A data-driven distributionally robust optimization model for multi-energy coupled system considering the temporal-spatial correlation and distribution uncertainty of renewable energy sources

Yachao Zhang,Yan Liu,Shengwen Shu,Feng Zheng,Zhanghao Huang

DOI: https://doi.org/10.1016/j.energy.2020.119171

IF: 9

2021-02-01

Energy

Abstract:<p>The increasing expansion of wind and gas turbine installation has intensified the interdependency between power system and natural gas network, which poses great challenges to the coordination scheduling of the multi-energy coupled system (MECS). The data-driven robust optimization (DDRO) model is proposed for the energy coupled system. In this model, the temporal-spatial correlation of wind power can be considered based on the minimum volume enclosing convex hull uncertainty set, and the confidence set about the probability distribution for wind power scenarios with the form of the norm-1 and norm-<em>inf</em> constraints is constructed to handle wind power uncertainty. Moreover, to describe natural gas transient characteristics, the hydrodynamic model for gas flow represented as a series of partial differential equations is transformed by the Wendroff difference scheme and linearization technique. And then the master-subproblem framework and tri-level duality-free decomposition method is developed to solve the above model. Finally, the proposed model and solving method are carried out on two test systems with different scale, and the robust optimization models and distributionally optimization models in the existing literatures are implemented for comparison. Simulation results demonstrate the effectiveness and superiority of the proposed model for solving the coordination scheduling problem of MECS.</p>

energy & fuels,thermodynamics

What problem does this paper attempt to address?

This paper aims to solve the coordinated scheduling problem caused by the uncertainty of renewable energy in the multi - energy coupling system (MECS). Specifically, with the continuous expansion of the installation scale of wind turbines and gas turbines, the interdependence between the power system and the natural gas network is increasing day by day, which poses a great challenge to the coordinated scheduling of the multi - energy coupling system. The paper proposes a data - driven distributionally robust optimization (DDRO) model to deal with the uncertainties of the wind energy and natural gas systems, especially considering the time - space correlation and distribution uncertainty of wind power generation.

### Main problems solved in the paper:

1. **Uncertainty of wind power generation**:
- There is a time - space correlation in the prediction error of wind power generation, that is, there is a correlation between the prediction errors of wind power generation at different times and locations.
- The distribution uncertainty of wind power generation, that is, the actual output of wind power generation may deviate from the predicted value, and the probability distribution of this deviation is uncertain.

2. **Coordinated scheduling of multi - energy coupling systems**:
- The operational differences between the power system and the natural gas network on different time scales need to be comprehensively considered in the scheduling model.
- As a coupling element, the gas turbine provides up/down regulation reserve capacity in the power system to deal with the uncertainty of wind power generation, and at the same time, it is necessary to ensure the operational feasibility of the natural gas system.

### Solutions:

1. **Constructing the minimum - volume - enclosing convex hull uncertainty set**:
- Use historical data to construct a high - dimensional ellipsoidal set, considering the time - space correlation of wind power generation prediction errors.
- Through orthogonal decomposition and scaling factor adjustment techniques, construct a modified convex hull uncertainty set.

2. **Establishing a data - driven distributionally robust optimization model**:
- Combine the time - space correlation and distribution uncertainty of wind power generation to construct a confidence set.
- Consider the operational differences between the power system and the natural gas network on different time scales and establish a joint optimization model.

3. **Solution methods**:
- Develop a three - level master - sub - problem framework based on the dual - free decomposition method and the column generation algorithm to solve the proposed DDRO model.
- By introducing auxiliary binary and continuous variables, linearize the nonlinear constraints and absolute value constraints, and convert the model into a mixed - integer linear programming problem.

### Main contributions:

1. **Comprehensively consider the time - space correlation and distribution uncertainty of wind power generation**:
- Construct the minimum - volume - enclosing ellipsoidal set and confidence set, reducing the conservatism of decision - making.

2. **Consider the operational differences of multi - energy coupling systems**:
- Through the joint optimization model, realize the coordinated scheduling of the power system and the natural gas network on different time scales.

3. **Develop an efficient solution method**:
- The proposed three - level master - sub - problem framework can effectively solve complex optimization problems and ensure the feasibility and effectiveness of the model.

### Summary:

This paper solves the coordinated scheduling problem caused by the uncertainty of wind power generation in the multi - energy coupling system by constructing a data - driven distributionally robust optimization model, improving the reliability and economy of the system.