JPWO2015079554A1 - Power system state estimation device, state estimation method thereof, and power system control system - Google Patents

Abstract

A power system state estimation device for estimating a power system state quantity, the power system measurement unit 110 receiving power distribution information input / output by a power source or a power consuming apparatus linked to the power system, and a reception The simulator 120 that calculates the state estimation of the power system based on the appearance distribution information of the power amount, the measurement data of the power system output from the power system measurement unit, and the simulator data output from the simulator unit A maximum likelihood state value calculation unit 130 that calculates a state value that is probabilistically probable using, and a parameter correction unit 140 that corrects the parameters of the simulator unit based on the output of the maximum likelihood state value calculation unit. The simulation is performed so that the difference between the power system measurement data of the power system measurement unit and the simulator data of the simulator unit is reduced. To correct sequentially the parameters of the data part.

Description

  The present invention relates to a power system state estimation device, a state estimation method thereof, and a power system control system using the same.

In the electric power system, a wide variety of devices that generate (generate power and supply) and consume (or store) electric power by electric power companies and customers (factories, facilities, homes, etc.) perform their own operations.
The electric power company introduces various control devices to the electric power system, maintains an electric power supply-demand balance, and operates to ensure electric power quality.
Conventionally, many power plants that generate electric power have been installed and operated in a centralized manner by electric power companies.
Recently, however, distributed power sources such as solar power generation and wind power generation have begun to spread in order to utilize natural energy. Many of these distributed power sources are installed on the customer side and are not operated or managed by the power company.

Some of these distributed power sources vary in power generation amount depending on weather conditions. For example, changes in the amount of power generation are caused by the amount of solar radiation, wind direction, and wind speed.
Further, in a consumer, a large-capacity storage battery may be mounted on an electric vehicle or the like, and may be connected to the grid while changing the connection point, the connection time, the charge / discharge capacity, and the like.
The fluctuation factors of these distributed power sources may cause, for example, voltage fluctuations in the power system, and deteriorate the quality of power.
An electric power company that supplies electric power obtains measurement data related to electric power at major locations in the electric power system in order to ensure a balance between supply and demand and quality of electric power. At the same time, the state of the power system is estimated by using an apparatus and method such as a power system analysis simulator to simulate the operation of the power system and calculate the state value of the power system, and the power system is stably controlled. .
However, the state of the power system changes and fluctuates from moment to moment, and the location and measurement time that the sensor acquires are generally limited.

When the state value of the target power system cannot be obtained sufficiently, the unknown state value is estimated using a simulator. However, there are errors and fluctuations in parameters set in the simulator.
In addition, since there are a plurality of variation factors in the power system, the operation of the power system is affected as a combination of a plurality of variation factors.
As a method for examining characteristics of variation factors and a state estimation method in such a case, there are a method in which changes in the time direction are measured by a sensor, and measurement data (sensor measurement data) and simulator data are handled as time-series signals.
In addition, there is a method of stochastically handling the variation factors of measurement data and simulator data.
Patent Document 1 incorporates system data of the power system and data defining the probability distribution of fluctuations in the amount of power generation, load or delivery voltage of the distributed power source connected to the system, and the probability of node voltage or line current. A technique for obtaining a distribution and calculating an average value is disclosed.

JP 2005-57821 A

As described above, in order to calculate the power flow state with the simulator, it is necessary to set in advance the characteristic values of the line impedance, distributed power source, load, etc., which are the elements constituting the target power system. is there. However, it is difficult to measure a characteristic value of an actual power system, and it is often set with an error. As a result, a difference (error) occurs between the operation result of the actual power system and the calculation result of the simulator.
In addition, the method of measuring the change in the time direction as one method for examining the characteristics of the variation factors described above involves observing all combinations of a plurality of variation factors. There is a problem that events may not be observable.

In addition, in a system that simulates the operation of the power system based on sensor measurement data, both the observation error (measurement data error) and the system error (simulator data error) are combined with the state value of the power system that is the simulation result. Therefore, there is a problem that the number of combinations becomes enormous.
In addition, since the probability distribution based on the actual measurement such as the amount of solar radiation described above has a complicated shape, it must be sampled over the entire random variable. As the number of variation factors increases, the number of combinations of probability distributions becomes enormous, which increases the calculation load.
As described above, there is a problem that the increase in the variation factor and the increase in the combination of the observation error and the system error cause a cost increase of the measurement device, the control device, and the simulator in the power system.

Patent Document 1 discloses a technique for calculating and displaying a node voltage or a line current based on input probability distribution data. The probability distribution data input here is data representing fluctuations in equipment (distributed power supply, load, etc.) connected to the grid. That is, it only considers the fluctuation (observation error) of the equipment connected to the power system, and does not consider the error (system error) included in the procedure for simulating the state.
The probability distribution assumes a normal distribution. In some cases, these errors cannot be expressed in a normal distribution. For example, the variation in the amount of solar radiation greatly depends on the weather, and as a result, the power generation amount of photovoltaic power generation cannot assume a normal distribution.
As described above, since Patent Document 1 is a technique for calculating power flow using observation noise assuming a normal distribution, there is a problem that a probability distribution other than the normal distribution and a combination of an observation error and a system error cannot be handled. .

  Then, this invention makes it a subject to provide the state estimation apparatus of the electric power system which improves the electric power quality of an electric power system and reduces the cost concerning stabilization control.

In order to solve the above problems, the present invention is configured as follows.
That is, the power system state estimation device of the present invention is a power system state estimation device that estimates the state amount of the power system, and the appearance of the amount of power input and output by a power supply or power consumption device linked to the power system. An electric power system measurement unit that receives distribution information, a simulator unit that calculates the state estimation of the electric power system based on the received distribution information of the electric energy, and an electric power system measurement output from the electric power system measurement unit A maximum likelihood state value calculating unit that calculates a state value that is probabilistically probable using data and simulator data output from the simulator unit, and based on the output of the maximum likelihood state value calculating unit, parameters of the simulator unit A parameter correction unit for correction, and the difference between the measurement data of the power system of the power system measurement unit and the simulator data of the simulator unit is reduced Sea urchin, wherein the modifying sequentially the parameter of the simulator.
Other means will be described in the embodiment for carrying out the invention.

  ADVANTAGE OF THE INVENTION According to this invention, the electric power system state estimation apparatus which improves the electric power quality of an electric power system and reduces the cost concerning stabilization control can be provided.

It is a figure which shows the structure of the state estimation apparatus of the electric power system which concerns on 1st Embodiment of this invention, and the state estimation method of an electric power system. It is a flowchart which shows the flow of the state estimation method of the electric power grid | system which concerns on 1st Embodiment of this invention. It is a figure which shows the structural example of an electric power grid | system. It is an example of the figure which shows the relationship between time and the amount of solar radiation as frequency distribution. It is an example of the figure which shows the relationship between the amount of solar radiation and the frequency as a time series signal. It is a figure which shows the relationship between the change of the solar radiation amount by time, and the voltage fluctuation | variation of the system | strain connected to PV. It is a figure which shows the relationship between the voltage fluctuation | variation and frequency distribution of the system | strain which carries out PV interconnection. In the state estimation method of the electric power system which concerns on 1st Embodiment of this invention, it is a schematic diagram which shows the procedure calculated sequentially according to transition of time. It is a schematic diagram which shows performing the electric current calculation for every partial system | strain in the state estimation method of the electric power system which concerns on 1st Embodiment of this invention. It is a schematic diagram which shows performing the accumulation calculation of the current along a track | line in the state estimation method of the electric power grid | system which concerns on 1st Embodiment of this invention. It is a schematic diagram which shows calculating the distribution of the voltage along a track | line in the state estimation method of the electric power grid | system which concerns on 1st Embodiment of this invention. It is a figure which shows an example of an electric power grid | system. It is a figure which shows the correction operation | movement of a simulator parameter in the state estimation method of the electric power grid | system which concerns on 1st Embodiment of this invention. It is a figure which shows the estimation result of the simulation which followed the temporal load fluctuation | variation in the state estimation method of the electric power system which concerns on 1st Embodiment of this invention. Normal distribution _{N 1 (μ 1, σ 1} ), N 2 (μ 2, σ 2) likewise normal distribution of the state value estimated by _{N 3 (μ 3, σ 3} ) is a schematic diagram showing the relationship between. It is a figure which shows typically the frequency characteristic of probability distribution. It is a figure which shows typically the frequency characteristic of a control apparatus. It is a figure which shows the structure of the state estimation apparatus of the electric power grid | system of this invention provided with GUI.

<First embodiment, part 1>
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram illustrating a configuration of a power system state estimation device and a power system state estimation method according to the first embodiment of the present invention.
Moreover, FIG. 2 is a flowchart which shows the flow of the state estimation method of the electric power grid | system which concerns on 1st Embodiment of this invention.
In FIG. 1, there are a power system measurement data unit 111 that acquires power system measurement data, a measurement data processing unit 112 that probabilistically processes the measurement data, or an error 114.
Prior to explaining these, a power system (configuration example of the power system) in which the state estimation device of the power system of the present invention is used and events / behavior related to the power system will be described first, and then again FIG. The first embodiment of the present invention will be described in detail with reference to FIG.

<Example of power system configuration>
FIG. 3 is a diagram illustrating a configuration example of a power system (distribution system).
In FIG. 3, a power system 300 includes a distribution line 302 from a distribution substation 301, and a voltage regulator 303 is provided in the middle of the distribution line 302.
In addition, a distributed power source 311 provided with wind power generation, a distributed power source 312 provided with a grid storage battery, and a distributed power source 313 provided with photovoltaic power generation (PV: Photovoltaics) are connected to the distribution line 302 as distributed power sources.
In addition, the distribution line 302 is converted to a low voltage via the pole transformer 304 and is supplied to consumers (factories, facilities, homes, etc.) 314A, 314B, 314C.

  The consumer 314A not only simply consumes the supplied power, but also includes a photovoltaic power generation facility (PV) and an electric vehicle (EV). In addition, the consumer 314B not only consumes the supplied power, but also includes a photovoltaic power generation facility (PV) and a consumer storage battery (storage battery). Further, the consumer 314C only consumes the supplied power and does not particularly have anything corresponding to the distributed power source.

As described above, the power system 300 includes the distributed power sources 311, 312, and 313 that not only supply power from the distribution substation 301 to the consumers 314 </ b> A, 314 </ b> B, and 314 </ b> C, but also generate or charge / discharge (storage battery) power. Existing.
In addition, the consumers 314A and 314B are also provided with one corresponding to the distributed power source (PV, EV, storage battery) as described above.
FIG. 3 shows a power distribution system (power system) close to the consumer. With the increase of the distributed power source due to the introduction of renewable energy, the variation factors of the power system become various, and the power system There is concern about the deterioration of quality related to stable voltage and power supply.
When grasping the state quantity regarding the power (voltage, current) of the power system 300 and calculating the power flow, it is necessary to grasp not only the fluctuation of the load amount but also the fluctuation of the power generation amount of the distributed power source.

When the distributed power source is solar power generation or wind power generation, the power generation amount varies depending on the weather and weather. Details of such fluctuation factors and how to deal with them will be described later.
In FIG. 3, the distribution line 302 is illustrated as being not branched in the middle, but the distribution line 302 may be branched at a plurality of locations.
Further, in FIG. 3, illustration of sensors for measuring voltage and current is omitted.

<First embodiment, part 2>
Next, the configuration of the power system state estimation device and the power system state estimation method according to the first embodiment of the present invention will be described in detail. Note that the description below also serves as a description of a power system state estimation device and a power system state estimation method.

As described above, FIG. 1 is a diagram illustrating a configuration of the power system state estimation device and the power system state estimation method according to the first embodiment of the present invention.
In FIG. 1, a power system state estimation apparatus (power system state estimation method) 100 includes a power system measurement unit (power system measurement unit) 110, a simulator unit (simulation calculation unit) 120, and a maximum likelihood state value calculation unit (maximum state). Likelihood state value calculating means) 130 and parameter correcting section (parameter correcting means) 140 are provided.

<About the power system measurement unit>
The power system measurement unit 110 includes a power system measurement data unit 111, a measurement data processing unit 112, and a summation unit 113.
The power system measurement data unit 111 shows the voltage and current at the location where the sensor (not shown) in the power system 300 (FIG. 3) is installed, and the appearance of the amount of power input / output by the power supply or power consuming device linked to the power system. Receive and acquire power related information (sensor measurement data) such as distribution information. Note that the measurement data (sensor measurement data) may include other elements such as a phase, and is not limited to the example of the measurement data.

  The data (power related information) received by the power system measurement data unit 111 is acquired by a sensor installed in the power system. There are various types of sensors. Even in the same type of sensors, the characteristics vary depending on the sensors. In addition, there are restrictions on the location and number of attachments. In addition, the data acquired by the sensor is not continuously measured data, but is so-called discontinuous data sampled, and there are restrictions on the sampling interval and the like. For example, while fluctuations in the amount of power generated by solar power generation may occur in units of seconds, measurement data from sensors may be transmitted at 30-minute intervals. In this way, fluctuation factors are introduced.

Errors due to these variations, restrictions, and various fluctuation factors are indicated by an observation error (error) 114 in FIG. This observation error (error) 114 is observation noise.
The data received by the power system measurement data unit 111 in the summation unit 113 and the data to which the observation error (error) 114 is added (summed) are actually the measurement data of the observed power system.
That is, the data of the power system measurement data unit 111 is ideal measurement data that does not include the observation error (error) 114, and the data that has passed through the summation unit 113 includes errors that are actually observed. It is the measurement data of the power system.

The measurement data processing unit 112 expresses the measurement data including the fluctuation factors and errors as described above as the probability distribution, and evaluates the influence on the power system as a combination of the probability distributions (probability / statistically). Arithmetic processing).
For example, by setting conditions based on probability distributions for a plurality of fluctuation factors and calculating the power flow, the power flow calculation considering the power system fluctuation factors is executed. Here, when the probability distribution can be expressed by a mathematical expression such as a normal distribution and a β distribution, the entire distribution can be reproduced by using only representative sample values. For example, the normal distribution can reproduce the entire distribution by performing calculations using several sample values defined by the mean and variance.

  The measurement data of the power system that has been arithmetically processed as described above is output to the maximum likelihood state value calculation unit 130 as an output of the power system measurement unit 110.

About the simulator
The simulator unit 120 includes a power system simulator 121, a simulator data processing unit 122, a parameter setting unit 125, and a summing unit 123.
The power system simulator 121 constitutes a system model reflecting the power system 300 (FIG. 3), and is based on various parameters related to the power system set by the parameter setting unit 125 and the input of sensor measurement data in real time. Simulate calculation by operation (simulation operation, simulation).
The various parameters include track / equipment characteristics, load data, power generation amount data, time, and the like, and sensor measurement data acquired by the power system measurement data unit 111 or the power system measurement unit 110 is also input.

However, the various parameters have errors due to factors such as device characteristics, calculation method, load and power generation amount in the system model.
In FIG. 1, errors of these various parameters are represented as system errors (errors) 124 as being added to the signal of the parameter setting unit 125 by the adding unit 123.
That is, the data that has passed through the summation unit 123 is various parameter data that reflects actual system errors (simulator errors, simulator noise).

The power system simulator 121 performs simulation calculation (simulation operation, simulation) as power flow calculation based on the various parameter data including the system error and the sensor measurement data including the observation error. Therefore, the simulation data of the power system simulator 121 includes an error.
The output of the power system simulator 121 including this error is input to the simulator data processing unit 122.
The simulator data processing unit 122 performs various arithmetic processes such as probability distribution and statistical processing, and outputs simulator data with reduced errors. The output of the simulator data processing unit 122 is also input to the maximum likelihood state value calculation unit 130 as the output of the simulator unit 120.

<< About Maximum Likelihood State Value Calculation Unit >>
The maximum likelihood state value calculation unit 130 receives the output of the power system measurement unit 110 and the output of the simulator unit 120.
Maximum likelihood state value calculation section 130 compares the output of power system measurement section 110 with the output of simulator section 120 and detects the difference (deviation).
Then, the output of this difference (or converted signal) is input to the parameter correction unit 140.
The maximum likelihood state value calculation unit 130 includes means for converting the state quantity of the difference into a code.
Note that the above operation of the maximum likelihood state value calculation unit 130 obtains a state value probabilistically probable using the measurement data of the power system output from the power system measurement unit 110 and the simulator data output from the simulator unit 120. It is also a calculation to calculate. Details will be described later.

<Parameter correction section>
The parameter correction unit 140 outputs a signal for changing various parameters in the power system simulator 121 based on the difference information in the output of the maximum likelihood state value calculation unit 130. That is, the parameters of the simulator are corrected so as to realize the target probability distribution obtained by the maximum likelihood state value calculation unit 130.
For example, in a power system, a relational expression between a change in voltage and active power and reactive power of a load is known. Therefore, the adjustment amount of the active power and reactive power of the load is calculated based on the relational expression of the power system so as to correct the difference in the voltage average value, and is set as a parameter of the simulator.
Alternatively, the resistance and impedance of the line can be specified as the correction target.
There are various selections of parameters to be corrected.

<< Cooperation of power system simulator, maximum likelihood state value calculation unit and parameter correction unit >>
The power system simulator 121 executes the simulation again and inputs the new output to the maximum likelihood state value calculation unit 130 via the simulator data processing unit 122.
Maximum likelihood state value calculation section 130 compares again the output of power system measurement section 110 and the output of simulator section 120 and outputs the difference (or the converted signal) to parameter correction section 140 again. .

In this way, the simulator operation of the power system simulator 121 is sequentially corrected so that the difference between the output of the power system measurement unit 110 and the output of the simulator unit 120 becomes small.
That is, the parameters of the simulator are set so that the sensor measurement data (data of the power system measurement unit 110) and the simulator output (output of the simulator unit 120) most probably match with each other (maximum likelihood state value calculation unit 130). It corrects sequentially (parameter correction part 140).
Therefore, when there is an observation error and a system error (simulator data error), a target setting for correcting the parameters of the simulator is stochastically performed.
This correction is sequentially processed along the time axis to follow the drift of the device characteristics.

This correction is realized by preparing in advance a correspondence relationship between the output fluctuation and the device characteristic fluctuation. This correspondence is difficult to associate in advance if it is a complex electric circuit.
However, the power system can often be broken down into partial systems. For example, in the case of a partial system with one input and one output and the adjustment target is one load, the association of fluctuations is simplified. The present invention does not limit the size of the partial system, and the partial system may be dynamically recreated.
When there is a voltage error between the simulator output and the sensor measurement data, the simulator output voltage can be adjusted by adjusting the active power or reactive power of the load set in the simulator.

  The above is the outline of the power system state estimation device and the state estimation method. Next, the procedure (flow) of the power system state estimation method will be described. Note that detailed descriptions of detailed functional operations and methods will be given later as appropriate.

<Flowchart of power system state estimation method>
The outline of the power system state estimation method will be described using a flowchart.
FIG. 2 is a schematic flowchart for explaining a power system state estimation method.
In FIG. 2, the flow of the power system state estimation method is started (S200). First, in step S201, system configuration and device characteristic parameters are set (parameter setting unit 125, power system simulator 121, FIG. 1).

  Next, in step S202, the acquired sensor measurement data of the power system (power system measurement unit 110, FIG. 1) is input to the simulator unit 120 (FIG. 1). The sensor measurement data is supplied to the simulator unit 120 (FIG. 1) and the maximum likelihood state value calculation unit 130 (FIG. 1).

  Next, in step S203, power system power flow calculation, that is, simulation is performed (power system simulator 121, parameter setting unit 125, simulator data processing unit 122, FIG. 1). The sensor measurement data is also used in this tidal current calculation.

In step S204, the sensor measurement data of the power system and the simulation data of the simulator unit 120 (FIG. 1) are compared and referred to, and a plausible estimated value (the most likely estimated value and the maximum likelihood state value) is calculated ( Maximum likelihood state value calculation unit 130, FIG. 1).
Further, a signal corresponding to the deviation (difference) between the sensor measurement data and the simulation data is output to the parameter correction unit 140 (FIG. 1).

  In step S205, the parameter correction unit 140 (FIG. 1) corrects the parameters used in the power system simulator 121 (FIG. 1) according to the signal of the maximum likelihood state value calculation unit 130 (FIG. 1).

  In step S206, power system power flow calculation (simulation) is performed again using the corrected parameters (power system simulator 121, simulator data processing unit 122, FIG. 1).

In step S207, the result of the tidal flow calculation (simulation) performed again is compared with the sensor measurement data, and whether the difference is within a predetermined range (convergence) or exceeds a predetermined range (unconvergence), (Convergence determination), (maximum likelihood state value calculation unit 130, FIG. 1).
In the case of “unconvergence”, the process returns to step S205 and the calculation from step S205 is repeated.
In the case of “convergence”, the process proceeds to the next step S208.

  In step S208, the converged state estimation value is output (maximum likelihood state value calculation unit 130, FIG. 1). Then, the process proceeds to next Step S209.

In step S209, it is determined whether it is “sequential calculation” or “offline calculation”. In the case of “sequential calculation”, in order to estimate the state of the electric power system at the next time, the process returns to step S202, and the operations after step S202 are repeated.
In the case of “offline calculation”, no more data is input, so the flow ends (S210).

In the above flow, steps S202 to S204 are steps for calculating the most likely state value (maximum likelihood state value) from the sensor measurement data.
In addition, steps S205 to S208 are steps for correcting the parameters of the simulator so as to output a state value that is most likely in terms of probability.
The flow shown above is an overview. The detailed processing in the individual steps will be described later as necessary.

<Details of fluctuation factors>
The details of the fluctuation factors in the power system and its simulation will be described.
In FIG. 1, a description will be given separately for a variation factor and error related to the measurement data of the power system acquired by the power system measurement unit 110 and a variation factor and error related to the simulation data in the simulator unit 120.

<Variation factors related to power system measurement data>
First, the variation factors related to the measurement data of the power system will be described.
As for the variation factors related to the measurement data of the power system, there are the variation factors related to the power system and the variation factors related to the sensor measurement system.
<1> Fluctuating factors related to the power system
<1-1> Fluctuating factors related to the power system include fluctuations in the load (power consumption) and the amount of power generated by distributed power sources. Consumer power consumption varies depending on the time of day, season and weather. In addition, the amount of power generation as a distributed power source varies depending on the weather, weather, and season of solar power generation and wind power generation.
Details regarding changes due to weather and time will be described later.
<1-2> In addition, other fluctuation factors related to the power system include a change in the system configuration in the power system and variations in device characteristics. The impedance of the line, which is one of the device characteristics, is expressed by a combination of the resistance R and the inductance X per unit distance, but there is an error in the specification and the actual line characteristics.
In addition, there are variations in characteristics of power equipment and distribution lines, and variations due to deterioration over time (aging).

<2> Fluctuating factors related to sensor measurement systems
<2-1> As a variation factor related to the sensor measurement system, the voltage value obtained by measuring the sensor is greatly influenced by the characteristics of the analog-to-digital converter of the sensor. For example, there are restrictions on the number of quantization bits, hold time, sampling interval, etc., and state values cannot be arbitrarily collected. These signal characteristics may cause measurement errors.
In addition, there are stability of measuring equipment, change or deterioration of time constant, and influence by noise. There are also temperature at the time of measurement, changes in weather, and changes in measurement time. In addition, there are also appropriate and unsuitable locations for sensor mounting.
<2-2> In addition, there are delays in data transmission, data loss and errors as fluctuation factors related to the sensor measurement system. There is also the influence of noise during transmission.

<Variation factors related to simulation data>
Next, the fluctuation factors related to the simulation data on the simulator side will be described.
As for the fluctuation factors relating to the simulation data on the simulator side, there are fluctuation factors relating to the system model. The simulator outputs a unique calculation result, but if there is a calculation method selection, parameter error, or variation, the result includes the error or variation.

<3> Fluctuating factors related to system model
<3-1> Fluctuation factors related to the system model include approximation by calculation method, calculation error, and convergence factor. The approximation, calculation error, and convergence vary depending on the calculation method used.
<3-2> In addition, when calculating the load / power generation amount by simulation, there are factors for setting the simulation of the load / power generation pattern (model) and the simulation of the weather conditions. The result calculated in the simulation differs depending on how these are selected and set.
<3-3> In addition, there is equipment characteristics in the power system as a variation factor related to the system model. The results calculated in the simulation differ depending on how the characteristic values of the various characteristics of the power equipment and distribution lines are set and how the time delay characteristics of each equipment are set in the simulation.

<Appearance distribution data and fluctuation factors of solar radiation>
As described above, there are various fluctuation factors that may cause an error. Next, as an example, simulation of the data on the distribution of the occurrence of solar radiation and the relationship between fluctuation factors and results (voltage fluctuation of the PV interconnected system) taking the solar radiation and system voltage as an example. Voltage fluctuation >>
First, the appearance distribution data of the solar radiation amount is shown, and then the factors and results of fluctuations in the solar radiation amount and the system voltage will be described.

《Data on the distribution of appearance of solar radiation》
FIG. 4A and FIG. 4B show data on the appearance distribution of solar radiation.
FIG. 4A is an example of a diagram showing the relationship between time and solar radiation as a frequency distribution.
FIG. 4B is an example of a diagram illustrating the relationship between the amount of solar radiation and the frequency thereof as a time-series signal.
4A and 4B are based on actually measured data on a day when weather conditions are sunny and cloudy.
In FIG. 4A, the horizontal axis is time, and the vertical axis is the amount of solar radiation. The characteristic line indicated by reference numeral 1S is data measured (measured) at intervals of 1 second, and the characteristic line indicated by reference numeral 30M is obtained by averaging data measured at 30 minutes.
In FIG. 4B, the horizontal axis represents the amount of solar radiation, and the vertical axis represents the appearance frequency (frequency, accumulated seconds) of the amount of solar radiation. The characteristic line indicated by reference numeral 1S is data measured at intervals of 1 second, and the characteristic line indicated by reference numeral 30M is expressed by accumulating data in units of 30 minutes. Further, the peak of the characteristic line indicated by reference character F corresponds to a sunny weather condition, and the small peak of the characteristic line indicated by reference character C corresponds to a cloudy weather condition.

As described above, this day is sunny and cloudy weather conditions, and in FIG. 4A, the amount of solar radiation has a large amplitude in about several seconds. In FIG. 4B, there are two peaks in the frequency distribution.
4A and 4B, the characteristic line indicated by reference numeral 1S measured at intervals of one second is larger than the characteristic line indicated by reference numeral 30M, which is 30-minute average data, of the actual solar radiation amount. It represents changes more faithfully. However, if the simulation is performed based on the data measured at intervals of 1 second, a huge amount of calculation is required.
On the other hand, when a simulation is performed based on the characteristic line indicated by reference numeral 30M, which is 30-minute average data, the amount of calculation at the time of the simulation is reduced, but it is difficult to cope with a short-term weather change. is there.

Solar power generation, which is one of the renewable energies, receives solar radiation radiated from the sun by a panel that is set on the ground and converts it into electric power.
There are many fluctuation factors in the process, long-period factors have an astronomical positional relationship between the sun and the earth, and short-cycle factors are caused by the flow of clouds as shown in FIGS. 4A and 4B. There is a temporal variation in the amount of solar radiation that occurs. The appearance distribution of the amount of solar radiation in the weather such as sunny and cloudy becomes the appearance distribution of the amount of solar radiation having two peaks of sunny and cloudy.

<< Voltage fluctuation of PV interconnected system >>
FIG. 5A and FIG. 5B show data of voltage fluctuations (simulation results) of a system interconnected with PV (the photovoltaic power generation facility PV is connected to the power system).
FIG. 5A is a diagram showing a relationship (node voltage time-series signal) between a change in the amount of solar radiation with time and a voltage fluctuation (node voltage) of a PV-linked system.
FIG. 5B is a diagram illustrating a relationship between voltage fluctuation (node voltage) and frequency distribution (probability, probability density function, probability distribution) of a PV-connected system.
In FIG. 5A, the horizontal axis represents time (the sunshine changes depending on the time), and the vertical axis represents the node voltage. Reference numerals N1 to N5 represent node voltage characteristics of the nodes 1 to 5, respectively.
Further, in FIG. 5B, it is a node voltage, and the vertical axis is a probability (frequency distribution, probability distribution, probability density function). Reference numerals N1 to N5 represent node voltage characteristics of the nodes 1 to 5, respectively.

FIG. 5A and FIG. 5B are the results of system power flow calculation using the weather conditions shown in FIG. 4A and FIG. 4B and the power distribution system (FIG. 8) described later. In order to simplify the calculation, the solar radiation amount is multiplied by a coefficient so that the maximum output of the PV power generation amount is 1 MW, and the system is connected to the system (node 5, symbol N5).
In the example shown in FIG. 5B, two frequency peaks appear in the node voltage in the same manner as the solar radiation frequency distribution. From this, the effectiveness of considering not only time-series signals but also probability distributions can be seen in the study of voltage stabilization control. Therefore, a method for improving the simulation accuracy of the simulator will be described for a power system having a variation factor.

<Probabilistic and statistical treatment of power system-related events and behaviors>
One method in the case of a combination of a plurality of variation factors is to measure a change in the time direction. However, this method observes all combinations of multiple fluctuation factors, which increases the observation time and may not allow observation of rare events.
Therefore, when considering the introduction of natural energy such as solar and wind power, the occurrence of disasters and accidents, and fluctuations in measurement and communication means (in other words, reliability) in the power system, these behaviors are treated stochastically and statistically. It is preferable.
When many variables are combined, the law of large numbers and the central limit theorem work as basic principles. By using these properties, it is possible to analyze a phenomenon assuming a normal distribution from the beginning without considering the probability distribution of each variation factor.

However, as described above, in the weather in which sunny and cloudy are mixed, the appearance distribution of the amount of solar radiation has two peaks that are largely separated. In such a system, the central limit theorem does not work sufficiently, and the phenomenon occurs while leaving the probability distribution of the variation factors described above.
Therefore, further ingenuity is required to perform the calculation in consideration of the above-mentioned variation factors in the power system. Next, those methods will be described in detail.

<First embodiment, part 3>
Next, the power system state estimation device and the power system state estimation method according to the first embodiment of the present invention will be described in more detail. Note that the description below also serves as a description of a power system state estimation device and a power system state estimation method.
As described above, the power system state estimation device and the power system state estimation method according to the first embodiment of the present invention include the following estimation process and correction process.
That is, the “estimation process” is a process of calculating a state value that is most probable from the sensor measurement data.
The “correction process” is a process for correcting the parameters of the simulator so as to output the state value that is most likely to be probable.
These processes will be described in further detail in order.

<Details of estimation process>
As described above, when there are a plurality of fluctuation factors, handling in calculation may be complicated. In order to reduce this complexity, the method described below will be used.
Note that the following method will be described assuming that the type and number of sensors are sufficiently secured in the power system. A method when the number and type of sensors are not sufficiently secured will be described later.

《Calculation of superposition of observation noise》
Noise in the power system and noise in the sensor measurement system are referred to as observation noise, and noise in the simulator is referred to as simulator noise.
As described above, when there is an observation error due to observation noise and a system error (simulator data error) due to simulator noise, a target for correcting the parameters of the simulator is set stochastically.
For simplicity, it is assumed that the probability distribution of the observation error and the system error is a normal distribution. At this time, the target probability distribution is set by convolution of the probability distribution of the observation error and the system error. This basis is based on the relationship between prior and posterior probabilities in Bayesian statistics. In particular, when assuming a normal distribution, the result of convolution integration is also a normal distribution. Therefore, the result of integration can be obtained only by calculating the mean and variance as shown in the following equation (1).
When both noises are normal distributions N ₁ (μ ₁ , σ ₁ ) and N ₂ (μ ₂ , σ ₂ ), the estimated state values are also represented by normal distribution N ₃ (μ ₃ , σ ₃ ). Then, the average and the variance are as follows.

FIG. 11 shows the relationship between the normal distributions N ₃ (μ ₃ , σ ₃ ) of the state values estimated from the normal distributions N ₁ (μ ₁ , σ ₁ ) and N ₂ (μ ₂ , σ ₂ ). It is a schematic diagram.
In FIG. 11, the horizontal axis is the state value, and the vertical axis is the probability density.
As shown in FIG. 11, when there are two normal distributions N ₁ and N ₂ , these distributions can be viewed as one normal distribution N ₃ by convolution integration. The above equation 1 expresses the relationship shown in FIG.

In the above, the case where both of the noises (observation error and system error) have a normal distribution has been shown, but details of cases other than the normal distribution will be described later.
Note that, as will be described later, the noise factor can be handled concisely by focusing on the partial system rather than handling the entire target system.

《Sequential calculation according to time transition》
FIG. 6 is a schematic diagram illustrating a procedure for performing sequential calculation according to a transition of time in the power system state estimation method according to the first embodiment of the present invention.
In FIG. 6, the horizontal axis is time (sampling time), and the vertical axis is the state value. Further, the sensor measurement data 61 on which the observation noise is superimposed is shown as a distribution 63 having a spread centering on the sample point. In addition, the most probable estimated values calculated by the above formula are superimposed.
The distribution 63 is a distribution that approximates a normal distribution in a short sampled time.
Thus, the estimated value 62 is sequentially calculated while inputting the sensor measurement data (corresponding to the feedback to step S202 in the case of sequential calculation in step S209 in FIG. 2), thereby following the operation of the power system with fluctuation. To do.

There are several challenges when targeting an actual power system. One is a normal distribution assumption. The above-mentioned solar radiation amount measurement data has a frequency distribution having two peaks. In order to handle rare events, it is desirable to faithfully use the probability distribution obtained by such measurement. On the other hand, to see typical behavior, it can be approximated as a mixed problem of two normal distributions (clear and cloudy). As a method for separating the normal distribution to be mixed, for example, an EM algorithm which is one of unsupervised learning methods can be used.
For example, when the purpose is voltage stabilization control, two normal distributions, sunny and cloudy, are assumed and the calculation is speeded up by switching between the two.
By switching between these two methods as needed for calculation, it becomes possible to use both high accuracy and high speed calculation.

  Another problem is how to set the distribution. In principle, dispersion can be obtained by statistically processing sensor measurement data, but it is not easy to measure many events in advance. On the other hand, distribution has the meaning of data reliability. The reliability of data with small variance is high, and vice versa. For example, when there is data loss and the measurement data from one cycle before is used as a substitute, the reliability of the data decreases. This difference can be expressed by the size of the variance.

<Details of revision process>
Next, the correction process will be described in more detail. A method for calculating the power flow and a method for setting the parameter correction amount will be described in order.

《Tidal current calculation method》
As a power flow calculation method, a method of calculating the power flow of the entire power system by dividing the power system into a plurality of partial systems and combining the calculation results of the individual partial systems is used. As an example, a backward forward method (BF method: Backward Forward) is used. The backward forward method repeats the following procedure for a branched radial power system. If the same division procedure is provided, the method is not limited to the BF method.
FIG. 7A is a schematic diagram showing that current calculation is performed for each partial system in the power system state estimation method according to the first embodiment of the present invention.
FIG. 7B is a schematic diagram showing that cumulative calculation of current along the line is performed in the power system state estimation method according to the first embodiment of the present invention.
FIG. 7C is a schematic diagram showing that the distribution calculation of the voltage along the line is performed in the power system state estimation method according to the first embodiment of the present invention.

(1) Current calculation for each partial system As shown in FIG. 7A, power is supplied from the distribution facility (distribution substation) 701 to the partial systems 702, 703, 704, and 705. Therefore, current flows through the partial systems 702, 703, 704, and 705. This current (I2, I3, I4, I5) is calculated for each of the partial systems 702, 703, 704, 705.
Here, the partial system is configured by a load, a power source, and the like that are connected to the node. Regardless of the scale of the power system, the relationship between the input, output, and internal parameters is simplified by using a partial system (one input, one output system) along the line as a unit of calculation.

(2) Cumulative calculation of current As shown in FIG. 7B, the calculated currents (I2, I3, I4, and I5) in the sub systems 702, 703, 704, and 705 are cumulatively calculated as currents along the line ( J1, J2, J3, J4, J5). For example, the current I4 of the partial system calculated by the partial system 704 in FIG. 7A and the current J4 in FIG. 7B are not necessarily equal. This is because the current J4 includes the influence of the current I5 (J5) of the partial system 705.

(3) Voltage distribution calculation As shown in FIG. 7C, the voltage distribution calculation at each location along the line is performed.

(4) Taking in estimated values In the above tidal flow calculation procedure, take in the estimated values calculated in the estimation process.

(5) Convergence determination The simulator parameters are modified so that the simulator calculation results asymptotically approach the estimated values. That is, feedback control for correcting the simulator parameter is configured in the partial system (corresponding generally to the parameter correction unit 140 in FIG. 1 and step S205 in FIG. 2).
When the difference between the estimated value calculated in the estimation process and the calculation result of the simulator falls within a predetermined range, the convergence is determined as having converged.

The backward forward method described above is a power flow calculation method that does not use the power equation of the entire power system. Compared to the method using the power equation of the entire power system, even when the type, number, and location of the connected photovoltaic power generation are updated, it is not necessary to recreate the power equation each time it is updated There is.
Further, by dividing into partial systems, the relationship between input, processing, and output is simplified, the causal relationship is easy to understand, and correction calculation can be realized stably.
This backward forward method corresponds to solving the equation by decomposing the admittance matrix (or impedance matrix) included in the power equation for each partial system. The admittance matrix is often a sparse matrix in which components other than near the diagonal are zero, which is why the power system can be divided into partial systems.
The present invention is not limited to the backward forward method, and can be used in the same manner as described above as long as the power system is divided into partial systems for calculation.

<Setting method of parameter correction amount>
Next, a parameter correction amount setting method will be described.
The parameter correction amount setting method depends on the configuration of the partial system. For example, consider a case where a simulator parameter (load) is corrected using voltage / current measurement data.
First, the difference (ΔV, ΔI) between the voltage / current measurement data and the simulator output is calculated, and then the load correction amount set in the simulator is calculated so as to reduce this difference. The parameter correction amount is calculated by multiplying the difference value calculated above by the coefficient k and a sign (±) (Equation 2A, Equation 2B). This is equivalent to replacing the sensitivity coefficient (proportional coefficient between state values) with a minute updated value. Instead of preparing sensitivity coefficients in advance, convergence is achieved by iterative calculation.
The coefficient k and the sign (±) may be set based on the system configuration and the system state. Or you may set based on the measurement data of a system state. The coefficient k and the sign (±) may be fixed, or may be changed according to the system state.

Here, the coefficient k and the sign (±) are determined based on the direction and speed of convergence in the target system configuration and system state.
Load P correction amount ΔP = sign (±) · ΔV · k (Equation 2A)
Load P correction amount ΔP = sign (±) · ΔI · k (Equation 2B)
The above procedure is summarized as the flowchart of FIG.
The above procedure corresponds to a 1-input 1-output system Kalman filter. Since the power equation is a nonlinear equation, a form of a nonlinear Kalman filter such as an extended Kalman filter (EKF) is used as the Kalman filter.
The main features of this method are as follows.
# Since it is a calculation procedure in units of partial systems, it can easily cope with changes in system configuration.
# Non-linear characteristics and program operation can be incorporated into the load.
# It is possible to follow the change of the target by calculating sequentially while changing the time.
# When sensor measurement data is not available, it operates as a definitive simulator.

<When the types and number of sensors are limited>
Next, the case where the types and number of sensors installed in the power system are limited will be described.
When the types and number of sensors installed in the power system are limited, there are cases where data available for parameter correction is insufficient.
As an example of a coping method in this case, a method of allocating sensor measurement data within a range between sensor measurement points is shown.

<1> It is assumed that voltage sensor measurement data V1 and sensor measurement data V2 can be collected at two points sandwiching N partial systems. In this case, the following procedure is performed.
<2> First, state values are calculated in the range of N subsystems by tidal current calculation using initial parameters.
<3> Next, the sensor measurement data V1 and the sensor measurement data V2 are distributed to N partial systems based on the calculated values.
<4> The N partial system parameters are corrected so as to approach the state value reflecting the distribution value.
<5> Repeat the above procedure and wait for convergence.

≪Power system calculation example≫
Next, a calculation example of the power system will be described with reference to the power system shown in FIG.
FIG. 8 is a diagram illustrating an example of a power system.
In FIG. 8, a power system 800 includes nodes 1 to 5 (N1 to N5), a distribution substation 801 supplies power to the node 1 (N1), and loads via distribution lines L12, L23, L24, and L45. Power is supplied to Loads 2, 3, 4, and 5.
Distribution lines L12, L23, L24, and L45 are distribution lines in node 1 and node 2, node 2 and node 3, node 2 and node 4, and node 4 and node 5, respectively.

Further, the distribution lines L12, L23, L24, and L45 each have a line impedance Z.
Loads 2, 3, 4, and 5 that are loads may include not only loads that consume power but also distributed power sources that generate power themselves.
A voltage / current sensor (M) 823 is provided on the distribution line L23 side of the output end of the node 2 (N2), and a voltage / current sensor (M) is provided on the distribution line L45 side of the output end of the node 4 (N4). M) There is 845.
At this time, the load on the nodes 2 (N2) to 5 (N5) is estimated using the sensor measurement data of the nodes 2 (N2) and 4 (N4).

Here, the simulator is used for two purposes. One is based on the power system itself and uses the simulator output as sensor measurement data.
The other is an original simulator and obtains a simulator output while correcting parameters.
For the initial value of the latter simulator, parameters different from the former (with errors) are set, and the state of convergence is observed. In the example shown in FIG. 8, the load of the node 2 (N2) to the node 5 (N5) is used as a parameter.

[Procedure 1]
Loads are set in the nodes 2 to 5 and the power flow is calculated to calculate the voltages and currents of the nodes 2 and 4. This is regarded as sensor measurement data collected from an actual power system. The loads on the nodes 2 to 5 are set to 500 kW, 300 kW, 500 kW, and −200 kW in order.
Note that the load value of the node 2 to the node 4 is positive while the load value of the node 5 is negative means that the load (distributed power supply) that generates power at the node 5 is connected. doing.

[Procedure 2]
The power flow calculation of the simulator is performed to obtain the voltage / current of node 2 and node 4. This is the simulator output. The initial values of loads including errors set in the nodes 2 to 5 are 400 kW, 700 kW, 700 kW, and 300 kW.

[Procedure 3]
The most likely estimated value is calculated using the sensor measurement data and the simulator output prepared in steps 1 and 2.
However, when adding random noise with an average noise of 0 and a fixed variance, the estimation value calculation procedure is equivalent to a noise removal filter (exponential filter).

[Procedure 4]
At nodes 2 and 4, voltage / current errors ΔV and ΔI of the power system and simulator are calculated. Then, a load correction amount ΔP is calculated so as to reduce the error. k is a coefficient (Equation 3).
ΔP ₂ = + ΔV ₂ · k
ΔP ₃ = −ΔI ₂ · k
ΔP ₄ = + ΔV ₄ · k (Equation 3)
ΔP ₅ = −ΔI ₄ · k

[Procedure 5]
Return to [Procedure 2]. In this example, since it is assumed that sequential calculation is performed over time, convergence determination is not performed.

《Simulator parameter correction operation》
Next, the simulator parameter correcting operation will be described.
FIG. 9 is a diagram illustrating a simulator parameter correcting operation.
In FIG. 9, the horizontal axis indicates the number of convergence calculation iterations, and the vertical axis indicates the load amount (load, kW). In addition, characteristic lines indicated by reference numerals N1 to N5 indicate the calculation results of loads at the nodes 1 to 5, respectively.
The initial values of the parameters at the start of the calculation include errors and the values fluctuate from one calculation to the next. As the calculation is repeated, the parameters gradually approach the load in the power system. Yes. As a result of repeated calculation, the error is within 1%.

<< Estimated results following temporal load fluctuations >>
Next, a simulation estimation result following the temporal load fluctuation is shown.
FIG. 10 is a diagram illustrating a simulation estimation result following the temporal load fluctuation.
In FIG. 10, the horizontal axis represents time (time transition), and the vertical axis represents the load amount (load, kW). Reference numerals N1 to N5 indicate load characteristic lines of the nodes 1 to 5, respectively.
The simulation result shown in FIG. 10 shows that the PV power generation amount is calculated from the solar radiation amount data shown in FIG. 4A and FIG. 4B and is connected to the node 5, and the remaining loads from the node 2 to the node 4 are fixed loads 500 kW and 300 kW. , 500 kW is set and calculated. In addition, voltage / current can be measured at node 2 and node 4.
Note that the load values of the nodes 2 to 4 show positive values, whereas the load value of the node 5 shows a negative value. ) Is done.

On the other hand, the simulator starts from a state in which each node load is unknown, and estimates the load of each node using the sensor measurement data (the voltage and current of the nodes 2 and 4). As described above, the load on the power system and the simulator is within 1% error.
Note that this procedure can be incorporated by simulating non-linear control operations such as output suppression of PV power generation.
Conversely, a target value may be set and a control amount for realizing the target value may be derived.

As described above, the power system state estimation device and the power system state estimation method according to the first embodiment of the present application improve the power quality of the power system and reduce the cost for stabilization control, and The state estimation method can be provided.
In addition, even when the fluctuation factor of the power system cannot be assumed to be a normal distribution, it can be applied and has an effect of high versatility.
In addition, not only observation errors, which are fluctuations in equipment connected to the power system, but also system errors (simulator data errors) of a simulator that simulates the state are taken into account, so that there is an effect of high accuracy.
Moreover, since the tidal current calculation that is most likely to be probabilistic is performed, there is an effect of realizing power control that is most likely to be probable.
In addition, since the factors and results of fluctuations occurring in the power system are handled stochastically, there is an effect that the cost for stabilization control can be reduced as compared with the case of handling with time series signals.
In addition, in an electric power system in which a large number of variable factors are combined, there is an effect that the configuration of a control device that deals with a rare event can be planned and dealt with on a probabilistic basis.

  In addition, the power system state estimation device according to the first embodiment of the present application is provided in the power system, and the power system is provided with reference to the most probable state value calculated by the power system state estimation device. If you configure a power system control system that controls power equipment including voltage distribution substations (voltage regulator SVR, reactive power compensator SVC <Static Var Compensator>, etc.), the power quality of the power system can be improved at low cost, Stabilization control is possible.

<Other embodiments>
The power system state estimation device and the state estimation of the first embodiment of the present application have been described above. Moreover, the power system control system provided with the power system state estimation device of the first embodiment in the power system has been described. However, the embodiment of the present invention is not limited to the embodiment described above.
Next, other embodiments will be described.

《Graphical user interface》
Further, the power system state estimation apparatus of the present invention may be provided with a GUI (graphical user interface).
The power system state estimation apparatus according to the present invention sets a probability distribution of fluctuation factors based on actual measurement data, and also sets an arbitrary probability distribution by GUI using the apparatus configuration as shown in FIG. Is provided.
FIG. 13: is a figure which shows the structure of the state estimation apparatus of the electric power grid | system of this invention provided with GUI.

In FIG. 13, an arithmetic device 903, a memory device 904, input data / probability distribution data 911, a power flow calculation result / error calculation result 912, a communication device 905, a display device 901, and an input device 902 are connected via a bus 913. Yes.
Information on the voltage and current of the power system is acquired via the communication device 905, and by referring to the data, the arithmetic device 903, the memory device 904, the input data / probability distribution data 911, the power flow calculation result / error calculation result 912 The most likely state estimation of the power configuration is calculated by the combination of. The display device 901 and the input device 902 display and input various data of the power system state estimation device.

The GUI is realized by a combination of the display device 901 and the input device 902, and has a function of editing and processing the probability distribution. For example, with regard to the probability distribution of the amount of solar radiation having two peaks of sunny and cloudy as described above, a distribution of only sunny or cloudy is extracted by a GUI operation to create a probability distribution when each appears independently, and The impact on the system can be evaluated.
Moreover, the calculation result based on the probability distribution can be displayed visually and easily by this GUI function. This function also has a small merit if a normal distribution is assumed, but when setting an arbitrary probability distribution, visualizing and displaying the probability distribution obtained as a calculation result is necessary to grasp the probabilistic features. It will be a big merit.

《Markov chain Monte Carlo method》
In the above, the fluctuation factors such as measurement data are described as probability distributions, and the influence on the power system is evaluated as a combination of probability distributions.
As this method, a combination of probability distributions of a plurality of fluctuation factors in the power system is set by MCMC (Markov Chain Monte Carlo), and the power flow is calculated based on the set conditions, and the state value of the power system is calculated. A probability distribution may be calculated.
The MCMC method is a method known in the field of probability statistics, and creates a procedure for sampling a probability distribution in a jump. By accumulating the results of simulating system operation using random variables sampled by the MCMC method, an output probability distribution corresponding to the input probability distribution is obtained. In a power system, for example, a probability distribution indicating fluctuations in distributed power sources is sampled by the MCMC method to set a power generation amount, and a procedure for calculating a power flow when the power generation amount is connected is repeated to calculate a calculated power flow The probability distribution of power flow can be obtained.

The advantage of this method is that the probability distribution can be approximated with a smaller number of samplings compared to the random number generation by the Monte Carlo method, which is effective in reducing the calculation load.
In addition, since the probability distribution of the power flow generated by the power system due to a plurality of variation factors in the power system is a combination of the probability distributions of the plurality of variation factors, reduction of the calculation load by the MCMC method is a great merit.
Further, since an arbitrary distribution is allowed without assuming a normal distribution as the probability distribution, the probability distribution of the amount of power generation of renewable energy (solar power generation) as shown in FIG. 4A can be taken in.

《Stochastic equation of state》
In the above, the case where the observation error and the system error are not a simple normal distribution has been mainly described. However, depending on the situation, there may be a combination of simple normal distributions. In this case, the calculation amount may be reduced by using a probability state equation described below.
When the observation error and the system error are normally distributed, it is easy to formulate using the mean and the variance. The state equation form is created by combining the observation equation including the observation error and the system equation including the system error. This can be called a stochastic equation of state because it incorporates the variance.

Each variable of this state equation is a scalar or a vector. By repeating the update procedure of the observation equation and the system equation based on the sensor measurement data according to the temporal transition, the state value becomes asymptotically asymptotically most probable.
A Kalman filter can be used as a solution to this state equation. Derivative methods such as extended Kalman filter and particle filter can be used. Since these are methods for obtaining a solution satisfying the most probabilistic likelihood, it is possible to calculate a state value that is most likely to be probabilistic among observation errors and system errors.
This calculation process can be called simulation, state estimation, Kalman filter, or the like, but it is also a method based on Bayesian statistics.

When this method is used, a physical model of the power system described by mathematical formulas and a probabilistic model such as fluctuation factors such as weather conditions or device characteristics that are difficult to set in advance are created. By combining as a state equation, calculation of a state value with a clear calculation basis is realized.
In the stochastic state equation, the magnitude of the variance indicates the reliability of the signal. Thus, since the degree of signal reliability can be adjusted by controlling the magnitude of dispersion, it can be used to arbitrarily set conditions for power system analysis.

<Method for considering frequency components>
Indicating state fluctuations with probability distribution is a widely used technique, but notation with probability distributions cannot reflect temporal characteristics of fluctuations, in other words, frequency characteristics. In an actual power system, there are many variations with different frequency characteristics. On the other hand, the capacitance and inductance of the power system set in the simulator have frequency characteristics.
Therefore, by considering the frequency component, it is possible to reduce the amount of calculation, stabilize the control, and reduce the energy or cost required to drive the control device.

As an example of this method, fluctuation factors occurring in the power system, observation errors included in measurement means such as sensors, and system errors included in a simulator that simulates the operation of the power system are represented as probability distributions that take frequency characteristics into account. Represent. This is obtained by inputting a variation factor as a time-series signal and decomposing it into frequency components and then calculating a probability distribution.
For example, the component that changes in a few seconds depending on the flow of the cloud and the component that changes depending on the solar altitude of the day are separated by the difference in frequency, such as fluctuations in the amount of power generated by the photovoltaic power generation described above. Can be treated as separate probability distributions. For stabilization control, control devices having different frequency characteristics can be combined.
For example, the above-mentioned variation in units of several seconds may be controlled by combining a storage battery, and the variation in units of one day may be controlled by combining a voltage regulator SVR (Step Voltage Regulator) on the system side. Specifically, the power conditioner PCS installed on the customer side performs the stabilization control using the storage battery only for the high frequency component of the fluctuation, and the remaining low frequency component is assigned to the system side stabilization control, This can be done based on the frequency characteristics of the probability distribution of variation.

In the above procedure, after calculating a control signal for stabilization control, the control signal is separated into frequency components, so that stabilization control can be realized by combining control devices having different frequency characteristics. As a result, it is possible to reduce the cost of voltage stabilization control by using a voltage regulator SVR (Step Voltage Regulator) instead of an expensive storage battery for daily fluctuations.
Also, as a preliminary preparation, prepare a time-series signal of fluctuations that occur in the power system, calculate the control signal to cope with this as a time-series signal, develop this control signal into frequency components, and then distribute the probability distribution Is obtained as a probability distribution with the frequency as a variable, in other words, the frequency characteristic of the probability distribution.

FIG. 12A is a diagram schematically illustrating frequency characteristics of the probability distribution.
FIG. 12B is a diagram schematically illustrating the frequency characteristics of the control device.
In FIG. 12A, the horizontal axis represents frequency, the vertical axis represents dispersion and average, and frequency characteristics of dispersion and average are shown.
As schematically shown in FIG. 12A, the mean (AVE) and variance (VAR) versus frequency can be plotted. Alternatively, although not shown, the frequency can be divided into several regions and tabulated as an average and variance for each divided frequency region.

In FIG. 12B, the horizontal axis represents frequency, the vertical axis represents control amount, and the frequency characteristics of the voltage regulator SVR, the storage battery BAT, and the reactive power compensator SVC are shown.
As shown in FIG. 12B, control devices in the power system, for example, the voltage regulator SVR, the storage battery BAT, the reactive power compensator SVC, and the like have frequency response characteristics that depend on the control method. It is desirable to drive the control device with a signal.
In addition, when the variance in the divided frequency domain is smaller than the variance in the entire frequency domain, the control signal amplitude is distributed to be small, which has the effect of reducing the energy or cost required to drive the control device. .

In addition, since the control device is not forcibly driven based on the control signal having the mismatched frequency characteristics, the energy or cost required for driving the control device can be reduced.
The control equipment and control method used for stabilization control of the power system should be selected from the combination of the frequency characteristics of the fluctuation factor probability distribution, the frequency response characteristics of the control equipment, as well as the control level range, installation location, equipment cost, etc. become. When the number of combinations of these conditions increases, it is possible to perform an equipment plan for a control device by preparing an evaluation function in advance and then performing a combination optimization calculation.

《Convolution integral》
In the maximum likelihood state value calculation unit 130 (FIG. 1), the observation error and the system error are compared. At this time, the respective errors are combined.
A probability distribution combining observation errors and system errors is obtained by convolution integration of both probability distributions. There are several methods for calculating the convolution integral as follows.

<1> Calculation method using mean and variance in the case of normal distribution When the probability distribution of both observation error and system error is expressed by normal distribution or approximated by normal distribution, the result of convolution integration also becomes normal distribution. Therefore, it can be handled by the calculation based on the relational expression between average and variance (Formula 1). It is a method that only needs to calculate the mean and variance. In this case, the calculation amount is small.

<2> Method of performing integral calculation based on the principle of convolution integration The probability distribution is not necessarily a normal distribution (for example, FIG. 4B). When the probability distribution is an arbitrary distribution, the result of convolution integration cannot be expressed in a form known in advance. In such a case, there is an integration calculation method based on the principle of convolution integration. This is a method in which random variables are represented discretely and all combinations are calculated and added. In this case, not only a normal distribution but also an arbitrary probability distribution can be calculated.

<3> Monte Carlo method The Monte Carlo method represents a random variable discretely, but uses a randomly selected combination instead of all combinations. In this case, there is a feature that the probability distribution can be handled even in an arbitrary distribution and the amount of calculation is reduced as compared with the method of performing the integral calculation based on the principle of convolution integration.

<4> Markov chain Monte Carlo method This method uses the Markov chain Monte Carlo method to solve the convolution integral of the probability distribution of observation errors and system errors. In this method, a convolution integral of an arbitrary probability distribution can be easily calculated. In addition, there is a feature that the influence of various fluctuation factors appearing in the power system can be evaluated.

<5> Method of approximating an arbitrary probability distribution with a combination of multiple normal distributions This method approximates an arbitrary probability distribution with a combination of multiple normal distributions, and performs a calculation equivalent to convolution integral by calculating the mean and variance. .
For example, the probability distribution of the solar power generation shown in FIGS. 4A and 4B is a combination of the power generation amount distribution in a clear time zone and a cloudy time zone, and is a normal distribution in two states, sunny and cloudy. It is a method to express by. This method is characterized in that the calculation load is light and that it is suitable for seeing the change in the average value.

<6> Method of narrowing the calculation range based on the voltage range specification This method is a method of reducing the calculation load by using constraints specific to the power system in the calculation process of convolution integration.
In the voltage estimation, the voltage range of the low voltage distribution system is defined as 101 ± 6V. If there are no rare fluctuation factors or accidents that rarely occur, the voltage does not deviate significantly from 101 ± 6V. Therefore, there is a method of considering that it is not necessary to calculate a voltage greatly deviating from 101 ± 6 V in the calculation process of convolution integration.
Assuming a normal distribution, the tail extends from infinity to infinity, so in principle it must be calculated from infinity to infinity. In contrast, the calculation load is reduced by narrowing the calculation range based on the voltage range specification. In addition, there is a feature that tends to converge.

Further, the above property is not limited to the voltage range, and the property that no more power than the supplied power flows can be used.
In addition, when the voltage range is defined and the upper limit of power is determined as described above, the property that the current that flows has an upper limit can also be used.
Moreover, the property regarding electric power, such as the property that a frequency is 50 Hz or 60 Hz, can be utilized.

<< Equation for power system >>
When the power system simulator expresses the entire power system or a partial system as a system model, a power equation can be used. However, there are various methods for describing the power system other than the power equation.
Since the power system is also an electric circuit, it can be described by an equation focusing on voltage or current. Still, load behavior may be described in terms of power. For example, there is a constant power operation by an inverter circuit. The inverter circuit performs feedback control so that constant power flows while measuring the voltage and current with sensors. Thus, the calculation of power system and load is described separately in two.

Procedure (1) Describe the power system with an electric circuit whose elements are voltage, current, and impedance.
Step (2) Describe the load that inputs and outputs power using the state quantities (voltage, current) determined in step (1). Since the calculation results of both interact, step (1) and (2) Is repeatedly calculated, and it is determined that convergence has occurred when the amount of change becomes sufficiently small.
This method is not a simple method, and the apparatus configuration for controlling a constant power load using an inverter circuit actually performs the above-described repetitive calculation. That is, the constant power load feedback-controls the voltage and current values so as to consume constant power (product of voltage and current). When the solar power generator sells surplus power via the inverter circuit, the surplus power flows to the grid side based on the sensor measurement data of the voltage and current at the connection point between the inverter circuit and the power grid. To feedback control.

In procedure (1), if the load is regarded as an impedance load, a definite solution can be obtained because it is a linear equation. Using the result, calculate the load behavior of (2). Repeat steps (1) and (2) to converge to a stable state.
This method does not require a procedure for locally linearly approximating a nonlinear equation like the method of solving a power equation by the Newton-Raphson method, and has a feature that it does not fall into a local solution.

《Sensor measurement data correction method》
Sensor measurement data may cause measurement failure, measurement time interval disturbance, time delay, or the like due to factors such as the characteristics of the sensor itself, communication delay, and noise.
When the target characteristic changes with time, if the asynchronous sensor measurement data is used, the target characteristic cannot be collected correctly. If the device characteristic is corrected using the sensor measurement data as it is, an error will occur in the correction result.
As a method of generating a synchronous signal from an asynchronous signal, a continuous signal interpolated in time from sensor measurement data collected discretely can be reproduced and resampled at an appropriate time interval.

Since the characteristics of the object change as time passes after the sensor measurement data is obtained, the reliability of the measurement data decreases. For this purpose, the variance of the sensor measurement data described above is expressed as a function of the passage of time from the measurement time.
In simple terms, the variance takes a minimum value at the measurement time, and the variance increases with time. In a stochastic equation of state, a state quantity with a large variance reduces the contribution to system operation.

With this method, it is not necessary to take the timing of solving the stochastic state equation in accordance with the collection timing of the sensor measurement data, and the timing can be managed separately.
When sensors with different sensor measurement periods coexist, it is not necessary to perform interpolation processing in the time direction based on the respective periods, and the dispersion value may be adjusted based on the period.
Similarly, dispersion can be set based on stability / unstableness of sensor operation, aging, temperature characteristics, number of quantization bits of AD conversion, frequency characteristics, and the like, and its contribution can be controlled.
In this method, the degree to which the state quantity included in the state equation contributes to the system operation is controlled by dispersion. This is expressed by reducing the variance of the state quantity of interest or the state quantity to which a weight is to be added in the target characteristics.
Thus, it is possible to estimate a state value at an arbitrary time at an arbitrary position where sensor measurement data cannot be obtained.

<< Method used to control power system of state value >>
There are methods such as centralized control and distributed control as methods for stabilizing control of the electric power system. Centralized control collects a wide range of state values of the power system and calculates a control signal for stabilization control, whereas distributed control is control with a narrowed range.
In any case, it is desirable that the current state value can be collected correctly, but in reality, there are restrictions on the type of sensor, installation location, data transmission, and the like.
As a method of removing this restriction, by operating the mathematical model in the power system using the sensor measurement data, it is possible to calculate an arbitrary state value of the power system at an arbitrary location and at an arbitrary time. This state value can be used for controlling the power system.

<Data compression>
As described above, there are methods such as centralized control and distributed control as methods for stabilizing the power system. In order to perform control, it is necessary to transmit data collected by sensors installed in the power system to the control device.
The more sensors there are, the more they can be used for control. However, on the other hand, the amount of data increases. Since the data transmission means has an upper limit on the amount of data that can be transmitted per unit time, an increase in equipment is required to increase the amount of data.
The power system state estimation apparatus (state estimation method) according to the first embodiment of the present invention calculates a probability distribution of state values of the power system. In the calculation process, the transmission data is converted into codewords based on the probability distribution of the state values of the power system to reduce the data amount.

The basic principle of data compression is to generate a code word based on the occurrence probability of an event. For example, there is a Huffman code generation procedure. The first embodiment of the present invention further includes means for generating or switching a code word used for data transmission or storage based on the occurrence probability of an event occurring in the electrical system.
By providing this means, there is an effect of improving the efficiency of transmission or storage.
Note that the codeword generation algorithm is not limited when the transmission data is converted into the codeword. In order to reproduce the original data, a means for preparing a common codeword or probability distribution for both compression and decompression is prepared. Several methods for this are shown.

(1) Method of preparing a plurality of codeword tables
(1-1) Select a codeword table based on the actual probability distribution and compress the data.
(1-2) Append the selected table number to the compressed data.
(1-3) Transmit or store the compressed data created by compression along with the table number.
(1-4) When decompressing, select the codeword used from the table number and decompress the compressed data.

(2) Centralized management of codeword table
(2-1) A codeword table to be used in advance is passed from the data management unit that receives the compressed data to the data transmission unit that transmits the compressed data.
(2-2) The data transmission means performs data compression using the received codeword table.
(2-3) Use a common codeword for the data management means and the data transmission means.
(2-4) When there are a plurality of data transmission means, different codewords may be assigned to each.
The data processing based on the probability distribution can perform variable control of the data transmission time interval in addition to data compression by codeword conversion. In measurement locations or time zones where fluctuations are estimated to be small in probability, communication traffic can be reduced by increasing the data transmission interval.

Conversely, in a measurement location or time zone where fluctuations are estimated to increase stochastically, the data transmission interval is shortened, the temporal resolution of stabilization control is increased, and the effect of improving power quality is obtained.
Furthermore, the sampling period of the sensor may be variably set based on the estimation result.
Moreover, the wattmeter installed at the connection point of the customer and the power system has a configuration in which the power is calculated based on the calculation formula of power = voltage × current based on the measurement result of the voltage and current by the sensor. The calculated power amount is transmitted to an electric power company for the purpose of charging, for example.
Such a wattmeter samples at a frequency higher than the data transmission interval in order to measure voltage and current waveforms. A cumulative value of about 30 minutes is calculated as the amount of power. Therefore, the wattmeter circuit has voltage and current waveform measurement data. When short-term fluctuations are estimated, transmission of voltage and current waveform measurement data collected by a power meter can be used to stabilize power quality.

DESCRIPTION OF SYMBOLS 100 State estimation apparatus of electric power system 110 Electric power system measurement part (electric power system measurement means)
111 Power System Measurement Data Unit 112 Measurement Data Processing Unit 113, 123 Summation Unit 114 Error, Observation Error 124 Error, System Error 120 Simulator Unit (Simulation Calculation Means)
121 Power System Simulator 122 Simulator Data Processing Unit 125 Parameter Setting Unit 130 Maximum Likelihood State Value Calculation Unit (Maximum Likelihood State Value Calculation Unit)
140 Parameter correction unit (parameter correction means)
300 Power system (distribution system)
301, 701, 801 Distribution substation, distribution facility 302, L12, L23, L24, L45 Distribution line 303 Voltage regulator 304 Pole transformer 311, 313 Distributed power supply 312 System battery 314A, 314B, 314C Consumer 702, 703, 704, 705 Partial system 800 Power system 823, 845 Voltage / current sensor

Claims (14)

A power system state estimation device for estimating a power system state quantity,
A power system measuring unit for receiving appearance distribution information of the amount of power input / output by a power source or a power consuming device linked to the power system;
A simulator unit that calculates the state estimation of the power system based on the received distribution information of the electric energy;
A maximum likelihood state value calculation unit for calculating a state value probabilistically using the measurement data of the power system output from the power system measurement unit and the simulator data output from the simulator unit;
Based on the output of the maximum likelihood state value calculation unit, a parameter correction unit for correcting the parameters of the simulator unit,
With
An apparatus for estimating a state of a power system, wherein parameters of the simulator unit are sequentially corrected so that a difference between power system measurement data of the power system measurement unit and simulator data of the simulator unit is reduced. In claim 1,
The power system measurement unit is
A power system measurement data unit for receiving and acquiring power related information measured by sensors in the power system;
A measurement data processing unit that probabilistically and statistically processes measurement data including variation factors and errors of the power system measurement data unit to reduce the influence of the variation factors and errors;
A state estimation device for a power system, comprising: In claim 1,
The simulator unit is
A parameter setting unit for setting various parameters related to the power system;
A power system simulator that reflects the power system, various parameters related to the power system set by the parameter setting unit, and a power system simulator that simulates based on measurement data of the power system,
A simulator data processing unit which reduces the influence of the variation factor and error by performing a stochastic and statistical calculation process on the simulator data including the variation factor and error of the power system simulator;
A state estimation device for a power system, comprising: In claim 1,
The simulator unit is configured to include a means for dividing the power system into a plurality of partial systems and calculating the state estimation for each partial system when calculating the state estimation of the power system. A power system state estimation device. In claim 1,
The simulator is configured to estimate any one of power, voltage, current, and phase of a power system in the calculation of state estimation of the power system. In claim 1,
The maximum likelihood state value calculation unit is configured to include means for convolving and integrating the probability distribution of the measurement data of the power system and the probability distribution of the simulator data. In claim 1,
The maximum likelihood state value calculating unit encodes the state quantity using the estimation result of the probability distribution of the state quantity obtained by comparing the measurement data of the power system and the simulator data output from the simulator unit. An apparatus for estimating the state of an electric power system, characterized in that it comprises means for converting into In claim 1,
The parameter correction unit increases the maximum likelihood of the probability distribution calculated by convolution integration between the probability distribution of the measurement data of the power system and the probability distribution of the simulator data in the maximum likelihood state value calculation unit, An apparatus for estimating a state of a power system, comprising means for correcting a parameter used for a simulator operation of the simulator unit. In claim 1,
In the step of sequentially correcting the parameters of the simulator unit so that the difference between the measurement data of the power system of the power system measurement unit and the simulator data of the simulator unit is small, the probability for each frequency component of the power system The state estimation apparatus for a power system according to claim 1, wherein state estimation is performed based on the distribution. In the electric power system state estimation device according to any one of claims 1 to 9,
Furthermore, the power system state estimation device is characterized by comprising means for displaying the received power amount appearance distribution information and the power system state estimation calculation result. In the electric power system state estimation device according to any one of claims 1 to 9,
The power system state estimation apparatus further comprises means for inputting in order to edit and process a probability distribution of variation factors used in the arithmetic processing by the power system state estimation apparatus. A power system state estimation method comprising power system measurement means for receiving appearance distribution information of power amount of power system and simulation calculation means for simulating calculation of power system state estimation,
In the simulation calculation means, the system configuration of the power system, the step of setting the parameters of the equipment characteristics,
The power system measuring means obtaining power system sensor measurement data; and
The simulation calculation means, based on the system configuration of the power system, parameters of equipment characteristics, referring to sensor measurement data of the power system, and performing a simulation calculation of the power system power flow;
The maximum likelihood state value calculating means compares and calculates the sensor measurement data of the power system and the simulation calculation of the power flow of the power system, and calculates a probabilistic likelihood estimated value;
A parameter correcting means referring to the likely estimated value calculated by the maximum likelihood state value calculating means, correcting parameters of the simulation calculating means;
Performing the simulation calculation of the power system power flow again by the simulation calculation means according to the corrected parameter;
Again, comparing the result of the simulation calculation of the tidal current of the power system performed with the sensor measurement data of the power system and determining whether the difference is within a predetermined range or exceeds a predetermined range; ,
If the compared difference exceeds a predetermined range, the parameter correction means refers to the stochastic likelihood estimated value calculated by the maximum likelihood state value calculation means, and sets the parameter of the simulation calculation means. The simulation calculation means performs again the simulation calculation of the power flow of the power system according to the corrected parameters, and the result of the simulation calculation of the power flow of the power system is performed again. Comparing with the sensor measurement data of the power system, repeatedly performing the difference until it falls within a predetermined range, and calculating the most probable state value from the sensor measurement data of the power system, To estimate the state of the power system. The power system state estimation method according to claim 12,
After calculating the most probable state value from the sensor measurement data of the power system,
The power system measuring means returns to the step of acquiring power system sensor measurement data,
State estimation of a power system characterized by repeatedly performing each step for calculating a state value that is most likely to be probabilistically according to the transition of time, and continuously calculating the most likely state value at each time Method. In the electric power system provided with the state estimating device of the electric power system according to any one of claims 1 to 9,
A power system control system for controlling a power device including a distribution substation provided in the power system with reference to a probabilistic most likely state value calculated by the power system state estimation device.

Images