CN-121998145-A - Startup optimization method and system for gas storage compressor unit

CN121998145ACN 121998145 ACN121998145 ACN 121998145ACN-121998145-A

Abstract

The invention discloses a starting-up optimization method and a starting-up optimization system for a gas storage compressor unit, which concretely comprise the following steps of S1, constructing a starting-up optimization model of the gas storage compressor unit according to historical data of the gas storage compressor unit, S2, training and optimizing the starting-up optimization model based on a reinforcement learning algorithm and a genetic algorithm, and S3, inputting real-time data of the gas storage compressor unit into the trained starting-up optimization model, and outputting an optimal starting-up scheme, namely the starting-up number per hour.

Inventors

Chen Jiexue
WANG BO
HE ZHIQIANG
TAN JIAN
CUI YUE
MA LIN

Assignees

中国石油天然气股份有限公司

Dates

Publication Date: 20260508
Application Date: 20241105

Claims (10)

1. The starting-up optimization method for the gas storage compressor unit is characterized by comprising the following steps of: S1, constructing a startup optimization model of the gas storage compressor unit according to historical data of the gas storage compressor unit; s2, training and optimizing the start-up optimization model based on a reinforcement learning algorithm and a genetic algorithm; And S3, inputting real-time data of the gas storage compressor set into the trained starting optimization model, and outputting an optimal starting scheme, namely the starting number per hour.
2. The method for optimizing the start-up of a gas storage compressor unit as set forth in claim 1, wherein in S1, an objective function of the start-up optimization model is: In the formula (1), The method is characterized by comprising the steps of representing optimization targets, wherein the optimization targets comprise compressor station electricity charge and starting loss charge, n i represents the number of compressors started in the ith hour, n i-1 represents the number of compressors started in the ith-1 hour, P represents single compressor power Kw, f pi represents the ith hour electricity price, yuan/Kw, Q g represents the single compressor starting and discharging natural gas quantity, and m 3 ;f g represents the natural gas price, yuan/m 3 .
3. The method for optimizing the start-up of a gas storage compressor unit as set forth in claim 1, wherein in S2, the reinforcement learning algorithm is a Q-learning algorithm.
4. The method for optimizing the start-up of a gas storage compressor unit as set forth in claim 1, wherein S2 specifically includes: s2-1, determining a reinforcement learning state, and equally dividing the state into m parts; S2-2, determining a reward value of reinforcement learning; S2-3, enhancing the mutation rate and the crossing rate of the learning optimization genetic algorithm; S2-4, updating the Q table according to the reinforcement learning state and the rewarding value; s2-5, selecting action in the current state by using the updated Q table.
5. The method for optimizing the start-up of a gas storage compressor unit as set forth in claim 4, wherein in S2-1, the reinforcement learning state is: S=w 1 *f_s+w 2 *d_s+w 3 *m_s; In the formula (2), S represents a state of reinforcement learning, w 1 、w 2 、w 3 represents a weight parameter, f_s represents a population fitness sum of interest, d_s represents an absolute deviation of population fitness of interest from a population average fitness, m_s represents a population fitness maximum of interest, fitv represents a current population fitness value, last_ fitv represents a previous population fitness value, and sum represents a summation function.
6. The method for optimizing the start-up of a gas storage compressor unit as set forth in claim 4, wherein in S2-2, the prize value is determined by the following formula: In the formula (3), r 1 represents a prize value when the action is a crossover value, and 2 represents a prize value when the action is a mutation rate.
7. A start-up optimization method for a gas storage compressor unit as defined in claim 4, wherein S2-3 comprises: S2-3-1, setting the minimum value and the maximum value of the mutation rate, and equally dividing the interval into n parts, wherein n is the action dimension; S2-3-2, then, using circulation to traverse each action dimension, and calculating a starting value and an ending value of each cell according to the specified minimum value and maximum value; s2-3-3, generating floating point numbers in the selected cells by a random method, and using the floating point numbers as variation rates of corresponding action dimensions.
8. The method for optimizing the start-up of a gas storage compressor unit as set forth in claim 4, wherein in S2-4, the Q table updating method is as follows: Q(s,a)=Q(s,a)+α[R+γmax a′ Q(s ′ ,a ′ )-Q(s,a)] (4) In the formula (4), Q (s, a) represents the Q value of the current action a in the current state s, α represents the learning rate, R represents the bonus value, γ represents the decay rate, and Q (s ′ ,a ′ ) represents the Q value of the next action a ′ in the next state s ′ .
9. The method for optimizing the start-up of the gas storage compressor unit according to claim 1, wherein in S3, constraint conditions in real-time data of the gas storage compressor unit are as follows: In the formula (5), n i represents the number of compressors started in the ith hour, Q represents the injection amount of a single compressor per hour, and Q z represents the total air injection amount required by the compressor per day, in case of square per hour.
10. A start-up optimization system for a gas storage compressor train based on the method of any one of claims 1-9, comprising: the data acquisition unit is used for acquiring historical data of the gas storage compressor unit and real-time data of the gas storage compressor unit from the server; The model construction unit is used for constructing a startup optimization model according to historical data of the gas storage compressor unit; the model training unit is used for training and optimizing the startup optimization model; and the output unit is used for outputting a starting scheme, namely the starting number per hour, according to the real-time data of the gas storage compressor unit and the trained starting optimization model.

Description

Startup optimization method and system for gas storage compressor unit Technical Field The invention relates to the technical field of data processing, in particular to a startup optimization method and system for a gas storage compressor unit. Background The underground gas storage is an artificial gas field or gas reservoir formed by reinjecting natural gas conveyed by a long-distance pipeline into an underground closed space. With the increasing pace of gas storage construction, the problem of optimizing operation of the gas storage is also more and more important. The compressor unit in the large-scale gas storage is composed of a plurality of compressors, the compressors are used for gas injection and production activities of the gas storage, the gas storage is different in electric charge in different time periods (peak time period, valley time period and valley time period) in the operation process, the starting-up quantity of the current compressor unit in different electric charge time periods is the same, so that the cost is not beneficial to reduction, the compressor unit needs to be operated in a peak-shifting mode on the premise of ensuring the injection and production quantity, namely, reasonable planning is required to be made on the starting-up quantity in different electric charge time periods on the premise of ensuring the injection and production quantity, and therefore the starting-up scheme of the compressor unit with the lowest consumption needs to be researched. At present, the optimization of the starting scheme of the compressor unit adopts genetic algorithms, wherein the genetic algorithms comprise genetic algorithms of a single group and genetic algorithms of multiple groups, the genetic algorithms can optimize the starting scheme of the compressor to a certain extent, the running cost of the compressor unit of the gas storage is reduced, but a common genetic algorithm optimization model has no self-learning capacity and cannot update parameters of the optimization algorithm in time, so that the optimization effect is not optimal, and the running cost of the compressor unit is higher. Disclosure of Invention Aiming at the technical problem of higher starting-up operation cost of a gas storage compressor unit in the prior art, the invention provides a starting-up optimization method and a starting-up optimization system for the gas storage compressor unit, which are used for jointly establishing an optimization model by a genetic algorithm and a reinforcement learning Q-learning algorithm, optimizing the starting-up quantity of the gas storage compressor unit at different times when the gas storage compressor unit operates, solving the problem of higher cost caused by the same starting-up quantity at different electric charge time periods. In order to achieve the above object, the present invention provides the following technical solutions: the starting-up optimization method for the gas storage compressor unit specifically comprises the following steps: S1, constructing a startup optimization model of the gas storage compressor unit according to historical data of the gas storage compressor unit; s2, training and optimizing the start-up optimization model based on a reinforcement learning algorithm and a genetic algorithm; And S3, inputting real-time data of the gas storage compressor set into the trained starting optimization model, and outputting an optimal starting scheme, namely the starting number per hour. Preferably, in the step S1, the objective function of the startup optimization model is: In the formula (1), The method is characterized by comprising the steps of representing optimization targets, wherein the optimization targets comprise compressor station electricity charge and starting loss charge, n i represents the number of compressors started in the ith hour, n i-1 represents the number of compressors started in the ith-1 hour, P represents single compressor power Kw, f pi represents the ith hour electricity price, yuan/Kw, Q g represents the single compressor starting and discharging natural gas quantity, and m 3;fg represents the natural gas price, yuan/m 3. Preferably, in the step S2, the reinforcement learning algorithm is a Q-learning algorithm. Preferably, the S2 specifically includes: s2-1, determining a reinforcement learning state, and equally dividing the state into m parts; S2-2, determining a reward value of reinforcement learning; S2-3, enhancing the mutation rate and the crossing rate of the learning optimization genetic algorithm; S2-4, updating the Q table according to the reinforcement learning state and the rewarding value; s2-5, selecting action in the current state by using the updated Q table. Preferably, in S2-1, the reinforcement learning state is: S=w1*f_s+w2*d_s+w3*m_s; In the formula (2), S represents a state of reinforcement learning, w 1、w2、w3 represents a weight parameter, f_s represents a population fitness sum of interest, d_s repre