Method and system for online decision making of generator start-up

Patent metadata

Field	Value
Publication number	US-11159018-B2
Application number	US-201816348190-A
Country	US
Kind code	B2
Filing date	Sep 29, 2018
Priority date	Mar 29, 2018
Publication date	Oct 26, 2021
Grant date	Oct 26, 2021

Abstract

A method and system for online decision making of generator start-up, determining the lines to be restored step by step based on the real-time power system data to guide the initial stage. The black-start unit is determined, and the units waiting to be restored are selected from all units. The training set including labeled samples to establish a value network. During generator start-up, the blackout area is judged, the state of power system is identified, the availability of equipment is judged, and the characteristics of generators are collected. The total generation capability is used as search objective for Monte Carlo tree search. Based on the value network, Monte Carlo is applied to search the next line to be restored during generator start-up. Parallel computation adopted to check the constraints for the restoration of alternative lines. The results of Monte Carlo are summarized to determine the line to be restored next.

First claim

The invention claimed is:

1. An online dynamic decision-making method of generator start-up, the method comprising steps of:
(1) selecting a black start generator unit and additional generator units that need to be started from all generator units;
(2) generating a labeled training data set that includes different potential states of generator start-up, and applying a deep learning algorithm based on the training data set to establish a trained deep neural network;
(3) obtaining real-time data of generator start-up including a blackout area, a state of a power system, an availability of equipment in the power system, and characteristics of generator units waiting to be restored of all of the generator units;
(4) searching and evaluating all alternative power lines to be restored in a next step using a Monte Carlo tree search algorithm and the trained deep neural network by checking voltage, frequency, and power flow variations caused by restoring the different alternative power lines;
(5) summarizing results of the Monte Carlo tree search algorithm to determine the corresponding transmission line to be restored of all alternative power lines; and
(6) restoring the determined transmission line based on the summarized results of the Monte Carlo tree search algorithm.

2. The method of claim 1 , wherein a hydro-power generator, a pumped-storage power generator, or a gas turbine generator is selected as the black-start generator unit based on a current state of the power system.

3. The method of claim 1 , wherein the generator units that need to be started are selected from all of the generator units based on:
selecting the generator units having a capacity between 300 MW and 600 MW;
selecting the generator units corresponding to large-scale capacity power plants; and
selecting the generator units located in an area of critical loads.

4. The method of claim 1 , wherein according to the obtained real-time data, the labeled training data set is generated by:
generating all possible sets of generator status through traversal, wherein N denotes a number of generator units in the power system, C denotes a number of possible generator status, such that the number of sets of generator status needing to be generated is represented by:
C N 1 +C N 2 + . . . +C N N−1 ;

setting a number of sets of line status when generator status is fixed and a number of sets of generator downtime when one of the generator units and the line status are fixed;
generating the line status randomly and verifying topological connectivity of the generated network based on a principle that all restored lines are required to be able to connect the restored generator units and the black-start generator unit, and adjusting the line status in a disconnected network;
calculating expected recovery time of all restored lines and downtime of each generator unit; and
executing a particle swarm optimization algorithm to optimize a maximum total generation capability in a generated power system situation for a corresponding total generation capability as a sample label, the power system situation being represented as a certain sample.

5. The method of claim 1 , wherein the trained deep neural network is implemented to evaluate an optimal total generation capability of the power system based on a power system situation.

6. The method of claim 5 , wherein the deep neural network is formed with three hidden layers based on sparse autoencoder to learn generated samples for the trained deep neural network, and input data to the trained deep neural network are status of the generator units, status of transmission lines, and downtime of generator units, and output data is the optimal total generation capability.

7. The method of claim 1 , wherein the Monte Carlo tree search algorithm is implemented to perform an optimal decision in artificial intelligence games including selection, expansion, simulation and back propagation, the implanting of the Monte Carlo tree search algorithm includes the following steps:
Selection: starting from a root node, after calculating modified upper confidence bound apply to tree (MUCT) value of each node, and selecting a node with a largest MUCT value for further expansion or simulation;
Expansion: eliminating the nodes which represent impossible generator unit start-up situations in reality with a move pruning technique;
Simulation: estimating subsequent optimal total generation capability with the trained deep neural network according to the state of the power system, and improving a selected probability of the alternative transmission lines with a higher total generation capability to guide the simulation process; and
Backpropagation: updating parameters of each node in the Monte Carlo tree search algorithm reversely after the simulation is completed.

8. The method of claim 7 , wherein the MUCT value of a node is determined by a top p % total generation capability of simulation results, visit times to the node, and visit times to a parent node of the node.

9. The method of claim 7 , wherein the move pruning technique includes marking the nodes which have the same parent node with a visited nodes by searching reversely from the node which is waited to be expanded to the root node, and avoiding the marked nodes in a new expansion until a new generator unit is linked to a backbone network.

10. The method of claim 1 , wherein a weighted total generation capability is maximized to determine the next alternative transmission line to be restored, and the weighted total generation capability of an mth alternative transmission line is equal to a sum of a ratio between the total generation capability of a simulation and a number of transmission lines to be restored in the simulation.

11. A system for network-based decision making of generator start-up, the system comprising at least one computer configured to:
determine a black-start generator unit for generator start-up;
select a plurality of generator units waiting to be restored from all generator units;
generate a labeled training data set that includes potential statuses of generator start-up;
execute a deep learning algorithm based on the training data set to establish a trained deep neural network;
obtain real-time data of generator start-up, and determine a blackout area, a state of a power system, an availability of equipment in the power system, and characteristics of generator units waiting to be restored of all of the generator units;
search and evaluate all alternative power lines to be restored in a next step using a Monte Carlo tree search algorithm and the trained deep neural network by checking voltage, frequency, and power flow variations caused by restoring the different alternative power lines;
summarize results of the Monte Carlo tree search algorithm to determine the corresponding transmission line to be restored of all the alternative power lines; and
restore the determined transmission line based on the summarized results of the Monte Carlo tree search algorithm.

Assignees

• Univ Shandong

Inventors

• Liu Yutian
• Sun Runjia

Classifications

• H02J2103/30

  Simulating, planning, modelling, reliability check or computer assisted design [CAD] of electric power networks · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/045

  Combinations of networks · CPC title

• H02J3/06 Primary

  Controlling the transfer of power between connected networks; Controlling load sharing between connected networks · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title