JPH09205731A - Power system computing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make effective use of operation independency in conducting computation of simultaneous ordinary differential equations on the internal condition of a generator and a load while ensuring preference of a parallel computer system of bus coupling type shared memory system in a transmission network computation. SOLUTION: This power system computing device 1 is provided with a bus 15, a plurality of processors 10 which consist of central processing units 10 and local memories 12, and a shared memory 16. In a computing process of a simultaneous ordinary differential equation on the internal condition of a generator and a load, the respective processors 10 conducts computing using data 14 on the internal condition of the generator and the load on the local memories 12. In a process where transmission network computation is computed by solving an simultaneous equation, the respective processors 10 conducts the allocated transmission network computation using data 17 for transmission network computation of the shared memory 16.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an electric power system calculation device, and more particularly to an electric power system calculation device for performing an analytical calculation regarding an electric power system including a plurality of generators, a plurality of loads, and a power transmission network interconnecting them.

[0002]

2. Description of the Related Art The electric power system is the core of the electric power business, and its contents are power generation facilities (hereinafter,
A generator), a transmission facility that distributes the electric power generated by the generator, a substation facility that sends load power (hereinafter referred to as a load), and a demand facility that consumes it. It is necessary to operate the power system efficiently without causing frequency changes or local voltage changes between these facilities.

The analysis calculation of the power system is performed by expanding the system scale,
The demand for high-speed processing is increasing due to the demand for immediate response.

On the other hand, in recent years, the progress of microprocessors has been remarkable. It is expected that it will be possible to construct a power system calculation device with much better cost performance than the conventional system by performing the power system calculation using a parallel computer system in which a large number of microprocessors are connected.

Generally, in the analysis calculation of the electric power system,
In order to solve the simultaneous ordinary differential equations related to the internal states of multiple generators and multiple loads, the data at each time of the simultaneous ordinary differential equations related to the internal states of generators and loads should be calculated and established between them. Repeat the grid calculation.

The percentage of time for these calculations depends on the algorithm used. Therefore, it is necessary to efficiently perform both of the simultaneous ordinary differential equation calculations concerning the internal states of the generator and the load, and the grid calculation.

In the latter transmission network calculation, the solution of simultaneous equations using LU decomposition is the main content. Generally, in the case of analytical calculation of a power system, since the number of couplings between the generator and the load is small, when the simultaneous equations to be solved are expressed as a matrix, the non-zero element rate of this matrix becomes very small.

When solving such simultaneous equations of a matrix having a very small non-zero element rate in parallel, the granularity of parallel processing is finer than that of a distributed memory type parallel computer system which requires explicit communication between processors. Is more efficient in the case of a bus-coupled shared memory parallel computer system in which a plurality of processors performs operations while referring to / updating data arranged in a shared memory connected to a bus (Hiroshi Kasahara:
System software for parallel processing, information processing Vo
l. 34, no. 9, pp. 1134-1138 (19
93)).

[0009]

On the other hand, in the calculation of simultaneous ordinary differential equations regarding the internal states of the generator and the load, the data are completely independent for each generator and the load. When such a calculation is performed by a bus-coupled shared memory system parallel computer system, even though the data is originally completely independent, when each processor refers to the shared memory via the bus, at the same time, one Since only the processor can refer to the shared memory, another processor cannot proceed with the calculation.

Further, in the bus-coupled shared memory system parallel computer system, there is a problem in that not only data but also a program (a group of instructions executed by a processor) may have to wait.

An object of the present invention is to ensure the superiority of a bus-coupled shared memory type parallel computer system in transmission network calculation, while maintaining the independence of operations in the calculation of simultaneous ordinary differential equations regarding the internal states of the generator and load. An object is to provide an electric power system calculation device that effectively utilizes the.

[0012]

A power system calculation apparatus according to the present invention performs an analytical calculation for a power system including a plurality of generators, a plurality of loads, and a power transmission network interconnecting them at every hourly step. In a power system calculation device that calculates simultaneous ordinary differential equations related to the internal states of a generator and a load, and then repeatedly solves the simultaneous equations to calculate the transmission network that holds between them. A bus that enables connection of clothing,
A central processing unit that is connected to the bus and performs arithmetic operations;
A plurality of processors each including a program executed by the central processing unit and a local memory, for holding data regarding the internal states of the generator and the load, which are assigned to the processor among the data regarding the internal states of the generator and the load. And a shared memory that is connected to the bus and holds data necessary for the power grid calculation, and in the calculation process of the simultaneous ordinary differential equations regarding the internal states of the generator and the load, each processor is In the step of performing a calculation using data on the internal states of the generator and the load on the local memory, and calculating the grid calculation by solving simultaneous equations, each processor uses the value of the shared memory. It is characterized in that the transmission network is assigned and calculated.

Further, in the electric power system calculation device according to the present invention, in the calculation step of the simultaneous ordinary differential equations regarding the internal states of the generator and the load, each processor is
After reading the connection terminal voltage data of the power grid relating to the generator and load assigned to the processor from the shared memory, the internal state of the generator and load assigned to the processor existing in the local memory Data is independently calculated based on the Runge-Kutta method, and the data of the internal induced voltage and admittance obtained as a result are transferred to the shared memory, and in the power grid calculation step, the processors share the shared data. The grid calculation assigned to the processor is executed while referring to and updating the data necessary for the grid calculation on the memory.

The power system calculation device according to the present invention is
In the calculation process of the simultaneous ordinary differential equations related to the internal states of the generator and the load, a modified equation of the variable indicating the internal state of the generator and the load based on the trapezoidal formula and a variable indicating the state of the power grid are used. In the case where each state variable solution is obtained by iterative correction using the Newton-Raphson method in parallel with the correction equation, the connection terminal voltage data of the transmission network regarding the generator and the load is read out from the shared memory, and the generator and The residuals of the equations relating to the internal state variables of the load and the elements of the modified equation coefficient matrix (Jacobi matrix) are independently calculated using the data relating to the internal states of the generator and the load assigned to the processor and stored in the local memory. Storage, residuals of equations related to the state variables of the grid and elements of the modified equation coefficient matrix (Jacobi matrix) The calculated using the data necessary to the grid computing assigned to the processor with data necessary for stored was power grid calculated local memory and the shared memory, stored the result in the shared memory,
In the step of calculating the correction equation, for the correction equation coefficient matrix relating to the generator and the load, refer to the LU decomposition and forward elimination operation of the element assigned to the processor, the residual of the local memory and the element of the Jacobian matrix. / Execute independently while updating, and further update the data on the shared memory when it is necessary to update the data of the part related to the power grid with the calculation of the generator and the load by the LU decomposition and the forward elimination operation. Regarding the modified equation coefficient matrix and residuals of the power transmission network, each processor executes LU decomposition and forward elimination operation of the elements assigned to the processor while referring to / updating data on the shared memory, Each processor uses the data in the shared memory to determine the correction amount related to the state variable of the power grid. It is determined by applying a backward substitution calculation to the residual subjected to forward elimination, and the correction amount of the state variable related to the generator and the load is corrected by each processor from the shared memory to the correction amount data of the state variable of the power grid. After reading, the Jacobian matrix in the local memory and the residual data that has been subjected to forward elimination and backward substitution are referenced to determine by backward substitution calculation to correct all state variables, and the correction amount of all state variables is small. Repeatedly calculate until.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the present invention will be described in detail with reference to the drawings.

[0016]

FIG. 1 is a block diagram showing the configuration of a power system computing device 1 according to an embodiment of the present invention. The power system calculation device 1 of the present embodiment has a form in which a plurality of processors 10 and a shared memory 16 are connected by a bus 15.

Each processor 10 includes a central processing unit 11
And a local memory 12.

The local memory 12 is a program 13 executed by the processor 10 and data concerning the internal states of the generator and the load in charge of the processor 10 (generator / load).
(Abbreviated as load data) 14 is held.

The shared memory 16 holds data 17 (abbreviated as power grid calculation data) necessary for power grid calculation.

FIG. 2 shows the flow of the entire processing when the transient stability calculation of the power system calculation is executed by the Runge-Kutta method in the power system calculation device 1 of this embodiment. ing. Referring to FIG.
This process is basically the calculation step 21 relating to the generator / load and the transmission network calculation step 2 at each time step.
2 is repeated 4 times.

The transient stability calculation described in the power system calculation device 1 of the present embodiment is basically a simultaneous ordinary differential including a grid calculation for M _g generators and M _l loads. It is a problem that solves an equation.

As a method of solving this in parallel, if there are n (positive integer) processors 10, M _g generators and M _l loads are evenly allocated to each processor 10.

Regarding the power transmission network calculation step 22, each processor 10 is assigned to each row.

In general, for one generator, the voltage as input

[Outside 1] _g and current

[Outside 2] There is a system of ordinary differential equations in response to _g, and as a result, the internal induced voltage of the generator is

[Outside 3] _g and internal impedance 1 /

[Outside 4] _g is obtained.

On the other hand, for one load, the terminal voltage

[Outside 5] Admittance with ₁ as input

[Outside 6] ₁ is obtained.

Since these generators and loads are connected to each other through the power transmission network, the relationship of the equation 1 is established.

[0027]

[Equation 1]

Here,

[Outside 7] Is the terminal voltage, in the case of a generator

[Outside 8] _g , for load

[Outside 9] _Is one. Also,

[Outside 10] Is the grid admittance matrix. By solving Equation 1, the terminal voltage

[Outside 11] Is obtained.

Further, the internal induced voltage of the generator

[Outside 12] _g and the terminal voltage of the generator

[Outside 13] _g and current

[Outside 14] _Since the relationship of 2 holds with _g , the current of the generator is

[Outside 15] _g is obtained.

[0030]

[Equation 2]

FIG. 3 is a flow chart for explaining the calculation step 21 relating to the generator / load in FIG. 2 in more detail. Each processor 10 executes the generator operation step 30 as many times as the number of allocated generators.

The following ordinary differential equations hold for the calculation step 30 of the generator (see Taiji Sekine: Power System Transient Analysis, Ohmsha (1984), pp. 377).

Regarding the rotor motion equation, the equations 3 and 4 hold.

[0034]

(Equation 3)

[0035]

(Equation 4)

Regarding the change equation of the interlinkage magnetic flux, the equations (5), (6), (7) and (8) are established.

[0037]

(Equation 5)

[0038]

(Equation 6)

[0039]

(Equation 7)

[0040]

(Equation 8)

Regarding the characteristic formula of the automatic voltage regulator,
Holds.

[0042]

[Equation 9]

Regarding the governor characteristic formula, the following equation 10 holds.

[0044]

(Equation 10)

Each processor 10 has a terminal voltage of the generator obtained from the shared memory 16 in the immediately preceding grid calculation step 22.

[Outside 16] _{Based on g} , the current of the generator

[Outside 17] The values of _g and internal variables in the generator are calculated based on the Runge-Kutta method, and finally the admittance is calculated.

[Outside 18] _g and internal induced voltage of generator

[Outside 19] The coefficient value on the left side and the value on the right side of Equation 1 calculated from _g are written in the shared memory 16.

After that, each processor 10 executes the load calculation step 31 by the number of assigned loads.

Regarding the load calculation step 31,
The ordinary differential equations of Equations 12 and 13 are established (Yamamoto Masato: Dynamic load model for system voltage stability analysis, The Institute of Electrical Engineers of Japan, B, Vol. 114-B, No.
12, pp. 1273-1279 (1994)).

[0048]

[Equation 11]

[0049]

(Equation 12)

[0050]

(Equation 13)

Each processor 10 has the voltage of the load obtained from the shared memory 16 in the immediately preceding grid calculation step 22.

[Outside 20] _{Based on 1} , the value of the internal variable in the load is calculated based on the Runge-Kutta method, and finally the admittance

[Outside 21] Writing the number 1 of the left side of coefficients calculated from ₁ to the shared memory 16.

The execution time required for the operation step 34 based on the Runge-Kutta method is the shared memory read step 32.
And much longer than the time required for the shared memory writing step 33.

Further, in the operation step 34 based on the Runge-Kutta method, each processor 10 can execute the operation completely independently.

FIG. 4 is a flow chart for explaining the power grid calculation step 22 in FIG. 2 in more detail.

In the transmission network calculation step 22, the equation 1 is used as the internal induced voltage of the generator obtained in the calculation step 34 based on the Runge-Kutta method.

[Outside 22] _g , generator admittance

[Outside 23] _g and load admittance

[Outside 24] Solve based on ₁ . That is, from Equation 1, the terminal voltage

[Outside 25] Ask for.

The solution of the simultaneous equations is occupied by the LU decomposition / forward elimination step 40 and the backward substitution step 41. Among these processes, the LU decomposition process has the longest processing time. Therefore, the parallel execution of LU decomposition will be described in more detail with reference to FIG.

In general, LU decomposition is performed in the following steps. Below, the n * n matrix that is the target of LU decomposition is (n is the size of the matrix)

[Outside 26] And

[Outside 27] The element in the i-th row and the j-th column of is represented by a _{i, j} . At this time, processing of the LU decomposition outer product format is expressed as follows (Haruo Tsuda: Numerical Processing Programming, pp. 1).
21-123, Iwanami Shoten).

For the kth row (k = 1 to n), the following is repeated. _{-A k, k = 1 / a} k, k - For the i-th row (i = k + 1~n), it repeats the following. * A _{i, k} = a _{i, k} * a _{k, k} * The following is repeated for the j-th column (j = k + 1 to n).・ A _{i, j} = a _{i, j} −a _{i, k} * a _{K, J}

In LU decomposition, each processor 10 takes charge of processing row by row. The process itself is performed by a process (50) of repeating the k-th row to be decomposed from k = 1 to n.

In the process, the processor 10 in charge of the k-th line indicates that the line has reached the line concerned.
Notify 0. The other processors 10 wait for processing until the notification comes.

After that, each processor 10 has the (k +
1) Regarding the i-th row in the row to the n-th row which the processor 10 is in charge of, a new value of each element in the row in charge is calculated based on the value of the kth row (step 51). That is, after updating the value of the element a _{i, k} in the i-th row and the k-th column, each element in the j-th column from the k-th column to the n-th column in the i-th row is calculated (step 52). _Find the value of a _{i, j} . In the actual calculation, since the matrix is a slant matrix, the calculation is performed only on the element having a non-zero value.

In the LU decomposition / forward elimination step 40, all the values of the elements a _{i, j} of the Jacobian matrix are arranged in the shared memory 16. In this part, communication between processors that originally occurs when reading the value of the pivot row is shared memory 16
It substitutes by using.

In the LU decomposing / forward erasing step 40 in which there are many data sharing processes, the superiority of the bus coupled shared memory system parallel computer system is secured.

FIG. 6 shows the flow of the entire processing when the transient stability calculation in the power system calculation device 1 of this embodiment is executed by the trapezoidal formula.

Basically, at each time step, the Newton-Raphson method process 6 is performed until the convergence condition is satisfied.
0, that is, a residual calculation step 61 relating to the state equations of the generator, the load, and the power grid, and a Jacobian matrix element calculation step 62 of the correction equation for obtaining the correction amount of the state variables of the power generator, the load, and the power grid. , The LU decomposition of the modified equation coefficient matrix (hereinafter referred to as the Jacobian matrix) of the load and the transmission network, the forward elimination and the backward substitution step 63 are repeated.

Here, a solution method based on the trapezoidal formula will be described. Ordinary differential equation for the generator (Equation 3 to Equation 10)
Then, an equation obtained by time-differentiating the ordinary differential equation (Equation 11) regarding the load using a trapezoidal formula is obtained. When this equation is combined with the algebraic equations regarding the generator and the load (Equation 2 and a similar equation not shown) and the algebraic equations regarding the power grid, the equation is expressed by Equation 14.

[0067]

[Equation 14]

Here,

[Outside 28] Shows all state variables of the generator and load and the grid.

The simultaneous equations of the equation (14) are non-linear equations, and when the Newton-Raphson method is used as the numerical solution, the modified equation becomes the equation (15).

[0070]

(Equation 15)

Here, 1, 2, ..., I, ..., N are identification numbers assigned to each equation of the simultaneous equations and each state variable, and the matrix on the left side is a coefficient matrix of the simultaneous equations, which is called a Jacobian matrix. is there. {Δx _i } indicates a correction amount vector added to each state variable provisional solution, and {r _i } indicates a residual vector when each state variable provisional solution is substituted into each equation.

FIG. 7 shows the process 60 of the Newton-Raphson method of FIG. 6 in more detail.

A generator and a load are mechanically assigned to each processor 10. Regarding the power grid,
The terminals for connecting to the generators or loads of the power transmission network are mechanically allocated to the processors 10 to which the generators and loads are allocated, by the number of nodes.

FIG. 8 shows the structure of the Jacobian matrix. Blocks in which nonzero matrix elements exist in the blocks surrounded by dotted lines are hatched, and blocks in which all matrix elements are zero. Is not hatched. However, even in the hatched blocks, not all matrix elements are non-zero.

Blocks not hatched do not require calculation in all cases. In the operation of the Newton-Raphson method, the structural features shown in FIG. 8 are used to arrange the data memory so that the time required for reading and writing the data described below is reduced. That is, the generator / load data 14 is stored in the local memory 12, and the power grid calculation data 17 is stored in the shared memory 16.

Each processor 10 relates to a generator and a load in charge thereof and has a terminal voltage of the generator.

[Outside 29] _g and load terminal voltage

[Outside 30] After reading ₁ from the shared memory 16 (step 32), using the local memory 12 of each processor 10, the residual calculation step 61 relating to the generator, load and state equation of the power grid in charge, and the generator, load, and A Jacobian matrix element calculation step 62 of a modified equation regarding a power transmission network and a LU decomposition and forward elimination calculation step 71 of a Jacobian matrix regarding a generator / load are performed. And the resulting generator
The step 33 of storing the update data of the Jacobian matrix element at the contact point of the load with the power transmission network in the shared memory 16 is performed.

The residual calculation step 61 relating to the state equations of the generator, the load and the transmission network in charge, the Jacobian matrix element calculation step 62 of the modified equation relating to the generator, the load and the transmission network, and the Jacobian matrix relating to the generator and the load. The execution time required for the LU decomposition and forward erasure calculation step 71 is much longer than the time required for the shared memory read step 32 and the shared memory write step 33.

Further, a residual calculation step 61 regarding the state equation of the generator, the load and the transmission network in charge, a Jacobian matrix element calculation step 62 of the modified equation regarding the generator, the load and the transmission network, and a Jacobian matrix regarding the generator / load. In the LU decomposition and forward erasure calculation step 71 of, each processor 10 can execute the operation completely independently.

Thereafter, the processors 10 refer to the shared memory 16 and frequently perform LU decomposition and forward elimination calculation step 72 of the Jacobian matrix related to the power transmission network (while communicating with each other by updating the data in the shared memory 16). , And the backward substitution calculation step 73 relating to the power transmission network.

For steps 72 and 73, LU
Processing is performed by the same method as the disassembly / forward elimination step 40 and the backward substitution step 41.

After that, each processor 10 reads out the correction amount of the state variable of the power transmission network on the shared memory 16 and then independently performs the backward substitution process 74 regarding the generator and the load on the local memory 12.

Thereafter, each processor 10 refers to the local memory 12 and performs the modifying step 75 on all of the state variables assigned to each processor 10.

As a result of the backward substitution for the power transmission network, the correction amount of the tentative state variable solution for the power transmission network is obtained and stored in the shared memory 16.

Subsequently, each processor 10 reads from the shared memory 16 the generators shared by the respective processors 10 and the correction amount of the tentative state variable solution of the contact point between the load and the power transmission network, and uses them to read the generators. Then, the backward substitution calculation step 74 regarding the load and the load is performed, and the correction amount of the temporary solution of the state variable regarding the generator and the load is obtained and stored in each local memory 12.

After that, each processor 10 tentatively solves the state variables relating to the generator and the load and the power grid.

[Outside 31] To the correction amount Δ

[Outside 32] The state variable solution is updated by the correction calculation step 75 according to the equation (16).

[0086]

(Equation 16)

This series of operations is repeated until the amount of modification of all state variables becomes sufficiently small, and the provisional solution at the end of the repetition is taken as the solution.

[0088]

As described above, according to the present invention,
Regarding the transmission network calculation process, there is an effect that the superiority of the bus-combined shared memory parallel computer system is secured.

On the other hand, in the process of calculating the simultaneous ordinary differential equations regarding the internal states of the generator and the load, each processor performs calculation by using the local memory of the processor, so that it is not necessary to refer to the shared memory on the bus. Therefore, there is an advantage that performance degradation due to bus reference contention does not occur.

Further, in both the power transmission network calculation step and the generator / load calculation step, since each processor reads the instruction of the processor from the local memory, there is an advantage in that the instruction read performance does not deteriorate.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a power system calculation device according to an embodiment of the present invention.

FIG. 2 shows the Runge in the power system calculation device according to the present embodiment.
It is a flowchart which shows an example of the electric power system calculation based on the Kutta method.

FIG. 3 is a flowchart for explaining in more detail the calculation process of simultaneous ordinary differential equations regarding the internal states of the generator / load in FIG.

FIG. 4 is a flowchart illustrating the power grid calculation process in FIG. 2 in more detail.

5 is a flowchart illustrating parallel execution of LU decomposition / forward erasing steps in FIG. 4. FIG.

FIG. 6 is a flowchart showing an example of a power system calculation based on a trapezoidal formula in the power system calculation device according to the present embodiment.

FIG. 7 is a flowchart for explaining the processing of the Newton-Raphson method in FIG. 6 in more detail.

FIG. 8 is a diagram showing a structure of a Jacobian matrix.

[Explanation of symbols]

1 Power System Calculator 10 Processor 11 Central Processing Unit 12 Local Memory 13 Program 14 Generator / Load Data 15 Bus 16 Shared Memory 17 Power Grid Calculation Data 21 Generator / Load Calculation Process 22 Power Grid Calculation Process 30 Generator Calculation process 31 Load calculation process 32 Shared memory reading process 33 Shared memory writing process 34 Calculation process based on Runge-Kutta method 40 LU decomposition / forward erasing process 41 Backward substitution process 42 Forward erasing process 50 Pivot row (k) from 1 Step 51 of repeating up to n 51 Updating the element of the row (i) in charge from the (k + 1) th row to the nth row 52 Updating the elements of the (k + 1) th column to the nth column of the i-th row Process 60 Treatment process of Newton-Raphson method 61 State variable of generator, load and power grid Residual calculation step of the equation to be performed 62 Jacobian matrix element calculation step of generator, load and transmission network 63 LU decomposition, forward elimination and backward substitution step of generator, load and transmission network Jacobian matrix 64 Correction and time of state variables Update process 71 LU decomposition of Jacobian matrix and forward elimination calculation process for generator / load 72 LU decomposition and forward elimination calculation process of Jacobian matrix for power transmission network 73 Backward substitution calculation process for power transmission network 74 Backward substitution calculation process for generator / load 75 Modification process of all state variables

Claims (3)

[Claims] 1. An analytical calculation for a power system including a plurality of generators, a plurality of loads, and a power transmission network interconnecting them, and a simultaneous ordinary differential equation for the internal states of the generator and the load is calculated at each time step. In the electric power system calculation device that calculates the power grid calculation that is established between them by repeatedly solving simultaneous equations, a bus that enables the connection of multiple pieces of hardware, and a bus that connects to the bus. A central processing unit that performs calculations,
A plurality of processors each including a program executed by the central processing unit and a local memory, for holding data regarding the internal states of the generator and the load, which are assigned to the processor among the data regarding the internal states of the generator and the load. And a shared memory that is connected to the bus and holds data necessary for the power grid calculation, and in the calculation process of the simultaneous ordinary differential equations regarding the internal states of the generator and the load, each processor is In the step of performing a calculation using data on the internal states of the generator and the load on the local memory, and calculating the grid calculation by solving simultaneous equations, each processor uses the value of the shared memory. An electric power system calculation device characterized by performing a transmission network calculation that is assigned according to the above. 2. In the step of calculating the simultaneous ordinary differential equations regarding the internal states of the generator and the load, each of the processors obtains the connection terminal voltage data of the transmission network regarding the generator and the load assigned to the processor. After reading from the shared memory, the data relating to the internal states of the generator and the load assigned to the processor, which are present in the local memory, are used to independently calculate based on the Runge-Kutta method, and the resulting internal Transferring the data of the induced voltage and the admittance to the shared memory, in the power grid calculation step, each processor,
The electric power system calculation device according to claim 1, which executes the electric power grid calculation assigned to the processor while referring to and updating the data necessary for the electric power grid calculation on the shared memory. 3. In the calculation process of simultaneous ordinary differential equations concerning the internal states of the generator and the load, a correction equation of variables indicating the internal states of the generator and the load based on a trapezoidal formula, and In the case of solving each state variable solution by iterative correction using the Newton-Raphson method by making simultaneous use of the modified equation of the variable indicating the state, the connection terminal voltage data of the transmission network regarding the generator and the load was read from the shared memory. Later, the residuals of the equations for the internal state variables of the generator and load and the elements of the modified equation coefficient matrix (Jacobi matrix) are independently calculated using the data for the internal states of the generator and load assigned to the processor. Stored in the local memory, the residuals of the equation relating to the state variables of the grid and the modified equation coefficient matrix Elements of the matrix) are calculated by using data necessary for grid calculation stored in the local memory and the shared memory and used for grid calculation allocated to the processor, and the result is calculated as Stored in a shared memory, in the step of calculating the correction equation, for the correction equation coefficient matrix related to the generator and the load, the LU decomposition and forward elimination operation of the elements assigned to the processor are performed and the residual and Jacobian of the local memory are stored. Executes independently while referring / updating the elements of the matrix, and further on the shared memory when it is necessary to update the data of the part related to the power grid with the calculation of the generator and the load by the LU decomposition and the forward elimination operation. Data for the grid and the modified equation coefficient matrix and residuals of the grid are assigned to the processor. Each processor executes LU decomposition and forward erase operation of the stored elements while referring / updating the data in the shared memory, and each processor executes the correction amount related to the state variable of the power transmission network in the shared memory. Data is used to perform a backward substitution calculation on the residuals that have been subjected to forward elimination, and the correction amount of the state variables related to the generator and the load is corrected by the respective processors from the shared memory to the state variables of the power grid. After the correction amount data of is read, the Jacobian matrix on the local memory and the residual data subjected to the forward elimination and the backward substitution are referenced to determine by the backward substitution calculation, and all the state variables are corrected. 2. The electric power system calculation device according to claim 1, wherein the calculation is repeated until the correction amount becomes small.