CN-121984131-A - Collaborative robust optimal scheduling method for day-ahead and day-in-day of hydrogen-containing multi-energy micro-grid

CN121984131ACN 121984131 ACN121984131 ACN 121984131ACN-121984131-A

Abstract

The disclosure provides a method for collaborative robust optimization scheduling in the daytime of a hydrogen-containing multi-functional micro-grid, and relates to the technical field of comprehensive optimization of micro-grids. The method comprises the steps of constructing a multi-stage robust optimal scheduling model containing an uncertain variable unexpected mechanism, performing daily offline training on the multi-stage robust optimal scheduling model to obtain a daily unit combination scheduling strategy corresponding to the worst case and a daily multi-stage robust optimal scheduling model, generating a control instruction according to the daily unit combination scheduling strategy corresponding to the worst case to control the working state of a daily unit combination element, performing daily online application on the daily multi-stage robust optimal scheduling model to obtain daily energy flow scheduling strategies of all stages, and generating the control instruction according to the daily energy flow scheduling strategy of all stages to control the working state of the daily energy flow element of all stages. According to the technical scheme, feasibility and robustness of scheduling strategies of each stage in a day of the hydrogen-containing multi-energy micro-grid can be improved.

Inventors

SHEN YU
QIAN BO
FENG YADONG
HE XIN
XIE QING
ZHENG DA
LI QIUHUA
LI CHUANJIANG
WANG GUANMING
ZHAO JINGYU
ZHANG RONG

Assignees

浙江清华长三角研究院
宁波中科孚奇能源科技有限公司
杭州协能科技股份有限公司
浙江氢邦科技有限公司
上海千贯节能科技有限公司
南京合智电力科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (7)

1. The method for collaborative robust optimal scheduling of the hydrogen-containing multi-energy micro-grid day-ahead and day-ahead is characterized by comprising the following steps of: constructing a multi-stage robust optimization scheduling model containing an uncertain variable unexpected mechanism for the hydrogen-containing multi-energy micro-grid system, wherein the range of the uncertain variable is defined by an uncertain set, the unexpected mechanism is that the multi-stage robust optimization scheduling model is in a sequential solving structure, and the multi-stage comprises a day-ahead stage and an intra-day multi-stage; Performing daily offline training on the multi-stage robust optimal scheduling model to obtain a daily combined scheduling strategy of a daily unit and a daily multi-stage robust optimal scheduling model which are in response to a worst case, wherein the worst case is a value scene that each daily unit has uncertain variables to maximize the scheduling cost of each daily unit; Generating a control instruction according to the worst day-ahead stage unit combination scheduling strategy, and controlling the working state of the day-ahead stage unit combination element; And carrying out on-line daily application on the multi-stage robust optimal scheduling model in the daily, obtaining energy flow scheduling strategies of each stage in the daily by combining the uncertain variable results actually measured by each stage in the daily, and generating control instructions according to the energy flow scheduling strategies of each stage in the daily to control the working states of energy flow elements of each stage in the daily.
2. The method of claim 1, wherein the uncertainty variables comprise a power generation side uncertainty variable and a load side uncertainty variable.
3. The method according to claim 1 or 2, wherein the pre-day offline training using a crew-combination robust dual dynamic programming algorithm comprises: Decomposing the multi-stage robust optimal scheduling model into an L-shaped nested structure of a combination problem of a unit in a day-ahead stage, a robust economic scheduling problem in a first stage in the day and a robust economic scheduling problem in other stages in the day; the objective function of the combination problem of the unit in the early stage is a total cost value function, and the combination problem comprises a combination scheduling cost value function of the unit in the early stage and a worst future cost value function of the unit in the first stage in the day; The objective functions of the first-stage robust economic dispatching problem in the day and the other-stage robust economic dispatching problem in the day are respectively a first-stage cost value function in the day and a other-stage cost value function in the day, and the objective functions comprise a current-stage dispatching cost value function in the day and a worst-case future cost value function in the next stage in the day.
4. The method of claim 3, wherein the pre-day offline training with a crew-combination robust dual dynamic programming algorithm further comprises: Initializing the unit combination robust dual dynamic programming algorithm to obtain initial upper bound approximation and initial lower bound approximation of a worst case future cost value function of each stage in the day; solving the combination problem of the day-ahead stage unit according to the upper-bound approximation and the lower-bound approximation of the worst future cost value function of each stage in the current day to obtain the upper-bound approximation and the lower-bound approximation of the current total cost value function; judging whether the unit combination robust dual dynamic programming algorithm converges or not according to the upper-bound approximation and the lower-bound approximation of the current total cost value function; if the judgment is that the set combination robust dual dynamic planning algorithm is not converged, carrying out forward process solution and backward process solution of the set combination robust dual dynamic planning algorithm, updating upper-bound approximation and lower-bound approximation of a worst case future cost value function of each stage in the day, further updating the upper-bound approximation and lower-bound approximation of the total cost value function, and then judging whether the set combination robust dual dynamic planning algorithm is converged or not again; and if the judgment is convergent, obtaining the unit combination scheduling strategy of the day-ahead stage and the multi-stage robust optimal scheduling model in the day according to the solving result of the unit combination robust dual dynamic programming algorithm.
5. The method of claim 4, wherein solving the upper bound approximation of the intra-day worst case future cost value function using an intra-relaxation approximation comprises constructing an upper bound approximation function of the intra-day future cost value function by introducing auxiliary relaxation variables and penalty coefficients, wherein the upper bound approximation function of the intra-day worst case future cost value function comprises a maximum-minimum structure.
6. The method of claim 5, wherein processing the maxima and minima structure using an alternating optimization method comprises: Performing dual processing on the inner layer minimum structure of the maximum minimum structure to obtain a single-layer maximum optimization problem comprising bilinear terms; Dividing the variables of the bilinear term into two groups, and decomposing the single-layer maximum optimization problem into two sub-objective functions based on the variable division; And solving the two sub-objective functions in an alternate iteration mode until the two sub-objective function values are equal, wherein when solving one sub-objective function, the variables in the sub-objective function come from the latest solving result of the other sub-objective function.
7. The method of claim 1 or 6, wherein the on-line daily application further comprises deriving an on-day scheduling total cost based on the on-day stage energy flow scheduling policy.

Description

Collaborative robust optimal scheduling method for day-ahead and day-in-day of hydrogen-containing multi-energy micro-grid Technical Field The disclosure relates to the technical field of comprehensive optimization of micro-grids, in particular to a collaborative robust optimization scheduling method for a hydrogen-containing multi-energy micro-grid day-ahead and day-ahead. Background In order to reduce fossil fuel pollution, with the continuous development of novel power systems, the installed proportion of renewable energy sources in China is improved year by year. As a system integrating various heterogeneous energy flows, a Multi-energy micro grid (MEMG) can absorb renewable energy sources such as wind photovoltaic and the like, and meanwhile, the whole energy utilization efficiency is improved by using an internal energy conversion element, so that the system is widely focused. The heat value of the hydrogen energy is high, and no pollution is caused. Therefore, it is necessary to study a multi-energy micro-grid (H-MEMG) that couples hydrogen energy. Because of the uncertainty of renewable energy output and load, higher requirements are placed on the scheduling of H-MEMG. The traditional Two-stage robust optimization method (Two-stage robust optimization, TSRO) ignores unexpected problems in the uncertain parameter implementation process, so that the solved scheduling strategy has low feasibility. Therefore, a collaborative robust optimal scheduling method for the hydrogen-containing multi-energy micro-grid day-ahead and day-ahead in consideration of the unexpected problem in the uncertain parameter implementation process becomes a technical problem to be solved in the field. Disclosure of Invention In a first aspect, the present disclosure provides a method for collaborative robust optimization scheduling in the day-ahead and day-ahead of a hydrogen-containing multi-energy micro-grid, including: And constructing a multi-stage robust optimization scheduling model containing an uncertain variable unexpected mechanism for the hydrogen-containing multi-energy micro-grid system, wherein the range of the uncertain variable is defined by an uncertain set, the unexpected mechanism is that the multi-stage robust optimization scheduling model is in a sequential solving structure, and the multi-stage comprises a day-ahead stage and an intra-day multi-stage. And performing daily offline training on the multi-stage robust optimal scheduling model to obtain a daily combined scheduling strategy of a daily unit and a daily multi-stage robust optimal scheduling model, wherein the worst case is a value scene in which uncertain variables of each daily stage maximize the scheduling cost of each daily stage. And generating a control instruction according to a combination scheduling strategy of the day-ahead stage unit in response to the worst case, and controlling the working state of the combination element of the day-ahead stage unit. And carrying out on-line daily application on the daily multi-stage robust optimal scheduling model, combining the uncertain variable results actually measured by each daily stage to obtain the daily energy flow scheduling strategy of each daily stage, and generating a control instruction according to the daily energy flow scheduling strategy of each daily stage to control the working state of the daily energy flow element of each daily stage. In a second aspect, the present disclosure provides a method for pre-day offline training of a multi-stage robust optimal scheduling model, comprising: The machine set combination robust dual dynamic programming algorithm decomposes a multi-stage robust optimal scheduling model into an L-shaped nested structure of a machine set combination problem of a day front stage, a machine set combination problem of a first stage in the day and a machine set combination problem of a rest stage in the day, wherein an objective function of the machine set combination problem of the day front stage is a total cost value function and comprises a machine set combination scheduling cost value function of the day front stage and a worst case future cost value function of the first stage in the day, the machine set combination robust economic scheduling problem of the first stage in the day and the machine set combination problem of the rest stage in the day are respectively a cost value function of the first stage in the day and a cost value function of the rest stage in the day, and the objective function of the machine set combination problem of the day front stage is a scheduling cost value function of the current stage in the day and a worst case future cost value function of the next stage in the day. And initializing a unit combination robust dual dynamic programming algorithm to obtain an initial upper bound approximation and an initial lower bound approximation of a worst case future cost value function of each stage in the day. And solving the combination pro