Search

This paper proposes an interval prediction technology of photovoltaic (PV) power based on parameter optimization of extreme learning machine (ELM) model. First, the weighted Euclidean distance is proposed as the evaluation index of PV power prediction interval. The historical sample units are screened and the ELM training set is optimized. Then, a hybrid optimization algorithm for ELM parameters is proposed. The hidden layer input and output weights and biases parameters of the ELM model are optimized by using the elitist strategy genetic algorithm and quantile regression, and the trained model is used to predict the PV power range. Finally, an actual calculation example is constructed based on the historical data of PV power plants and weather stations. The PV power interval is predicted, and the results are compared with those obtained by other methods. The results of the calculation example show that the method proposed can greatly improve the accuracy of interval prediction while increasing the reliability of interval prediction.

Aimed at the problem of uncertainties in the regional development stage and the difficulties in quantifying regional investment demand in different development stages, a robust optimization method for power grid investment decision-making considering regional development stage uncertainties is proposed to promise the matching degree between power grid investment decisions and development needs, and to improve the ability of decision-making results to deal with portfolio risks and uncertainties in regional development stage. First, investment risk constraints are constructed based on the modern portfolio theory. Then, a box uncertainty set is used to characterize uncertainties in regional development stage, and a robust optimization model for power grid investment decision-making considering uncertainties in development stage is established. In the optimization model, the outer minimization problem is used to solve the uncertain variables in regional development stage in the worst scenario, while inner maximization problem is used to obtain the decision-making plan that can maximize investment return in the worst scenario. Furthermore, according to the strong duality theory, the double-layer optimization model is transformed into a single-layer model that can be solved directly, and the big-M method is used to solve the model proposed. Finally, an actual example of 13 cities in an eastern coastal province verifies the applicability and effectiveness of the power grid investment decision-making model.

With the increasing scale of wind power installation and grid connection, the interconnection of power grids through tie lines can absorb wind power in a wider range, but the uncertainty of wind power output leads to an increase in the risk of system operation. Conditional value-at-risk (CVaR) was introduced in multi-regional unit commitment to quantify the risk, and a multi-regional unit commitment model considering conditional value at risk was proposed, so as to balance the operation economy and risk loss of multi-regional unit commitment with an interconnected power grid. Besides, the tie line operation mode was modeled in detail to coordinate the economy and flexibility of tie line operation. By introducing auxiliary variables, the tie line operation mode model was transformed into a mixed-integer linear constraint. Finally, the proposed multi-region unit commitment model was transformed into a mixed-integer linear programming form for the solution. With the improved two area system as a simulation example, the results demonstrate that this method can effectively control the tail risk of the system and improve the ability to resist the wind power output uncertainty while ensuring the operation economy of multi-area internet power grid. [ABSTRACT FROM AUTHOR]