Author: Junfei Wang(junfeiw@yorku.ca)
Introduction:
This dataset is generated based on the IEEE 118-Bus System [1]. Power demands are sampled uniformly in the range of [80%-120%] of the default demands in the system snapshot corresponding to each bus as is common practice in existing literature [2-4]. The branch flow constraints are adopted from [5] as these are not natively available in benchmark case files. The dataset released contains 70,000 feasible scenarios on IEEE 118 bus system. These data points emulate real datasets containing feasible load flow data recorded by system operators.
Furthermore, renewable energy resources are considered in this dataset. As of 2020, wind and solar power sources compose 8.4% and 2.3% of overall generation in the United States[6]. The same proportion is applied to model the penetration of renewable generation systems in our experiments. For example, amongst the 64 buses that are defined initially to be load buses in the 118-bus system, 24 buses are randomly selected to be connected to wind farms, and the remaining 40 buses are associated with solar generation systems. Each wind farm is assumed to be composed of turbines representing a total of 1800 m^2 swept area. Each solar farm consists of 10,000 photovoltaic panels (1.67m^2 per panel). Uncertainties in wind speed and solar generation are modeled using the Weibull probability distribution with the shape and location parameters taking values of (2,5) and (2.5,6), respectively, as is common practice in the literature (e.g. [7,8]). DGs such as wind and/or photovoltaics are incorporated into power flow equations where power injection will decrease.
How to use:
There are 6 files in this dataset:
1. grid_118_demand.mat is the demand file, contains active and reactive power demand at each bus
2. grid_118_solution.mat is the solution file. 1-118, 119-236 columns are voltage magnitudes and phase angles for each bus. 237-290 columns are active power generations, 291-344 are reactive power generations.
3. case118_gencost.mat is the coefficient of the cost function from Matpower. This can be used for Optimal Power Flow problem. Please check details at: https://matpower.org/docs/ref/matpower5.0/case118.html
4.grid118_line_constraints.mat is line constraints from [5]
5. IEEE118_gen.mat contains attributes of all generators, also from Matpower. Upper/lower value limits are in this data structure.
6. Y_bus118.mat the admittance matrix, which can be used for physics equations.
How to cite:
For any academic research usage, please cite our paper as follows:
The journal paper is under review, please cite this page for now.
Links:
References:
[1] https://icseg.iti.illinois.edu/ieee-118-bus-system/
[2] Pan, X., Chen, M., Zhao, T., & Low, S. H. (2020). Deepopf: A feasibility-optimized deep neural network approach for ac optimal power flow problems.
[3] Fioretto, F., Mak, T. W., \& Van Hentenryck, P. (2020, April). Predicting ac optimal power flows: Combining deep learning and lagrangian dual methods. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 34, No. 01, pp. 630-637).
[4] Huang, W., & Chen, M. DeepOPF-NGT: Fast No Ground Truth Deep Learning-Based Approach for AC-OPF Problems.
[5] http://motor.ece.iit.edu/data/ROSCUC_118.xls
[7] Parajuli, A. (2016). A statistical analysis of wind speed and power density based on Weibull and Rayleigh models of Jumla, Nepal. Energy and Power Engineering, 8(7), 271-282.
[8] Afzaal, M. U., Sajjad, I. A., Awan, A. B., Paracha, K. N., Khan, M. F. N., Bhatti, A. R.,& Tlili, I. (2020). Probabilistic generation model of solar irradiance for grid connected photovoltaic systems using weibull distribution. Sustainability, 12(6), 2241.