pandapower is a Python based, BSD-licensed power system analysis tool aimed at automation of static and quasi-static analysis and optimization of balanced power systems. It provides power flow, optimal power flow, state estimation, topological graph searches and short circuit calculations according to IEC~60909. pandapower includes a Newton-Raphson power flow solver formerly based on PYPOWER, which has been accelerated with just-in-time compilation. Additional enhancements to the solver include the capability to model constant current loads, grids with multiple reference nodes and a connectivity check. The pandapower network model is based on electric elements, such as lines, two and three-winding transformers or ideal switches. All elements can be defined with nameplate parameters and are internally processed with equivalent circuit models, which have been validated against industry standard software tools. The tabular data structure used to define networks is based on the Python library pandas, which allows comfortable handling of input and output parameters. The implementation in Python makes pandapower easy to use and allows comfortable extension with third-party libraries. pandapower has been successfully applied in several grid studies as well as for educational purposes. A comprehensive, publicly available case-study demonstrates a possible application of pandapower in an automated time series calculation.

Combining switching state optimization (SSO) and network expansion planning (NEP) in AC systems results in a mixed-integer non-linear optimization problem. Two methodically different solution approaches are mathematical programming and heuristic methods. In this paper, we develop a hybrid optimization method combining both methods to solve the combined optimization of SSO and NEP. The presented hybrid method applies a DC programming model as an initialization strategy to reduce the search space of the heuristic. A greedy heuristic ensures that the obtained solutions are AC feasible. We compare the hybrid method with other heuristic methods and three mathematical programming models on the same set of planning problems. We show optimization results for four realistic sized power system study cases. Evaluation criteria are convergence, solution cost, and run time. Results show that the hybrid method is able to find a higher number of valid AC-solutions in comparisons to the mathematical programming methods. Furthermore, the obtained solutions have lower expansions costs and are obtained in a shorter run-time compared to the remaining methods for the analyzed study cases. As an addition to this paper, the hybrid implementation and the defined benchmark cases are available as open-source software.

Fast approximations of power flow results are beneficial in power system planning and live operation. In planning, millions of power flow calculations are necessary if multiple years, different control strategies, or contingency policies are to be considered. In live operation, grid operators must assess if grid states comply with contingency requirements in a short time. In this paper, we compare regression and classification methods to either predict multivariable results, for example, bus voltage magnitudes and line loadings, or binary classifications of time steps to identify critical loading situations. We test the methods on three realistic power systems based on time series in 15 and 5 minutes resolution of 1 year. We compare different machine learning models, such as multilayer perceptrons (MLPs), decision trees, k‐nearest neighbors, gradient boosting, and evaluate the required training time and prediction times as well as the prediction errors. We additionally determine the amount of training data needed for each method and show results, including the approximation of untrained curtailment of generation. Regarding the compared methods, we identified the MLPs as most suitable for the task. The MLP‐based models can predict critical situations with an accuracy of 97% to 98% and a very low number of false negative predictions of 0.0% to 0.64%.

Integrating active power curtailment (APC) of renewable energy sources (RES) in power system planning reduces necessary investments in the power system infrastructure. In current target grid planning methods, APC is considered by fixed curtailment factors without considering the provided flexibility to its full extent. Time-series-based planning methods allow the integration of the time dependency of RES and loads in power system planning, leading to substantial cost savings compared to the worst-case method. In this paper, we present a multi-year planning strategy for high-voltage power system planning, considering APC as an alternative investment option to conventional planning measures. A decomposed approach is chosen to consider APC and conventional measures in a long-term planning horizon of several years. The optimal investment path is obtained with the discounted cash flow method. A case study is conducted for the SimBench high-voltage urban benchmark system. Results show that the time-series-based method allows for reducing investments by up to 84% in comparison to the worst-case method. Furthermore, a sensitivity analysis shows the variation in total expenditures with changing cost assumptions.

Fast approximations of power flow results are beneficial in power system planning and live operation. In planning, millions of power flow calculations are necessary if multiple years, different control strategies, or contingency policies are to be considered. In live operation, grid operators must assess if grid states comply with contingency requirements in a short time. In this paper, we compare regression and classification methods to either predict multi-variable results, e.g., bus voltage magnitudes and line loadings, or binary classifications of time steps to identify critical loading situations. We test the methods on three realistic power systems based on time series in 15min and 5min resolution of one year. We compare different machine learning models, such as multilayer perceptrons (MLPs), decision trees, k-nearest neighbors, gradient boosting, and evaluate the required training time and prediction times as well as the prediction errors. We additionally determine the amount of training data needed for each method and show results, including the approximation of untrained curtailment of generation. Regarding the compared methods, we identified the MLPs as most suitable for the task. The MLP-based models can predict critical situations with an accuracy of 97-98% and a very low number of false negative predictions of 0.0-0.64%.