Researches On Intelligent Optimization Dispatching Model And Algorithm In Internet Of Things About Coal Logistics

China is a large country in coal production and consumption, the feature————"transport coal from north to south"————makes coal production, consumption and transport become resources which are most worthy of integration and optimization in the coal supply chain. As a big province of coal natural resources, Shanxi is the source of coal supply chain. How to coordinate the coal production, transportation, supply and demand balance has been the hard task. With the promotion and application of the Internet of Things technology, the full application of the Internet of Things technology in the coal industry has been put on the agenda in our province. The application of Internet of Things technology in coal supply chain can make all the nodes with a single function and isolated from each other achieve resource sharing and collaborative process, promote fusion among the nodes of the coal supply chain, integrate information and functions, complete resource dispatching at the level of coal supply chain, make the coal mines’production associate with the customer’s coal consumption, and alleviate the drawbacks of coal inventory either excess or shortage in the supply chain. Around the above problems, this paper discusses the coal logistics supply chain integration framework based on Internet of Things and constructs the coal logistics optimization dispatching model, which is very significant for enhancing the competitiveness of Shanxi province’s coal supply chain, reducing logistics costs of coal and balancing the supply and demand of coal.First of all, this paper constructs the coal logistics supply chain intelligent system senses framework based on Internet of Things, in which the various sectors of the logistics are regarded as the intelligent sensors, and these nodes can real-time automatically sense changes of inventory and transportation and can provide timely and then provide accurate and timely data for integrated optimization dispatching of inventory and transportation based on the coal logistics supply chain.Secondly, this paper establishes the optimization model of coal logistics. The model is characterized by:①The model breakthroughs single inventory optimization model only for local inventory optimization, establishes multi-echelon overall inventory optimization model, calculates the inventory of all nodes in coal supply chain by using coal trading order and contract information, determines inventory levels of each node, and then realizes the lowest cost of the total inventory and the highest satisfaction of the customer.②The model not only considers the inventory optimization, but also considers optimization of the whole supply chain transportation dispatching to adapt this situation which the dynamic and sudden demand lead to the frequent changes of inventory and transportation planning, thus the agility and flexibility requirements for production are satisfied.③The model runs in finite periodic level. Based on stochastic elements such as stochastic demand, stochastic transportation environment and stochastic production and so on, coal logistics supply chain multi-echelon inventory control and transportation integration model seeks the optimal target which is to minimize the cost of production planning and the best decision which is customer inventory strategy and transport strategy.④Due to the random variable, the model does not have the complete mathematical significance; therefore, the model is transformed into Stochastic Expected Value Model and Stochastic Chance-Constrained Programming Model.To solve the logistics optimization dispatching stochastic programming model with NP-hard, this paper constructs an improved particle swarm optimization algorithm under complex constraint conditions based on stochastic simulation and neural network. First, in order to deal with the complexity of the constraint in the actual model, multiplier penalty function method for handling constraints conditions is combined with particle swarm optimization and an improved hybrid particle swarm optimization algorithm for solving nonlinear constrained optimization problems is proposed. The two characteristics of the algorithm include:①Multiplier penalty function value is regarded as a particle fitness;②Hybrid algorithm redefines particles personal best and group best in the particle swarm updating formula, using the particle’s location of previous generation as individual optimal of the current generation and group optimal of the previous generation as the global optimum of the current generation.On the foundation of above algorithms, this paper establishes the particle swarm optimization algorithm for solving Stochastic Expected Value Model and Stochastic Chance-Constrained Programming Model under complex constraints. The basic ideas of the algorithm is:①It constructs multiplier penalty function as the objective function of the stochastic programming to transform constrained problem into a unconstrained optimization problem;②As the model contains random variables, the methods of stochastic simulation and BP neural network is proposed to produce the output of uncertain function;③Particle swarm algorithm is improved to solve stochastic programming model with the multiplier penalty function as a new goal function. The innovation of the algorithm is:①The multiplier method is introduced to stochastic programming model to transform the constrained Stochastic Programming Model into unconstrained Stochastic Programming Model. Constraint conditions not only include constraints without random variables, but also include those constraints with random variable. According to the wishes of the decision-makers, the latter can be divided into stochastic expected value constraint and stochastic chance constrained. Since the objective function also contains random variable, it can also be divided into stochastic expected value goal function or stochastic chance constrained programming objective function based on the decision makers’will; thus Stochastic Expected Value Model and Stochastic Chance Constrained Programming Model are formed based on multiplier penalty function.②Stochastic simulation and BP neural network method are used to generate the output function of the Stochastic Expected Value Model and the Stochastic Chance-Constrained Programming Model which are based on the multiplier penalty function.③During the process of solving the Stochastic Expected Value Model and the Stochastic Chance-Constrained Programming Model based on the multiplier penalty function, in order to better follow the optimal particle fly to the optimum of the feasible region, the particle individual position of previous iteration is as the particles personal best of the current generation and the best position of the group’s previous iteration is as the particle groups best position of the current generation while the particle is updated, and finally the particle swarm will eventually converge to the optimal position in the feasible domain.Finally, some experiments with different complexity are designed to validate the model and algorithm proposed in this paper. The simulation results show that the model and the algorithm are effective.