Harbor district flexible load adjustable potential assessment method based on hybrid modeling and multi-objective optimization

Abstract

The invention discloses a harbor district flexible load adjustable potential evaluation method based on hybrid modeling and multi-objective optimization, and particularly relates to the technical field of power system optimization, comprising the steps of S1, constructing a load hybrid model fusing a physical mechanism and data driving, S2, establishing a multi-objective interval optimization model considering peak clipping and valley filling, blockage mitigation and new energy consumption, and S3, establishing a multi-objective interval optimization model considering peak clipping and valley filling, blockage mitigation and new energy consumption. The method comprises the steps of S1, obtaining a Pareto optimal solution set, S3, solving the Pareto optimal solution set by adopting an improved multi-target particle swarm algorithm, S4, carrying out real-time potential evaluation based on the solution set, and introducing a rebound effect constraint, and S5, generating a multi-main-body collaborative operation instruction through a collaborative decision-making module. According to the method, the problems that flexible load assessment is inaccurate and multi-target conflict coordination is difficult are solved, precise quantification and optimal regulation and control of harbor area flexible resource potential are realized, and the economical efficiency and new energy consumption capability of a harbor area energy system are improved.

Description

Technical Field

[0001] This invention relates to the field of power system optimization technology, and more specifically, to a method for assessing the adjustable potential of flexible loads in port areas based on hybrid modeling and multi-objective optimization. Background Technology

[0002] As a key node in energy consumption, ports are constantly improving their electrification levels. Flexible loads such as electric-driven port machinery, electric transfer vehicles, and ship shore power systems are becoming increasingly common. These loads have significant spatiotemporal adjustability characteristics and can theoretically provide services such as peak shaving and valley filling, alleviating local congestion, and promoting the consumption of new energy sources for the power grid.

[0003] However, existing methods for assessing the adjustable potential of flexible loads in port areas often suffer from limitations in modeling: pure physical mechanism models are difficult to accurately characterize the random fluctuation characteristics of loads, while pure data-driven models are difficult to maintain stable generalization ability when data is insufficient. This deficiency in modeling leads to an inaccurate characterization of load dynamics and adjustable boundaries, which cannot meet the needs of accurate potential assessment and optimized regulation under complex operating conditions. Therefore, this invention provides a method for assessing the adjustable potential of flexible loads in port areas based on hybrid modeling and multi-objective optimization. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for assessing the adjustable potential of port area flexible load based on hybrid modeling and multi-objective optimization, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for assessing the adjustable potential of flexible loads in port areas based on hybrid modeling and multi-objective optimization, comprising the following steps: S1. Construct a hybrid model of flexible load in the port area. The hybrid model includes a physical mechanism model and a data-driven model. The physical mechanism model is based on the electrical characteristics of the equipment, thermodynamic equations and operating constraints to construct a steady-state power model. The data-driven model is based on a time-series deep learning algorithm to construct a dynamic fluctuation component model. The outputs of the physical mechanism model and the data-driven model are fused by adaptive weights to form a unified load model. S2. Based on the hybrid model, a multi-objective interval optimization model is established. The multi-objective interval optimization model includes multiple conflicting objective functions and constraints. The objective functions include at least peak shaving and valley filling objectives, congestion mitigation objectives, and new energy consumption objectives. The constraints include equipment operation boundary constraints, power grid security constraints, and multi-stakeholder game equilibrium constraints. The multi-objective interval optimization model uses interval variables to represent uncertainty parameters. S3. The improved multi-objective particle swarm optimization algorithm is used to solve the multi-objective interval optimization model to obtain the Pareto optimal solution set. The improved multi-objective particle swarm optimization algorithm includes chaotic mapping initialization, dynamic inertia weight adjustment and Pareto front pruning mechanism. S4. Based on the Pareto optimal solution set, perform real-time potential assessment and feedback regulation, including a two-stage prediction-optimization framework and rebound effect constraints, to generate the adjustable potential curve and economic indicators of the port area's flexible load. S5. Based on the adjustable potential curve and economic indicators, the collaborative decision-making module generates collaborative operation instructions and power allocation schemes for power grid dispatching systems, port operation control systems and ship shore power systems.

[0006] Preferably, in step S1, the physical mechanism model is constructed for electric port machinery, electric transfer vehicles and ship shore power systems respectively. For cold chain transportation loads, the physical mechanism model includes the thermal inertia characteristics, temperature control fluctuation range and thermal buffering capacity of refrigeration equipment, and describes the load characteristics through static power characteristic curves and dynamic time series models.

[0007] Preferably, in step S1, the data-driven model uses an improved Mogrifier long short-term memory network. The input data includes historical power data, weather data, ship berthing schedules, production shift plans, and new energy output forecast data. The model captures the nonlinear dynamic fluctuations of the load through feature extraction and sequence modeling.

[0008] Preferably, in step S1, the adaptive weights are dynamically adjusted by an online learning algorithm. The online learning algorithm optimizes the weight parameters using gradient descent based on real-time prediction errors, changes in external conditions, and equipment operating status, in order to minimize model fitting errors.

[0009] Preferably, in step S2, the objective function of the multi-objective interval optimization model is specifically defined as: Peak shaving and valley filling objective: Minimize the difference between peak and valley loads, expressed as: ; Congestion mitigation objective: Minimize the risk of line overruns, expressed as: ; The target for renewable energy consumption is to minimize the renewable energy curtailment rate, expressed as: ; Where P(t) is the load power, I l (t) represents the line current, I l,max P is the upper limit of the line current. available (t) represents the available power of new energy sources, P used (t) represents the actual power used by the new energy source.

[0010] Preferably, in step S2, the interval variable is used to represent the uncertainty of new energy output, load demand and equipment performance, and the uncertainty parameter is modeled as an interval number through interval optimization theory to construct a robust optimization model to handle volatility and prediction error.

[0011] Preferably, in step S3, the chaotic mapping initialization of the improved multi-objective particle swarm optimization algorithm uses Logistic mapping to generate the initial population, the dynamic inertia weight is adaptively reduced according to the number of iterations and the convergence state, and the Pareto front pruning mechanism removes inferior solutions based on crowding distance and convergence index, thereby improving algorithm efficiency and convergence accuracy.

[0012] Preferably, in step S4, the two-stage prediction-optimization framework includes: The first stage involves predicting the upper and lower boundaries of the load power based on the hybrid model and estimating the uncertainty using confidence intervals. The second stage involves dynamically adjusting the objective weights and constraints of the multi-objective interval optimization model based on real-time power grid dispatch instructions and changes in external conditions, and then resolving the optimization problem.

[0013] Preferably, in step S4, the rebound effect constraint is defined as the upper limit of the electricity rebound rate after the load regulation ends, and the electricity rebound rate is calculated using the following formula:

[0014] Among them, P post (t) represents the load power during a specific period after the control is completed, P base (t) represents the unregulated baseline load power, T post and T base These are the durations of the period after the regulation ends and the baseline period, respectively.

[0015] Preferably, in step S5, the operation of the collaborative decision-making module is based on the Stackelberg game framework, in which the optimization objectives and constraints of the power grid company, port group and ship owner are mapped to the payoff function and strategy space of the game, and the collaborative operation instructions are automatically generated by solving the equilibrium state of the game.

[0016] The technical effects and advantages of this invention are as follows: 1. This invention deeply integrates physical mechanism models and data-driven models, which not only utilizes the interpretability and stability of physical models, but also fully leverages the ability of data-driven models to capture nonlinear and stochastic dynamic characteristics. It overcomes the limitations of single modeling and can more accurately characterize the operating characteristics and response boundaries of complex equipment such as electric machinery, cold chain loads, and shore power systems. This provides a high-precision basic model for subsequent optimization and control, and significantly improves the accuracy and reliability of potential assessment. 2. The multi-objective interval optimization model constructed in this invention and solved using an improved multi-objective particle swarm optimization algorithm can simultaneously handle multiple conflicting objectives such as power grid peak shaving, congestion mitigation, and renewable energy consumption. By introducing interval variables to characterize the uncertainty of renewable energy output and load demand, the robustness of the optimization scheme is enhanced. The final output set of Pareto optimal solutions provides decision-makers with a variety of feasible trade-offs, realizing the maximization of the potential of flexible resources and the optimization of the overall system operation efficiency in complex and uncertain environments. 3. By embedding a collaborative decision-making module based on Stackelberg game theory into the technical framework, the interests and economic behaviors of multiple stakeholders, such as power grid companies, port groups, and ship owners, are mapped into systematic technical parameters and constraints. This module can automatically generate collaborative operation instructions and power allocation schemes, realizing a closed loop from potential assessment to collaborative control. It effectively stimulates the enthusiasm of all stakeholders to participate in demand response and promotes the transformation of the port area's energy system from traditional single management to an intelligent, friendly, and interactive multi-stakeholder collaborative operation mode while ensuring the core interests of all parties. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the overall steps of the present invention.

[0018] Figure 2 This is a flowchart illustrating the overall process of the method of the present invention.

[0019] Figure 3 This is a schematic diagram of the hybrid modeling and two-stage prediction-optimization framework of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] This embodiment takes a typical coastal container port as an example and applies the port area flexible load adjustment potential assessment method based on hybrid modeling and multi-objective optimization of the present invention. The main electrification loads of the port area include: electric quay cranes, electric yard cranes, electric trucks (partially used for cold chain container transportation), and ship shore power systems. The overall implementation process of this method includes three core components: hybrid modeling, multi-objective optimization solution, and real-time evaluation and collaborative decision-making. The specific steps are as follows: S1. Construct a hybrid model of the port area's flexible load. This step aims to establish a unified model that can accurately describe the static and dynamic characteristics of the port area's flexible load. S1.1 Construction of physical mechanism model: For electrically driven port machinery (such as quay cranes and yard cranes): its steady-state power model is mainly related to the lifting load, operating speed, and mechanical efficiency. The steady-state power model is established as follows:

[0022] Where load is the lifting weight, speed is the speed of the running mechanism, and mechanical efficiency is the efficiency of the transmission system. This functional relationship can be determined through the equipment nameplate parameters and engineering manual. For electric transport vehicles (especially cold chain transport vehicles): In addition to drive power, it is necessary to focus on modeling the heat load of its refrigeration system. Steady-state power includes drive power and refrigeration power. Refrigeration power is closely related to the temperature difference between the inside and outside of the container, insulation performance, heat capacity of the goods, and the set temperature control range. At the same time, by utilizing the thermal inertia characteristics of the cold chain load, its thermal buffer capacity is defined, that is, the potential to allow temperature control to fluctuate within a certain range or to experience short-term power outages while ensuring the quality of the goods. This constitutes an important part of its adjustable potential. For ship shore power systems: model the constant power demand after connection, and consider the power consumption characteristics of ship auxiliary systems (such as refrigerated container sockets and ventilation systems). The steady-state power mainly depends on the ship type and the total capacity of electrical equipment during berthing.

[0023] S1.2 Data-Driven Model Construction To capture the random fluctuations and nonlinear characteristics of load, an improved Mogrifier Long Short-Term Memory network (an advanced recurrent neural network capable of extracting temporal features more deeply through iterative interactions between inputs and hidden states) is employed as the data-driven model. Input data includes historical power sequences, real-time meteorological data (temperature and humidity), ship berthing schedules, port operation shifts, and wind / solar power output forecasts. This network utilizes its unique multi-iterative tensor product operation to achieve deep information interaction between the input and forget gates, thereby more effectively capturing complex temporal dependencies under the influence of multiple factors and outputting dynamic fluctuation components. .

[0024] S1.3, Model Fusion The steady-state power output by the physical mechanism model Dynamic fluctuation components output by the data-driven model Weighted fusion is performed to form a unified mixed load model:

[0025] The adaptive weights are implicit in the model structure. The parameters of the fusion model are dynamically adjusted through an online learning algorithm. This algorithm uses the model prediction error within the most recent time window as feedback and uses gradient descent to iteratively optimize the parameters. The goal is to minimize the overall model fitting error, so that the model can adapt to changes in port conditions.

[0026] S2. Establish a multi-objective interval optimization model. This step is based on the hybrid model established in step S1, and constructs a multi-objective optimization model that considers uncertainties to quantify the adjustable potential of flexible loads. Specifically: The objective function is: Peak shaving and valley filling objective (F1): Minimize the difference between peak and valley loads, expressed as ; Congestion mitigation objective (F2): Minimize the risk of line overruns, expressed as ; Renewable energy consumption target (F3): Minimize the renewable energy curtailment rate, expressed as: ; Where P(t) is the load power, I l (t) represents the line current, I l,max P is the upper limit of the line current. available (t) represents the available power of new energy sources, P used (t) represents the actual power used by the new energy source; The constraints are: Equipment operating boundary constraints: including upper and lower power limits for various loads, state of charge constraints for energy storage systems, and rigid temperature control constraints for cold chain transportation, etc. Power grid security constraints include node voltage deviation and line transmission capacity limitations. Multi-agent game equilibrium constraints: implicitly achieved through the subsequent game framework, ensuring that the interests of all parties are taken into account; Uncertainty handling: Powering new energy sources Uncertain parameters such as load base demand are defined as interval variables, for example denoted by , which represents the range of possible fluctuations, an interval optimization model is constructed based on this, with the goal of seeking a robust solution set that is acceptable under all possible uncertainty scenarios.

[0027] S3. Solve using an improved multi-objective particle swarm optimization algorithm. This invention employs an improved multi-objective particle swarm optimization algorithm to efficiently solve the aforementioned complex interval multi-objective optimization model, as detailed below: Chaotic mapping initialization: The initial particle swarm is generated using the Logistic mapping. Compared with random initialization, it can obtain a more uniform and diverse initial population and avoid premature convergence. Dynamic inertia weight: Inertia weight The update rate decreases adaptively with increasing iteration count, initially encouraging global exploration and later promoting refined local search. The update formula is as follows:

[0028] Where K is the current iteration number, K max This represents the maximum number of iterations. Pareto front pruning mechanism: During the iteration process, the distribution density of individuals in the Pareto solution set is evaluated by calculating the crowding distance. While keeping the solution set size unchanged, individuals with small crowding distances (i.e. inferior solutions surrounded by other solutions in the solution space) are periodically eliminated, thereby ensuring that the final output Pareto front is evenly distributed and has wide coverage. The algorithm yields a set of Pareto optimal solutions, each representing a potential control scheme that achieves different trade-offs among multiple objectives.

[0029] S4. Real-time potential assessment and feedback control Stage Prediction-Optimization Framework: Phase 1 (Prediction): Based on the hybrid model in step S1, predict the load power in the short term (e.g., the next 24 hours) and output its upper and lower boundaries (e.g., 95% confidence interval), quantifying the uncertainty of the prediction. The second stage (optimization): The predicted boundary information is input into the interval optimization model in step S2. When a real-time power grid dispatching instruction (such as emergency peak shaving) is received or external conditions (such as changes in shipping schedules) change, the weights of each objective in the optimization model are dynamically adjusted and the solution is re-solved to generate the optimal control instruction that adapts to the new scenario. Rebound effect constraint: To prevent a new peak in electricity consumption caused by a concentrated recovery of load after the control period ends, a rebound effect constraint is introduced, defining the electricity rebound rate R. rebound The calculation formula is as follows:

[0030] Among them, P post (t) represents the load power during a specific period after the control is completed, P base(t) represents the unregulated baseline load power, T post and T base These are the durations of the period after the control measures ended and the baseline period, respectively. In the optimization model, the upper limit of the rebound rate is used as a hard constraint to ensure that the selected scheme will not lead to serious rebound problems; Finally, this step outputs an intuitive adjustable potential curve (showing when and how much power can be adjusted for the load) and economic indicators (such as energy saving benefits, subsidy benefits, etc.).

[0031] S5, Collaborative Decision Making and Instruction Generation The collaborative decision-making module of this invention operates based on the Stackelberg game framework. In this framework, the power grid company is set as the leader, whose strategy is to formulate electricity price incentive signals and subsidy policies. The port group and ship owners are set as followers, whose strategy is to adjust their own electricity consumption behavior (such as shifting operation time and adjusting charging plans) according to the incentive signals of the power grid to maximize their own benefits. By mapping the optimization objectives and operational constraints of each party to the payoff function and strategy space in the game model, multiple rounds of iterative solutions are performed until a Nash equilibrium is reached. In this equilibrium state, no party can gain more benefits by unilaterally changing its strategy. The game solution process eventually automatically generates collaborative operation instructions and power allocation schemes that can be directly issued to the power grid dispatch system (for adjusting the output plan), the port operation control system (for arranging the mechanical operation sequence), and the ship shore power system (for controlling the power supply), thereby realizing the overall optimized operation of the port area energy system.

[0032] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A port flexible load adjustable potential evaluation method based on hybrid modeling and multi-objective optimization, characterized in that: The method comprises the following steps: S1, constructing a hybrid model of port flexible load, the hybrid model comprising a physical mechanism model and a data-driven model, wherein the physical mechanism model constructs a steady-state power model based on equipment electrical characteristics, thermodynamic equations and operation constraints, the data-driven model constructs a dynamic fluctuation component model based on a time series deep learning algorithm, and the outputs of the physical mechanism model and the data-driven model are fused through adaptive weights to form a unified load model; S2, based on the hybrid model, a multi-objective interval optimization model is established, the multi-objective interval optimization model comprising a plurality of conflicting objective functions and constraint conditions, wherein the objective functions at least comprise a peak load shifting and valley filling objective, a congestion mitigation objective and a new energy consumption objective, the constraint conditions comprise equipment operation boundary constraints, power grid safety constraints and multi-agent game equilibrium constraints, and the multi-objective interval optimization model uses interval variables to represent uncertain parameters; S3, the multi-objective interval optimization model is solved by using an improved multi-objective particle swarm optimization algorithm to obtain a Pareto optimal solution set, wherein the improved multi-objective particle swarm optimization algorithm comprises a chaotic mapping initialization, a dynamic inertia weight adjustment and a Pareto front pruning mechanism; S4, based on the Pareto optimal solution set, real-time potential assessment and feedback control are performed, including a two-stage prediction-optimization framework and a rebound effect constraint, to generate an adjustable potential curve and an economic index of the port flexible load; S5, based on the adjustable potential curve and the economic index, a collaborative decision-making module is used to generate collaborative operation instructions and power distribution schemes for a power grid dispatching system, a port operation control system and a ship shore power system.

2. The method for port flexible load adjustable potential evaluation based on hybrid modeling and multi-objective optimization according to claim 1, characterized in that: In step S1, the physical mechanism model is constructed for electric-driven port machinery, electric-driven transfer vehicles and ship shore power systems respectively, wherein for cold chain transportation load, the physical mechanism model comprises thermal inertia characteristics, temperature control fluctuation range and thermal buffering capacity of refrigeration equipment, and describes load characteristics through static power characteristic curves and dynamic time series models.

3. The method for port flexible load adjustable potential evaluation based on hybrid modeling and multi-objective optimization according to claim 1, characterized in that: In step S1, the data-driven model uses an improved Mogrifier long short-term memory network, input data comprising historical power data, weather data, ship berthing schedules, production shift plans and new energy output prediction data, and captures nonlinear dynamic fluctuations of the load through feature extraction and sequence modeling.

4. The method for port flexible load adjustable potential evaluation based on hybrid modeling and multi-objective optimization according to claim 1, characterized in that: In step S1, the adaptive weight is dynamically adjusted by an online learning algorithm, the online learning algorithm based on real-time prediction error, external condition changes and equipment operation state, using gradient descent method to optimize weight parameters to minimize model fitting error.

5. The method for port flexible load adjustable potential evaluation based on hybrid modeling and multi-objective optimization according to claim 1, characterized in that: In step S2, the objective functions of the multi-objective interval optimization model are specifically defined as: Peak shaving objective: Minimize load peak valley difference, expressed as: ; Congestion mitigation targets: Minimize the risk of line overruns, expressed as: ; The new energy consumption target is: minimize the new energy curtailment rate, expressed as: ; Wherein, P(t) is the load power, I l (t) is the line current, I l,max is the line current upper limit, P available (t) is the new energy available power, P used (t) is the new energy actual use power.

6. The port flexible load adjustable potential evaluation method based on hybrid modeling and multi-objective optimization according to claim 1, characterized in that: In step S2, the interval variables are used to represent the uncertainty of new energy output, load demand and equipment performance, and the uncertainty parameters are modeled as interval numbers through interval optimization theory to construct a robust optimization model to handle volatility and prediction errors.

7. The port flexible load adjustable potential evaluation method based on hybrid modeling and multi-objective optimization according to claim 1, characterized in that: In step S3, the improved multi-objective particle swarm optimization algorithm uses the Logistic mapping to generate the initial population in the chaos mapping initialization, the dynamic inertia weight is adaptively decreased according to the iteration number and the convergence state, and the Pareto front pruning mechanism removes the inferior solutions based on the crowding distance and the convergence index, thereby improving the algorithm efficiency and the convergence accuracy.

8. The port flexibility load adjustable potential evaluation method based on hybrid modeling and multi-objective optimization according to claim 1, characterized in that: In step S4, the two-stage prediction-optimization framework includes: The first stage: predicting the upper and lower boundaries of the load power based on the mixed model and estimating the uncertainty through the confidence interval; The second stage: dynamically adjusting the objective weight and the constraint condition of the multi-objective interval optimization model according to the real-time power grid scheduling instruction and the external condition change, and re-solving the optimization problem.

9. The port flexibility load adjustable potential evaluation method based on hybrid modeling and multi-objective optimization according to claim 1, characterized in that: In step S4, the rebound effect constraint is defined as the upper limit of the electricity rebound rate after the end of the load regulation, and the electricity rebound rate calculation formula is: ； P post (t) is the load power in a specific period after the end of regulation base (t) is the reference load power without regulation post and T base are the durations of the period after the end of regulation and the reference period, respectively.

10. The method for port flexible load adjustable potential evaluation based on hybrid modeling and multi-objective optimization according to claim 1, characterized in that: In step S5, the operation of the collaborative decision module is based on the Stackelberg game framework, in which the optimization objectives and constraints of the power grid company, the port group and the ship owner are mapped into the payoff function and the strategy space of the game, and the collaborative operation instruction is automatically generated by solving the equilibrium state of the game.

Images