JPH0869490A - System and method for interchanging distribution load - Google Patents

Abstract

PURPOSE: To fast process all load interchange plans by defining a made plan as an optimum candidate when the absolute value of the difference between the object function value of the plan and that of the optimum solution candidate satisfies the prescribed conditions. CONSTITUTION: A distribution monitor controller 9 fetches the load current, etc., of a feeder from a distribution system network as the process input information. A distribution load interchange plan making device 1 decides an optimum system switching operation procedure based on the process information and the input information received from an input device 3. Then the device 1 connects the elements whose array is changed based on the prescribed change data to a 1st plan consisting of an array of elements showing the load sections based on the distribution system network information, etc., and also makes a 2nd plan where the elements are arrayed in the order of connection. Then the object function value is calculated to show the item that maximizes/ minimizes the 2nd plan, and the 1st plan is decided based on the difference between the preceding object function value and the object function value of the 1st plan and on the prescribed installation value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system and method for providing an optimal plan for switching a system due to an accident, overload, construction or the like in a distribution system, and particularly to a huge load for a wide range of system switching targets. The present invention relates to a means for quickly solving an accommodation problem with a simple configuration.

[0002]

2. Description of the Related Art A distribution load accommodation plan is to re-transmit power to a section to be switched (hereinafter referred to as a concession section) when switching to a distribution system due to an accident, overload, construction, etc. in a distribution system ( Hereinafter, there is a problem of obtaining an optimum switching operation procedure (hereinafter, referred to as load accommodation procedure) for performing the accommodation.

In other words, in the event of an accident in the distribution system, when the power is supplied from the surrounding system to the downstream power failure section, or when the distribution system is overloaded. , How to load the downstream section from the surrounding system, or when constructing a section with a distribution system, how to switch the downstream section from the surrounding system Or etc.
It is a distribution load accommodation planning problem to find the optimum system switching operation procedure.

This distribution load accommodation planning problem generally has the following constraint conditions and objective functions.

[Restriction Conditions] (1) A flexible structure of a dendritic structure having a substation outlet (hereinafter referred to as a feeder) as a root. (Do not form a loop.) (2) Do not exceed the supply reserve capacity of the feeder. (Do not overload.) (3) Due to accommodation, the passing current of the switch and distribution line should not exceed the allowable value.

(4) The voltage drop should not fall below the lower limit due to accommodation.

[Objective Function] (1) Minimize supply hindrance.

(2) The operation of the manual switch is minimized.

(3) Different feeders (hereinafter,
Minimize flexible operations using a heterogeneous feeder).

(4) The remaining power supply to the feeder is leveled.

Conventionally, in the load accommodation planning problem in a distribution system, the maximum or minimum item that satisfies the above objective function is determined, and the optimum solution for system switching that maximizes or minimizes the item is determined. Various methods have been proposed for finding out in a realistically acceptable time.

[0012] For example, in the document “Algorithm for load accommodation of distribution system, National Conference of the Institute of Electrical Engineers of Japan No. 1003, Showa 63”.
, April "," Expert system for load accommodation of distribution system, IEEJ National Convention No. 989, April 1989, "
"Application of Hopfield neural network to load accommodation problem of power system, IEEJ Transactions B1
Vol. 11, No. 1, January 1991 "," Application of fuzzy reasoning to distribution accommodation problem, IEEJ Transactions on B 113, No. 5, May 1993 ", etc. Method,
By applying an expert system, a neural network, a fuzzy reasoning, etc., it was intended to realize a means for optimizing a distribution load accommodation plan problem within a practically acceptable time.

In addition, the document "HANDBOOK OF GENETIC ALGORI
As described in "THM (Van Nostrand Reinhold)," genetic algorithms have been attracting attention as a promising means for solving optimization problems by simulating the evolution of living things.

[0014]

As for the so-called "enumeration method", which considers all possible combinations of accommodation modes for a given distribution load accommodation planning problem, the heuristic method, expert system,
The common idea of the optimization means of the distribution load accommodation planning problem including application examples such as neural networks and fuzzy inference is that a certain initial solution is first prepared, and the contents of the initial solution are gradually and effectively applied. Change (that is, do not plan at random, but try to obtain an optimal solution with as few planning times as possible), and finally evaluate the value of the objective function that represents the item you want to maximize or minimize. Then, an excellent solution is adopted as a next candidate solution, and finally an optimal solution is obtained in a short time.

FIG. 2 is a diagram showing an actual model of a distribution network using a node-branch method in which a load section is a node and a switch for separating or connecting the load section is a branch, which is represented by a solid line. Indicates connection, and broken line indicates separation. Correspondence between nodes and branches is reversed compared to the usual expression method, but since the power load amount is assigned to each node and the feeder can be connected, it is suitable for carrying out the distribution load accommodation plan. There is. The load is discretely distributed along the distribution line, but it is treated as a node.

Further, the feeders F1 to F18 are roots of a power distribution system having a capacity of remaining supply reserve capacity, and are directional branches having only one end as a node.

The constraint condition of this distribution network is
The following are (1) to (3).

(1) Consider the "no-voltage open" state of the switch (fix the switch that operates "off → on" by the voltage of the upstream node to "off") (2) Consider the characteristics of the switch (3) Consideration of distant control / field controllability in switch operation Fig. 3 shows that a power failure due to a local accident that occurred in two locations, section 120 and 181, in Fig. 2 spreads to a wide area system. In order to prevent it, this is an example of the load accommodation result in which a tree rooted at several feeders in the vicinity is reconfigured and corresponding. If the load cannot be accommodated, an accident in section 120 will result in 120
The sections 73, 74, 78, 77, 72, which are further downstream,
116,65,64,69,110 became a power failure, and in 181 accidents 179,178,130,174,12
9,134,177,135,182,181,18
5,186,183,248,280,281,28
2,254,255,249,237,231,24
4,250,238,239,261,229,262
Will cause a power outage.

In the distribution load accommodation plan, it is intended to cover the concession system as broadly as possible by the rooted tree rooted at the link entrance from the adjacent distribution line. Therefore, it is possible to solve the combination problem with permutation on the graph in the Euclidean space. Come back.
Therefore, even in the case of a simple accident as shown in FIG.
The 181 neighborhood search alone is not sufficient, and a search for the entire space consisting of 18 feeders and 294 concomitant sections is required. Considering 18 feeders as sections, the number of sections is 29.
4 + 18 = 312. The distribution load accommodation plan generated by this section is 312! (>> ^{10 1000} ) Because there are streets,
The enumeration method that examines all of these cannot be solved within the time required for restoration.

Further, in the conventional methods other than the enumeration method, (1)
For each new solution planning, a process for planning a solution that is at least 3 to 5 times the number of concession sections, which is a parameter at the time of solution planning for optimizing the objective function, is required. (2) Since the optimality is empirically evaluated, there is a problem that it takes a long time to reach the optimal solution or the suboptimal solution.

Therefore, in a large-scale load accommodation plan problem, the concealed section is divided into, for example, each distribution line to reduce the load accommodation plan problem for each distribution line, that is, the number of concealed sections is reduced. By doing so, a means for reducing the number of combinations and shortening the processing has been taken. But,
With such means, there is a problem that the optimality is inferior because the supply obstacle remains and the number of manual switches to be operated increases, as compared with the load accommodation problem for the entire to-be-contracted section.

On the other hand, since the genetic algorithm has a large processability, especially the initial convergence to the optimal solution, the calculation time is short and there is a high possibility of improvement. However, as stated in the Operations Research July 1993 issue "Genetic Algorithms and Optimization" (pp. 347-351), this method has not reached the stage of maturity, so it is standard for various problems. Alternatively, a universal method has not been established and it is extremely difficult to apply it to the actual industrial field.

Further, in recent years, the present invention intends to solve the problem,
Although many attempts have been made to solve the combination explosion problem by using a massively parallel processing device in which a large number of processing devices are combined, an appropriate method or device for improving the efficiency of parallel processing and increasing the optimality is being developed. It did not exist.

The present invention can solve the problems of processability and optimality, and from among the distribution load accommodation planning problems in which the number of combinations is enormous, the accommodation mode for maximizing or minimizing the value of the objective function is extremely short. It is intended to provide a device for selecting by time.

[0025]

A feature of the present invention for solving the above-mentioned problems is that a preset plan array is provided for a first plan including an array of elements indicating a load section. The arrangement of the elements of the first plan is changed based on the changed data to be changed, and the elements are set based on connection information such as data of distribution system network status, load status of each section, and accident status of each element set in advance. Of the first plan whose arrangement has been changed and elements are arranged in the order in which they are connected to each other to create a second plan, and minimize or maximize the second new plan. A value of an objective function representing an item to be sought is calculated, and a first plan is determined based on a difference between the calculated value of the objective function and the value of the objective function of the first plan and a preset value. .

A feature of the present invention for solving the above-mentioned problems is that a plurality of first plans in which elements indicating a plurality of load sections are arranged are generated, and the plurality of first plans are predetermined. The array of each first plan is changed based on the array conversion data, and a plurality of second plans are generated based on the conditions of the elements predetermined for each first plan with the changed array, and the second plan is generated. A value of an objective function that represents an item to be minimized or maximized for each of a plurality of plans is obtained, and at least one of the second plans is selected based on the obtained value of the objective function and the selected second plan is selected. Of the plan, the array of the second plan selected based on the array conversion data set in advance is changed, and a plurality of third plans are set based on the conditions of the elements set in advance for the second plan in which the array is changed. Generate and third The value of the objective function that represents the item to be minimized or maximized for each number plan is obtained, and at least one third plan is selected based on the obtained value of the objective function. To find the optimal solution.

[0027]

According to the present invention, the difference between the generated objective function value of the distribution load accommodation plan and the objective function value of the optimal load candidate distribution load accommodation plan is obtained, and the absolute value of this difference is compared with a predetermined variable. However, by making the distribution load accommodation plan generated when a certain condition is satisfied the optimal solution candidate, it is possible to speed up the process without having to obtain the target difference function value for all distribution load accommodation plans. it can.

Further, when parallel processing is performed on a plurality of distribution load accommodation plans, the objective function value of each distribution load accommodation plan is compared with the objective function value of the optimal solution candidate, and a distribution load accommodation plan is selected to newly By creating a new distribution load accommodation plan, the one close to the desired value remains as the optimum solution candidate, and thus the solution converges quickly to the optimum solution.

[0029]

EXAMPLE An example of the present invention will be described below. FIG.
FIG. 1 is a configuration diagram of a distribution load accommodation planning system of the present invention. This system uses the input device 3 for inputting input information such as commands from the operator and the load current of the feeder, the on / off operation state of the automatic switch, etc. as process input information from the distribution system network shown in the lower right of the figure. A distribution monitoring control device 9 for inputting and outputting to the distribution load accommodation planning device 1, and conversely outputting the optimum system switching operation procedure from the distribution load accommodation planning device 1 to the distribution network as a process output signal, and input information. And a distribution load accommodation plan planning device 1 for obtaining an optimum system switching operation procedure from process input information, and an output device 2 for outputting the planning result to a display, a printer or the like.

Note that F1 to F3 in the power distribution network at the lower right of FIG.
tsw3 and sw35 represent switches, and the power supply state is shown by a forest (forest) which is a set of load section trees whose roots are feeders depending on whether the switch group is turned on or off.

The distribution load accommodation plan making apparatus 1 may be a conventional sequential execution type computer if the accommodation section is about 100 or less, but if it is more than that, the processing apparatuses are connected in parallel and the same time is used. A parallel computer that can perform multiple processes is required. In the present invention, an embodiment using a single sequential execution type computer will be described first, and then an example of using a parallel computer will be described.

FIG. 4 shows a processing flow of the distribution load accommodation plan making apparatus 1. First, in step I1, network configuration data indicating network connection, switch characteristics, etc., and in step I2 network status data indicating network load status, switch on / off, etc.
In step I4, one initial plan data is stored, and in step I4, various constants indicating genetic manipulation constants, objective value function evaluation constants, etc. are stored in the memory in the distribution load accommodation plan planning device 1 before the main plan is prepared. And the optimization process A is repeatedly executed by using the set data while performing the genetic operation until the optimum solution of the distribution load accommodation plan is obtained. The number of times this is repeated is called the number of generations.

The processing result at this time is sequentially output to the output device 2 of FIG. The output data includes distribution system network diagram plan evaluation FBOUT (k, nn), accommodation tree configuration TREE (k), and concession section chromosome XX (g), and also system voltage drop value, work man-hours, loss cost, etc. It may be defined arbitrarily.

FIG. 6C shows the plan evaluation FBOUT (k, n
It shows an example of the output of n). Here, nn is an evaluation item, nn = 1 is the total load value, nn = 2 is the number of switch operations, and nn = 3 is the number of load accommodation sections. The flexible tree structure TREE (k) indicates a parent node to which the node k is connected, and TREE = (17, 17,
2,1,2,3,11,9,10,13,19,13,
15, 11, 18, 15). For example TRE
E (3) = 2 means that the parent node to which the node 3 is connected is the node 2. Incompatibility section chromosome XX
(G) is the final result when the redundancy of the forest generated from the loaded section chromosome X is 0, and in this example, X
X (g) = (1,2,3,4,5,6,11,14,1
5,16,7,13,10,12,9,8).

The processing of FIG. 4 will be described in detail below, but in order to help understanding of the processing flow, a distribution load accommodation plan in the event of an accident using the simple network shown in FIG. The planning will be explained. Figure 5
20 of load sections (hereinafter referred to as "nodes") 1 to 16, 99, 100, 199, 200, each of which has five feeders F1 to F5 and has different supply reserve power.
In the section, the solid line shows the connected state and the broken line shows the separated state.
The parentheses of each feeder indicate the reserve power (<pre-accident reserve power> → <post-accident reserve power>).

If an accident occurs in the section 100 and the section 200, electric power cannot be supplied from the feeders F4 and F5, so that the sections 3, 7 and the sections 8, 12 are out of power. Therefore, these power failure sections must be restored using the feeders F1, F2 and F3. In this case, it is possible to make a load accommodation plan for a narrow range including the power outage section or the nearby section and restore it. However, limiting the range may leave a power outage in some situations. If you can plan at high speed, it is better to plan with as wide a system as possible. Therefore, in this embodiment, a load accommodation plan for all 16 sections including the power failure section and the surrounding charging section is prepared.

First, in step I1 of FIG. 4, as shown in FIG.
As shown in (b), the network configuration data after the occurrence of the accident is set. The network configuration data is defined by the network configuration index NIX (k) and the network configuration table NDEF (k, l). Here, k represents a node number. FIG. 6A shows the network configuration index NIX (k), and shows the number that can be linked to the adjacent node. Therefore, for example, the node 5 can be linked with the nodes 2, 4, 6, 9 and four nodes, and thus NIX (5) = 4. Also, node 3 could be linked to nodes 2, 6, 200, but node 20
Since it is possible to link only to nodes 2 and 6 due to the 0 accident, NIX (3) = 2. It should be noted that here, each of the feeders F1, F2, F3 that can be effectively used after the occurrence of the accident is also 1
Two nodes. 6B shows the network configuration table NDEF (k, l), which is a pointer indicating the maximum path value that the node k can take (the maximum number of branches, that is, the maximum value that an adjacent node can take). The value of l is set so that it can take the maximum number of the maximum number of adjacent nodes of each node regardless of whether the switch is opened or closed. For example, the adjacent nodes of the node 3 are the nodes 2, 6 and 200,
After the occurrence of the accident at the node 200, the two nodes, the nodes 2 and 6, become NDEF (3, l) = (2,6,0,
0). Here, NDEF (k, l) is represented by four elements because the other nodes 2, 5, 6, 9, 10,
This is because each of 11 and 13 has four maximum adjacent nodes. (Note that 0 means that the gene is empty.) Next, in step I2, the network status data is set. The network status data is a vector LOAD that represents the actual load amount of each node shown in FIG. 5, feeder information FBINF (k), and data that indicates switching on / off of the network. The vector LOAD is, for example, LOAD
= (20,40,20,20,50,30,40,3
0,10,20,30,40,30,40,30,1
0) is defined. What is feeder information? Feeders F1, F
2 and F3 are the amounts of standby power that can be supplied, for example, FBINF (17) = 180, FBINF (18) =
Let 300 and FBINF (19) = 200. FIG. 7 shows an example of ON / OFF data of the switch of the network. In this figure, 1 indicates that the switch is "ON", 0 indicates that the switch is "OFF", and empty indicates that there is no connection.

Next, at step I3, an initial plan is generated using a random or empirical method.

In the initial plan, the load section forward vector X having the concession section number represented by a node as an element is called a load section chromosome, and the concession section number ii (element ii) is called a gene, and is defined as follows. To do.

X = (1,2, ..., ii, ... n) Since there are 1 to 16 genes in the conserved section in FIG. 5, the initial value X ₀ of the loading section chromosome X is X _0. For example X ₀
= (1,9,8,7,6,5,4,3,2,10,1
1, 12, 13, 14, 15, 16).

Next, in step I4, various constants such as a genetic constant and an objective function value evaluation constant are set.

FIG. 8 (a) shows an example of genetic constants stored in the memory of the distribution load accommodation plan making apparatus 1. Tables J (i) and K (i) are as follows. Set a number that has a uniform random distribution of positive integers that satisfies the condition of. However, i is a generation. (1) 1 ≦ i ≦ n ² (1) (2) J (i) <K (i) (2) (3) 1 ≦ J (i), K (i) ≦ n (3) Purpose The evaluation function value is "minimize total blackout load",
It represents the priority or weight of the objective function when the objective function such as “minimize the number of manual switches” is used for evaluation. Since many data can be set before the accident occurs, data such as genetic manipulation constants may be set in advance and the data after the accident may be inserted later.

The optimization process A is performed using the various data set in this way.

In the optimization process A, as shown in FIG. 9, the objective function difference value when the plan provided to the distribution load accommodation plan making apparatus 1 is changed by a predetermined genetic operation is foreseen, If the distribution load accommodation plan after the change is better, the distribution load accommodation plan after the change is made the optimal distribution load accommodation plan candidate, and if not, the given distribution load accommodation plan is made the optimal distribution load accommodation plan candidate. Find the optimal solution.

Next, the prediction of the result of the genetic operation of the process A10 of FIG. 9 will be described. FIG. 10 shows the specific processing contents of the processing A10. Tree root hotnode
(k), TREE (k), and XX (k) are generated.

First, the process for the generation number 1 which is the process for the initial value X ₀ will be described.

In step a, genetic manipulation is performed. In the present embodiment, an inversion operation will be described as an example of a genetic operation, but this may be an operation such as mutation. The inversion operation is performed based on the load section chromosome X (k) based on the genetic operation constants J (i) and K (i) set in step I4 of FIG.
That is, the genetic manipulation constant of the number of generations 1 is J (1) = 2, K
When (1) = 9, the initial plan given in step I3 of FIG. 4 is X ₀ = (1,9,8,7,6,5,4,3,2,2)
(10,11,12,13,14,15,16), so that the second gene to the ninth gene are arranged in reverse and X = (1,2,3,4,5,6,7, 8, 9, 11,
12, 13, 14, 15, 16).

In step b, as shown in FIG. 11B, the load section chromosome X (k) is set to the node-accepted section forward vector w.
Move to x (k). Below, the node conserved section forward vector wx
Based on (k), the processing from step c to step i is repeated to link each node in order.

In step c, the nodes which can be linked to the nodes of the node conserved section forward vector wx (k) are extracted using the network configuration table. In this case, the extracted node extracts any feeder or a node whose root is any feeder (hereinafter, referred to as “HOT parent node”). For example, when k = 1, the node that can be linked to the node 1 of the node conserved section forward vector wx (1) is NDEF (1, l) = (17,2,4,0) from the network configuration table of FIG. 6 (b). ), And since the node 17 is the feeder F1, it becomes the HOT parent node and is extracted. For nodes 2 and 4,
Since it is not linked to any node and does not have a feeder at the root, it is not extracted as a HOT parent node.

At step d, HOT at step c
It is determined whether the parent node is extracted. If the HOT parent node is not extracted, the process of step g is performed. If the HOT parent node is extracted, the process of step e is performed. Here, since the node 17 is extracted as the HOT parent node, the process moves to step e.

In step e, the remaining reserve power amount rpower of each node extracted in step c is compared to obtain the node having the maximum remaining reserve power amount. Where the remaining reserve power rpower
Is a value obtained by subtracting the actual load amount LOAD of all the nodes linked to the feeder from the FBINF reserve power amount that can be supplied by the feeder that is the root of the extracted node. Then, the node with the maximum remaining power reserve and the node-accommodated section forward vector wx linked to this node
The actual load amount of the node (k) is compared to determine whether the node having the maximum remaining reserve power amount can be the HOT parent node. That is, the node extracted in step c is node 17,
The remaining reserve power amount rpower = 180. Then, since there is no other node extracted in step c, the node 17 is set as the node having the maximum power reserve. Here, the actual load amount of the node 1 of the node conserved section forward vector wx (1) is LO.
Since AD (1) = 20, the node 1 can be linked to the node 17, so the node 17 is extracted as the HOT parent node.

In step f, the node extracted in step e and the node-to-condense-section forward vector wx (k) are linked, and the tree root hotnode (k), the accommodation tree configuration TREE (k), and the accommodation-section chromosome xx ( k), generate and output the plan evaluation FBOUT. Plan evaluation FBOU
T is the sum of the load electric energy linked to the feeder as described above, the number of switching operations compared with the state of the switch between the nodes before the planning set in step I2 of FIG. 4, and linked to the feeder. Output the number of connected nodes. In other words, as shown in the system diagram of FIG. 11A, the node 17 is linked to the node 1 of the node conserved section forward vector wx (1),
As shown in FIG. 11B, hotnode (1) = 17,
TREE (1) = 17 and xx (1) = 1. At this time, only the feeder F1 is linked with the nodes, so FB
The output of OUT is the output of only the feeder F1. Since only the node 1 is linked to the feeder F1, the load power amount is 20 and the number of links is 1. Also, the feeder F
The link between 1 and node 1 does not change the state of the switch between the nodes before planning, so the number of switching operations is 0. When the process of step f is completed, the process returns to step c again, and the process is performed for the next node conserved section forward vector wx (k).

Node conserved section forward vectors wx (2) to wx (6), that is, when the processes from node 2 to node 6 are repeated, nodes 1 to 6 all have the same feeder F1 as a root. You can link to hotno
de = (17,17,17,17,17,17,17), and a flexible tree configuration TREE = (17,17,2,1,4,5) indicating the parent node to which the nodes 1 to 6 are connected. .
Further, since only nodes of the feeder F1 are linked to the plan evaluation FBOUT, only the evaluation value of the feeder F1 is output.

When the process is advanced for the node 7 which is the node conserved section forward vector wx (7), the node 6 is once extracted in step c, but the remaining reserve power of the node 6 is already 0 and it becomes the HOT parent node. I can't.

In step g, it is judged whether the node-accommodated section forward vector wx (k) is deadlocked, and if it is deadlocked, the processing is stopped. The deadlock is a state in which the area including the node of the node conserved section forward vector wx (k) is isolated from the feeder, and a power failure will occur in this area. The judgment as to whether or not the deadlock state is established is performed as follows. If there is a node of the conserved section forward vector wx (k) that cannot be linked, the node conserved section forward vector wx (k) is extended in step j, and the node that cannot be linked after the last gene is duplicated. To do. Then, in the extended section, when the node returns to the original node without contributing to the generation of the tree even once, it is determined that the deadlock state is set. Therefore, since the node 7 has not finished checking the deadlock, it proceeds to step h and extends the node conserved section forward vector wx (k) as shown in FIG. To duplicate. Here, ndmax in step i is initially the same as the number of elements of the loaded section chromosome X, but the node-accommodated section forward vector wx
Each time (k) is extended, the value is incremented by 1. afterwards,
The process proceeds from step c for the next node conserved section forward vector wx (8).

FIG. 13 shows the processing results from node 8 to node 11. Similar to node 7, node 8,
9 and 10 cannot be registered as a tree, and the conserved section order vector wx
Has been extended. Only when it becomes the node 11, it is connected to the feeder F3, and the tree root hotnode = (1
7, 17, 17, 17, 17, 17, 0, 0, 0, 0,
19), flexible tree configuration TREE = (17, 17, 2,
1, 4, 5, 0, 0, 0, 0, 19). Then, a new node can be added to the conserved section chromosome XX only after becoming the node 11, and XX = (1, 2, 3, 4, 5, 6,
11). Further, as shown in FIG. 13C, when connecting to the feeder F1, the number of times of operating the switch which is different from the situation before the accident is 3 times before the processing up to the node 11.
There are 6 nodes connected to the feeder, the number of switch operations different from the situation before the accident when connecting to the feeder F3 is 0, and the number of nodes connected to the feeder F3 is 1.

Similarly, the nodes are linked to the tree to increase the number of chromosomes XX in the concession section, or the forward vector wx (k) in the concession section is extended. A forest made up of three trees shown in FIG. 14 is generated, and one load accommodation plan is generated. From the evaluation plan FBOUT at this time, it is understood that the load power amount of the feeder F3 is 110 and the remaining backup power amount is reduced to 200−110 = 90, for example.
The number of switch operations is 1, the number of concession sections is node 7,
It turns out that there are three, 11 and 14. Load section chromosome X
The length of 16 is 16
It extends from x to a length of 29. And the conserved section chromosome XX is XX = (1, 2, 3, 4, 5, 6, 1
1, 14, 15, 16, 7, 13, 10, 12, 9,
Although it is displaced to 8), it is equivalent to the load section chromosome X.
That is, the forests generated by the loaded section chromosome X and the conserved section chromosome XX are the same. This indicates that when a forest is generated from the loaded section chromosome X, it requires 29 preliminary studies, but 16 studies from the concession section chromosome XX are sufficient, and it is possible to repeat optimization of multiple individuals. It becomes a very important property in chemical logic. In other words, this means that from the next generation, this conserved section chromosome XX is used as a plan, and if a displacement is applied to this, step A10 in FIG. 7 “foreseeing the result by genetic manipulation” can be speeded up. This result is the evaluation plan F in FIG.
As indicated by BOUT, some constraints are satisfied before being output.

Next, step A2 in the process A of FIG.
0 to A30 will be described. In step A20, it is judged whether or not the solution obtained in step A10 is closer to the optimal solution than the optimal solution candidates up to the previous time. If it is closer to the optimal solution than the optimal solution candidates up to the previous time, step A30 Replace the solution with the optimal solution candidate.

In this way, one generation ends, and the next generation starts the process from step A10 for this optimal solution candidate. Then, this process is repeated until the optimum solution is obtained.

Here, step A20 and step A30
Will be specifically explained.

FIG. 15 shows the transition of the objective function value and its peculiar property. In FIG. 15 (a), the load section chromosome X _i obtained for the number of planning times _i is plotted on the horizontal axis. The function F (X _i ) is represented on the vertical axis. The arrangement of the load section chromosomes X _i is changed little by little, and the target function values at that time are compared and replaced with the load section chromosomes X _i close to the optimum solution. In this case, even when the extreme point (minimum point in this embodiment) such as X _i is reached as shown in the figure, the optimal load is loaded after shifting to the next load segment chromosome and planning a sufficient number of times. Reach interval chromosome Xopt. That is, with this empirical or brute force method of randomly displacing chromosomes, it takes a huge amount of time to reach the optimum solution Xopt when the number n of nodes in the plan is large. .

Therefore, attention is paid to the differential value ΔF (X _i ) with respect to X _i . In FIG. 15B, the load section chromosome X _i is plotted on the horizontal axis,
The vertical axis represents the differential value ΔF of the objective function F (X _i ) with respect to X _i .
It is expressed by taking (X _i ). Where the differential value Δ
F (X _i ) is represented by equation (4).

ΔF (X (i)) = dF (X _i ) / dX _i = F (X _{i + 1} ) −F (X _i ) ... (4) From the expression (4), the load section chromosome X _{i + 1} When the objective function value of is smaller than the objective function value of the previous load section chromosome X _i, it is negative, and otherwise it is zero or positive. Although the transition corresponds to ΔF as shown in the figure, the waveform becomes a waveform similar to the damping vibration (that is, a waveform in which the cycle changes as the vibration becomes smaller) with zero as the center. In the actual simulation result, in the case of a general load section chromosome X _i , such a shape is obtained regardless of the contents of the chromosome, and the attenuation degree of the amplitude when X _i is considered to be time is C / log (i + 2) ( In this embodiment, C = 10
It is similar to 0).

This means that when ΔF (X _i ) becomes negative, that is, even if the objective function value of the new load section chromosome X _{i + 1} becomes worse than the objective function value of the previous load section chromosome X _i (that is, If it is a maximization problem, it decreases, if it is a minimization problem, it increases), and its absolute value is C / log (i + 2)
If it is smaller, it means that the new load interval chromosome X _{i + 1} is replaced with the optimum solution candidate, so that the optimum solution is always reached even for the load interval chromosome X _i having a sufficiently large n.

This property does not require the processing from step A10 to step A30 for all the new load section chromosomes X _{i + 1} created at each iteration of the planning, and thus contributes greatly to the improvement of the performance of the optimization processing. To do. Thereby, the transition of the objective function value can be examined without depending on the optimization node number n, and the loaded interval chromosome problem can be solved at high speed.

From the above, in step A20, the objective function difference value ΔF (X (i)) is obtained and compared with C (i) shown in the following equation.

C (i) = A · a (i) / log (i + 2) (5) where a (i) is a uniform random number value from 0 to 1, and
It is sufficient that the values are within the diagonal lines of the transition of the distribution area of C (i) when A = 100 and a (i) = 1 shown in (b). A is determined by the value of ΔF (1).

When the absolute value of the objective function difference value ΔF (X (i)) is smaller than C (i), it is determined as an optimum solution candidate in step A30, and the absolute value of the objective function difference value ΔF (X (i)) is obtained. If is greater than or equal to C (i), the process of step A10 is executed for the previous optimal solution candidate.

The above is step A2 in the process A of FIG.
This is a processing method of optimal planning when a single sequential execution computer is used in the processing of 0 to A30. Next, a case where the parallel computer is used to speed up the processing will be described.

FIG. 16 is a detailed configuration of the distribution load accommodation plan making device 1, which is an input device control unit 8 for receiving input information from the input device 3, an output device control unit 7 for outputting an output result to the output device 2, A shared memory 6 for storing constants or variables, a cell processor massively parallel control unit 4 for controlling the entire apparatus, a shared memory 6 and a plurality of cell processor massively parallel control units 4 capable of executing respective processes at the same time Cell processor 5 of FIG. In this example, the cell processor 5 has 82
Although the number is 5, the number may be increased or decreased depending on the required processability.

At present, the capacity of a single processor that can be manufactured by general semiconductor devices and production equipment is about 1,300 (MIP.
S), it is possible to realize 1,300 × 825≈10 ⁶ (MIPS) = 1 (TIPS) from the configuration of this embodiment. That is, 10 ¹² instructions can be executed per second.

FIG. 17 shows a detailed configuration example of the cell processor 5. The main processing unit (CPU) 10 performs processing such as integer calculation and real number calculation, and a local storage device used only in the corresponding cell processor. 11 and a parallel processing scheduler 12 for synchronizing processing with other cell processors,
A channel control device 13 for transmitting / receiving information to / from the shared memory 6 and the cell processor massively parallel control unit 4 and for synchronizing processing timing, the main processing device 10, a local storage device 11, a parallel processing scheduler 12, and a channel control device 13. Is formed by a bus mechanism 14 that couples with each other.

FIG. 18 is a processing flow of the distribution load accommodation plan planning apparatus 1 using a parallel computer, which operates so that the optimization processing A and the processor allocation control processing B which are processed in parallel are repeated for the number of generations. The data used in the optimization process A and the processor allocation control process B are set in the shared memory 6 in advance, as in the case of using a single sequential processing type computer.

After the processing A, the processor allocation control processing B allocates a next-generation concession section chromosome to be processed by each cell processor 5 according to the objective function values of a plurality of defined load section chromosomes. Here, the number of generations is n ² / Pn, where n is the number of conserved sections and Pn is the number of cell processors. Therefore, if the number of cell processors 5 is the same as the number of genes n in the load section chromosomes, the number of generations is n ^2. / N = n.

At the initial stage, a random or empirical method is used as in the case of using a single sequential processing type computer, but a possible plan is generated by the number of cell processors 5 (step I3). ). The load section chromosome X _ipr (pr = 1 to Pn) is a multidimensional vector whose elements are the concession section numbers represented by nodes,
The initial value is, for example, X ₀₁ = (1,2, ..., i, ... n) X ₀₂ = (2,3, ..., i, ... 1): X _0Pn = (n, 1) , ..., i, ... N-1).

The load section chromosome X _0pr is the cell processor p.
The optimization process A is performed according to the process flow shown in FIG. In step A10, the inversion operation (invers
Ion) results are the same as when a single sequential processing type computer is used. However, in the processing of FIG. 10, individual cell processors 5 are used instead of the single sequential processing type computer. The points of implementation differ.

That is, FIG. 8C shows an example of genetic constants stored in the shared memory 6 of the cell processor 5 in advance. The table J (i, p) for the cell processor 5 is shown in FIG.
r) and K (i, pr) are set to numerical values having a uniform random number distribution of positive integers satisfying the following conditions from the first generation to the n ² generation. However, i is a generation and pr is a cell processor number. The constant C (i) is the same as that when a single sequential processing type computer is used.

(1) 1 ≦ i ≦ n ² (6) (2) J (i, pr) <K (i, pr) (7) (3) 1 ≦ J (i, pr), K ( i, pr) ≦ n (8) The operation of the process A of FIG. 18 for speeding up using a parallel computer has been described above. Pn = 1,
That is, if the number of the cell processors 5 is one, it becomes the case where the above-mentioned independent sequential processing type computer is used.

FIG. 19 (a) shows the processor allocation control process B of FIG. Which plan candidate is assigned to which cell processor can be solved by applying the evolutionary idea of living things. That is, it is possible to assign each plan candidate to each cell processor as follows.

In step B10 of FIG. 19A, the evaluation grade of the objective function value F (X _ipr ) and the bias BS for determining the maximum / minimum are determined. The bias BS is the maximum objective function value F (X _ipr ) +1 among a plurality of plans.

That is, in the example of FIG. 19B, the i-th generation plan (individual) is a for each of 10 cell processors.
(I) to j (i), and its objective function value F
(X _ipr ) is required. Bias BS is F (X _ipr ).
Since the maximum value in (pr = 1 to 10) is F (X _i6 ), BS = F (X _i6 ) + 1 = 320 + 1 = 321.

Next, at step B20, the adaptive value y (p
r) is obtained as y (pr) = BS-F (X _ipr ).

That is, in the example of FIG. 19B, y (pr) is y (1) = BS-F (X _i1 ) = 321−288 = 33 y (2) = 21 :: y (10) = 29 Is required as.

Next, the sum T of the adaptive values of the plan in all cell processors is

[0085]

[Equation 1]

It is calculated by.

Next, in step B30, the ratio z (pr) of each plan proposal adaptive value to the total sum T is obtained as z (pr) = y (pr) / T × 100%. That is, in the example of FIG. 19B, z (pr) is Z (1) = y (1) / T × 100% = 33/178 × 1
00% ≈19% Z (2) ≈12% :: Z (10) ≈16% Then, from this ratio, a plan (individual) is assigned to each cell processor in proportion to the number of cell processors. That is, since the number of processors to be allocated to a (i) next generation is 10 total processors × 19% ≈2, a (i + 1) is allocated to the cell processors 1 and 2, but f () has a low ratio. i), h (i), i (i)
Are not assigned to cell processors and have been culled. The above processing indicates that a plan whose objective function value is close to a desired value is handed down to the next generation with priority.

FIG. 20 shows the transition of the above-mentioned processing for each generation. The distribution load accommodation plan is assigned to 10 cell processors in the first generation from genus a to genus. It shows the transition when there is. Here, for example, the subscript of the genus a means the current generation, and
The superscript of the genus indicates the processor that has passed from the first generation to the current generation. For example, the subscript 112277 is
It has been shown that the cell processors have been processed in the order of 1 → 1 → 2 → 2 → 7 → 7.

FIG. 21 shows the state of proliferation by the generation of genus a in FIG. Since there are 6 descendants in the 7th generation, it can be seen that there were 6 types of optimization transitions. And the update transition of each plan clearly follows the route to reach the optimal solution. Furthermore, a plurality of different plans are mutated, evaluated, and propagated by the cell processors 5 assigned to each, and those having evaluation values close to the desired values for each generation remain as optimal solution candidates. Therefore, it converges to the optimum solution much faster than the conventional neural network or simulated annealing. The above is the description of the operation of the hardware and the optimization logic for performing the optimal distribution load accommodation plan at high speed.

FIG. 22 shows the evaluation items corresponding to the respective axes for the minimum total power outage load amount, the minimum number of manual switches, the maximum residual reserve power worst, and the minimum load interchange plan of the number of different system operations. It was obtained by changing the objective function on the radar graph. There is no plan that satisfies all objective functions at the same time,
It can be seen from the figure that it is necessary to appropriately establish a plan according to the situation.

FIG. 23 shows a result of planning an optimum solution in the allowable space by designating an allowable planning space in order to satisfy the above. For four kinds of objective functions, total blackout load amount ≦ 0 The number of times of manual switch operation ≦ 14 The minimum value of residual reserve power ≧ 50 The number of times of operation from different system feeders ≦ 14 is given, and the desired objective function is maximized or minimized in the space satisfying these conditions. This is an example of a load accommodation plan. Since all of these results satisfy all practical requirements, they can be immediately implemented.

FIG. 24 is a processing flow for realizing FIG. 23, which is obtained by adding processing step A15 to FIG.
It is only necessary to check whether or not the plan created by adding the genetic operation to the processing step A10 is in the designated space, and if it is outside the designated space, the optimal solution candidate may not be selected. By the above multi-purpose evaluation and the search method in the designated space, an extremely realistic plan can be obtained.

FIG. 25 shows the processing result of the distribution load accommodation plan making apparatus 1 having the configuration shown in FIG.
In the case of the 00 segment, the conventional method requires a processing time of several seconds to several minutes to find a possible solution which is a tentative feminine process, whereas the present invention provides an optimal solution of only 10 ms.
There is a big difference between the feasible solution and the optimal solution as well as the calculation speed.

On the contrary, FIG. 26 is a graph showing to what extent the number of sections can be optimized by giving a processing time allowable value.
In the case of minutes, it shows that the optimum solution can be obtained for a wide distribution network area having 8,800 sections. It is understood that all the distribution system network areas under the control of one system are about 8,000 sections and can be sufficiently covered in this embodiment.

[0095]

As described above, according to the present invention, by comparing the difference value of the objective function values of the load accommodation plan with a preset variable and replacing it with an optimal solution candidate, a large load is applied. The optimum solution of the load accommodation plan problem consisting of accommodation sections can be found at high speed. In addition, the load accommodation plan problem consisting of a sufficiently large load accommodation section is processed in parallel, and by determining the load accommodation plan to perform the next processing based on the objective function value of each processing, the evaluation value for each generation is desired. A value close to the value remains as an optimal candidate, and a limited processor can be effectively used, so that it converges to an optimal solution much faster than a conventional neural network or the like.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an embodiment of a distribution load accommodation system of the present invention.

FIG. 2 is a diagram showing modeling of a distribution system network.

FIG. 3 is a diagram showing an example of load accommodation when an accident occurs.

FIG. 4 is a processing flow chart of a distribution load accommodation planner.

FIG. 5 is a diagram showing a specific example when it is necessary to draw up a distribution load accommodation plan. It is the figure which showed the detailed structural example of a cell processor.

FIG. 6 is a diagram showing input / output information of distribution load interchange pattern generation evaluation processing. It is a processing flow figure of a distribution load accommodation plan device.

FIG. 7 is an example of a state of switches between nodes before planning.

FIG. 8 is an example of a genetic manipulation constant. It is a processing flow figure of distribution load accommodation pattern generation and evaluation.

FIG. 9 is an example of an optimization processing flow.

FIG. 10 is a diagram showing a flow of load accommodation pattern generation and evaluation processing.

11 is a diagram showing generation and evaluation transition of a load accommodation pattern for the load section chromosome X. FIG.

FIG. 12 is a diagram showing the generation and evaluation transition of a load accommodation pattern for the load section chromosome X.

FIG. 13 is a diagram showing generation and evaluation transition of a load accommodation pattern for the load section chromosome X.

FIG. 14 is a diagram showing generation and evaluation transition of a load accommodation pattern for the load section chromosome X.

FIG. 15 is a diagram showing a transition of an objective function value.

FIG. 16 is a diagram showing a distribution load accommodation planning device.

FIG. 17 is a diagram showing a configuration of a cell processor.

FIG. 18 is a diagram showing the processing of the distribution load accommodation planning apparatus in the case of parallel processing.

FIG. 19 is a diagram showing a processing flow chart of processor allocation control, and an individual of a generation and adaptive value transition for each cell processor.

FIG. 20 is a diagram showing a transition of a draft plan for each generation.

FIG. 21 is a diagram showing a state of proliferation of layer a.

FIG. 22 is a diagram showing an example of a multi-purpose evaluation graph.

FIG. 23 is a diagram showing an example of a multi-purpose evaluation graph when a planning space is designated.

FIG. 24 is a flow chart of an optimization process when a design space is specified.

FIG. 25 is a processing time graph required for optimization when a parallel processing device of 1 (TIPS) is used.

FIG. 26 is a processing time graph required for optimization when a parallel processing device of 1 (TIPS) is used.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Distribution load accommodation planning device, 2 ... Output device, 3 ... Input device, 4 ... Massively parallel control unit, 5 ... Cell processor, 6 ... Shared memory, 7 ... Output device control unit, 8
... Input device control unit, 9 ... Distribution monitoring control device, 10
... main processing unit, 11 ... local storage device, 12 ... parallel processing scheduler, 13 ... channel control device, 14 ... bus mechanism.

Front page continuation (72) Inventor Mitsuo Hayami 5-2-1 Omika-cho, Hitachi-shi, Ibaraki Hitachi Ltd. Omika factory (72) Inventor Minoru Kanai 7-1-1, Omika-cho, Hitachi-shi, Ibaraki Hitachi, Ltd. Hitachi Research Laboratory, Ltd. (72) Inventor, Yuzuru Imamura 7-1, 1-1 Omika-cho, Hitachi City, Ibaraki Prefecture Hitachi Research Laboratory, Hitachi Ltd.

Claims (17)

[Claims] 1. An array of at least one plan consisting of at least one element is set as an array of elements, and the array of elements of the set plan is changed based on predetermined array change data. A new draft is generated from the element information determined in advance for the plan with the array changed, and the objective function value of the plan is obtained, and the objective function value of the plan before the array is changed and the new plan. A method of planning, characterized in that a plan to be set for generating a new plan is determined based on an evaluation result of a result obtained from the objective function value of and a predetermined evaluation variable. 2. The planning method according to claim 1, wherein the difference between the objective function value of the plan before changing the array and the objective function value of the new plan is equal to or more than a predetermined evaluation variable. At this time, the plan planning method is characterized in that the plan set for carrying out a new plan is the plan before changing the arrangement. 3. The planning method according to claim 1, wherein the difference between the objective function value of the plan before changing the array and the objective function value of the new plan is less than or equal to a predetermined evaluation variable. A plan planning method, wherein the generated plan is set when a value is taken. 3. The planning method according to claim 1 or 2, wherein the change data is a set of two values, and the change of the array of elements of the plan is an element to which the two values correspond. A planning method characterized by arranging spaces in reverse. 4. Change data for changing at least one first plan consisting of at least one array and changing an array of elements of the first plan in advance to change the array of the plan. Generate a second plan based on the connection information, and connect the elements of the generated second plan based on the connection information of each preset element in the arranged order, and arrange the elements in the connected order. And generate a third plan, and based on the objective function value of the third plan, the objective function value of the first plan and the preset evaluation variable, the third plan or the first plan. At least one plan is selected from the plans, the selected plan is set as the first plan, and the second plan is selected based on the change data from the first plan.
The third plan is obtained based on the element information, and the third plan is calculated based on the objective function value of the third plan, the objective function value of the first plan, and the preset evaluation variable. 3. A planning method, comprising selecting at least one plan from the plan 3 or the first plan and setting the selected plan as the first plan. 5. The planning method according to claim 4, wherein the absolute value of the difference between the value of the objective function of the third planning and the value of the first objective function is less than or equal to a preset value. A method of planning, characterized in that the third plan is set as a new first plan. 6. The planning method according to claim 4 or 5, wherein the change data is data having two values as a set, and the change of the array of elements of the plan is the two values. A planning method characterized by arranging elements having corresponding values in reverse order. 7. The first power distribution based on the preset change data for the number of times of changing the change data for changing the arrangement of the plan for presetting the first distribution load accommodation plan including an array of a plurality of elements. Change the array of elements of the load accommodation plan,
The first distribution load accommodation plan in which the arrangement of the elements is changed is connected based on the connection information of each element, and the first
Of the new distribution load accommodation plan, and a second distribution load accommodation plan composed of an array of a plurality of new elements is created.
Of the distribution load accommodation plan, the value of the objective function representing the item to be minimized or maximized is obtained, and the third distribution load accommodation plan is obtained from the obtained value of the objective function and the preset value of the objective function. The third distribution load accommodation plan based on the preset change data, the arrangement of the elements of the third distribution load accommodation plan is changed, and the connection information of each preset element is changed. A fourth distribution connecting a plurality of elements of the distribution load accommodation plan, in which the arrangement of the elements is changed based on the above, and a fourth arrangement of a plurality of elements.
Of the distribution load accommodation plan, the value of the objective function representing the item to be minimized or maximized with respect to the fourth distribution load accommodation plan is obtained, and the value of the obtained objective function and the third A distribution load accommodation plan planning method, characterized in that a third distribution load accommodation plan is determined from the value of the objective function of the distribution load accommodation plan. 8. The distribution load accommodation plan planning method according to claim 7, wherein the difference between the value of the objective function of the fourth distribution load accommodation plan and the value of the objective function of the third distribution load accommodation plan. A method for planning a distribution load accommodation plan, wherein the fourth plan is a third distribution load accommodation plan when the absolute value of the value is equal to or less than a preset value. 9. The distribution load accommodation plan planning method according to claim 8, wherein the preset value is the same number as the number of times of change in the change data. Planning method. 10. An initial plan generation step of generating a predetermined number of parent distribution load accommodation plans expressed by a plurality of first-generation elements, and an objective function value corresponding to the plurality of parent distribution load accommodation plans. A calculation step of calculating, a plan rearrangement step of arranging the plurality of parent distribution load accommodation plan plans in descending order or ascending order of the objective function value, and the plurality of parent distribution load accommodation plan plans are given to the objective function. When the selection number of each parent distribution load accommodation plan is assigned according to the value, the parent distribution load accommodation plan is set as a number indicated by a predetermined number of constants that are predetermined for each generation repetition. A predetermined number of parent distribution load accommodation plan selection steps, and any two elements are replaced with the selected parent distribution load accommodation plan draft to generate a child distribution load accommodation plan. A child distribution load accommodation plan generating step to form, the parent distribution load accommodation plan and the generated child distribution load accommodation plan, the calculation step and the distribution load accommodation plan rearrangement step, the rearranged, A generation alternation step of selecting a predetermined number of distribution load accommodation plans as a new parent distribution load accommodation plan according to the rearrangement order of the distribution load accommodation plan, and from the second generation to a predetermined number of generations, the parent The distribution load accommodation plan selection step, the child distribution load accommodation plan generation step, and the generation alternation step are repeatedly activated, and one distribution load accommodation plan is selected from a plurality of distribution load accommodation plans selected based on the value of the objective function. Optimal distribution load accommodation plan for selecting a plan as a distribution load accommodation plan draft. Flexible planning method. 11. The distribution load accommodation plan drafting method according to claim 10, wherein a new distribution load accommodation plan is newly generated from the distribution load accommodation plan by a predetermined operation, and the objective function values of the two distribution load accommodation plans are generated. The difference between the objective function values is smaller than the value of the predetermined variable, the parent distribution load accommodation plan is set to the new distribution. A distribution load accommodation plan planning method, characterized in that an optimal distribution load accommodation plan is replaced with the load accommodation plan. 12. An input device for inputting information necessary for solution of a plan, a plan making device for generating a plan made up of a plurality of elements, and an output device for outputting a plan made by the plan making device. In the planning system, the planning device sets a storage device for storing the information and the plan, which are input from the input device, and the plan as an array of elements, and sets the setting based on predetermined array change data. Before changing the array, the array of the elements of the proposed plan is changed, a new plan is generated from the element information determined in advance for the changed plan, and the objective function value of the plan is obtained. At least one or more new plans are generated based on the evaluation result of the objective function value of the plan and the result of the objective function value of the new plan and a predetermined evaluation variable. And a processor allocation unit that allocates a processor for processing based on the objective function value of the plan generated by the at least one or more processors. 13. An input device for inputting information necessary for solving a distribution load accommodation plan, a distribution load accommodation plan drafting device for generating a distribution load accommodation plan draft consisting of a plurality of elements, and the distribution load accommodation plan drafting device. In the distribution load accommodation plan planning system including an output device that outputs the distribution load accommodation plan, the distribution load accommodation plan planning device stores information input from the input device and the distribution load accommodation plan. A new distribution load accommodation plan is generated by a predetermined operation from the storage device and the parent distribution load accommodation plan, the difference between the objective function values of the two distribution load accommodation plans is obtained, and the difference between the objective function values and the generation A new distribution load accommodation plan is selected based on the value of the variable determined for each of the at least one processor and the at least one processor. A distribution load accommodation plan planning system, comprising a processor allocation unit for allocating a processor for processing based on an objective function value of the distribution load accommodation plan. 14. The distribution load accommodation system according to claim 13, wherein the distribution load accommodation plan planning device stores a plurality of parent distribution load accommodation plans for a given distribution load accommodation plan. And, a new distribution load accommodation plan is generated by a predetermined operation from the parent distribution load accommodation plan,
The difference between the objective function values of these two distribution load accommodation plans is calculated, the values of the variables previously determined for each generation are compared, and when the difference between the objective function values is smaller than the value of the previously determined variable, A plurality of cell processors that simultaneously perform the operation of replacing the parent power distribution load accommodation plan with the new power distribution load accommodation plan as an optimum power distribution load accommodation plan for the plurality of parent power distribution load accommodation plans, and the plurality of cell processors. A cell processor massively parallel control unit that allocates a next-generation processing processor corresponding to the order in which the objective function value of the optimal distribution load accommodation plan generated by the cell processor is close to a predetermined value, and outputs a planning result or a variable value to the input device. A distribution load accommodation system using a distribution load accommodation planning device, comprising an output device control unit. 15. The distribution accommodation system according to claim 14, wherein the distribution load accommodation planning device changes an objective function difference when a planning given to each of the cell processors is changed by a predetermined operation. If the distribution load accommodation plan after the change is excellent is calculated, the distribution load accommodation plan is changed to make the optimal distribution load accommodation plan candidate, and if not, the given distribution load accommodation plan is optimized. According to the optimization process as the load accommodation plan candidate and the objective function values of the plurality of optimum distribution load accommodation plan candidates defined above, the distribution load accommodation plan to be processed by each cell processor of the next (next generation) The distribution load accommodation system using the distribution load accommodation planning device is characterized in that it is configured to repeat the number of generations of cell processor allocation control processing for allocating Beam. 16. The distribution load accommodation system according to claim 14, wherein the generation and evaluation of the load accommodation pattern when a given plan is changed by a predetermined operation is at least a given number of sections to be accommodated. For the control step repeating the processing steps of minutes or less, and the number of concealed sections adjacent to each concealed section for each repetition, the upstream section of the corresponding concealed section has already been load accommodating, and from the reserve power amount of the system. If it is large, repeat the step of connecting to the upstream section, and if unable to connect to the upstream section, copy the corresponding conserved section at the end of the given plan and extend the plan. A distribution load accommodation system characterized by having. 17. The distribution load accommodation plan planning system according to claim 13, 14 or 15, wherein said output device sequentially displays a distribution system diagram.