This paper presents the application of an improved ant colony optimization algorithm called mixed integer distributed ant colony optimization to optimize the power flow solution in power grids. The results provided indicate an improvement in the reduction of operational costs in comparison with other optimization algorithms used in optimal power flow studies. The application was realized to optimize power flow in the IEEE 30 and the IEEE 57 bus test cases with the objective of operational cost minimization. The optimal power flow problem described is a non-linear, non-convex, complex and heavily constrained problem.

This paper provides a method for simplified description of a regional power grid model aimed to deliver a grid reduction, and improve grid performance observability. The derived power grid model can be used to analyze the regional allocation of the decentralized energy generation and consumption. The expansion of wind and solar generation in the power system affects the residual load. The power balance between electricity consumption and generation was calculated and analyzed based on the temporal and spatial scales. The proposed grid clustering method is a useful approach for performance analysis in systems with a growing share of renewable generation.

In recent years, due to the increasing number of renewable energy sources, which are characterised by the stochastic nature of the generated power, interest in energy storage has increased. Commercial installations use simple deterministic methods with low economic efficiency. Hence, there is a need for intelligent algorithms that combine technical and economic aspects. Methods based on computational intelligence (CI) could be a solution. The paper presents an algorithm for optimising power flow in microgrids by using computational intelligence methods. This approach ensures technical and economic efficiency by combining multiple aspects in a single objective function with minimal numerical complexity. It is scalable to any industrial or residential microgrid system. The method uses load and generation forecasts at any time horizon and resolution and the actual specifications of the energy storage systems, ensuring that technological constraints are maintained. The paper presents selected calculation results for a typical residential microgrid supplied with a photovoltaic system. The results of the proposed algorithm are compared with the outcomes provided by a deterministic management system. The computational intelligence method allows the objective function to be adjusted to find the optimal balance of economic and technical effects. Initially, the authors tested the invented algorithm for technical effects, minimising the power exchanged with the distribution system. The application of the algorithm resulted in financial losses, €12.78 for the deterministic algorithm and €8.68 for the algorithm using computational intelligence. Thus, in the next step, a control favouring economic goals was checked using the CI algorithm. The case where charging the storage system from the grid was disabled resulted in a financial benefit of €10.02, whereas when the storage system was allowed to charge from the grid, €437.69. Despite the financial benefits, the application of the algorithm resulted in up to 1560 discharge cycles. Thus, a new unconventional case was considered in which technical and economic objectives were combined, leading to an optimum benefit of €255.17 with 560 discharge cycles per year. Further research of the algorithm will focus on the development of a fitness function coupled to the power system model.