Iron And Steel Enterprise Gas System Prediction And Optimal Operation Research

Gas system of iron and steel enterprises enjoys close coupling with main production system. With the change of production conditions, occurrence and consumption of gas will change accordingly but with small regularity. Personal experiences can not predict gas consumption accurately, which leads to the backwardness of management. Given that most production condition are controlled after the change, especially when it takes long time to control it, it will cause huge waste of energy, and bring high pressure on the operation of gas system. In order to solve the above problems, starting from the half life circle of gas, this paper systematically defines and evaluates factors affecting gas utilization. It adopts the thinking of gradual optimization and establishes a prediction model of classifying gas system. With this prediction model, this paper takes a perspective of overall situation of gas system, targeting at the minimum consumption of fixed and variable users. Taking the characteristics of potential users’using gas into consideration, a model of optimized scheduling is constructed. Gas utilization of enterprises is explored based on the scheduling model. Detailed investigation is as follows.(1) Based on the identification of factors affecting gas occurrence and consumption in iron and steel enterprises, a SVC-HP-ENN-LSSVM-MC classification prediction model is established to cope with the reality of little prediction accuracy of the mechanistic model. The whole modeling process adopts the recognition of pre-prediction model, that is, data is predicted based on their qualities and then modified correspondingly. The occurrence and consumption of gas in iron and steel enterprises fluctuate frequently. SVC is used to classify the production condition and different prediction models are built to deal with different production conditions. HP wave filtering is used to divide original data sequence into trend and volatility series. Elman neural network and advantages of least square support vector machine are combined to build a prediction model. After prediction, Markov transition matrix is introduced to modify residual series. Wilcoxon rank test is conducted to examine data of iron and steel enterprises. Results show that:based on offline classification model, prediction models reflecting all production conditions are established, which not only reduces online training time, but also improves the accuracy of prediction, especially when the production condition changes, the outcome is more satisfactory.(2) Based on consumption characteristics of gas users and prediction results of gas occurrence and consumption in iron and steel enterprises, the paper builds a optimized scheduling model of gas system based on factors like heat demand by gas users, energy loss in gas mixing entropy and users’ interruptible capacity. It aims at the minimum fuel consumption. In line with its quality, energy cascade utilization is realized. Meanwhile, in order to reduce the computational complexity, fixed users, variable users and interruptible users are taken into calculation except transferred users. This model compensates the drawback of ignoring interruptible users.(3) In order to realize the utmost utilization of gas as a whole, this paper presents two modeling approaches for interruptible users to adjust the consumed amount of gas. Method1:according to the working features of captive power plant boilers, a model reflecting the relation between fuel consumption and boiler load is built to acquire boilers’ economic area. Based on this model, optimal scheduling rules to ensure gas operates with safety within the economic area is set up for interruptible users. Method2:in the objective function of interruptible users’economic benefits, variable weight penalty function is adopted to deal with the relation between energy cost and operational risks. It presents probability model coordinating the relationship between economic quality of scheduling and operational risks for interruptible users. Boilers’ variable weight penalty function and gas tank counter variable weight penalty function, are constructed to scientifically and reasonably depict scheduling situation, which effectively avoids the shortage of conventional multi-objective optimization that artificially fixed weights is far from the actual gas fluctuation, leading to large deviation of penalty estimation and making hcorrect decisions.The paper establishes prediction and optimized scheduling models of gas system in iron and steel enterprises. The prediction model enjoys high accuracy than other models, and has the capability to adapt to production condition’s change. The average relative error of gas consumption prediction is less than or equal to2.3%, and endures Wilcoxon rank test, which meets the need of industrial production. For typical operating conditions, the scheduling scheme is reasonable and practical. Results show that:if the prediction and scheduling models apply in iron and steel enterprises with normal production conditions, it will save gas about16.63kgce/t steel and save197956.29tce per year, which enjoys broad prospects for energy saving.