JP7278343B2 - POWER PLANT OPERATION MANAGEMENT DEVICE, METHOD AND PROGRAM - Google Patents

Description

An embodiment of the present invention relates to an operation management technique for a power plant in which partial load operation is introduced.

In recent years, there has been an increasing trend toward using renewable energy sources that do not emit greenhouse gases, such as solar power, wind power, geothermal power, small- and medium-sized hydropower, and biomass power sources. Such renewable energy is likely to be affected by factors such as output fluctuations due to time of day, season, and weather, and uneven distribution of resource distribution areas. Therefore, there is a need for a separate regulated power supply that mitigates the effects of such factors and contributes to optimizing the balance between power supply and power supply and demand. As such a regulated power source, a power plant using a gas turbine whose power generation output can be adjusted relatively easily or a combined cycle power plant in which a gas turbine and a steam turbine are combined is expected to play a role.

JP-A-1-271602 JP-A-2004-54808

These power plants are generally designed to maximize power generation efficiency at maximum output. Therefore, maximum efficiency cannot be obtained at intermediate output due to partial load operation that optimizes the power supply and demand balance. In addition, the range of operating conditions in partial load operation is wide, and it is difficult to accurately manage power generation efficiency from design data for all operating conditions. Furthermore, when partial load operation was introduced, it was difficult to envision fuel plans and financial plans compared to continuous maximum power operation.

The embodiment of the present invention has been made in consideration of such circumstances, and is a power plant that can accurately manage the power generation efficiency when partial load operation is introduced and also facilitates the assumption of fuel plans and financial plans. The purpose is to provide operation management technology.

In the power plant operation management device according to the embodiment, an accumulation unit that accumulates a plurality of process signals composed of time-series numerical data, a selection unit that arbitrarily selects the process signals having different attributes, and each attribute a setting unit for setting discrimination classes corresponding to the numerical divisions of the numerical data, a conversion unit for converting the numerical data of the process signal into the corresponding discrimination classes, and the selected process signal that is time-synchronized with a generation unit that generates groups of discrimination classes in chronological order.

The embodiment of the present invention provides an operation management technique for a power plant that accurately manages the power generation efficiency when partial load operation is introduced and facilitates the assumption of a fuel plan and a financial plan.

1 is a block diagram of a power plant operation management device according to a first embodiment of the present invention; FIG. A table describing process signals and weather information. FIG. 4 is an explanatory diagram of functions of an input unit, a selection unit, a conversion unit, and a generation unit according to the embodiment; FIG. 2 is a block diagram of a power plant operation management device according to a second embodiment of the present invention; 4 is a flowchart for explaining the procedure of the power plant operation management method and the algorithm of the operation management program according to the embodiment;

(First embodiment)
An embodiment of the present invention will be described below with reference to the accompanying drawings. FIG. 1 is a block diagram of a power plant operation management device 10A (10) (hereinafter simply referred to as "operation management device 10A") according to the first embodiment of the present invention. FIG. 2 is a list for explaining process signals R _m (m=1 to M) and weather information K (K _a , K _b , . . . ).

Thus, the operation management device 10A (10) accumulates a plurality of process signals R _m (m=1 to M) composed of time-series numerical data r _m (t _n ) (n=1, 2, . . . ). a selection unit 12 for arbitrarily selecting _process signals R _m having different attributes 13; and a discrimination _class S _m ^x (x = 1 to d _m ), a conversion unit 17 for converting the numerical data r _m (t _n ) of the process signal R _m into the corresponding discrimination class S _m ^x , and the selected process signal R _m and a generation unit 18 that generates groups G _n (n=1 _, 2, ^.

The power plant 20 performs algorithm processing on process signals _Rm such as pressure, flow rate, temperature, and position output from a sensor (not shown), determines the operation amounts of actuators and mechanical/electrical switches, and repeats the routine. The entire plant is automatically controlled.

The receiving unit 21 accumulates process signals R _m (m=1 to M) output from many sensors provided in the power plant 20 in the accumulation unit 11 . The numerical data r _m (t _n ) of the process signal R _m may be the physical quantity detected by the sensor, or may be a normalized quantity that is dimensionless so that physical quantities with different unit systems can be compared with each other. be.

The acquisition unit 26 acquires the weather information K (K _a , K _{b . .} . ) in time synchronization with the process signal R _m and stores it in the storage unit 11 . The weather information K includes, but is not limited to, temperature, humidity, atmospheric pressure, and the like. Further, the weather information K may be acquired by a dedicated measuring device (not shown), but it is also possible to utilize WEB information.

The selector 12 arbitrarily selects process signals R _m (m=1 to M) and/or weather information K (K _a , K _{b .} . . ) having different attributes 13 from the stored ones. This selection is performed by an operator's operation via the input unit 22. FIG. Furthermore, the input unit 22 can specify the state value J of the power plant 20, and the selection unit 12 selects the process signal R _m (m=1, 2) for calculating this state value J in the calculation unit 27. can be selected.

As shown in FIG. 2, _the setting unit 15 _sets a discrimination _class S _m ^x (x = 1 to d _m ). Furthermore, when the weather information K (K _{a ,} K _{b .} . . ) is selected in the selection unit 12, the identification class S ( S _a , S _{b .} . . ) are set. Furthermore, when the state value J is selected in the selection unit 12, the discrimination class S corresponding to the numerical value division 16 of the numerical data (not shown) is similarly set for each attribute 13 thereof.

FIG. 3 is an explanatory diagram of functions of the input unit 22, the selection unit 12, the conversion unit 17, and the generation unit 18 according to the embodiment. Here, it is assumed that the operator designates the process signal R (R ₂ , R ₅ , R ₈ ) through the input unit 22 . The selector 12 stores numerical data r ₂ (t _n ), r ₅ (t _n ), r ₈ (t _n ) {n=1, 2, . 11 to choose from.

An upper limit value 31, a lower limit value 32 and a number of divisions 33 are separately set for these numerical data r _m (t _n ) {m=2, 5, 8}. Here, when the number of divisions 33 is α, β, and γ, the classification class S ₂ of the numerical data r ₂ (t _n ) is the total of S ₂ ¹ , S ₂ ² . . . S ₂ ^α . The discrimination class _S5 ^of the data _r5 ( _tn ) is set ^to β types by the total of _S21 , _S22 ... ^S2β , and the discrimination class _S8 of _the numerical data _r8 ( _tn ⁾ is _S21 , S ₂ ² . . . S ₂ ^γ , the γ types are set. The setting of the discrimination class S is not limited to the case based on the upper limit value 31, the lower limit value 32 and the number of divisions 33 described above, and can be set according to the operator's sense.

In the conversion unit 17, numerical data _{r 2} (t _n ), r ₅ (t _{n )} , r ₈ (t _n ) _{ n=1, 2, . Each is converted into a corresponding discrimination class S (S ₂ ^x , S ₅ ^x , S ₈ ^x ). In the example of FIG. 3, the numerical data r ₂ (t ₁ ) of the process signal R ₂ is converted to the discrimination class S ₂ ³ , r ₂ (t ₂ ) is converted to the discrimination class S ₂ ⁵ , and r ₂ (t _n ) has been converted to ^the discriminant class S ₂₄ . Numerical data r ₅ (t ₁ ) of process signal R ₅ is converted into discrimination class S ₅₁ , r ₅ (t ₂ ) into discrimination class S ₅₂ , r ₅ ( _{t n} ⁾ into discrimination ^class It has been converted to S ₅ ³ . Numerical data r ₈ (t ₁ ) of process signal R ₈ is converted into discrimination class S ₈₄ , r ₈ (t ₂ ) into discrimination class S ₈₇ , r ₈ ( _{t n} ^{) into} discrimination class It has been converted ^to _S82 .

_The generation unit 18 generates groups G _n ⁽ n=1 _{, 2 ,} . The generated groups G _n (n=1, 2 . . . ) are stored in the storage unit 19 .

Here, group G ₁ consists of ^{a combination of discrimination classes S (S 23 , S 51 , S 84 )} _{, and group} ^{G 2} _consists of a combination of discrimination classes S (S ₂₅ , ^S ₅₂ , S ₈₇ ⁾ . and ^the group G _n consists of ^{a combination of discrimination classes S (S 24 ,} _{S 53} , S ₈₂ ). Note that here, each of the discrimination classes S (S ₂ ^x , S ₅ ^x , S ₈ ^x ) in the numerical data r _m (t _n ) {m=2, 5, 8} are α types, β types, γ Since it is set to the type, it can be said that there are a total of α×β×c combination patterns for the group G _n .

Returning to FIG. 1, the description continues.
The extraction unit 29 can extract the group G _n (n=1, 2, . . . ) stored in the storage unit 19 using the attribute 13 and time t _n of the process signal R _m as information keys. This makes it possible to easily extract the time series of groups G in which various process signals R, state values J, and weather information K are combined.

By configuring the operation management device 10 in this way, by describing the operation state by combining the classified process signal R, the state value J, and the weather information K, the partial load operation of the power plant 20 with a wide operating range can be achieved. can be comprehensively analyzed over the entire area.

(Second embodiment)
Next, a second embodiment of the present invention will be described with reference to FIG. FIG. 4 is a block diagram of a power plant operation management device 10B (10) (hereinafter simply referred to as "operation management device 10B") according to the second embodiment of the present invention. In FIG. 4, parts having configurations or functions common to those in FIG. 1 are denoted by the same reference numerals, and overlapping descriptions are omitted.

As with the operation management device 10A of the first embodiment, the operation management device 10B (10) of the second embodiment includes an accumulation unit 11 that accumulates the process signal R and a selection unit 12 that arbitrarily selects the process signal R. , a setting unit 15 for setting the discrimination class S, a conversion unit 17 for converting the numerical data r(t) into the corresponding discrimination class S, and a generation unit 18 for generating the group G time series.

Furthermore, the operation management device 10B includes a machine learning unit 45 that creates a learning model 46 using the time series of the generated group G as at least one of input and output teacher data 41 and 42, a virtual process signal R and weather information K and a prediction unit 47 that inputs the time series of the virtual group V generated based on to the learning model 46 and outputs prediction data 48 .

The machine learning unit 45 can create a learning model 46 by using the extracted time series of the group G as the teacher data 41 on the input side and power generation efficiency, fuel consumption, etc. as the teacher data 42 on the output side. Alternatively, the learning model 46 can be created by using the time series of the two groups G with different combinations of the attributes 13 as the teacher data 41 on the input side and the teacher data 42 on the output side. In order to create a learning model 46 with high validity, it is preferable to increase the data sets of the teacher data 41 on the input side and the teacher data 42 on the output side and perform machine learning.

A virtual process signal R can be simulated based on the power generation plan of the power plant 20 . Furthermore, by combining the weather information K in this power generation plan, the time series of the virtual group V can be generated. By inputting the virtual group V _n and the learning model 46 to the prediction unit 47, it is possible to output the prediction data 48 such as the power generation efficiency and fuel usage amount in this power generation plan.

By outputting the prediction data 48 from various combinations of the virtual process signals R and the like in this way, it is possible to use the data for analysis of factors of deterioration in power generation efficiency and correction of the simulation model when designing the power plant 20 . That is, assuming that the common virtual process signal R is input, the deviation between the output data of the simulation model and the prediction data 48 in the second embodiment is evaluated. Then, the simulation model is corrected so that the output data of this simulation model approximates the prediction data 48 .

Based on the flowchart of FIG. 5, the procedure of the power plant operation management method and the algorithm of the operation management program according to the embodiment will be described. First, the process signal R _m (m=1 to M) of the power plant 20 is received and stored (S11). A process signal _Rm having a different attribute 13 is arbitrarily selected from the stored signals (S12).

Then, a discrimination class S _m ^x (x=1 to d _m ) corresponding to the numerical division 16 of the numerical data r _m (t _n ) is set for each attribute 13 (S13). Subsequently, the numerical data r _m (t _n ) of the process signal R _m are converted into the corresponding discrimination class S _m ^x (S14). ^Furthermore _, a group G _n (n=1 _, 2, .

Next, a learning model 46 is created using the generated time series of the group G as at least one of input and output teacher data 41 and 42 (S16). Then, a virtual process signal R and weather information K are created according to the operation plan of the power plant 20 (S17), and a time series of the virtual group V is generated therefrom (S18). Subsequently, the time series of this virtual group is input to the learning model 46, and the forecast data 48 is output (S19 END).

According to the power plant operation management device of at least one embodiment described above, numerical data is converted into identification classes, and a group of identification classes that are time-synchronized in some selected process signals is generated in time series. As a result, it is possible to accurately manage the power generation efficiency when partial load operation is introduced, and to facilitate the assumption of a fuel plan and a financial plan.

While several embodiments of the invention have been described, these embodiments have been presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, changes, and combinations can be made without departing from the scope of the invention. These embodiments and their modifications are included in the scope and spirit of the invention as well as the scope of the invention described in the claims and equivalents thereof.

The power plant operation management device described above includes a control device that highly integrates a processor such as a dedicated chip, FPGA (Field Programmable Gate Array), GPU (Graphics Processing Unit), or CPU (Central Processing Unit), Storage devices such as ROM (Read Only Memory) and RAM (Random Access Memory), external storage devices such as HDD (Hard Disk Drive) and SSD (Solid State Drive), display devices such as displays, mice and keyboards, etc. and a communication I/F, and can be realized with a hardware configuration using a normal computer. Therefore, the constituent elements of the power plant operation management device can be realized by a computer processor, and can be operated by the power plant operation management program.

Further, the operation management program for the power plant is preinstalled in a ROM or the like and provided. Alternatively, this program is stored in a computer-readable storage medium such as a CD-ROM, CD-R, memory card, DVD, flexible disk (FD) as an installable or executable file. You may make it

Further, the power plant operation management program according to the present embodiment may be stored on a computer connected to a network such as the Internet, and may be downloaded via the network and provided. Also, the power plant operation management device can be configured by connecting and combining separate modules that independently perform the respective functions of the constituent elements by connecting them via a network or a dedicated line.

DESCRIPTION OF SYMBOLS 10 (10A, 10B)... Operation management apparatus, 11... Accumulation part, 12... Selection part, 13... Attribute, 15... Setting part, 16... Numerical classification, 17... Conversion part, 18... Generation part, 19... Storage part, 20 Power plant 21 Receiving unit 22 Input unit 26 Acquisition unit 29 Extraction unit 31 Upper limit value 32 Lower limit value 33 Number of divisions 41 Teacher data for input 42 Teacher data for output 45 Machine learning unit 46 Learning model 47 Prediction unit 48 Prediction data r _m (t _n ) Numerical data R _m Process signal K Weather information J State value, _Smx ...Distinguished class, _Gn ...Group ^, _Vn ...Virtual group.

Claims (8)

an accumulation unit for accumulating a plurality of process signals composed of time-series numerical data;
a selection unit that arbitrarily selects the process signals having different attributes;
a setting unit for setting an identification class corresponding to the numerical division of the numerical data for each of the attributes;
a conversion unit that converts the numerical data of the process signal into the corresponding discrimination class;
and a generation unit that generates groups of the identification classes time-synchronized in the selected process signal in time series. In the power plant operation management device according to claim 1,
a storage unit that stores the group;
An operation management device for a power plant, comprising: an extraction unit that extracts the group using the attribute and time of the process signal as information keys. In the power plant operation management device according to claim 1 or claim 2,
An operation management device for a power plant that generates a time series of the group including the identification class of the acquired weather information. In the power plant operation management device according to any one of claims 1 to 3,
An operation management device for a power plant, wherein the identification class is also set for a state value calculated from the process signal. In the power plant operation management device according to any one of claims 1 to 4,
An operation management device for a power plant in which the identification class is set based on the upper limit value, the lower limit value, and the number of divisions of the numerical data. In the power plant operation management device according to any one of claims 1 to 5,
a machine learning unit that creates a learning model using the time series of the generated groups as at least one of input and output teacher data;
an operation management device for a power plant, comprising: a prediction unit that inputs a time series of virtual groups generated based on the virtual process signal and weather information to the learning model, and outputs prediction data. Using the power plant operation management device according to claim 1,
a step of accumulating a plurality of process signals composed of time-series numerical data in an accumulator ;
arbitrarily selecting the process signals having different attributes by a selection unit ;
setting a discrimination class corresponding to the numerical division of the numerical data for each attribute by a setting unit ;
converting the numerical data of the process signal into the corresponding discrimination class by a conversion unit ;
A power plant operation management method, comprising the step of generating in chronological order groups of the identification classes that are time-synchronized in the selected process signal by a generation unit . to the computer,
accumulating a plurality of process signals composed of time-series numerical data;
arbitrarily selecting the process signals having different attributes;
setting a discrimination class corresponding to the numerical division of the numerical data for each of the attributes;
converting the numerical data of the process signal into the corresponding discrimination class;
A power plant operation management program causing execution of the step of chronologically generating groups of the discrimination classes that are time-synchronized in the selected process signal.

Images