Power system control device and method

Abstract

To provide a power system control device capable of properly controlling a power system designed considering fluctuations in load demand.SOLUTION: The power system control device includes: a prediction unit that predicts a prediction base value that is the component of a load demand that is longer than the specified period after the forecast target time of the power system and a predicted fluctuation range that represents the degree of fluctuation from the prediction base value of the load demand after the forecast target time based on the actual load demand, which represents the actual load demand of the power system; and an operating value calculation unit that calculates the operation value for controlling the power system based on the prediction base value and predicted fluctuation range.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a technique for predicting the state of a power system and stably controlling the power system. [Background technology]

[0002] Patent Document 1 discloses a phase modifying equipment operation planning device and a phase modifying equipment operation planning method that create an effective phase modifying equipment operation plan so that an operator can grasp future situations in which a VQC (Voltage and Reactive Power Control) device will control the phase modifying equipment. In the phase modifying equipment operation planning device, a phase modifying equipment operation planning means inputs a total demand forecast value for each time slot of the next day from a second memory, then searches the total demand actual record in the first memory to detect the date, time and time slot of the past total demand that is closest to the input total demand for each time slot, reads the phase modifying equipment operation record and tap operation record for the time slot of the detected past date and time, and writes the read phase modifying equipment operation record and tap operation record into a third memory as a phase modifying equipment operation plan and tap operation plan for the corresponding time slot of the next day. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2010-183761 Summary of the Invention [Problem to be solved by the invention]

[0004] When renewable energy power generation facilities such as solar and wind power generation are connected to the load at a consumer's site, the local load demand will show a value that includes the amount of power generated by the renewable energy. The amount of power generated by renewable energy fluctuates irregularly. For example, the amount of power generated by solar power generation during a cloudy day changes irregularly and suddenly. Therefore, when viewed from the upper side of the power system, the fluctuations in load demand appear irregular. For example, there may be multiple sudden increases or decreases in load demand. This creates periods of large fluctuations in load demand and periods of small fluctuations, making it difficult to predict local load demand.

[0005] The method disclosed in Patent Document 1 calculates a total demand forecast value for each time slot for the next day, identifies a past time slot that showed an actual total demand value closest to the calculated total demand forecast value, and determines a power plant control plan based on the operational performance of the identified time slot. However, if the load demand fluctuates widely, the difference between the total demand forecast value for the next day and the actual total demand value in the past becomes large. In this case, the power plant may not perform necessary or sufficient control, resulting in a power surplus or shortage. In a power grid, a power surplus or shortage can lead to voltage deviations or overloads.

[0006] One object of the present disclosure is to provide a technology that enables appropriate control of a power system taking into account fluctuations in load demand. [Means for solving the problem]

[0007] A power system control device according to one aspect of the present disclosure includes a prediction unit that predicts a forecast base value, which is a component of the load demand of the power system having a period equal to or greater than a predetermined period after a target prediction time, and a forecast fluctuation range that represents the degree of fluctuation from the forecast base value in the load demand after the target prediction time, based on a load demand record that represents the actual load demand of the power system, and an operating value calculation unit that calculates an operating value for controlling the power system based on the forecast base value and the forecast fluctuation range. [Effects of the Invention]

[0008] According to one aspect of the present disclosure, it is possible to appropriately control a power system in consideration of fluctuations in load demand. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a block diagram of a power system control device according to a first embodiment. [Figure 2] 1 is a block diagram showing the overall configuration of a power system including the configuration of its control equipment. [Figure 3] FIG. 2 is a block diagram showing a hardware configuration of the power system control device. [Figure 4] 3 is a flowchart showing an example of the overall processing by the power system control device of the first embodiment. [Figure 5] FIG. 1 is a diagram schematically illustrating a series of processes from learning to prediction. [Figure 6] 1 is a graph of actual and predicted load demands. [Figure 7] FIG. 10 is a block diagram of a power system control device according to a second embodiment. [Figure 8] 10 is a flowchart illustrating an example of an overall process performed by a power system control device according to a second embodiment. [Figure 9] FIG. 10 is a diagram illustrating ensemble prediction. DETAILED DESCRIPTION OF THE INVENTION

[0010] The following description of the embodiments will be made with reference to the drawings. Note that the embodiments described below do not limit the invention, and not all of the elements and combinations thereof described in the embodiments are necessarily essential to the invention. [Example]

[0011] FIG. 1 is a block diagram of a power system control device according to a first embodiment.

[0012] The power system control device 10 has a machine learning prediction unit 101, an operating value calculation unit 102, and an output unit 39. The machine learning prediction unit 101 has a filter processing unit 31, a base value machine learning unit 32, a base value prediction unit 33, a fluctuation range machine learning unit 34, and a fluctuation range prediction unit 35. The operating value calculation unit 102 has a system state calculation unit 36, a verification unit 37, and a predicted value correction unit 38.

[0013] The power system control device 10 stores, as input data, load demand record data D1, set value data D2, deviation condition data D3, and system equipment data D4 in a database. The power system control device 10 calculates operation values for controlling the power system using an operation value calculation unit 102, and outputs the operation values to an output unit 39. In addition to the operation values, the data output to the output unit 39 may include one or more of a predicted value of a base value (predicted base value), a predicted value of a fluctuation range (predicted fluctuation range), and a correction value. The output unit 39 displays or outputs various data including the operation values to the outside.

[0014] In the first embodiment, the following series of processes are generally executed to realize appropriate control of the power system.

[0015] First, the filter processing unit 31 performs a filter process on the load demand record data D1, which is time-series information of power system information acquired in the past. This filter process separates the load demand record into a base value (record base value) and a fluctuation range (record fluctuation range). The base value is a component contained in the load demand that has a period equal to or greater than a predetermined reference period. The fluctuation range is an index that represents the degree of fluctuation from the base value of the load demand. The index that represents the fluctuation range is not particularly limited, and may be determined based on the standard deviation of the load demand, for example.

[0016] Then, the base value machine learning unit 32 and the fluctuation range machine learning unit 34 perform machine learning on the base value data and fluctuation range separated from the load demand record, respectively, to generate respective prediction models (base value prediction model, fluctuation range prediction model).

[0017] Next, the base value prediction unit 33 and the fluctuation range prediction unit 35 predict the base value and fluctuation range of the future load demand using these prediction models.

[0018] Then, the system state calculation unit 36 calculates the system state using the predicted value of the base value, the system equipment data, and the power system control information, and calculates the operation values for controlling the power system based on the result. The system equipment data is information related to the configuration of each facility included in the power system. The power system control information is information related to the control of control devices.

[0019] Next, the verification unit 37 uses the operational value and the predicted value of the fluctuation range to verify whether the system state satisfies a predetermined deviation condition, for example, whether the state in which the index value representing the system state (state evaluation index value) exceeds a predetermined threshold value continues beyond a predetermined set time.

[0020] When the grid state satisfies the deviation condition, the predicted value correction unit 38 determines a correction value for correcting the predicted base value using the deviation amount, the predicted base value, and the set value data. The set value data includes a value that is set in advance to be used in a calculation to determine the correction value. For example, if the correction value is calculated from the deviation amount that indicates how much the index value deviates from a threshold, the calculation formula may be included as a set value in the set value data.

[0021] The process of correcting the predicted value of the base value using the correction value, calculating the system state again to calculate the operating value, and verifying whether the system state satisfies the deviation condition using the new operating value is repeated until the system state no longer satisfies the deviation condition or until a predetermined termination condition is satisfied. The predetermined termination condition may be, for example, that the above process is repeated a predetermined maximum number of times. The setting value for the maximum number of times may be included in the setting value data described above. Thereafter, the verification unit 37 outputs the operating value to the output unit 39.

[0022] The first embodiment will be described in more detail below.

[0023] FIG. 2 is a block diagram showing the overall configuration of the power system including the configuration of its control devices.

[0024] 2, the power system control device 10 is connected to the measurement devices 44 and individual control devices 45 of the power system 100 via a communication network 300. The power system control device 10 receives data measured by the measurement devices 44, calculates operation values using the data, and transmits the operation values to the individual control devices 45. The data measured by the measurement devices 44 reflects the system state of the power system 100.

[0025] The power system 100 includes various types of equipment, including a power source 110, nodes (buses) 120 and 121, branches (lines) 140, a load 150, a power capacitor 160 (a static condenser (SC)), and a shunt reactor (ShR) 170. The power system 100 may also include one or more other measurement devices and controllable devices (not shown). Examples of the controllable devices include batteries, rechargeable secondary batteries, storage batteries for electric vehicles (EVs), flywheels, and other phase modifying equipment (e.g., static var compensators (SVCs), static var generators (SVGs), and loop power controllers (LPCs)).

[0026] Here, the power source 110 may be, for example, a rotary power source such as a thermal power generator, a hydroelectric power generator, or a nuclear power generator, or may be an inverter-connected power source in which a distributed power source such as a solar power generator or a wind power generator is connected to the power grid 100 via an inverter.

[0027] The measuring device 44 is a device that measures one or more index values of the system state, such as node voltage V, branch current I, active power P, reactive power Q, power factor cos(Φ), tap value, and on / off information of switches such as nodes, branches, transformers, SCs, and ShRs. Devices that measure index values of the system state include voltage transformers (VTs), potential transformers (PTs), and current transformers (CTs).

[0028] The measuring device 44 has a function of transmitting data including the measured index value, a data measurement location ID, and / or a built-in timestamp of the measuring device. The data may include, for example, measurement information (telemeter (TM)) and / or display information (supervision (SV)). The measuring device 44 may also be a device that measures power information (voltage phasor information) with absolute time using the Global Positioning System (GPS).

[0029] The measurement device 44 may also be a phasor measurement unit (PMU), or may be other measurement equipment.

[0030] Here, the measuring device 44 is assumed to be located within the power system 100, but it may also be installed on a bus or line connecting the power source 110, the transformer 130, the tapped transformer 131, the load 150, the power capacitor 160, the shunt reactor 170, the measuring device 44, and the individual control device 45.

[0031] The load demand actual value data D1 and the measurement values used in the processing by the system state calculation unit 36 and the verification unit 37 are one or more data items among the power factor cos(Φ), active power P, and reactive power Q measured by the measurement device 44. These data items are sent from the measurement device 44 to the power system control device 10 via the communication network 300.

[0032] However, instead of transmitting the measurement value data directly from the measuring device 44 to the power system control device 10, the measurement value data from the measuring device 44 may be temporarily collected in a monitoring device (not shown) and then transmitted from the monitoring device to the power system control device 10 via the communication network 300. The measurement value data may include, in addition to the measurement value, a unique number for identifying the data and a timestamp.

[0033] Here, the monitoring device is, for example, a central load dispatching center, a system stability monitoring server, etc. Furthermore, measurement value data transmitted directly from the measuring device 44 to the power system control device 10 and measurement value data once collected by the monitoring device and then transmitted to the power system control device 10 may be mixed.

[0034] FIG. 3 is a block diagram showing the hardware configuration of the power system control device.

[0035] 3, in a power system control device 10, a display unit 11, an input unit 12 such as a keyboard or a mouse, a communication unit 13, a computer or computer server (CPU: Central Processing Unit) 14, a memory 15, and various databases are connected to a bus line 43. The display unit 11 is configured as, for example, a display device. The various databases include a load demand record database 21, a deviation condition database 22, a set value database 23, and a program database 51. Note that while FIG. 3 shows a program database 51 that stores software programs for realizing the processing of each unit, and databases 21, 22, and 23 that store various data used in the processing, the power system control device 10 may further include a database that stores data generated as a result of the processing.

[0036] The load demand record database 21 is a database in which load demand record data D1 is stored. The deviation condition database 22 is a database in which deviation condition data D3 is stored. The set value database 23 is a database in which set value data D2 is stored. The program database 51 is a database in which data of software programs executed by the power system control device 10 is stored.

[0037] The display unit 11 is, for example, a display device that displays text and images on a screen, or may be configured to include a printer device and / or an audio output device in addition to the display device.

[0038] The input unit 12 is configured to include at least one of a keyboard switch, a pointing device such as a mouse, a touch panel, a voice instruction device, and the like.

[0039] The communication unit 13 includes a circuit for connecting the power system control device 10 to the communication network 300, and is capable of executing a communication protocol for realizing communication.

[0040] The CPU 14 reads and executes a predetermined computer program from the program database 51. The CPU 14 may be configured with one or more semiconductor chips, or may be a computer device such as a calculation server.

[0041] The memory 15 is, for example, a RAM (Random Access Memory), and stores computer programs read from the program database 51, as well as calculation result data and image data required for each process. The screen data stored in the memory 15 is sent to the display unit 11 and displayed on the screen.

[0042] The various programs stored in the program database 51 shown in FIG. 3 will be described below.

[0043] The program database 51 stores, for example, a filter processing program P31, a base value machine learning program P32, a base value prediction program P33, a fluctuation range machine learning program P34, a fluctuation range prediction program P35, a system state calculation program P36, a verification program P37, a predicted value correction program P38, and an output program P39.

[0044] The CPU 14 executes various programs read from the program database 51 to the memory 15. The filter processing program P31 executes the programs to implement the filter processing unit 31 shown in FIG. 1. The base value machine learning program P32 executes the programs to implement the base value machine learning unit 32 shown in FIG. 1. The base value prediction program P33 executes the programs to implement the base value prediction unit 33 shown in FIG. 1. The fluctuation range machine learning program P34 executes the programs to implement the fluctuation range machine learning unit 34 shown in FIG. 1. The fluctuation range prediction program P35 executes the programs to implement the fluctuation range prediction unit 35 shown in FIG. 1. The system state calculation program P36 executes the programs to implement the system state calculation unit 36 shown in FIG. 1. The verification program P37 executes the programs to implement the verification unit 37 shown in FIG. 1. The predicted value correction program P38 executes the programs to implement the predicted value correction unit 38 shown in FIG. 1. The output program P39 executes the programs to implement the output unit 39 shown in FIG. 1.

[0045] Temporary data and calculation result data in calculations by various programs, and image data of a screen for accepting user instructions required for calculations or a screen for displaying calculation results are temporarily stored in memory 15. The image data stored in memory 15 is then displayed on a display screen or the like by display unit 11. The screen displayed by power system control device 10 may be a simple screen only for rewriting various programs and databases.

[0046] The load demand record database 21 stores, for example, time-series values of the active power load of each substation as load demand record data D1. An example of load demand is shown in FIG. 6. The load demand record is shown in the portion of FIG. 6 before the target time of prediction. As described above, the load demand record data D1 is measured value data. The load demand record data D1 may be calculated, for example, by proportionally dividing total demand data held by the central load dispatching center. Each substation may be stored for each node. If there is no measuring device for each node, the estimated load demand of each node may be calculated from the actual measured value measured by the measuring device. The power system control device 10 may also obtain load demand record data stored in other systems, such as a supervisory control device, a central load dispatching center, or an EMS. The load demand record data may also be manually input into the power system control device 10. The data may be manually input via the input unit 12. When manually inputting data, the power system control device 10 may display a data input screen on the display unit 11. Furthermore, in order to reduce the work of data input, the power system control device 10 may be semi-manualized so that a large amount of data can be easily input using a function for supplementing data.

[0047] Here, the active power load PL and the reactive power load QL can be calculated using existing methods. For example, the method described in "Predictive Voltage and Reactive Power Control Method for Bulk Power Systems Using the LP Method" by Takaharu Ishida et al., Transactions on Electrical Engineering, Vol. 117, No. 8, 1997, pp. 3216-1120, may be used to calculate the active power output of each generator using economic load dispatch (ELD) from total demand data and generator fuel consumption characteristic data. Alternatively, the individual active power load of a substation may be calculated using equation (1) from total demand data and substation load-to-total demand ratio data.

[0048] P li =P all ×ratio i …(1)

[0049] In equation (1), P li is the predicted active power load of substation i. Pall is the total demand estimate. i is the ratio of the active power load of substation i to the total demand.

[0050] Furthermore, the individual substation reactive power load may be estimated and calculated using equation (2) from the estimation result of the individual substation active power load and the substation load PQ correlation data. Q li =f i (P li ) …(2)

[0051] In equation (2), Q li is the predicted reactive power load of substation i. i is the PQ correlation of substation i.

[0052] The deviation condition database 22 stores, as deviation condition data D3, criteria for determining whether the system state of the power system 100 is unstable. Here, the term "deviation" refers to the system state becoming unstable. The deviation conditions, which are the criteria for determining whether a deviation occurs, include, for example, threshold values for index values indicating the system state, such as voltage, heat capacity, and current, and a set period during which a deviation is determined to have occurred if the index values continue to exceed the thresholds, or a set number of times a deviation is determined to have occurred if the index values continue to exceed the thresholds.

[0053] The deviation condition data D3 may be set in the power system control device 10 remotely from a supervisory control device, a central load dispatching center, or an EMS. The deviation condition data D3 may also be manually input into the power system control device 10. In the case of manual input, data may be manually input from the input unit 12. When inputting data manually, the power system control device 10 may display a data input screen on the display unit 11. In order to reduce the work of data input, the power system control device 10 may be semi-manual, allowing large amounts of data to be easily input using a data complementing function.

[0054] The set value database 23 stores, as set value data D3, a correction range (correction value) indicating how much the base forecast value is corrected for the deviation amount as shown in Fig. 1. The correction value may be calculated from the deviation amount using arithmetic operations, linear expressions, quadratic expressions, etc., or may be calculated using a discrete value expression, a nonlinear expression, etc. These data may be stored using the input unit 12 of the power system control device 10, or may be stored from another monitoring device.

[0055] The set value database 23 stores, as set value data D2, set values for determining how much to correct the base forecast value when the system state corresponds to a deviation. The correction value may be calculated from the deviation amount by a predetermined calculation such as four arithmetic operations, a linear expression, or a quadratic expression. Alternatively, a discrete value expression or a nonlinear expression may be used for the calculation. In such cases, the calculation method may be stored as set value data D2 in the set value database 23. The set value data D2 may be input to the power system control device 10 from the input unit 12, or may be input from another monitoring device.

[0056] Next, the calculation process performed by the power system control device 10 will be described.

[0057] FIG. 4 is a flowchart illustrating an example of the overall processing by the power system control device according to the first embodiment. First, the outline of the flow will be explained.

[0058] First, in step S30, load demand record data D1, which is time-series information of power system information, is read. Next, in step S40, filter processing is performed using the load demand record data D1 to separate the load demand record into a base value and a fluctuation range. In steps S50 to S54, the base value and fluctuation range for a set period are each machine-learned, and the base value and fluctuation range for a predetermined future period are predicted. Next, in step S60, the system state is calculated using the predicted base value, system equipment data, and power system control information, and an operating value is determined based on the calculation result. Next, in step S70, the system state is calculated taking the fluctuation range into consideration using the predicted fluctuation range, operating value, and system equipment data, and it is verified whether this calculation result satisfies the deviation condition set in the deviation condition data D3. In step S71, it is determined whether the calculation result satisfies the deviation condition set in the deviation condition data D3. If the deviation condition is met (Yes), in step S80, the deviation amount, the predicted value of the base value, and the setting value data D2 are used to determine a correction value for the predicted value of the base value, and the process returns to step S60, where the system state is calculated again using the corrected predicted value of the base value to calculate the operating value.

[0059] If the determination in step S71 indicates that the deviation condition is no longer met (No), then in step S90, one or more of the load demand actual data D1, the predicted base value, the predicted fluctuation range, the operating value, the deviation condition data D3, the set value data D2, and the correction value are output. Even if the determination in step S71 indicates that the deviation condition is met, if the processes in steps S60 to S80 have been repeated a predetermined number of times, the process proceeds to step S90 and the data is output. Here, "output" includes displaying data such as the operating value on a screen in a format that can be viewed by an operator or planner, outputting a control command based on the operating value to the controlled object, and transmitting to a planning system (not shown).

[0060] In addition, the calculation result data at each step and the data recorded in the memory during the calculation may be output to another monitoring device and displayed sequentially on the screen, thereby enabling the operator or planner to easily grasp the operating status of the power system control device 10.

[0061] The processing of each step will be explained in more detail below.

[0062] First, in step S30, the power system control device 10 reads the load demand record data D1, which is time-series information of the power system information.

[0063] Next, in step S40, the power system control device 10 performs filtering using the load demand record data D1 to separate the load demand record into a base value and a fluctuation range. An example of a filtering method will now be described. By passing the time-series waveform of the load demand record through a filter with each preset fluctuation period, the components can be separated into time-series waveforms with each fluctuation period. This has the effect of separating the main base fluctuation component from components that fluctuate in short periods, such as the output of solar power generation on a cloudy day. The filters used here are devices such as high-pass filters, low-pass filters, and band-pass filters that pass components that fluctuate with a set period or remove components outside the set period. This has the effect of decomposing the load demand record into components that fluctuate with a desired period.

[0064] In step S50, the power system control device 10 determines whether machine learning of the base value and fluctuation range is necessary. If machine learning is necessary, the power system control device 10 proceeds to step S51. If machine learning is not necessary, the power system control device 10 proceeds to step S53. Whether machine learning is necessary is determined based on whether machine learning of the base value and the fluctuation range has been completed. When each machine learning is completed, a learned flag may be assigned to the learning data, and whether machine learning is necessary may be determined based on whether the learned flag is assigned to the learning data. Alternatively, the learned flag may be removed after a certain time has passed since the completion of machine learning, and machine learning may be performed again a certain time after the previous machine learning to periodically update the prediction model. Furthermore, if machine learning is not possible due to insufficient learning data, an alarm may be output to warn of this. Furthermore, if the base value prediction unit 33 cannot properly execute a prediction calculation using a prediction model based on machine learning in the base value machine learning unit 32, or if the fluctuation range prediction unit 35 cannot properly execute a prediction calculation using a prediction model based on machine learning in the fluctuation range machine learning unit 34, an alarm may be output to warn of this. This has the effect of informing the operator or planner whether there is any problem with the operation of the power system control device 10.

[0065] In step S51, the power system control device 10 performs machine learning using data of base values separated from the load demand record, and generates a prediction model for predicting the base value.

[0066] FIG. 5 is a diagram schematically illustrating a series of processes from learning to prediction.

[0067] An example of training data is shown in the upper part of Figure 5. Here, data is available for the past few days, i.e., (d-3), (d-2), and (d-1), and the graph shows the time variations in active power P and temperature T for each day. Time is used as the explanatory variable, and P or T is used as the dependent variable. Using this training data, a prediction formula for predicting the dependent variable P or T at each point in time (t to t+55) is obtained using a regression model. The prediction formula for t is shown in Equation 3, and the prediction formula for t+55 is shown in Equation 4.

[0068] P(d,t)=a1(t)·T(d,t-60) +a2(t)·P(d-1,t) +a3(t)·{P(d-1,t-60)-P(d,t-60)} +a4(t) …(3)

[0069] (d, t) indicates the target date and time. d is the target date and t is the target time. d-1 is the day before the target date. t-60 is 60 minutes before the target time.

[0070] For example, in equation (3), T(d, t-60) on the right side is the temperature T of the weather data 60 minutes before the target time on the target day. P(d-1, t-60) is the active power P measured 60 minutes before the target time on the day before the target day. P(d, t-60) is the active power P measured 60 minutes before the target time on the target day.

[0071] Equation (3) is a regression equation that calculates the predicted value of active power P at the target time on the target day from weather data for that day and measurement data of active power on the previous day and that day.

[0072] P(d,t+55)=a1(t+55)·T(d,t-60) +a2(t+55)·P(d-1,t+60) +a3(t+55)·{P(d-1,t-60)-P(d,t-60)} +a4(t+55) …(4)

[0073] d is the target date, and t is the target time. d-1 is the day before the target date. t-60 is 60 minutes before the target time. t+55 is 55 minutes after the target time.

[0074] For example, in equation (4), P(d-1, t+60) on the right side is the active power P measured 60 minutes after the target prediction time on the day before the target prediction day.

[0075] For both equations (3) and (4), data measured up to 60 minutes before the target time of prediction is used for the data on the day.

[0076] Note that the regression model used here may be linear regression or shrinkage regression (Ridge, Lasso, Elastic Net, etc.). Using shrinkage regression can reduce the effects of multicollinearity and improve prediction accuracy.

[0077] In step S52, the power system control device 10 performs machine learning using the fluctuation range data separated from the load demand record, and generates a prediction model for predicting the fluctuation range. The fluctuation range may be, for example, a fluctuation of ±σ, where σ is the standard deviation, that is, 2σ. The machine learning method for the fluctuation range is basically the same as the machine learning method for the base value.

[0078] In step S53, the power system control device 10 calculates a predicted value of the base value for a predetermined set period from the prediction target date and time using the prediction model obtained in step S51. The lower part of Fig. 5 shows a graph representing the measured or predicted active power (objective variable) around the prediction target date and time. The predicted value of active power P from the prediction target date and time (d, t) to one hour later (d, t+60), which is the set period, is represented by a white circle.

[0079] In step S54, the power system control device 10 calculates a predicted value of the fluctuation range for a predetermined set period from the prediction target date and time using the prediction model obtained in step S52.

[0080] FIG. 6 is a graph of the actual measured and predicted values of load demand. FIG. 6 shows the actual measured values of load demand before the target prediction time and the predicted values of load demand after the target prediction time. For the time before the target prediction time, a graph showing the actual measured values of load demand and a graph showing a base value separated from the actual measured values of load demand are shown. For the time after the target prediction time, a graph showing the predicted value of the base value and graphs showing the fluctuation range above and below it are shown. Here, the fluctuation range is shown as a graph of base value + standard deviation σ × 2 and a graph of base value - standard deviation σ × 2. Here, the fluctuation range is displayed as standard deviation ±σ2, but is not limited to this. The fluctuation range may be another value, such as standard deviation ±σ2, or may be changeable by settings.

[0081] By displaying the fluctuation range on a screen as a graph, operators and planners can easily visually grasp the extent of fluctuation that may occur during that time period. For example, it is easy to distinguish between time periods when conservative operation is required and time periods when proactive operation is required, which has the effect of maintaining the stability of the power system and improving its economic efficiency.

[0082] In step S60, the power system control device 10 calculates the system state of the power system 100 using the predicted value of the base value, the system equipment data, and the power system control information, and calculates an operating value to be used for controlling the power system 100 based on the calculation result. For example, the predicted value of the base value is used to calculate an operating value (Vref(t)) for voltage and reactive power control. This operating value is, for example, a voltage target value of the individual control device 45 for maintaining the voltage and reactive power balance in a certain area. The individual control device 45 receives one or more pieces of fluctuation range prediction result data D13 at a preset cycle and performs voltage and reactive power control.

[0083] In step S70, the power system control device 10 verifies the system state using the predicted value of the fluctuation range, the operating value, and the system equipment data. Here, it is verified whether the index value of the system state deviates from the threshold value when the load demand reaches the predicted value of the fluctuation range when the power system 100 is operated with the operating value obtained in step S60. This has the effect of enabling the operator or planner to easily confirm in advance whether the operating value was appropriate.

[0084] In step S71, the power system control device 10 determines whether the system state corresponds to the deviation condition indicated in the deviation condition data D3 as a result of the verification in step S70. If the system state corresponds to the deviation condition, the power system control device 10 proceeds to step S80. If the system state does not correspond to the deviation condition, the power system control device 10 proceeds to step S90.

[0085] In step S80, the power system control device 10 determines a correction value for the predicted base value using the deviation amount indicating how much the system state has deviated from the threshold, the predicted base value, and the setting value data D2, and then returns to step S60. This makes it possible to correct the operating value so as to reduce the deviation amount, thereby achieving the effect of enabling the system stability of the power system to be maintained.

[0086] Returning from step S80 to step S60, the power system control device 10 again calculates the system state of the power system 100 using the predicted value and the corrected value of the base value, recalculates the operating value, and proceeds to step S70. Then, in steps S70 and S71, the power system control device 10 again verifies and determines whether the system state satisfies the deviation condition. When the system state does not satisfy the deviation condition or the process has been repeated a preset number of times, the power system control device 10 proceeds to step S90. This makes it possible to prevent a phenomenon in which system stability deteriorates drastically over time, even if the output of renewable energy fluctuates due to weather, and improves supply reliability.

[0087] In step S90, the power system control device 10 outputs one or more of the load demand actual data D1, the base value forecast, the fluctuation range forecast, the operating value, the deviation condition data D3, the setpoint data D2, and the correction value. The term "output" here includes displaying the operating value and the like in a format viewable by the operator or planner, outputting a control command to the control target, and outputting the information to the planning system. By displaying this information on the screen, the operator or planner can visually and easily grasp the amount of fluctuation that may occur during the displayed time period. For example, by displaying the graph shown in FIG. 6 on the screen, the operator or planner can visually and easily grasp how the load demand will change and to what extent. Furthermore, for example, it is possible to easily distinguish between time periods when conservative operation is required and time periods when aggressive operation is required, thereby maintaining the stability of the power system and improving its economic efficiency. Furthermore, it is also possible to reduce the effort required by the operator or planner to set the operating value. Furthermore, by using this appropriate information to output control commands to the control target or output data to the planning system, it is possible to maintain the system stability of the power system 100 and improve its economic efficiency.

[0088] As described above, in this embodiment, the power system control device 10 uses the machine learning prediction unit 101 to predict a forecast base value, which is a component of the load demand of the power system 100 with a period equal to or greater than a predetermined period, and a forecast fluctuation range, which represents the degree of fluctuation from the forecast base value of the load demand of the power system 100 after the target prediction time, based on the load demand record that represents the actual load demand of the power system 100, and then uses the operation value calculation unit 102 to calculate an operation value for controlling the power system 100 based on the forecast base value and the forecast fluctuation range. As a result, the operation value is calculated based on the forecast base value and the forecast fluctuation range, and appropriate control of the power system is possible that takes fluctuations in load demand into consideration.

[0089] Furthermore, the operating value calculation unit 102 calculates the operating value based on the predicted base value and corrects the operating value based on the predicted fluctuation range. Therefore, since the operating value is calculated based on the predicted base value and corrected based on the predicted fluctuation range, it is possible to calculate an appropriate operating value based on the base value of the load demand and taking the fluctuation range into consideration.

[0090] Furthermore, the operating value calculation unit 102 calculates a state evaluation index value for evaluating the state of the power system 100 based on the predicted fluctuation range and the operating value, determines whether the state evaluation index value satisfies a predetermined deviation condition, and corrects the operating value if the state evaluation index value satisfies the deviation condition, repeating this process until the state evaluation index value no longer satisfies the deviation condition. As a result, the operating value is corrected until the state evaluation index value no longer satisfies the predetermined deviation condition, making it possible to calculate an appropriate operating value.

[0091] Furthermore, the operating value calculation unit 102 predetermines a set period and a set number of times, and if the number of times that the state evaluation index value exceeds a predetermined threshold during the set period is equal to or exceeds the set number of times, it determines that the state evaluation index value meets the deviation condition. This provides protection for the correction of the operating value, making it possible to suppress sensitive fluctuations in the operating value.

[0092] The operating value is a voltage target value or a reactive power target value to be given to a control device that controls the power system 100, including one or more of a battery, a rechargeable secondary battery, a storage battery for an electric vehicle, a flywheel, a static reactive power generator, and a transformer with a phase shifter. In this way, an operating value for controlling the equipment included in the power system 100 is determined, and the corrected operating value can be used to control the power system.

[0093] The output unit 39 further displays on a screen a graph that represents the actual load demand up to the target prediction time and the predicted base value and predicted fluctuation range after the target prediction time.

[0094] The machine learning prediction unit 101 calculates, based on the actual load demand, an actual base value, which is a component included in the actual load demand and has a period equal to or greater than a reference period, and an actual fluctuation range, which represents the degree of fluctuation of the actual load demand from the actual base value, generates a base value prediction model by learning the actual base value, generates a fluctuation range prediction model by learning the actual fluctuation range, calculates a predicted base value using the base value prediction model, and calculates the predicted fluctuation range using the fluctuation range prediction model.By using machine learning that separates the base value and the fluctuation range, it is possible to calculate operating values that are suitable for controlling a power system with irregular load demand fluctuations.

[0095] Furthermore, the machine learning prediction unit 101 extracts components equal to or greater than the reference period from the actual load demand using one or a combination of a high-pass filter, a low-pass filter, and a band-pass filter, and sets the extracted components as the actual base value. Because the actual base value is separated from the load demand actual value using a filter, the actual base value can be easily calculated.

[0096] Furthermore, the machine learning prediction unit 101 outputs an alarm when it is unable to perform a normal calculation for at least one of the calculations for calculating a predicted base value using a base value prediction model or the calculation for calculating a predicted fluctuation range using a fluctuation range prediction model. In machine learning predictions, there are cases where normal prediction calculations cannot be performed for input data in a range far removed from the learning data. By issuing an alarm when a normal prediction calculation cannot be performed, it is possible to detect an abnormal state that deviates far from the learning data, which is past performance, and issue an alarm.

[0097] Furthermore, the machine learning prediction unit 101 generates a base value prediction model and a fluctuation range prediction model by learning a regression equation. By using the regression equation, it becomes possible to appropriately predict a time-series load demand. [Example]

[0098] In Example 1, the load demand was separated into a base value and a fluctuation range, and each was subjected to machine learning to generate respective prediction models, and each prediction model was used to calculate predicted values of the future base value and fluctuation range.

[0099] In this respect, the second embodiment differs from the first embodiment in the method of calculating the predicted values of the base value and the fluctuation range. In the second embodiment, the predicted values of the base value and the fluctuation range are calculated using ensemble prediction, which simplifies the configuration and processing.

[0100] The following describes a system control device according to the second embodiment, but the description of the same configuration or processing as in the first embodiment will be omitted.

[0101] Fig. 7 is a block diagram of a power system control device according to Example 2. The power system control device 10 of Example 2 shown in Fig. 7 differs from that of Example 1 shown in Fig. 1 in that it has an ensemble prediction unit 103 instead of the machine learning prediction unit 101.

[0102] Fig. 8 is a flowchart showing an example of the overall processing by the power system control device of Example 2. The flowchart of the overall processing by the power system control device 10 of Example 2 shown in Fig. 8 differs from that of Example 1 shown in Fig. 4 in that it has step S55 instead of steps S40 to S54.

[0103] In step S55, the power system control device 10 calculates a predicted value of the base value and a predicted value of the fluctuation range by ensemble prediction. Ensemble prediction is a prediction method that predicts the future using multiple scenarios, each of which is given a different initial value, and calculates a prediction result by statistically processing the prediction results of the multiple scenarios.

[0104] 9 is a diagram showing the state of ensemble forecasting. After the target forecast time, a graph of the predicted value (forecast result) of the load demand for each scenario is shown.

[0105] The ensemble prediction unit 103 sets the average value of the multiple scenario prediction results, i.e., the ensemble average, as the predicted value of the base value.Then, the width of the range in which the multiple scenario prediction results are included with a predetermined probability is set as the predicted value of the fluctuation range.The predetermined probability may be, for example, 50%.

[0106] The power system control device 10 of the second embodiment may display the predicted value of the base value of the ensemble forecast and the predicted value of the fluctuation range in a graph. Furthermore, the power system control device 10 may further display the forecast results of each scenario in a graph. Operators can easily visually grasp the amount of fluctuation that may occur in each time period, and can immediately determine time periods when conservative operation is required and time periods when proactive operation is required, which has the effect of enabling the maintenance of system stability and improvement of economic efficiency of the power system. Furthermore, it has the effect of reducing the effort required by operators and planners to set operating values. Furthermore, by outputting information on the predicted values and operating values as commands or to a planning system, it has the effect of enabling the maintenance of system stability and improvement of economic efficiency of the power system.

[0107] As described above, in this embodiment, the power system control device 10 uses the ensemble prediction unit 103 to predict load demand forecast values, which are forecast values of load demand after the target forecast time in multiple scenarios, by ensemble prediction based on load demand records, and sets the average value of the load demand forecast values in the multiple scenarios as the forecast base value, and sets the width of the range in which the load demand forecast values in the multiple scenarios are included with a predetermined probability as the forecast fluctuation width. With simple calculations, it is possible to calculate operating values suitable for controlling a power system with irregular load demand fluctuations. [Explanation of symbols]

[0108] 10...power dispute control device, 11...display unit, 12...input unit, 13...communication unit, 14...CPU, 15...memory, 21...load demand record database, 22...deviation condition database, 23...set value database, 31...filter processing unit, 32...base value machine learning unit, 33...base value prediction unit, 34...fluctuation range machine learning unit, 35...fluctuation range prediction unit, 36...system state calculation unit, 37...verification unit, 38...prediction value correction unit, 39...output unit, 43...bus line, 44...measuring device, 45...individual control device, 51...program database, 100...power system, 101...machine learning prediction unit, 102...operation value calculation unit, 103...ensemble prediction unit, 110...power source, 130...transformer, 131...transformer, 150...load, 160...power capacitor, 170...shunt reactor, 300...communication network

Claims

1. a prediction unit that predicts, based on a load demand record that indicates an actual load demand of the power system, a prediction base value that is a component of the load demand of the power system having a period equal to or greater than a predetermined period after a target time of prediction, and a prediction fluctuation range that indicates a degree of fluctuation of the load demand from the prediction base value after the target time of prediction; an operation value calculation unit that calculates an operation value for controlling the power system based on the forecast base value and the forecast fluctuation range; and the management value calculation unit calculates the management value based on the forecast base value and corrects the management value based on the forecast fluctuation range; Power system control device.

2. the operating value calculation unit calculates a state evaluation index value for evaluating the state of the power system based on the predicted fluctuation range and the operating value, determines whether or not the state evaluation index value satisfies a predetermined deviation condition, and corrects the operating value if the state evaluation index value satisfies the deviation condition, repeating this process until the state evaluation index value no longer satisfies the deviation condition. The power system control device according to claim 1 .

3. the operation value calculation unit predetermines a set period and a set number of times, and determines that the state evaluation index value satisfies the deviation condition if the number of times that the state evaluation index value exceeds a predetermined threshold during the set period is equal to or greater than the set number of times. The power system control device according to claim 2 .

4. the operating value is a voltage target value or a reactive power target value to be given to a control device that controls the power system, the control device including one or more of a battery, a rechargeable secondary battery, a storage battery of an electric vehicle, a flywheel, a static reactive power generator, and a transformer with a phase shifter; The power system control device according to claim 1 .

5. An output unit that displays on a screen a graph representing the actual load demand up to the target prediction time and the forecast base value and the forecast fluctuation range after the target prediction time. The power system control device according to claim 1 .

6. the prediction unit calculates, based on the load demand actual result, an actual result base value which is a component having a period equal to or greater than a reference period included in the load demand actual result, and an actual result fluctuation range which represents the degree of fluctuation of the load demand actual result from the actual result base value, generates a base value prediction model by learning the actual base value, generates a fluctuation range prediction model by learning the actual result fluctuation range, calculates the predicted base value using the base value prediction model, and calculates the predicted fluctuation range using the fluctuation range prediction model; The power system control device according to claim 1 .

7. The prediction unit extracts components having a period equal to or greater than the reference period from the actual load demand data using one or a combination of a high-pass filter, a low-pass filter, and a band-pass filter, and sets the extracted components as the actual data base value. The power system control device according to claim 6.

8. the prediction unit outputs an alarm when it is unable to execute normal calculations for at least one of the calculation for calculating the predicted base value using the base value prediction model and the calculation for calculating the predicted fluctuation range using the fluctuation range prediction model. The power system control device according to claim 6.

9. the prediction unit generates the base value prediction model and the fluctuation range prediction model by learning a regression equation. The power system control device according to claim 6.

10. the prediction unit predicts load demand forecast values, which are forecast values of load demand after the forecast target time in a plurality of scenarios, by ensemble forecasting based on the load demand actual results, sets an average value of the load demand forecast values in the plurality of scenarios as the forecast base value, and sets a range in which the load demand forecast values in the plurality of scenarios are included with a predetermined probability as the forecast fluctuation range. The power system control device according to claim 1 .

11. based on a load demand record representing an actual load demand of a power system, predicting a forecast base value which is a component of the load demand of the power system having a period equal to or greater than a predetermined period after a target time of prediction, and a forecast fluctuation range which represents a degree of fluctuation of the load demand from the forecast base value after the target time of prediction; calculating an operation value for controlling the power system based on the forecast base value; correcting the operational value based on the predicted fluctuation range; The power system control method is implemented by a computer.

Images