a set of the learning data d1 has the following format:
X
t
={x
i,1
, . . . ,x
1,n} (t=1, . . . ,N) (Equation 9)
where t is a serial number for each diagnosis section c, and N corresponds to the number of the diagnosis sections c. Moreover, n is a device ID or specifically is an ID of the smart meter, and x is a voltage value obtained from the corresponding smart meter.
The diagnosis data d2 is a representative value obtained by calling the measurement data from the measurement data storage unit 5 for the voltage data observed during a diagnosis target period set in advance and by performing calculation from the data. The diagnosis data d2 can be expressed as follows:
d={d
1
, . . . ,d
n} (Equation 10)
where di is a representative value of the diagnosis target period of the i-th device.
The failure sign detection unit 13 fetches failure occurrence information 3 of the electrical power grid C at the same time and determines presence of occurrence of a failure in the electrical power grid C on the basis of the failure occurrence information 3. The failure sign detection unit 13 fetches supply amount information 4 from the outside and determines in which of a high load state and a low load state the system is at present on the basis of the supply amount information 4.
The failure sign detection unit 13 fetches in the normal subspace from either one of the normal subspace storage units 11 and 12 classified based on the load. That is, it fetches the high-load time normal subspace Z_HIGH from the high-load time normal subspace storage unit 12, while fetching the low-load time normal subspace Z_LOW from the low-load time normal subspace storage unit 11.
Moreover, the failure sign detection unit 13 fetches the diagnosis data d2 from the diagnosis data calculation unit 6 and calculates a distance f (d, Z_HIGH) between the high-load time normal subspace Z_HIGH fetched from the high-load time normal subspace storage unit 12 and the diagnosis data d2. Moreover, the failure sign detection unit 13 calculates a distance f (d, Z_LOW) between the low-load time normal subspace Z_LOW fetched from the low-load time normal subspace storage unit 11 and the diagnosis data d2. These distances f are differences between the diagnosis data d2 and the normal subspace Z_HIGH or Z_LOW.
As a value of the distance f, a distance function between a point and a plane in Euclidean geometry can be used. However, in the case of high-dimensional data, a method using the reciprocal of cos θ by inner product calculation is more resistant to an error.
The failure sign detection unit 13 detects presence of a failure sign of the electrical power grid by comparing the distance f obtained by the calculation and a threshold value TH1 or TH2 determined in advance. The threshold value TH1 corresponds to the high-load time normal subspace Z_HIGH, and the threshold value TH2 corresponds to the low-load time normal subspace Z_LOW. If the distance f is larger than the threshold value TH1 or TH2, the failure sign detection unit 13 determines that a failure sign occurs in the electrical power grid C and transmits the diagnosis data d2 and a failure sign occurrence signal e (illustrated in
When receiving the diagnosis data d2 and the failure sign occurrence signal e from the failure sign detection unit 13, the estimating unit 15 reads the FMEA data g (illustrated in
In the first embodiment having the above-described configuration, the data measured by the smart meter S is obtained from the customer D via the AMI 1, the learning data d1 is calculated by using the obtained measurement data, and a normal subspace to be normal data is calculated on the basis of the learning data d1 and stored. A process up to this shall be a learning process.
Then, a representative value of the diagnosis target period set in advance is calculated as the diagnosis data d2 for the measurement data, a difference between this diagnosis data d2 and the normal subspace obtained by the learning process is acquired, and a failure sign of the electrical power grid C is statistically diagnosed from the difference. This is called a diagnosis process. These two processes will be described by using flowcharts in
The learning process is a process performed before the diagnosis process is performed, and normal data to be a diagnosis reference of the diagnosis process is created. In this process, data considered to be normal is defined from the observed data and characterized. For characterizing the normal data, a non-linear function is used in some cases, but in the first embodiment, a subspace is used by putting an emphasis on estimation of a cause of the failure sign.
The learning process will be specifically described by using the flowchart in
At S2, the high-load time slot calculation unit 18 calls the supply amount information of the electrical power grid C from the supply amount information storage unit 10, calculates the high-load time slot, and transmits it to the normal subspace calculation unit 7 (a load-time slot calculation step).
At S3, the normal subspace calculation unit 7 fetches the normal operation period and the high-load time slot calculated in the normal operation period calculation unit 8 and the high-load time slot calculation unit 18 and specifies the “normal and high-load period” and the “normal and low-load period” from the data as the diagnosis section c (a diagnosis section specifying step).
At S4, the diagnosis data calculation unit 6 divides the received data sent from the normal subspace calculation unit 7 for each diagnosis section c, calls the measurement data corresponding to each diagnosis section c from the measurement data storage unit 5 and calculates the learning data d1 for each diagnosis section c (a learning data calculation step).
At S5, the normal subspace calculation unit 7 calculates the high-load time normal subspace Z-HIGH and the low-load time normal subspace Z_LOW (a normal subspace calculation step).
At S6, the high-load time normal subspace Z_HIGH and the low-load time normal subspace Z_LOW are stored in the high-load time normal subspace storage unit 12 and the low-load time normal subspace storage unit 11, respectively. As a result, the failure sign diagnosis system 2 of the electrical power grid holds the normal subspaces of both the high load and the low load.
Subsequently to the above-described learning process, the diagnosis process of the failure sign illustrated in
In the diagnosis process, initially at S7, the diagnosis data calculation unit 6 calls the measurement data from the measurement data storage unit 5 and calculates the diagnosis data d2 corresponding to the period (a diagnosis data calculation step).
At S8, the failure sign detection unit 13 checks the real-time failure occurrence information 3 in the electrical power grid C and determines presence of failure occurrence (a failure sign detection step). When the failure occurrence information 3 is confirmed, a failure has occurred in the electrical power grid C (Nat S8), the failure sign as preventive maintenance does not have to be detected. Thus, the routine proceeds to S22 and the diagnosis is discontinued to finish this process.
As the result of check of the failure occurrence information 3, if a failure has not occurred in the electrical power grid C (Y at S8), at S9, the failure sign detection unit 13 checks the supply amount information 4 and determines in which of the high load state and the low load state the system is at present. If the present system is at a high load (Y at S9), the routine proceeds to S10, while if the present system is at a low load (N at S9), the routine proceeds to S11.
At S10, the failure sign detection unit 13 reads the high-load time subspace Z_HIGH from the high-load time normal subspace storage unit 12 and calculates the distance f (d, Z_HIGH) between the high-load time subspace Z_HIGH and the diagnosis data d2. That is, the failure sign detection unit 13 statistically diagnoses the failure sign for the electrical power grid C in a high-load state from a difference between the high-load time normal subspace Z_HIGH and the diagnosis data d2
At S12, it is determined whether or not a value of the distance f (d, Z_HIGH) is larger than the threshold value TH1 set in advance. If the value of the threshold value TH1 is not exceeded, the routine proceeds to S14 to complete the diagnosis, and the system according to this embodiment determines that there is no abnormality in the electrical power grid C to finish the diagnosis process. Alternatively, if the threshold value TH1 is exceeded, the failure sign detection unit 13 transmits the diagnosis data d2 and the failure sign occurrence signal e to the estimating unit 15.
At S11, the failure sign detection unit 13 reads the low-load time subspace Z_LOW from the low-load time normal subspace storage unit 11 and calculates the distance f (d, Z_LOW) between the low-load time subspace Z_LOW and the diagnosis data d. That is, the failure sign detection unit 13 statistically diagnoses the failure sign for the electrical power grid C in a low-load state from a difference between the low-load time normal subspace Z_LOW and the diagnosis data d2.
At S13, it is determined whether or not a value of the distance f (d, Z_LOW) is larger than the threshold value TH2 set in advance. If the value of the threshold value TH2 is not exceeded, the routine proceeds to S15 to complete the diagnosis, and the system determines that there is no abnormality to finish this process. Alternatively, if the threshold value TH2 is exceeded, the failure sign detection unit 13 transmits the diagnosis data d2 and the failure sign occurrence signal e to the estimating unit 15.
At S16, the estimating unit 15 receives the diagnosis data d2 and the failure sign occurrence signal e from the failure sign detection unit 13 and estimates a failure sign occurrence spot in the electrical power grid C. An estimation process of the failure sign occurrence spot will be described in detail with reference to the flowchart illustrated in
The estimation process of the failure sign occurrence spot at the above-described S16 will be described by using the flowchart in
At S18, the estimating unit 15 determines a point d′ obtained by projecting the diagnosis data d to the normal subspace. The data of the normal subspace at this time is assumed to be the same as the subspace used immediately before at S12 or S13.
At S19, a norm of each component of a vector d-d′ is calculated.
At S20, a maximum norm component is specified, and its suffix i is selected as a candidate of the failure sign occurrence spot.
At S21, the estimation unit 15 reads out the FMEA data g from the FMEA storage unit 14. If the suffix having the maximum norm acquired at S20 is 199, for example, the failure mode of an ID 3 is selected, and the corresponding failure mode “switch does not operate” and damage details “power outage” are outputted as the estimation result.
According to the first embodiment as described above, the data measured by the smart meter S is obtained from the customer D via the AMI 1, and the normal subspace to be the normal data is calculated by using the obtained measurement data. Then, the failure sign of the electrical power grid C can be statistically diagnosed from differences between the normal subspaces Z_HIGH, Z_LOW and the diagnosis data d2.
In the first embodiment, since the subspace effective for reduction of the number of dimensions is used for characterizing the normal data, the failure sign of the electrical power grid C can be diagnosed with high accuracy even if the measurement data obtained from the AMI 1 is high dimensional data amounting to several thousands of dimensions. As a result, the AMI 1 can be effectively used as a monitoring sensor of the electrical power grid C.
In the first embodiment, when the measurement data is to be obtained from the smart meter S, the AMI 1 is used and thus, even an extremely large amount of data can be used or managed easily, and reliability is improved. As a result, social and economic values in the AMI 1 can be further improved, and quality improvement of the AMI 1 can be realized.
Moreover, in the first embodiment, the diagnosis data d2 corresponding to the “normal and high-load section” is compared with the high-load time normal subspace Z_HIGH, and the diagnosis data d2 corresponding to the “normal and low-load section” is compared with the low-load time normal subspace Z_LOW. Thus, at the high load time and the low load time, the differences between the diagnosis data d2 and the normal subspaces Z_HIGH and Z_LOW are respectively derived after comparison with the corresponding normal subspaces.
According to the first embodiment as above, occurrence of a failure sign can be detected with high accuracy, and exactly only the device which might fail can be replaced. As a result, a device with no concern of a failure is not replaced wastefully, and cost reduction by preventive maintenance can be achieved reliably.
Moreover, in this embodiment, since the maintenance information storage unit 9 and the supply amount information storage unit 10 are provided, the maintenance information and the supply amount information of the electrical power grid C having been stored can be called quickly, and the normal operation period calculation unit 8 and the high-load time slot calculation unit 18 can execute quick calculation processing.
Moreover, since the storage units 11 and 12 for separately storing the normal subspaces different depending on the load of the electrical power grid C are provided, the failure sign detection unit 13 can retrieve the normal subspaces Z_HIGH, Z_LOW immediately from each of the storage units 11 and 12. Moreover, since the failure occurrence information 3 and the supply amount information 4 are fetched from the outside on a real-time basis, presence of a failure and the load state of the electrical power grid C can be grasped accurately.
Moreover, in this embodiment, since the estimation unit 15 is provided, a spot where a failure occurs in the electrical power grid C can be estimated from the differences between the diagnosis data d2 and the normal subspaces Z_HIGH, Z_LOW. Moreover, the failure mode and assumed damage are estimated by using the FMEA data g, and the result notification unit 16 outputs the estimation result h of the estimation unit 15. Therefore, a service staff notified of the estimation result h can quickly and reliably handle the devices in which a failure is predicted to occur.
The above-described embodiment is presented herein as an example and is not intended to limit the scope of the invention. That is, it can be put into practice in other various forms and is capable of various omissions, replacements, and changes within a range not departing from the scope of the invention. These embodiments and their variations are included in the scope and the gist of the invention and are also included in the invention described in the scope of claims and the range of its equivalents.
For example, this embodiment is mainly based on a computer system, but as an embodiment, not only a single piece of hardware but a plurality of pieces of hardware may be used. Moreover, in this embodiment, only a voltage value is described as the measurement data obtained from the AMI, but a current value or an effective or ineffective power value and the like may be used.
A threshold value determining whether the load of the electrical power grid is high or low can be changed as appropriate and also, discrimination of the time slots with different loads is not limited to two, that is, the high-load time slot and the low-load time slot. For example, phased standards such as introduction of a type of middle loads between the high load and the low load depending on the operation of the system may be set. Moreover, as a method for notifying an estimation result in the result notification unit, displaying by using a display, acoustic alarms, pilot lamps and the like, for example, can be selected as appropriate. The starting frequencies of the learning process and the diagnosis process can be selected as appropriate. Moreover, regarding the diagnosis section c, the time width is set to several seconds to approximately 1 hour but this can be also changed as appropriate.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2014-188575 | Sep 2014 | JP | national |
