WO2020250423A1

WO2020250423A1 - Storage control system, storage control method and a non-transitory computer readable medium

Info

Publication number: WO2020250423A1
Application number: PCT/JP2019/023667
Authority: WO
Inventors: Alexander Viehweider
Original assignee: Nec Corporation
Priority date: 2019-06-14
Filing date: 2019-06-14
Publication date: 2020-12-17

Abstract

An objective of the invention is to appropriately control storage with a simple uncertainty description of the storage. A system control unit (5) provides an overall system command (OSC). A load measurement unit (4) measures a current local load of the local storage unit in each storage site (S1-S3). A load prediction unit (3) predicts a future local load of the local storage unit in each storage site (S1-S3). An uncertainty description unit (1) generates uncertainty description (UD) of the storage sites (S1-S3) based on environment variables, the overall system command (OSC), the load measurement (LM), the load prediction (LP), and information indicating operational states (OSF) of the storage sites (S1-S3). An optimization unit (2) configured to optimize performance of the storage sites (S1-S3) with reference to the uncertainty description (UD) and to output the information indicating the operational states (OSF) to the uncertainty description unit (1).

Description

STORAGE CONTROL SYSTEM, STORAGE CONTROL METHOD AND A NON-TRANSITORY COMPUTER READABLE MEDIUM

　　The present invention relates to a storage control system, a storage control method and a non-transitory computer readable medium.

　　Methods for operation of distributed storage control exist (e.g. NPTLs 2 and 4). They assume the exact knowledge of load (and/or generation) prediction or do not use, in an optimal way, the knowledge about the load or generation prediction uncertainty. Methods for dealing with uncertainty sets of different shapes exist (e.g. NPTL 3). However, their complexity does not allow for use in large scale problems since they lead to a kind of problem for which the optimization is too expensive to compute.

　　State of the art solutions consider a fixed, flexible or adjustable uncertainty set which does not consider the peculiarities of the single site or the single time instant together with the limits on the common occurrence or uncertainty sets or methods that are too complex for large scale systems (>1000 units).

　　Besides that, the impact of an uncertainty on the quality of the system service provided by the distributed storage operation system does not depend only on the size of the deviation from the prediction only, but also from operational aspect. For example, if uncertainty of some loads is high while the overall expected storage system command is low, deviations of the prediction can be tolerated.

　　The technique described in NPTL 3 can be considered as one of the most advanced techniques at the moment. It allows for a very flexible and general uncertainty set for optimization. The optimization algorithm itself (and not only the parameters of the optimization algorithm) adapts to the peculiarities (shape) of the uncertainty sets. A successful application example has been reported in the chemical process domain. However, due to the small number of considered parameters and the curse of dimensionality, which this type of algorithms are especially prone to, the application range is limited and it is not expected to overcome this problem soon.

　　An example of a general augmented optimization system is described in PTL 1. In PTL 1, strategic decision making is considered for robust optimization. However, this contains strategies applicable to technical distributed storage system and the uncertainty is described in a different way than the description introduced in PTL 1. The method mentioned to deal with uncertainty is to generate a plurality of uncertainty models by sampling a probability distribution, using scenarios or the uncertainty models are implemented by a neural network.

　　PTL 1: International Patent Publication No. WO 01/55939

Non-Patent Literature

　　NPTL 1: Chen Jian Wang, Feng Liu, "Robust Risk-Constrained Unit Commitment with Large-scale Wind Generation: An Adjustable Uncertainty Set Approach", January 2017, Volume 32, Issue 1, IEEE Transactions on Power Systems, pp.723-733.
　　NPTL 2: Alexandre Velloso, Alexandre Street, David Pozo, Jose M. Arroyo, Noemi G. Cobos, "Scenario-Based Uncertainty Set for Two-Stage Robust Energy and Reserve Scheduling: A Data Driven Approach", June 12, 2018, Cornell University Archive, arXiv:1803_06676v2.
　　NPTL 3: C. Ning, F. You, "A Data-Driven Multistage Adaptive Robust Optimization Framework for Planning and Scheduling under Uncertainty", May 12, 2017, John Wiley and Sons.
　　NPTL 4: Hisato Sakuma, Ryo Hashimoto, Hitoshi Yano, Alexander Viehweider, Koji Kudo, "Novel Demand Response scheme for frequency regulation using consumers' distributed energy storages", ISGT 2014, IEEE.

　　A first problem is the generation of an appropriate uncertainty description by means of available data. A second problem is the most compact and useful description of the load and generation uncertainty affecting a local site and the integration of the uncertainty description into an overall optimization algorithm. Accordingly, it is desirable that the uncertainty affecting a distributed storage operation system has a proper representation with a low number of parameters.

　　The reason for the occurrence of the first problem is that the uncertainty description generation may not fit the problem to be dealt with, maybe not general enough and no method available to calculate. Therefore, a proper architecture is needed to calculate the appropriate uncertainty description. The concrete shape of the uncertainty description (how many numbers and how do they have to be interpreted to be a valid description) is the second problem.

　　The reason for the occurrence of the second problem is that the uncertainty description has to be in such a way that it can be included into an optimization algorithm while guaranteeing good performance. The associated problem is the re-use of an already existing optimization solver for this type of problem.

　　The present invention has been made in view of the above-mentioned problem, and an objective of the present invention is to appropriately control storage with a simple uncertainty description of the storage.

　　An aspect of the present invention is a storage control system including: a plurality of storage sites, each storage site including local storage means; system control means configured to provide an overall system command; load measurement means configured to measure a current local load of the local storage means in each storage site; load prediction means configured to predict a future local load of the local storage means in each storage site; uncertainty description means configured to generate uncertainty description of the storage sites based on environment variables, the overall system command, the load measurement, the load prediction, and information indicating operational states of the storage sites; and optimization means configured to control the storage sites to optimize performance of the storage sites with reference to the uncertainty description and to output the information indicating the operational states of the storage sites to the uncertainty description means.

　　An aspect of the present invention is a storage control method including: measuring a current local load of local storage means in each storage site included in a plurality of storage sites; predicting a future local load of the local storage means in each storage site; generating uncertainty description of the storage sites based on environment variables, an overall system command, the load measurement, the load prediction, and information indicating operational states of the storage sites; and controlling the storage sites to optimize performance of the storage sites with reference to the uncertainty description.

　　An aspect of the present invention is a non-transitory computer readable medium storing a storage control program to cause a computer to execute processes of: measuring a current local load of local storage means in each storage site included in a plurality of storage sites; predicting a future local load of the local storage means in each storage site; generating uncertainty description of the storage sites based on environment variables, an overall system command, the load measurement, the load prediction, and information indicating operational states of the storage sites; and controlling the storage sites to optimize performance of the storage sites with reference to the uncertainty description.

　　According to the present invention, it is possible to appropriately control storage with a simple uncertainty description of the storage.

Fig. 1 is a block diagram schematically illustrating an overall structure of a distributed storage control system according to a first example embodiment; Fig. 2 is a block diagram schematically illustrating a configuration of the uncertainty description unit 1 according to the first example embodiment; Fig. 3 illustrates prediction errors of load prediction for each storage site; Fig. 4 illustrates a table of the errors of the load prediction for each storage site at each time step; Fig. 5 is a concrete example of the table of the errors of the load prediction for each storage site at each time step; Fig. 6 is a block diagram schematically illustrating an overall structure of a distributed battery control system according to a second example embodiment; Fig. 7 illustrates a configuration of a battery site group according to the second example embodiment; Fig. 8 is a block diagram schematically illustrating an overall structure of an water tank control system according to a third example embodiment; Fig. 9 illustrates a configuration of a water tank site group according to the third example embodiment; Fig. 10 schematically illustrates an example configuration of a system configuration including a server and a computer; and Fig. 11 schematically illustrates an example configuration of the computer.

　　Example embodiments of the present invention will be described below with reference to the drawings. In the drawings, the same elements are denoted by the same reference numerals, and thus a repeated description is omitted as needed.

First example embodiment
　　A distributed storage control system according to a first example embodiment will be described. Fig. 1 is a block diagram schematically illustrating an overall structure of the distributed storage control system 100 according to the first example embodiment.

　　The distributed storage control system 100 includes an uncertainty description unit 1, a optimization unit 2, a load prediction unit 3, a load measurement unit 4, a system control unit 5, and a storage site group 10. The storage site group 10 includes a plurality of distributed storage sites. The distributed storage control system 100 is configured to optimally control each storage site included in a storage site group 10.

　　In the present example embodiment, each storage site in the storage site group 10 includes a local storage unit, a local generation (a local production, a local input, or a local supply), and a local load (a local demand, or a local output). The local storage unit may be a battery or the like. The local generation may be a local generator such as a photovoltaic cell. The local load may be power consumption in general household or a factory, for example. Each storage site is controlled to fulfill a pre-specified (predetermined) service or task by the optimization unit 2. This fulfillment needs to be guaranteed to a high degree by the optimization unit 2.

　　The storage sites S1 to S3 may receive control commands C1 to C3 that control operations of the storage sites S1 to S3 and output operation states OS1 to OS3 to the optimization unit 2, respectively.

　　The uncertainty description unit 1 receives environment variables from the outside environment. For example, the environment variables are detected by sensing devices and the detected environment variables are input to the uncertainty description unit 1. The environment variables include, for example, outside temperature, day of the week, time of the day, and other variables affecting directly or indirectly all or a part of an overall storage system condition, future storage service command and other relevant variables of the storage system.

　　The uncertainty description unit 1 also receives other information from the optimization unit 2, the load prediction unit 3, and the load measurement unit 4, and the system control unit 5. The system control unit 5 outputs an overall system command OSC to the uncertainty description unit 1. The optimization unit 2 outputs operational status and feedback OSF to the uncertainty description unit 1. The detail of the operational status and feedback OSF will be described below.

　　The load prediction unit 3 predicts a future local net load that is a difference between the local load and the local generation (i.e. the local load - the local generation) for each storage site. Various general methods such as machine learning can be used for predicting the local net load and a load prediction error for each storage site. Note that the method of the prediction is not limited to the specific method. The load prediction unit 3 outputs the load prediction LP including the local net loads to be expected in the future and prediction error confidence intervals to the uncertainty description unit 1.

　　The load measurement unit 4 measures the current local net load affecting each storage site and outputs the measurement result as the load measurements LM to the uncertainty description unit 1. The load measurements LM are necessary to evaluate the quality of the prediction values and the quality of the prediction error bounds over time.

　　The uncertainty description unit 1 generates a compact uncertainty description UD by computing the received information described above and outputs the generated uncertainty description UD to the optimization unit 2.

　　The optimization unit 2 optimizes performance of each storage site in the storage site group 10 based on the received information. The system control unit 5 also outputs the overall system command OSC to the optimization unit 2. Further, the load prediction unit 3 outputs the load prediction LP to the optimization unit 2. In accordance with the overall system command OSC, the optimization unit 2 optimizes the performance of each storage site in the storage site group 10 based on the load prediction LP and the uncertainty description UD. Accordingly, the performance of each storage site can match the future load indicated by the load prediction LP while taking into account the uncertainty description UD.

　　As described above, the optimization unit 2 outputs the operational status and feedback OSF including the information indicating the optimization of the performance of each storage site to the uncertainty description unit 1. For example, the operational status and feedback OSF may be information, for example, indicating how many storage units are in high SoC (State of Charge) mode or low SoC mode, respectively, when each storage site includes a battery. Therefore, the uncertainty description unit 1 can reflect the state of each storage site that affects the occurrence of the uncertainty on the generated uncertainty description UD.

　　Here, the uncertainty description unit 1 will be described in detail. Fig. 2 is a block diagram schematically illustrating a configuration of the uncertainty description unit 1 according to the first example embodiment. The uncertainty description unit 1 includes a command prediction unit 1A, a state vector generator 1B, a vector compression unit 1C, an action decision unit (also referred to as ADU) 1D, an uncertainty set generation unit (also referred to as USCU) 1E, and a load prediction uncertainty common occurrences evaluation and learning unit (also referred to as LPUCOELU) 1F.

　　At the current time instant, the command prediction unit 1A receives the overall system command OSC and predicts how the overall system command OSC is expected to change in future in time series. Thus, the command prediction unit 1A may predict what degree of uncertainty needs to be considered in time series. The command prediction unit 1A outputs the prediction result PR to the state vector generator 1B.

　　The state vector generator 1B also receives the overall system command OSC, the environment variables, and the operational status and feedback OSF. Then, the state vector generator 1B generates a full state vector V from the prediction result PR output from the command prediction unit 1A and the received information (the overall system command OSC, the environment variables, and the operational status and feedback OSF). For example, the full state vector V may be interpreted as a summary of the status of the overall storage system (the overall storage sites) without using the single SoC (State of Charge) information of each storage site. The generated full state vector V is output to the vector compression unit 1C.

　　The vector compression unit 1C includes all necessary logic to reduce the full state vector V to its essentials. This compression may be achieved by using techniques such as auto-encoder networks or other neural network inspired technologies. By the compression, redundancy is eliminated in the full state vector V. The compressed full state vector CV is output to the ADU 1D and the LPUCOELU 1F.

　　The ADU 1D computes the compressed full state vector CV to determine an appropriate action for changing the uncertainty description based on a rough estimation and information condensation of the overall storage system condition. For example, the ADU 1 may determine the appropriate action DA from three types of actions described below, for example by pattern matching.

Type 1: Relax Uncertainty Description
　　The overall storage system operation does not consider so much occurring uncertainty since their occurrence conditions are relaxed, a lower number of possible worst cases are considered.

Type 2: Keep Uncertainty Description
　　The uncertainty description is kept as it is computed by another unit.

Type 3: Turn More Restrictive or Conservative
　　The uncertainty description due to the internal condition of the overall storage system, a higher number of worst cases of prediction errors are considered.

　　The ADU 1D outputs the determined action DA to the USCU 1E. Note that the actions are not limited to the three types of actions described below, and the ADU 1 may determine the appropriate action DA from four or more actions including the all or a part of the above three actions or four or more actions other than the above three actions.

　　In parallel to the computation of the ADU 1D, the LPUCOELU 1F computes a basic uncertainty description BUD based on the operational status and feedback OSF by use of statistical methods, and outputs the computed basic uncertainty description BUD to the USCU 1E.

　　The USCU 1E modifies the computed the basic uncertainty description BUD by the determined action DA (e.g. any one of the three types of the actions described above), and outputs the modified uncertainty description UD to the optimization unit 2. For example, the USCU 1E may perform weighting of the parameters in the basic uncertainty description BUD to modify it.

　　The load prediction of the load prediction unit 3 will be described in detail. Fig. 3 illustrates prediction errors of the load prediction for each storage site. In Fig. 3, and in the following description and drawings, N denotes the number of storage sites, where N is an integer equal to or more than two. At a certain time instant for each storage site, a bar indicating the load uncertainty extends along the vertical direction between a higher bound and a lower bound. Thus, each bar indicates the range of the uncertainty of the load prediction for each storage site at one time instant.

　　Fig. 4 illustrates a table of the errors of the load prediction for each storage site at each time step. In Fig. 4, and in the following description and drawings, M denotes the number of the considered time steps, where M is an integer equal to or more than two. In a rectangle corresponding to i-th time step and j-th storage site, a higher bound of the uncertainty of the load prediction +Δz_ij and a lower bound of the uncertainty of the load prediction -Δz_ij for the j-th storage site at the i-th time step, where i is an integer from 1 to M and j is an integer from 1 to N.

　　Fig. 5 is a concrete example of the table of the errors of the load prediction for each storage site at each time step. Each time step can be typically 15 minutes or less in the case of a battery storage system. A characteristic time range is a range from the first time step to the M-th time step. For example, the characteristic time range can be one day with sampling every 15 minutes so that one day includes 96 time steps (i.e. M=96).

　　A rectangle (i, j) is a case corresponding to the i-th time step and the j-th storage site. Further, the hatched rectangle denotes the occurrence of the worst-case uncertainty. Here such case is merely referred to as "the worst case".

　　A number H_{T_i} indicates the highest (or maximal) one of higher numbers of the uncertainty of the net load in the worst cases. A number L_{T_i} indicates the lowest (or minimal) one of lower numbers of the uncertainty of the net load in the worst cases. In Fig. 5, a concrete example in which some worst case occurs is illustrated. In this example, in the first time step, the case (1, 2) and the case (1, N-1) fall into the worst case.

　　In this case, the highest one of the higher number of the uncertainty of the net load in the case (1, 2) and the higher number of the uncertainty of the net load in the case (1, N-1) is adopted as the number H_{T_i.} The lowest one of the lower number of the uncertainty of the net load in the case (1, 2) and the lower number of the uncertainty of the net load in the case (1, N-1) is adopted as the number L_{T_i}.

　　A number H_{S_j} indicates the highest one of higher numbers of the uncertainty of the net load for j-th storage site over the characteristic time range that is from the first time step to the M-th time step. For example, at the second storage site (j=2), the case (1, 2), the case (3, 2), and the case (M, 2) fall into the worst case.

　　In this case, the highest one of the higher numbers of the uncertainty of the net load in the case (1, 2), the higher number of the uncertainty of the net load in the case (3, 2), and the higher number of the uncertainty of the net load in the case (M, 2) is adopted as the number H_{S_j.} The lowest one of the lower numbers of the uncertainty of the net load in the case (1, 2), the lower number of the uncertainty of the net load in the case (3, 2), and the lower number of the uncertainty of the net load in the case (M, 2) is adopted as the number L_{S_j.}

　　Therefore, the uncertainty description unit 1 may devise the uncertainty description UD relying on additional 2N+2M parameters H_{T_1} to H_{T_M}, L_{T_1} to L_{T_M}, H_{S_1} to H_{S_N}, and L_{S_1} to L_{S_N} in order to appropriately describe the occurrence of the uncertainty.

　　According to the present configuration, the uncertainty description unit 1 continuously recalculates and updates the uncertainty description UD. Therefore, it is possible to ensure that the uncertainty description of the net load is kept always as informative (i.e. useful) and to reflect the current status from the uncertainty perspective of the overall distributed storage system.

　　Further, according to the present configuration, it is possible to have and compute the most accurate and the most compact description of the uncertainty. The compactness of the description can achieve the low number of parameters describing the uncertainty while guaranteeing that the major characteristic and quality of the uncertainty is captured by the chosen description.

　　Furthermore, the present configuration allows the use of the above-described uncertainty description in the general optimization system and the method to use the available solver without major system changes. Therefore, it is possible to achieve robustness of the system capable of appropriately controlling the operation of the storage system.

　　Next, operation of the distributed storage control system 100 will be described. Firstly, the distributed storage control system 100 operates with sufficient margin for each storage site in a calibration mode. Next, the generation and updating of the uncertainty description UD as described above. This results in improvement of the uncertainty description UD and the reduction of the margin for each storage site. Finally, the operation of the distributed storage control system 100 is changed from the calibration mode into a regular operation mode.

　　According to the present configuration, only the small number of extra parameters (i.e. the additional 2N+2M parameters) with high expressiveness are derived and used for calculating the uncertainty description. Therefore, it is possible to ensure an appropriate and lean description of the uncertainty to integrate it in the optimization unit.

　　For guaranteeing the overall service that the present system is expected to provide, it is desirable that the storage site has relatively larger capacity in order to correspond to deviation between the predicted local generation and local load. However, from the viewpoint of cost reduction and system simplification, the overcapacity of the storage site is not desirable. Thus, according to the present configuration, the overcapacity of the storage site can be prevented by appropriately optimizing the performance of the storage site while guaranteeing the common service.

Second example embodiment
　　A distributed battery control system according to a specific example embodiment will be described. Fig. 6 is a block diagram schematically illustrating an overall structure of the distributed battery control system 200 according to the second example embodiment. The distributed battery control system 200 according to the second example embodiment is configured as a modified example of the distributed storage control system 100 according to the first example embodiment.

　　The distributed battery control system 200 has a configuration in which the storage site group 10 in the distributed storage control system 100 is replaced with a battery site group 20. The battery site group 20 includes a plurality of the battery sites. For simplicity, Fig. 6 illustrates the battery site group 20 including three battery sites BS1 to BS3.

　　Fig. 7 illustrates the configuration of the battery site group 20 according to the second example embodiment. The battery sites BS1 to BS3 include batteries B1 to B3, respectively. The batteries B1 to B3 are connected to a power grid 22. Each of the batteries B1 to B3 receives power supply from the local generation LG and supplies power to the local load LL. The batteries B1 to B3 are also connected to a demand site 21 and provide the demand site 21 with the power supplies P₁ to P₃ through the same grid or separated grid, respectively.

　　The batteries B1 to B3 may receive the control commands C1 to C3 that control the operations of the batteries B1 to B3 and output the operation states OS1 to OS3 to the optimization unit 2, respectively.

　　Since the other configurations of the distributed battery control system 200 are similar to those of the distributed storage control system 100, the description of those will be omitted.

　　In the present configuration, it is necessary to guarantee the service provided to the demand site 21 and to compensate deviation from predicted values by disposing additional spare capacity to the battery of the each battery site. The local batteries or the local sites are connected to the power grid and fulfills local targets while guaranteeing the provision of the service by using the combination of all batteries.

　　The local site has a local controller that gets parameters from the higher level distributed battery operation control systems, for example, from the system control unit 5. The optimization of these parameters is assisted by the optimization unit 2. Thus, the overall optimization takes into account the uncertainties of local load and generation prediction by special optimization techniques, e.g. augmentation such as an enhanced LP (linear programming) problem (by using standard parameters a matrix A vectors b and c) with quite larger describing matrices and vectors A_enh, b_enh, and c_enh. The augmentation can be achieved as an enhanced optimized program such a LP program or other kind of program for processing the enhanced problem as described above. In this embodiment, the LP program is automatically translated into an augmented LP program by using relaxation, duality, …and the information of the minimal and maximal uncertainty as described with reference to Fig. 5 to the extent possible.

It means that the problem:
min c^Tx
Ax <= b
may be transformed to:
min [c^T 0^T] x_enh
A_enh x_enh <= b_enh.
with x^T _enh =[ x^Tx_add ^T]
or similar descriptions.

Third example embodiment
　　A water tank control system according to a third example embodiment will be described. Fig. 8 is a block diagram schematically illustrating an overall structure of the water tank control system 300 according to the third example embodiment. The water tank control system 300 according to the third example embodiment is configured as a modified example of the distributed battery control system 200 according to the second example embodiment.

　　The water tank control system 300 has a configuration in which the battery site group 20 in the distributed battery control system 200 is replaced with the water tank group 30 including a plurality of water tank sites. Fig. 9 illustrates a configuration of the water tank site group 30 according to the third example embodiment. For simplicity, Figs. 8 and 9 illustrate the water tank site group 30 including six water tank sites WS1 to WS6. The water tank sites WS1 to W6 include water tanks W1 to W6, respectively. The water tanks W1 to W6 are connected to main flow line 32. Each of the water tank sites WS1 to WS6 receives water supply from a local supply LS that corresponds to the local generation LG and supplies water to a local demand LD that corresponds to the local load LL.

　　Thus, the water tank sites WS1 to WS6 can satisfy overall demand by supplying water and receive an overall water supply through the main flow line 32 while corresponding to the local demand LD and the local supply LS.

　　The water tank sites WS1 to WS6 may receive the control commands C1 to C6 that control the operations of the water tank sites W1 to W6 and output the operation states OS1 to OS6 to the optimization unit 2, respectively.

　　The water tank control system 300 can be operated in the similar way of the distributed battery control system 200 by replacing the current, the voltage, and the charge of the battery or storage with flow, pressure and an amount of the water, respectively.

　　The uncertainty description unit 1 accesses the environment variables (humidity, pressure, and others …). Water use is predicted using the time-series method to understand what the future water needs are. Operational status could be the overall level of all tanks, and the number of tanks close to the upper limit and the number of tanks close to the lower limit. Thus, each water tank may serve the water to stabilize the local demand and the supply of water and all tanks may supply a certain predefined amount of water in cooperation.

Other example embodiments
　　Note that the present invention is not limited to the above example embodiments and can be modified as appropriate without departing from the scope of the invention. For example, in the above example embodiments, the present invention is described as a hardware configuration, but the operations of the uncertainty description unit 1, the optimization unit 2, the load prediction unit 3, the load measurement unit 4, and the system control unit 5 by causing a CPU (Central Processing Unit) to execute a computer program. The program can be stored and provided to a computer using any type of non-transitory computer readable media. Non-transitory computer readable media include any type of tangible storage media. Examples of non-transitory computer readable media include magnetic storage media (such as floppy disks, magnetic tapes, hard disk drives, etc.), optical magnetic storage media (e.g. magneto-optical disks), CD-ROM (Read Only Memory), CD-R, CD-R/W, and semiconductor memories (such as mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (Random Access Memory), etc.). The program may be provided to a computer using any type of transitory computer readable media. Examples of transitory computer readable media include electric signals, optical signals, and electromagnetic waves. Transitory computer readable media can provide the program to a computer via a wired communication line, such as electric wires and optical fibers, or a wireless communication line.

　　An example in which the distributed storage control system 100 is configured by a server and a computer will be described. Fig. 10 schematically illustrates an example configuration of a system configuration 1000 including a server 1001 and a computer 1002. In this case, the server 1001 executes a program to perform the operation of the system control unit 5. The computer 1002 executes a program to perform the operations of the uncertainty description unit 1, the optimization unit 2, the load prediction unit 3, and the load measurement unit 4.

　　Fig. 11 schematically illustrates an example configuration of the computer 1002. The computer 1002 includes a CPU 1002A, a memory 1002B, an input/output interface (I/O) 1002C and a bus 1002D. The CPU 1002A, the memory 1002B and the input/output interface (I/O) 1002C can communicate each other through the bus 1002D. The CPU 1002A achieves functions of the uncertainty description unit 1, the optimization unit 2, the load prediction unit 3, and the load measurement unit 4 by executing the program. The memory 1002B may store the program. The computer 1002 may communicate with the server 1001 and the storage site group 10 through the I/O 1002C.

　　The server 1001 may also have a similar configuration to the computer 1002. In this case, the CPU 1002A achieves functions of the system control unit 5 by executing the program.

　　Further, the single computer having the similar configuration to the computer 1002 may function as the uncertainty description unit 1, the optimization unit 2, the load prediction unit 3, the load measurement unit 4, and the system control unit 5 by executing the program.

　　While the present invention has been described above with reference to example embodiments, the present invention is not limited to the above example embodiments. The configuration and details of the present invention can be modified in various ways which can be understood by those skilled in the art within the scope of the invention.

　　While the present invention has been described above with reference to exemplary embodiments, the present invention is not limited to the above exemplary embodiments. The configuration and details of the present invention can be modified in various ways which can be understood by those skilled in the art within the scope of the invention.

1　　UNCERTAINTY DESCRIPTION UNIT
1A　　COMMAND PREDICTION UNIT
1B　　STATE VECTOR GENERATOR
1C　　VECTOR COMPRESSION UNIT
1D　　ACTION DECISION UNIT (ADU)
1E　　UNCERTAINTY SET GENERATION UNIT (USCU) 1E
1F　　LOAD PREDICTION UNCERTAINTY COMMON OCCURRENCES EVALUATION AND LEARNING UNIT (LPUCOELU)
2　　OPTIMIZATION UNIT
3　　LOAD PREDICTION UNIT
4　　LOAD MEASUREMENT UNIT
5　　SYSTEM CONTROL UNIT
10　　STORAGE SITE GROUP
20　　BATTERY SITE GROUP
21　　DEMAND SITE
22　　POWER GRID
30　　WATER TANK GROUP
32　　MAIN FLOW LINE
100　　 DISTRIBUTED STORAGE CONTROL SYSTEM
200　　DISTRIBUTED BATTERY CONTROL SYSTEM
300　　WATER TANK CONTROL SYSTEM
1000　　SYSTEM CONFIGURATION
1001　　SERVER
1002　　COMPUTER
1002A　　CPU
1002B　　MEMORY
1002C　　INPUT/OUTPUT INTERFACE (I/O)
1002D　　BUS
B1 TO B3　　BATTERIES
BS1 TO BS3　　BATTERY SITES
BUD　　BASIC UNCERTAINTY DESCRIPTION
C1 TO C6　　CONTROL COMMANDS
CV　　COMPRESSED FULL STATE VECTOR
DA　　DETERMINED ACTION
LD　　LOCAL DEMAND
LG　　LOCAL GENERATION
LL　　LOCAL LOAD
LM 　　LOAD MEASUREMENTS
LP　　LOAD PREDICTION
LS　　LOCAL SUPPLY
OS1 TO OS6　　OPERATION STATES
OSC　　OVERALL SYSTEM COMMAND
OSF　　OPERATIONAL STATUS AND FEEDBACK
PR　　PREDICTION RESULT
S1 TO SN　　STORAGE SITES
TS1 TO TSM　　TIME STEPS
UD　　UNCERTAINTY DESCRIPTION UD
V　　FULL STATE VECTOR
W1 TO W6　　WATER TANKS
WS1 TO WS6　　WATER TANK SITES

Claims

　　A storage control system comprising:
　　a plurality of storage sites, each storage site comprising local storage means;
　　system control means configured to provide an overall system command;
　　load measurement means configured to measure a current local load of the local storage means in each storage site;
　　load prediction means configured to predict a future local load of the local storage means in each storage site;
　　uncertainty description means configured to generate uncertainty description of the storage sites based on environment variables, the overall system command, the load measurement, the load prediction, and information indicating operational states of the storage sites; and
　　optimization means configured to control the storage sites to optimize performance of the storage sites with reference to the uncertainty description and to output the information indicating the operational states of the storage sites to the uncertainty description means.
　　The storage control system according to Claim 1, wherein
　　the uncertainty description means performs steps of:
　　determining the minimal and maximal uncertainty for each storage site in a characteristic time range including a plurality of time steps, and
　　determining the minimal and maximal uncertainty for each time step in the all storage sites.
　　The storage control system according to Claim 2, wherein
　　the uncertainty description means comprises:
　　command prediction means configured to receive the overall system command, predict how the overall system command is expected to change in future, and output the prediction result;
　　state vector generation means configured to generate a full state vector from the prediction result, the overall system command, the environment variables, and the information output from the optimization means;
　　vector compression means configured to compress the full state vector to its essentials;
　　action decision means configured to compute the compressed full state vector to determine an appropriate action for changing the uncertainty description based on an estimation and information condensation of an overall system condition;
　　load prediction uncertainty common occurrences evaluation and learning means configured to compute a basic uncertainty description based on the information output from the optimization means; and
　　uncertainty set generation means configured to modify the basic uncertainty description computed by the determined action and output the modified uncertainty description to the optimization means.
　　The storage control system according to any one of Claims 1 to 3, wherein
　　each storage site is configured as a battery site including a battery,
　　each battery supplies power to the local load and receives power from a local generation, and
　　the battery sites supply power to an external demand site through one grid or separated grids.
　　The storage control system according to any one of Claims 1 to 3, wherein
　　each storage site is configured as a water tank site including a water tank,
　　each water tank supplies water to a local demand and receives water from a local supply, and
　　the water tank sites supply water to overall demand and receive water through from overall supply through a main flow line.
　　The storage control system according to Claim 2, wherein
　　a LP program of the storage control system is automatically translated into an augmented LP or other kind of program by using relaxation, duality properties, and the information of the minimal and maximal uncertainty to the extent possible.
　　A storage control method comprising:
　　measuring a current local load of local storage means in each storage site included in a plurality of storage sites;
　　predicting a future local load of the local storage means in each storage site;
　　generating uncertainty description of the storage sites based on environment variables, an overall system command, the load measurement, the load prediction, and information indicating operational states of the storage sites; and
　　controlling the storage sites to optimize performance of the storage sites with reference to the uncertainty description.
　　A non-transitory computer readable medium storing a storage control program to cause a computer to execute processes of:
　　measuring a current local load of local storage means in each storage site included in a plurality of storage sites;
　　predicting a future local load of the local storage means in each storage site;
　　generating uncertainty description of the storage sites based on environment variables, an overall system command, the load measurement, the load prediction, and information indicating operational states of the storage sites; and
　　controlling the storage sites to optimize performance of the storage sites with reference to the uncertainty description.