Distributed optimization method for lightweight port integrated energy system

Abstract

A lightweight port comprehensive energy system distributed optimization method, constructs and trains the port logistics predictor through the historical operation data, models the port logistics-energy system distributed optimization problem, and uses the output of the trained port logistics predictor as the initial solution of the local sub-problem of the logistics system in the optimization problem model, solves the port logistics-energy system distributed optimization problem model through the ADMM algorithm iteration, and obtains the globally optimal joint scheduling scheme. The application uses a convolutional neural network to predict the initial solution of the logistics system sub-problem, reduces the solving time of the MIQP sub-problem, and guarantees the convergence of the global optimal solution of the system, which can significantly improve the efficiency of the port energy and logistics system scheduling, realize the lightweight and intelligent solving process, and adapt to the efficient management demand of large-scale port comprehensive energy system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of port control, and particularly relates to a lightweight port integrated energy system distributed optimization method. BACKGROUND

[0002] In the existing port operation scheduling, the energy system and the logistics system are usually managed independently, and there is a lack of data and decision interaction. This separated management mode leads to resource waste, increased operation cost and carbon emissions. With the increasing demand for integrated management, the scheduling problem of the energy-logistics system is converted into a large-scale mixed integer programming problem, which is complex and time-consuming to solve, affecting the efficiency of actual application. SUMMARY

[0003] The present application is directed to the problem of too long calculation time in the existing port integrated energy-logistics system scheduling optimization, and proposes a lightweight port integrated energy system distributed optimization method, which uses a convolutional neural network to predict the initial solution of the logistics system sub-problem, reduces the solving time of the MIQP sub-problem, and guarantees the convergence of the global optimal solution of the system, which can significantly improve the efficiency of port energy and logistics system scheduling, realize a lightweight and intelligent solving process, and adapt to the efficient management needs of large-scale port integrated energy systems.

[0004] The present application is implemented by the following technical solutions:

[0005] The present application relates to a lightweight port integrated energy system distributed optimization method, which constructs and trains a port logistics predictor using historical operation data, models the port logistics-energy system distributed optimization problem, uses the output of the trained port logistics predictor as the initial solution of the local sub-problem of the logistics system in the optimization problem model, iteratively solves the port logistics-energy system distributed optimization problem model through the ADMM algorithm, and obtains a globally optimal joint scheduling scheme.

[0006] Technical effects

[0007] The application optimizes the predictor through embedded learning, quickly predicts the initial solution of the logistics system in the early iteration stage of ADMM, bypasses the cumbersome solution of the mixed integer programming problem, uses the predictor based on the convolutional neural network in the early stage to accelerate the solution, and switches to the existing optimization method in the later stage to ensure the global optimal solution. Compared with the prior art, the application can quickly predict the initial solution of the logistics system in the early iteration stage of ADMM by using the predictor model based on the convolutional neural network, bypass the cumbersome solution of the mixed integer programming problem, and flexibly switch the solution method according to the iteration stage through the mixed strategy. The strategy makes the method adaptable, can dynamically adjust the solution process according to the actual demand, optimizes the utilization of computing resources, significantly reduces the calculation time, significantly improves the solution efficiency, and is particularly suitable for real-time scheduling of large-scale port systems. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 is a flowchart of the application;

[0009] Figure 2 and Figure 3 is an effect schematic diagram of the embodiment. DETAILED DESCRIPTION

[0010] As shown in Figure 1 , the embodiment relates to a lightweight port comprehensive energy system distributed optimization method, which comprises the following steps:

[0011] Step 1, training a port logistics predictor, specifically comprising:

[0012] 1.1 Collecting the ship arrival plan and logistics equipment scheduling situation of the port.

[0013] 1.2 Based on the collected historical operation data, integrating the interactive variables of the energy system and the logistics system to construct a complete training data set.

[0014] 1.3 Taking the convolutional neural network (CNN) as the core structure to construct the port logistics predictor and training the predictor through the training data set obtained in step 1.2.

[0015] The port logistics predictor comprises a one-dimensional convolutional layer, a pooling layer, a flattening layer, a fully connected layer and an output layer, wherein: the one-dimensional convolutional layer extracts the local features of the input data and captures the key patterns in the interactive data of the energy and logistics system; the pooling layer performs down-sampling on the features, thereby reducing the data dimension and reducing the calculation complexity; the flattening layer maps the multi-dimensional features to a one-dimensional vector, providing input for the subsequent fully connected layer; the fully connected layer extracts the complex nonlinear relationship in the data; and the output layer generates the predicted value of the interactive variable of the logistics system, obtaining the required logistics system predictor.

[0016] Step 2, modeling of the distributed optimization problem of port logistics-energy system, specifically including:

[0017] 2.1 For the logistics system including shore crane equipment, ship scheduling and berth allocation links, an equipment management control model is established to optimize the scheduling and operation scheme of logistics equipment.

[0018] The equipment management control model takes the minimization of logistics system cost as the objective function, specifically: Where: c PG,t is the purchase price of electricity at time t, e PG , c CO2 are the carbon emission coefficient and carbon emission price of the power grid, P QC,t , P SR,t are the power consumption of the shore crane and shore power at time t.

[0019] The equipment management control model meets the requirements of port entry and exit time and equipment operation capacity constraints, wherein the equipment operation capacity constraints include berth allocation restrictions and shore crane task restrictions.

[0020] The berth allocation restrictions include: total length constraint of the ship when stopping non-overlapping constraint of the ship when stopping Where: h i,j and z i,j are auxiliary variables. When ship i is docked before ship j, h i,j = 1, otherwise h i,j = 0. When ship i is docked on the left side of ship j, z i,j = 1, otherwise z i,j = 0. L berth , L ship,j are the total length of the dock and the length of ship j, in meters (m). b i is the berth allocated to ship i.

[0021] The shore crane task restrictions include: quantity constraint of shore crane tasks constraint that each shore crane can only perform one task at a time constraint that shore crane tasks are executed in sequence

[0022] power consumption constraint of the shore crane power consumption constraint of the ship Where: C QC,i,j,t is the number of containers handled by shore crane j for ship i at time t, is the rated power of the shore crane, P B,j is the basic power requirement of ship j, in megawatts (MW).

[0023] 2.2 Model the common operational constraints part of the distributed solution problem, i.e. the power interaction between the logistics system and the energy system is coordinated by Lagrange multipliers, which are solved alternately in each iteration, specifically: the calculation method of the interaction variables of the logistics system and the energy system in the k+1 iteration process is as follows:

[0024] Wherein: Φ LS and Φ ES represent the interaction variables of the logistics system and the energy system respectively, L ρ (Φ LS ,Φ ES ,λ) is the augmented Lagrange function, λ is the Lagrange multiplier, and ρ is the penalty parameter.

[0025] 2.3 An energy management control model is established for photovoltaic power generation, energy storage system, gas turbine, electric boiler and shore power system involved in the port energy system.

[0026] The energy management control model aims to minimize the energy consumption cost ESS , wherein: c represents the total cost of the energy storage system, and represent the charging and discharging efficiency of the energy storage system at time t, G(Ω s,t ) is the other operation cost of the energy system under scenario s and time t, and p s is the probability of scenario s.

[0027] The energy management control model meets the energy equipment constraints and power balance constraints.

[0028] The energy equipment constraints include: power output limit of photovoltaic panel power output and energy capacity limit of energy storage system (ESS)

[0029] energy state conversion equation of ESS

[0030] initial energy state of ESS gas supply limit of gas turbine (GT) and boiler (GB) equation of gas conversion to electric energy and heat energy of GT

[0031] gas conversion to heat energy of GB power output limit of electric boiler (EB) electric energy conversion to heat energy of EB Wherein: PPV,t and These represent the photovoltaic output power and predicted available power at time t, respectively. and These represent the charging / discharging power of the energy storage system at time t; and These represent the maximum charging / discharging power of the energy storage system at time t; E ESS,t and These represent the capacity and maximum capacity of the energy storage system at time t, respectively. and For the charge / discharge efficiency of the energy storage system; G GT,t G GB,t and These represent the gas consumption and maximum gas input of the gas turbine and gas boiler at time t, respectively, both in km². 3 / t;P GT,t P GB,t P EB,t and These represent the output power and maximum output power of the gas turbine, gas boiler, and electric boiler at time t, respectively; H GT,t H GB,t and H EB,t These represent the output heat of the gas turbine, gas boiler, and electric boiler at time t, respectively, in megawatts (MW). For gas turbine power generation efficiency, and These represent the thermal efficiencies of the gas turbine, gas-fired boiler, and electric boiler, respectively; T represents the total dispatch time; and the unit for all power outputs is megawatts (MW).

[0032] The aforementioned power balance constraints include: restrictions on the supply of electricity and natural gas from the upstream power grid. Heat balance equation Natural gas balance equation Where: P G,t and G represents the electrical power purchased from the upstream power grid at time t and the maximum purchased electrical power, respectively, in megawatts (MW); G,t and The gas volume purchased from the superior gas network at time t and the maximum supply from the superior gas network. The natural gas load forecast value at time t is given, and the unit is km². 3 / t;P SR,t and P QC,t These are the shore power consumption and the terminal crane power consumption at time t, respectively, in megawatts (MW). and These are the predicted power and heat load values ​​at time t, in megawatts (MW).

[0033] Step 3, using the output of the port logistics predictor trained in step 1 As an initial solution of the local sub-problem of the logistics system, the distributed optimization problem model of the port logistics-energy system established in step 2 is iteratively solved by the ADMM algorithm, which specifically includes:

[0034] 3.1 Obtain the ship arrival and departure times and load information from the ship arrival plan, and collect scenario prediction data to provide input for the optimization solution of the logistics system and the energy system.

[0035] 3.2 Solve the local energy management control problem of the energy system by the GUROBI solver

[0036] to optimize the operation scheme of the energy system.

[0037] 3.3 In the initial stage of distributed optimization, determine whether the current iteration number is less than the set threshold (i.e., whether it is in the early iteration stage). If it is greater than the threshold, use the commercial solver to directly solve the local equipment management control problem of the logistics system to obtain the initial solution. If it is less than the threshold, call the logistics port logistics predictor trained in step 1 to quickly solve the local equipment management control problem of the logistics system, thereby accelerating the iteration process.

[0038] 3.4 After each iteration, determine whether the iteration termination condition is met, i.e., whether the primal residual and the dual residual are both less than a given value. If the condition is met, end the iteration and form the final logistics-energy system joint scheduling scheme; if the condition is not met, repeat steps 3.2 and 3.3 until the condition is met, specifically: the primal residual the dual residual wherein: when calculating the dual residual, the iteration number k should be greater than or equal to 2.

[0039] 3.5 When the residual obtained in step 3.4 is less than the given value, end the iteration. Integrate the scheduling results of the logistics system and the energy system to form the globally optimal joint scheduling scheme.

[0040] Through specific actual experiments, taking a low-carbon port as an example, through the coordinated scheduling of the energy system and the logistics system, the port ship plan is shown in Table 1, and the simulation parameter settings of the port are shown in Table 2.

[0041] Table 1

[0042]

[0043] Table 2

[0044]

[0045] Learning Optimizer Predictor Training Set Generation and ADMM Parameter Setting: To generate a dataset that can be used to train a learning optimizer predictor, a series of energy system net interaction power and ship arrival schedules are randomly set as shown in Figure 2 and the corresponding logistics system net interaction power is pre-solved using the Gurobi optimizer. Then the training set and test set are created. The solving process can be performed with a gap in the mixed integer programming of the actual distributed scheduling algorithm. Although the process of generating a single sample is time-consuming, these samples can be generated in parallel by multi-process management, thereby reducing the time required to generate the dataset. A convolutional neural network is trained using the PyTorch framework, and 600 pairs of samples are generated, with a ratio of 6:4 between the training set and the test set. The convergence threshold ∈ of ADMM is set to 0.01, and ρ is set to 5.

[0046] As shown in Figure 3 , the error distribution of the convolutional neural network learning optimizer predictor on the test set is shown, where the error is defined as the gap between the predicted operating cost of the logistics system and the operating cost calculated by the actual net interaction power. It can be seen that the well-trained convolutional neural network has most of the errors concentrated around 0, and the deviation of the predicted logistics system operating cost from the actual cost is mainly within 100 dollars, indicating the effectiveness of the convolutional neural network as a learning optimizer predictor.

[0047] As shown in Figure 3 , the convergence speed curves of the original and dual residuals of different methods are compared, and it can be seen that the initial steps of the existing ADMM iteration are usually time-consuming, but the convergence speed is fast. The CNN-ADMM solving method effectively bypasses the process of solving the logistics scheduling MIQP problem in the initial iteration, reducing the total time required for distributed optimization while ensuring convergence speed.

[0048] Compared with the prior art, the method embeds a convolutional neural network learning optimizer predictor into an ADMM optimization framework, ensuring the accuracy of the learning and the convergence of the algorithm. Compared with the prior art, this strategy improves computational efficiency without sacrificing convergence, and is suitable for large-scale port energy-logistics systems.

[0049] The above specific embodiments can be adjusted in different ways by those skilled in the art without departing from the principles and purposes of the present application, and the protection scope of the present application is subject to the claims and is not limited by the above specific embodiments, and each implementation within the scope is subject to the constraints of the present application.

Claims

1. A method for distributed optimization of a lightweight port integrated energy system, characterized in that, After the port logistics predictor is constructed and trained by historical operation data, the port logistics-energy system distributed optimization problem is modeled, and the output of the trained port logistics predictor is used as the initial solution of the local sub-problem of the logistics system in the optimization problem model, and the ADMM algorithm is used to iteratively solve the port logistics-energy system distributed optimization problem model to obtain the globally optimal joint scheduling scheme; The port logistics predictor is obtained by the following method: 1.1 Collecting the ship arrival plan and logistics equipment scheduling of the port; 1.2 Based on the collected historical operation data, integrating the interactive variables of the energy system and the logistics system to construct a complete training data set; 1.3 Using a convolutional neural network (CNN) as the core structure to build a port logistics predictor and training it by the training data set obtained in step 1.2; The port logistics-energy system distributed optimization problem is modeled, and the globally optimal joint scheduling scheme is obtained by the following method: 2.1 For the links of shore-based crane equipment, ship scheduling and berth allocation in the logistics system, establish an equipment management control model to optimize the scheduling and operation scheme of logistics equipment to minimize the cost of the logistics system To meet the ship arrival plan and equipment operation capacity constraints at the same time; 2.2 Model the common operational constraints part of the distributed solution problem, i.e. the power interaction between the logistic system and the energy system coordinated by Lagrange multipliers, by solving alternately in each iteration, in particular: the way the interaction variables are calculated for the logistic system and the energy system in the k+1 iteration is: where: and represent the interaction variables of the logistic system and the energy system, respectively, is the augmented Lagrangian function, is the Lagrange multiplier, is the penalty parameter; 2.3 For photovoltaic power generation, energy storage system, gas turbine, electric boiler and shore power system involved in port energy system, an energy management control model is established, which takes energy consumption cost Minimization as the goal, while ensuring the feasibility of energy equipment operation and dynamic balance of port energy demand; 3.1 Obtaining the ship arrival and departure time and load information from the ship arrival plan, and collecting scene prediction data to provide input for the optimization solution of the logistics system and the energy system; 3.3 In the initial stage of distributed optimization, it is judged whether the current iteration number is less than the set threshold, that is, whether it is in the early iteration stage, and when it is greater than the threshold, a commercial solver is used to directly solve the local equipment management control problem of the logistics system to obtain the initial solution, and when it is less than the threshold, the logistics port logistics predictor trained in step 1 is called to quickly solve the local equipment management control problem of the logistics system, thereby accelerating the iteration process; 3.2 Solving the local energy management control problem of the energy system by the GUROBI solver solving, optimizing the operation scheme of the energy system; 3.5 When the residual error obtained in step 3.4 is less than a given value, the iteration is ended, and the scheduling results of the logistics system and the energy system are integrated to form the globally optimal joint scheduling scheme; 3.4 After each iteration, judge whether the iteration termination condition is met, i.e. whether the primal residual and the dual residual are both less than a given value, and end the iteration to form the final joint scheduling scheme of the logistics-energy system when the condition is met; if the condition is not met, repeat steps 3.2 and 3.3 until the condition is met, which is specifically: the primal residual , the dual residual , and the iteration number should be ensured when the dual residual is calculated. The port logistics predictor comprises a one-dimensional convolutional layer, a pooling layer, a flattening layer, a fully connected layer and an output layer, wherein: the one-dimensional convolutional layer extracts local features of the input data and captures key patterns in the interactive data of the energy and logistics systems; the pooling layer down-samples the features to reduce the data dimension and reduce the computational complexity; the flattening layer maps the multi-dimensional features to a one-dimensional vector to provide input for the subsequent fully connected layer; the fully connected layer extracts complex nonlinear relationships in the data; and the output layer generates the predicted value of the interactive variables of the logistics system to obtain the required logistics system predictor; The equipment management control model satisfies the requirements of port entry and exit time and equipment operation capacity constraints, wherein the equipment operation capacity constraints include berth allocation restrictions and shore crane task restrictions; The device management control model takes minimizing the cost of the logistics system as an objective function, and specifically comprises: Wherein, is the purchase electricity price at time t, , are respectively a carbon emission coefficient and a carbon emission price of the power grid, , is the power consumption of the quay crane and the shore power at time t.

2. The lightweight port integrated energy system distributed optimization method according to claim 1, characterized in that, The energy management control model satisfies the energy equipment constraints and the power balance constraints. The aforementioned berth allocation restrictions include: overall length constraints of vessels when berthed. The constraints do not overlap when the vessel is berthed. , , ,in: , , and It is an auxiliary variable, when the ship On the ship Previously docked, ,otherwise When the ship On the ship If parked on the left, then ,otherwise , , For the total length of the pier and the ships length, For ships Length, in meters (m). It is allocated to the ship berths; The aforementioned constraints on quay crane tasks include: constraints on the number of quay crane tasks. Each quay crane can only perform one task at a time. The constraint that quay crane tasks are executed sequentially. Power consumption constraints of quay cranes Power consumption constraints of ships ,in: for shore bridge In time Ships handled The number of containers, The rated power of the quay crane, For ships The basic power requirement is expressed in megawatts (MW).

3. The lightweight port integrated energy system distributed optimization method according to claim 1, characterized in that, The energy management control model aims to minimize the total cost of energy consumption minimize the total cost of energy consumption represents the total cost of the energy storage system, and represents the charging and discharging efficiency of the energy storage system at time t, represents the other operating cost of the energy system at scenario s and time t, represents the occurrence probability of scenario s.

4. The lightweight port integrated energy system distributed optimization method according to claim 1 or 3, characterized in that, ​ The energy source device constraints include: power output limits of photovoltaic panels ; power output and energy capacity limits of energy storage systems (ESS) ; , ; energy state transition equations of ESS ; initial energy state of ESS ; gas supply limits of gas turbines (GT) and boilers (GB) , ; equations of gas conversion to electrical and thermal energy of GT , ; gas conversion to thermal energy of GB ; power output limits of electric boilers (EB) ; electrical energy conversion to thermal energy of EB , wherein: and are the output power and predicted available power of photovoltaic at time t, respectively; and are the charge / discharge power of energy storage system at time t; and are the maximum charge / discharge power of energy storage system at time t; and are the capacity and maximum capacity of energy storage system at time t, respectively; and are the charge / discharge efficiency of energy storage system; , and , are the gas consumption and maximum gas input of gas turbine, gas boiler at time t, respectively, with units of ; , and are the output power of gas turbine, gas boiler, electric boiler at time t, respectively, with units of is the maximum output power; , and are the output heat of gas turbine, gas boiler and electric boiler at time t, with units of megawatt (MW); is the gas turbine power generation efficiency, , and are the thermal efficiency of gas turbine, gas boiler and electric boiler, respectively; T is the total scheduling time length; the units of the above powers are all megawatt (MW); The power balance constraints include: upper grid power and natural gas supply limit , ; heat balance equation ; natural gas balance equation , wherein: and The maximum value of the purchased electric power and the purchased electric power from the upper grid at time t is megawatts (MW) ; and The gas purchase amount and the maximum supply amount of the upper gas grid at time t, The natural gas load prediction value at time t is , The heat load prediction value at time t is megawatts (MW).

Images