Power prediction of hybrid power scraper based on conditions recognition and Markov chain

Hybrid power scraper has the characteristics of harsh working environment, special operating conditions, complex power system and large variation range of demand power.The control strategy is difficult to achieve an optimal power distribution.Therefore, it is necessary to effectively predict the demand power of hybrid power scraper, while the traditional single prediction model is difficult to accurately predict its demand power.In order to solve the above problems, this paper presents a method for predicting the demand power of hybrid power scraper based on working condition identification and Markov chain.Firstly, the structure and overall parameters of the power system of the hybrid power scraper are introduced.According to the working characteristics of the hybrid power scraper, the cyclic working conditions are divided into five typical working conditions.On this basis, according to the actual working conditions data collected from the field operation of the hybrid power scraper, the dimensionality of the characteristic parameters is reduced using principal component analysis (PCA), and six special parameters are extracted.The possibility C-means clustering algorithm is used to identify five typical operating conditions of hybrid power scraper.Then, according to the principle of Markov chain, the output power of hybrid power scraper is taken as the state, and the state space is determined according to the range of power variation.Finally, the state transition probability matrix is established based on the identified operating output power data to predict the future demand power of hybrid power scraper.The results show that the Markov chain model based on condition identification can significantly improve the accuracy of power prediction of hybrid power scraper compared with the model without distinguishing the working conditions.At the same time, it can maintain a good stability in the face of the rapidly changing output power, which verifies that the model has high prediction accuracy and robustness.