US 10,338,542 B2Granted

Method for computer-aided installation control optimization using a simulation module

AssigneeSiemens Aktiengesellschaft

InventorsSiegmund Düll, Alexander Hentschel, Jatinder P. Singh, Volkmar Sterzing, Steffen Udluft

Life of this patent

Estimating the term…

Loading patent…

Abstract

An interactive assistance system and method for computer-aided control optimization for a technical system is provided. For example, a gas or wind turbine, in particular for optimizing the action sequence or the control variables of the plant (e.g. gas supply, compression), wherein an input terminal is provided for reading at least one status parameter providing a first system status of the technical system, and at least one setting parameter for adapting a reward function. A simulation module having a pre-trained neuronal network, simulating the plant, serves to simulate an action sequence on the technical system, starting from the first system status and to the prediction of the resulting statuses of the technical system.

Claims13

Dependent claims

An interactive assistance system for computer-aided control optimization for a technical system, having

a) an input terminal for reading in at least one state parameter, which indicates a first system state of the technical system, and at least one setting parameter for setting up a reward function suitable for training a neural network, wherein said at least one setting parameter is a weighted vector for said at least one state parameter of the technical system,
b) a simulation module for simulating performance of an action sequence on the technical system starting from the first system state and for predicting resulting successor states of the technical system, wherein the simulation module includes a neural network for simulation of the technical system,
c) an optimization module, coupled to the input terminal and to the simulation module, and configured to:
i) set up the reward function on the basis of the setting parameter (W), said reward function is a summation of a product of each state parameter and each respective setting parameters for said each state parameter,
ii) generate a multiplicity of action sequences for the first system state (S0),
iii) transmit the action sequences to the simulation module and to receive the resulting successor states,
iv) determine rewards which can be expected for the resulting successor states by means of the reward function which has been set up, and
v) determine a reward-optimizing action sequence as a function of said first system state and said at least one setting parameter, where different selections of said first system state and said at least one setting parameter thereby allowing said reward-optimizing action sequence to be independently optimized, and
d) an output terminal for outputting a system state resulting from the reward-optimizing action sequence,
wherein said reward-optimizing action sequence and said multiplicity of state sequences are transmitted from the optimization module to the output terminal, and
wherein the simulation of the technical system by the simulation module is independent of the reward function.

8 more elements

A method for interactive computer-aided control optimization for a technical system,

a) an input terminal reading in at least one state parameter, which indicates a first system state of the technical system, and at least one setting parameter for setting up a reward function suitable for training a neural network, where said at least one setting parameter is a weighted vector for said at least one state parameter of the technical system,
b) the state parameter and the setting parameter being transmitted to an optimization module which sets up the reward function on the basis of the setting parameter, said reward function is a summation of a product of each state parameter and each respective setting parameters for said each state parameter,
c) the optimization module generating a multiplicity of action sequences for the first system state and transmitting them to a simulation module,
d) the simulation module respectively simulating performance of the transmitted action sequences on the technical system starting from the first system state (S0) and predicting resulting successor states of the technical system, wherein the simulation module uses a neural network for simulation of the technical system,
e) the successor states being transmitted to the optimization module,
f) the optimization module determining rewards which can be expected for the successor states by means of the reward function which has been set up and determining a reward-optimizing action sequence as a function of said first system state and said at least one setting parameter, wherein different selections of said first system state and said at least one setting parameter thereby allowing said reward-optimizing action sequence to be independently optimized, and
g) a system state resulting from the reward-optimizing action sequence being output by an output terminal,
wherein said reward-optimizing action sequence and said multiplicity of state sequences are transmitted from the optimization module to the output terminal, and
wherein the simulation of the technical system by the simulation module is independent of the reward function.

6 more elements

Dependencies are inferred from the claim text.

Facts

Application: 15/321,307Filed 5 May 2015
Priority: DE 10 2014 212 7471 Jul 2014
Published: US 2017/0160706 A18 Jun 2017
Granted: 2 Jul 2019
Term ends: Not estimatedThe source records don’t support an estimate.

Classification17 codes

G05B 13/042
Involving the use of models or simulators: in which a parameter or coefficient is automatically adjusted to optimise the performance
G05B 13/027
The criterion being a learning criterion: using neural networks only
G05B 13/048
Involving the use of models or simulators: using a predictor
G06N 3/006
Artificial life, i.e. computing arrangements simulating life: based on simulated virtual individual or collective life forms, e.g. social…