Wind field electric heavy truck cluster joint optimization scheduling method and system

Abstract

The invention belongs to the technical field of intelligent scheduling, and provides a wind field electric heavy truck cluster joint optimization scheduling method and system, and the method comprises the steps: delimiting a fan polymer and an electric heavy truck polymer, and respectively constructing a wind field power model and an electric heavy truck charging and discharging model; the method comprises the following steps of: constructing a target function by taking minimization of operation cost of a power distribution network and maximization of new energy consumption as targets, and setting a target constraint condition; and according to the wind field power model, the electric heavy truck charging and discharging model, the target function and the target constraint condition, guiding the fan polymer and the electric heavy truck polymer to carry out information interaction through a hierarchical communication topological structure to obtain scheduling decision data of the wind field electric heavy truck cluster. According to the scheme, the wind field power model and the electric heavy truck charging and discharging model are matched with the target function, the target constraint condition and the hierarchical communication topological structure, accurate model solving can be achieved, the feasibility of the whole scheduling process is higher, and the scheduling accuracy and reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent scheduling technology, and in particular to a joint optimization scheduling method and system for wind farm electric heavy truck clusters. Background Technology

[0002] Against the backdrop of energy transition and low-carbon development, the joint dispatch of wind farm clusters and electric heavy-duty truck clusters has become an important direction for improving the absorption capacity of new energy and optimizing the operating efficiency of power distribution networks. Electric heavy-duty trucks, as energy storage units with fast response, high mobility, and high flexibility, offer an innovative solution to address the fluctuations in wind turbine output and the mismatch between energy production and load demand over time. Through the joint dispatch of wind farms and electric heavy-duty truck clusters, multiple objectives can be achieved. For power distribution network operators, this reduces overall operating costs; from an environmental perspective, it promotes the low-carbon transformation of the energy system; and simultaneously, it reduces energy consumption costs for electric heavy-duty truck users through reasonable charging and discharging strategies.

[0003] However, existing solutions have several limitations in the joint scheduling of wind farm electric heavy-duty truck clusters. First, the model construction is incomplete. Existing solutions typically only use wind power-related data directly for simulation examples, making it difficult to accurately reflect the impact of complex characteristics such as wake effects and spatial distribution among wind turbines on power output. Second, the characteristics of electric heavy-duty trucks are not adequately considered. The travel chain information of electric heavy-duty trucks is not fully integrated, resulting in low adaptability of charging and discharging strategies to actual operating scenarios. At the same time, the loss of batteries during charging and discharging is ignored, affecting the rationality of the scheduling scheme. Third, the system correlation and algorithm adaptability are lacking. Existing solutions do not fully consider the collaborative relationship between the regional power grid and electric heavy-duty trucks and wind farms on a spatial scale. Moreover, single-agent deep reinforcement learning algorithms are difficult to characterize the real-world scenario of multi-agent interactions, resulting in high computational complexity and non-unique decomposition results. Fourth, the communication mechanism needs optimization. In large-scale wind farm and electric heavy-duty truck cluster scenarios, communication difficulties become a key bottleneck restricting scheduling efficiency. Traditional communication methods cannot balance communication costs with the effectiveness and accessibility of information transmission.

[0004] In summary, the existing joint scheduling scheme for electric heavy-duty truck clusters in wind farms has technical problems such as inaccuracy, unreliability, and low feasibility. Summary of the Invention

[0005] This invention provides a method and system for joint optimization scheduling of electric heavy-duty truck clusters in wind farms, which addresses the shortcomings of existing joint scheduling schemes for electric heavy-duty truck clusters in wind farms, such as inaccuracy, reliability, and low feasibility.

[0006] On one hand, this invention provides a joint optimization scheduling method for wind farm electric heavy-duty truck clusters, comprising: delineating wind turbine clusters and electric heavy-duty truck clusters, and constructing wind farm power models and electric heavy-duty truck charging and discharging models respectively; constructing an objective function with the goal of minimizing distribution network operating costs and maximizing renewable energy consumption, and setting objective constraints; and guiding the wind turbine clusters and electric heavy-duty truck clusters to interact through a hierarchical communication topology based on the wind farm power model, the electric heavy-duty truck charging and discharging model, the objective function, and the objective constraints, thereby obtaining scheduling decision data for the wind farm electric heavy-duty truck clusters.

[0007] On the other hand, the present invention also provides a joint optimization scheduling system for wind farm electric heavy-duty truck clusters, comprising: a modeling module for defining wind turbine clusters and electric heavy-duty truck clusters, and constructing wind farm power models and electric heavy-duty truck charging and discharging models respectively; a setting module for constructing an objective function and setting objective constraints with the goal of minimizing distribution network operating costs and maximizing renewable energy consumption; and a decision-making module for guiding the wind turbine clusters and the electric heavy-duty truck clusters to interact through a hierarchical communication topology based on the wind farm power model, the electric heavy-duty truck charging and discharging model, the objective function, and the objective constraints, to obtain scheduling decision data for the wind farm electric heavy-duty truck clusters.

[0008] The present invention provides a joint optimization scheduling method and system for wind farm electric heavy-duty truck clusters. This method defines wind turbine clusters and electric heavy-duty truck clusters, and constructs wind farm power models and electric heavy-duty truck charging / discharging models respectively. With the objectives of minimizing distribution network operating costs and maximizing renewable energy absorption, an objective function is constructed, and objective constraints are set. Based on the wind farm power model, electric heavy-duty truck charging / discharging model, objective function, and objective constraints, the wind turbine clusters and electric heavy-duty truck clusters are guided to interact through a hierarchical communication topology to obtain scheduling decision data for the wind farm electric heavy-duty truck clusters. This scheme uses the wind farm power model and electric heavy-duty truck charging / discharging model to coordinate the control of wind turbines in the wind farm. While maximizing wind farm output, it performs orderly charging / discharging scheduling of the electric heavy-duty truck clusters, which can improve renewable energy output, stabilize grid load, and, with the objective function, objective constraints, and hierarchical communication topology, achieve accurate model solutions, making the entire scheduling process more feasible and improving the accuracy and reliability of joint scheduling of wind farm electric heavy-duty truck clusters. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating the joint optimization scheduling method for wind farm electric heavy-duty truck clusters provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the distribution structure of the wind turbine cluster in a wind farm with ten wind turbines; Figure 3 This is a schematic diagram of the hierarchical communication topology in the wind farm and electric heavy truck aggregation system; Figure 4 This is a schematic diagram of the control framework for a wind farm and an electric heavy-duty truck cluster. Figure 5 This is a schematic diagram of the reward curve during the joint scheduling training process of electric heavy truck aggregates and wind turbine aggregates; Figure 6 This is a schematic diagram of wind speed data for a standard day with normal wind speeds; Figure 7 This is a schematic diagram showing the power output of a standard intraday wind farm and the power output of different wind turbine clusters; Figure 8 This is a schematic diagram of the fan location distribution and flow field in fan aggregate 1; Figure 9 This is a schematic diagram of the fan location distribution and flow field in fan aggregate 2; Figure 10 This is a schematic diagram of the fan location distribution and flow field in fan aggregate 3; Figure 11 This is a schematic diagram of the fan location distribution and flow field in fan aggregate 4; Figure 12 This is a schematic diagram of the power distribution of the power distribution network under normal wind speed conditions; Figure 13 This is a schematic diagram showing the relationship between standard daytime wind speed and time, where the wind speed is at its maximum. Figure 14 This is a schematic diagram showing the power output of the wind farm and the power output of different wind turbine clusters; Figure 15 This is a schematic diagram of the fan location distribution and flow field of fan aggregate 1 at a wind speed of 10.01 m / s; Figure 16 This is a schematic diagram of the fan location distribution and flow field of fan aggregate 2 at a wind speed of 10.01 m / s; Figure 17 This is a schematic diagram of the fan location distribution and flow field of fan aggregate 3 at a wind speed of 10.01 m / s; Figure 18 This is a schematic diagram of the fan location distribution and flow field of fan aggregate 4 at a wind speed of 10.01 m / s; Figure 19 This is a schematic diagram illustrating the control of the power distribution network load by a cluster of electric heavy-duty trucks under maximum wind speed conditions. Figure 20 This is a schematic diagram of the uncontrolled load and optimized load curves of the distribution network under normal and maximum wind speeds; Figure 21 This is a schematic diagram of the structure of the wind farm electric heavy truck cluster joint optimization scheduling system provided in the embodiment of the present invention. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0012] The following is combined Figures 1 to 21 This invention describes the detailed scheme of the wind farm electric heavy truck cluster joint optimization scheduling method and system provided in the embodiments of the present invention.

[0013] like Figure 1 As shown in the figure, the joint optimization scheduling method for wind farm electric heavy truck clusters provided in this embodiment of the invention mainly includes the following steps: Step 110: Define the wind turbine aggregate and the electric heavy truck aggregate, and construct the wind farm power model and the electric heavy truck charging and discharging model respectively.

[0014] In this embodiment, a wind turbine cluster refers to treating multiple wind turbines as a whole. When controlling the power output of a large-scale wind farm, dividing the wind farm into wind turbine clusters enables joint control of the wind farm. Similarly, defining electric heavy-duty truck clusters enables cluster management of large-scale electric heavy-duty trucks.

[0015] Understandably, the wind farm power model is used to characterize the power output characteristics of wind turbines in a wind farm and related information about the overall power generation capacity of the wind farm. The electric heavy-duty truck charging and discharging model is used to characterize information such as changes in battery power, time patterns, battery wear, and interaction with the power grid during the charging and discharging process of electric heavy-duty trucks.

[0016] Step 120: With the goals of minimizing the operating cost of the distribution network and maximizing the absorption of new energy sources, construct the objective function and set the objective constraints.

[0017] In this embodiment, the setting of the objective function and objective constraints can provide an important basis for subsequent model solving.

[0018] Step 130: Based on the wind farm power model, electric heavy truck charging and discharging model, objective function, and objective constraints, guide the wind turbine cluster and the electric heavy truck cluster to interact through a hierarchical communication topology to obtain the scheduling decision data of the wind farm electric heavy truck cluster.

[0019] In this embodiment, a multi-agent constrained strategy optimization algorithm based on hierarchical communication is adopted. Through the interaction process between multiple agents and the environment, the scheduling decision data of the wind farm electric heavy truck cluster can be obtained, thereby providing accurate and effective data basis for the scheduling process.

[0020] In one embodiment, defining the wind turbine aggregate and the electric heavy-duty truck aggregate includes: On the one hand, based on the spatial distribution of wind turbines and the wake effect, the wind farm is divided into multiple wind turbine clusters, each of which contains multiple wind turbines.

[0021] In this embodiment, the wind turbines in the wind farm are divided into M wind turbine clusters based on their spatial location and wake effect, and each wind turbine cluster contains N wind turbines.

[0022] On the other hand, based on the user's travel chain, the electric heavy truck cluster is used as a mobile energy storage unit to form an electric heavy truck aggregation.

[0023] In this embodiment, a large-scale electric heavy-duty truck cluster is formed into an electric heavy-duty truck agglomerate, integrating the charging and discharging characteristics of individual heavy-duty trucks, so that it can participate in the power distribution network scheduling as a whole, and can achieve coordinated response with wind farm output and load demand.

[0024] In one embodiment, constructing a wind farm power model specifically includes: First, the steady-state output power of the wake region is determined based on the factors affecting the power of a single wind turbine.

[0025] In this embodiment, the factors affecting the power of a single wind turbine include yaw angle, tilt angle, and axial sensing factor.

[0026] Then, the steady-state output power corresponding to each wind turbine aggregate is superimposed to obtain the aggregate output power.

[0027] Finally, the output power of all aggregates is superimposed to calculate the wind field output power, thus obtaining the wind field power model.

[0028] In this embodiment, the nth fan in the mth fan assembly can be represented as: Wind turbines in wind farms The coordinates are Therefore, wind turbines should be considered. Yaw angle Inclination angle and axial sensing factor The impact of wind turbines Steady-state output power in the wake region It can be represented as: (1) In the formula, Indicates air density; Indicates wind turbine The impeller area is defined as ; Indicates wind turbine Impeller plane diameter; It is a fan Wind speed ahead; power coefficient Due to axial sensing factor The empirical formula for the effect of yaw angle on power output was determined to be... replace, Obtained from statistical experimental data through fitting; tilt coefficient Power coefficient obtained by fitting experimental data. Defined as: (2) In this embodiment, when controlling the power output of a large-scale wind farm, the wind farm is divided into wind turbine clusters to achieve joint wind farm control. The architecture of the wind turbine cluster in a wind farm with ten wind turbines can be found in [reference needed]. Figure 2 ,like Figure 2 As shown, there are two high-grade wind turbines and eight low-grade wind turbines among the ten wind turbines, which are divided into three wind turbine clusters.

[0029] Except for the Mth wind turbine cluster, the Nth wind turbine in each wind turbine cluster is considered an advanced wind turbine. In other words, except for the first wind turbine cluster, the first wind turbine in each wind turbine cluster is an advanced wind turbine. Each advanced wind turbine is located between two adjacent wind turbine clusters and can transmit information between these two adjacent wind turbine clusters.

[0030] Considering the coupling relationship between wind turbine aggregates in the wind farm, the wind farm output power It can be represented as: (3) In the formula, This represents the aggregate output power of the m-th fan aggregate. ; This represents the steady-state output power of the first fan in the m-th fan assembly.

[0031] It is understandable that the first wind turbine in the m-th wind turbine aggregate... and the Nth wind turbine in the (m-1)th wind turbine aggregate. They are the same advanced wind turbines. The advanced wind turbines shared between the two wind turbine clusters are used for communication between adjacent wind turbine clusters.

[0032] Based on the wind power model presented in formula (3) above, the optimization problem describing the wind power control of the wind turbine aggregate can be expressed as: (4) The constraints are as follows: (5) (6) (7) In large-scale wind field multivariate joint control, yaw angle Inclination angle and axial sensing factor All of these adjustments were made continuously to reduce the wake effect on downstream wind turbines and maximize the wind farm's output power.

[0033] In one embodiment, constructing an electric heavy-duty truck charging and discharging model specifically includes: First, based on battery loss factors, the subsidy cost coefficient is calculated using battery cost, battery capacity, and depth of discharge, and the boundary values ​​of charge and discharge power are determined.

[0034] Then, based on the subsidy cost coefficient and the charge / discharge power boundary value, the energy interaction relationship between the electric heavy-duty truck aggregate and the external system is established, and the charge / discharge model of the electric heavy-duty truck is obtained.

[0035] In one embodiment, an objective function is constructed with the goals of minimizing distribution network operating costs and maximizing renewable energy consumption. Specifically, this function includes: On the one hand, the operating and maintenance costs of new energy power generation, the network loss costs of distribution network lines, the cost of purchasing electricity, and the dispatch subsidy costs of electric heavy trucks are summed. The difference between the summation and the revenue from selling electricity is used to obtain the operating cost of the distribution network. The minimum value of the operating cost of the distribution network is then obtained to establish the first sub-function.

[0036] On the other hand, the net load of the distribution network is determined, and the minimum value of the net load of the distribution network is obtained to establish the second sub-function.

[0037] Finally, the first and second sub-functions are used as the objective function.

[0038] In this embodiment, the objective function primarily represents the operating costs of each unit in the distribution network, mainly including the operation and maintenance costs, electricity purchase costs, network loss costs, and charging and discharging interaction costs with electric heavy trucks for each renewable energy output unit. Simultaneously, for the distribution network operator, the objective is to minimize its own operating costs (i.e., minimize expenses) and maximize the proportion of clean energy consumption, i.e., minimize net load. Therefore, the first sub-function can be expressed as: (8) In the formula, Indicates the operating cost of the power distribution network. This indicates the operation and maintenance costs of wind turbines and photovoltaic systems within the system scheduling cycle; The cost of electrical energy lost during the operation of distribution network lines is mainly obtained through power flow calculations; This represents the cost of electricity purchased by the distribution network for discharging to the upstream power grid, distributed photovoltaic power, wind power, and electric heavy trucks; This indicates the subsidy provided by the power distribution network to electric heavy-duty truck users for their cooperation in dispatching. This represents the revenue generated by the power distribution network from selling electricity to users.

[0039] The specific expressions for the above parameters are as follows: (9) (10) (11) (12) (13) In the formula, Indicates the scheduling step size; Indicates the scheduling period; , These represent the operation and maintenance coefficients for wind turbines and solar power, respectively, and are 0.01 yuan / kWh. , express The sum of wind turbine power generation and photovoltaic power generation at any given moment; This indicates the active power loss generated by the lines during the operation of the distribution network; This indicates the electricity price for power purchase transactions in the distribution network; This indicates the power purchased by the distribution network, including the power purchased from the upstream power grid and the power purchased from wind farms, photovoltaic systems, and electric heavy trucks. This indicates the electricity price for power sales transactions on the distribution network; This indicates the power sold by the power distribution network excluding the charging load of electric heavy trucks; , They represent the first An electric heavy truck assembly in The charging and discharging power during a given period; This represents the subsidy cost coefficient for electric heavy-duty truck users participating in the dispatch strategy of the distribution network.

[0040] Considering battery aging during the charging and discharging process of electric heavy-duty trucks, the subsidy cost coefficient can be specifically expressed as: (14) In the formula, This indicates the cost of purchasing batteries. This indicates the battery capacity of the electric heavy-duty truck. Indicates the depth of discharge.

[0041] The second sub-function can be represented as: (15) In the formula, It represents the net load of the distribution network and is used to measure the absorption of new energy sources; This represents the net load power of the distribution network at time t.

[0042] Furthermore, the net load power of the distribution network at time t can be specifically expressed as: (16) In the formula, Indicates the base load power; This represents the photovoltaic power generation at time t; This represents the wind turbine's power generation at time t; This represents the charging and discharging power of the electric heavy-duty truck polymer at time t.

[0043] In one embodiment, setting target constraints specifically includes: On the one hand, based on the range of purchased power, the upper and lower limits of the power seller's ramp-up, and the limitations of reserve capacity, sub-conditions for power purchase constraints in the distribution network are established.

[0044] In this embodiment, the power purchase constraint sub-conditions of the distribution network consider the power purchase range, the upper and lower limits of the power seller's ramp-up, and the downward reserve capacity constraints in the power sales direction, specifically expressed as follows: (17) (18) (19) In the formula, This indicates the real-time power purchase capacity of the power distribution network; , These represent the maximum and minimum power output of the electricity seller, respectively. , This indicates the upstream reserve power value and the downstream reserve power value of the electricity seller; , These are the upper and lower limits for the electricity seller's ramp-up.

[0045] On the other hand, based on the principle that the output of new energy sources does not exceed the maximum limit, output-reserve-ramp constraint sub-conditions are established.

[0046] In this embodiment, the output-standby-climbing constraint sub-condition can be specifically expressed as: (20) (twenty one) In the formula, express Power output of electricity sellers during specific time periods , These represent the maximum and minimum power output of the electricity seller, respectively. Indicates the reserve factor. Indicates the number of load nodes. Represents a node exist The load demand at any given moment.

[0047] On the other hand, based on the power balance equation and the fact that voltage, active power, and reactive power all meet the safety range, nodal power and voltage constraint sub-conditions are established.

[0048] In practical applications, the power distribution network should satisfy the power balance equation, that is: (twenty two) In the formula, , and They represent At any given moment, the photovoltaic power generation capacity, wind turbine power generation capacity, and power purchased by the distribution network; express The load power at any given time includes the charging power of electric heavy trucks and the base load power; express The active network loss at all times.

[0049] At the same time, the voltage and power at each node of the power grid should meet the operating limits, namely: (twenty three) (twenty four) (25) (26) (27) In the formula, Represents a node The voltage; , Representing nodes respectively The maximum and minimum voltage values; Represents a node The active power output of the power supply; , Representing nodes respectively The upper and lower limits of the active power output of the power supply; Represents a node The reactive power output of the power supply; , Representing nodes respectively The upper and lower limits of the reactive power output of the power supply; Represents a node For nodes The absolute value of the phase angle difference; Indicates from node Transmit to node The capacity; Indicates from node Transmit to node The maximum capacity.

[0050] On the other hand, based on the principle that the output of new energy sources does not exceed the maximum possible value in real time, a constraint sub-condition for the output of new energy sources is established.

[0051] In this embodiment, the power output of new energy sources mainly refers to the power output of wind turbines and photovoltaic power, which can be specifically expressed as follows: (28) (29) In the formula, express Solar power output at all times; This indicates the maximum output of photovoltaic power. for The wind turbine output at all times; This indicates the maximum output power of the fan.

[0052] Finally, the power purchase constraint sub-condition, the output-reserve-ramp constraint sub-condition, the node power and voltage constraint sub-condition, and the new energy output constraint sub-condition are used as target constraint conditions.

[0053] After obtaining the objective function and objective constraints, this embodiment uses a multi-agent constrained strategy optimization algorithm based on hierarchical communication to solve the model.

[0054] Considering the constraints of the power distribution network and the electric heavy-duty truck cluster, this embodiment models the cluster control problem of the electric heavy-duty truck cluster and the wind turbine cluster as a problem with tuples. Restricted Markov decision processes, including: state space Action space ; has conditional transition probability The transfer dynamics distribution satisfies the Markov property, i.e. Reward Space ,Right now Cost space ,Right now Cost constraint value .

[0055] exist At any given moment, the intelligent agent observes the system state. Based on this information, the intelligent agent will select an action; that is, the electric heavy-duty truck aggregator will select a charging or discharging action, thus obtaining the electric heavy-duty truck aggregator's... The wind turbine aggregate will select yaw angle, tilt angle, and axial sensing factor to maximize the wind farm's output power, based on the energy that is about to be charged or discharged. After performing the above actions, the intelligent agent can observe the new system state. , and select New actions at any moment .

[0056] Transforming the multivariate control problem into a partially observable Markov model to address... Individual agent requirements. (Targeting) For each intelligent agent, this embodiment defines a corresponding state space. Observation space and action space Among them, the state space It can describe the operating status of the power grid; observation space For the first The observation space of an agent Each intelligent agent at any time The observations are used as the current state. Part of; action space It is the first The action space of an agent. Given a state... In the case of the first Individual agents utilize strategies According to its corresponding observation Select an action from its action space.

[0057] In practical applications, the environment consists of power grid interconnections, photovoltaics, base loads, wind farms, and electric heavy-duty truck aggregators. All electric heavy-duty truck aggregators and wind turbine aggregators interact as intelligent agents. Furthermore, adjacent intelligent agents can communicate and transmit information, and wind turbine aggregators and electric heavy-duty truck aggregators can also communicate with each other.

[0058] In this embodiment, the wind turbine aggregate and the electric heavy truck aggregate together constitute an intelligent agent, which accumulates experience through repeated interactions with the power distribution network environment and gradually learns how to improve its strategies.

[0059] In large-scale distributed wind farm and electric heavy-duty truck distribution network systems, applying hierarchical communication topologies can improve the efficiency of information acquisition and full understanding for large-scale wind turbines and electric heavy-duty trucks. For example... Figure 3As shown, local information represents the power and time transmitted from a lower-level wind turbine or electric heavy-duty truck within a wind turbine aggregate or electric heavy-duty truck aggregate to a higher-level wind turbine or electric heavy-duty truck. Aggregate information represents the observations of the wind turbine aggregate or electric heavy-duty truck aggregate, specifically the local observations of the agent. Proximity information represents the information exchanged between adjacent wind turbine aggregates or electric heavy-duty truck aggregates, which is then transmitted to the lower-level wind turbine or electric heavy-duty truck within the wind turbine aggregate or electric heavy-duty truck aggregate.

[0060] Combination Figure 3 The hierarchical communication topology is established through the following process: On the one hand, lower-level nodes transmit local information to higher-level nodes, and the higher-level nodes aggregate to form aggregate perception, which serves as the aggregate topology within the group.

[0061] Under the intra-group aggregation topology, the lower-level wind turbines or electric heavy-duty trucks embed their local information into each wind turbine aggregator or electric heavy-duty truck aggregator and send it to the higher-level wind turbines or electric heavy-duty trucks. The higher-level wind turbines or electric heavy-duty trucks aggregate the information of all relevant lower-level wind turbines or electric heavy-duty trucks and obtain the perception of the wind turbine aggregator or electric heavy-duty truck aggregator.

[0062] On the other hand, information is exchanged between high-level nodes of adjacent aggregates as a shared topology between groups.

[0063] In the shared topology between groups, advanced wind turbines or electric heavy-duty trucks use wind turbine aggregators or electric heavy-duty truck aggregators to sense and communicate with adjacent advanced wind turbines or electric heavy-duty trucks. They can obtain global perception without further aggregating all the information received from other wind turbines or electric heavy-duty trucks, which can reduce the pressure on communication.

[0064] On the other hand, higher-level nodes feed back global information to lower-level nodes as a shared topology within the group.

[0065] In a shared topology within a group, each advanced wind turbine or electric heavy-duty truck shares all its features with its associated low-level wind turbine or electric heavy-duty truck, while the low-level wind turbine or electric heavy-duty truck aggregates the information received from the advanced wind turbine or electric heavy-duty truck to update the embedded features of the advanced wind turbine or electric heavy-duty truck and the low-level wind turbine or electric heavy-duty truck.

[0066] The intra-group aggregation topology, inter-group shared topology, and intra-group shared topology are used as hierarchical communication topology structures.

[0067] The above-described hierarchical communication topology does not require agents to conduct global communication and will not cause interference from unnecessary information, thus ensuring communication efficiency.

[0068] Combination Figure 4 In this embodiment, since the intelligent agent has the ability to exchange information with neighboring intelligent agents, the first... An intelligent agent in Observations at time It can be described as: (30) (31) In the formula, Indicates the first An intelligent agent in The messages received at any time represent the output power and output time of other intelligent agents.

[0069] state space Includes The observations of the i-th agent, the i-th The state at each time step is a vector. . No. An intelligent agent in state of time It can be described as: (32) According to the current strategy Given the corresponding observations, each agent selects an action from its action space. An intelligent agent at time The action, namely It can be described as: (33) (34) In the formula, the action Indicates the first An intelligent agent in Control variables and messages at specific times; Including the The yaw angle, tilt angle, and axial sensing factor of each wind turbine assembly, and the first The charging and discharging power of an electric heavy-duty truck assembly; Indicates the first An electric heavy truck assembly in The charging and discharging power at any given moment; Indicates the first An intelligent agent in Local messages sent at any time are always the first one. The output power and time of each intelligent agent.

[0070] Penalties are typically a mechanism used to constrain agent behavior, punishing undesirable actions or violations of constraints to avoid unstable or undesirable policies. In this embodiment, penalties... This includes penalties for exceeding power purchase limits in the distribution network, penalties for exceeding power and voltage limits at nodes, and penalties for exceeding energy limits in electric heavy-duty truck aggregates. If any of these constraints are not met, a penalty will be imposed. When all constraints are satisfied, .

[0071] Rewards are the goal of agent optimization. To avoid sparse reward signals during agent learning, the rewards and penalties for each agent's actions during scheduling are defined as follows: (35) In the formula, Indicates the first The reward value of each agent is determined by maximizing the reward, which means minimizing the operating cost of the distribution network, maximizing the output of new energy sources, and minimizing the penalty.

[0072] In one embodiment, the above-mentioned joint optimization scheduling method for wind farm electric heavy-duty truck clusters may further include: First, in a pre-defined simulation environment, the wind turbine aggregate and the electric heavy truck aggregate are controlled to execute scheduling decision data to obtain simulation data.

[0073] Then, based on the simulation data, the scheduling decision data is optimized and updated.

[0074] In this embodiment, the system can be based on the IEEE 33-node system, with input of known experimentally obtained wind speed and photovoltaic data, and training parameters set, such as a learning rate of 5×10⁻⁶. -4 The training rounds are 15,000, and the parallel environment accelerates sampling.

[0075] By iteratively updating strategies through algorithms, network losses can be reduced. For example, network losses can be reduced by 4.5% under normal wind speeds and by 7.2% under maximum wind speeds, thereby reducing the operating costs of the distribution network.

[0076] In one specific implementation, the scheduling decision data is optimized and updated based on simulation data, specifically including: On the one hand, based on simulation data, the output enhancement index of the wind turbine aggregate under normal wind speed and maximum wind speed scenarios is determined, and based on the output enhancement index, the adjustment range and step size of yaw angle, tilt angle and axial sensing factor in the scheduling decision data are iteratively optimized.

[0077] On the other hand, based on the simulation results of power distribution in the simulation data, the magnitude of the reduction in network loss and the change in user costs are determined, and the charging and discharging power boundaries and time windows in the scheduling decision data are adjusted according to the magnitude of the reduction in network loss and the change in user costs.

[0078] In some embodiments, in the event of a wind farm equipment failure, the faulty wind turbine cluster can be identified in real time based on a complete wind farm power model. The output gap can be compensated by dynamically adjusting parameters such as the yaw angle and tilt angle of the remaining wind turbines. At the same time, the backup discharge capacity of the electric heavy truck cluster can be called up, and its rapid response characteristics can be used to fill the power fluctuation.

[0079] In the event of a communication outage in electric heavy-duty trucks, a redundant design based on a hierarchical communication topology can be used to switch the outage cluster to a regional backup communication channel. Combined with a pre-set emergency template for the travel chain, a minimum charge / discharge strategy (such as prioritizing the maintenance of critical line load balance) can be automatically executed while ensuring controllable battery loss. An independent safety module can also block emergency actions exceeding safety boundaries. Furthermore, the distributed decision-making capabilities of multi-agent deep reinforcement learning algorithms can enable coordinated scheduling between faulty and normal areas, ensuring that the goals of power distribution network economy, carbon emission control, and cost optimization are still achieved even under abnormal conditions.

[0080] Furthermore, to further enhance the resilience of the dispatching scheme, a risk assessment strategy can be introduced. Specifically, a risk assessment model for the uncertainty of wind farm output can be established, combining historical wind speed data with real-time monitoring information to quantify the probability and impact of output deviation under different wind speed fluctuation scenarios. At the same time, a randomness risk model for electric heavy truck travel can be constructed, based on big data analysis of the travel chain to analyze the probability distribution of deviation from the preset path and assess the impact of failed charging and discharging plans on the grid balance.

[0081] In the simulation system, the upper bound of the constraint is set to 0.1. The algorithm uses the Adam optimizer to optimize the value network and the penalty value network, with a learning rate of 0.96 and a reward discount rate of 0.96. The policy network is updated iteratively 10 times using conjugate gradients and backtracking search, with a step size of 0.1. The algorithm has 15 training epochs, a pruning parameter of 0.2, a batch size of 1, an entropy coefficient of 0.01, and a value loss coefficient of 0.5.

[0082] The model was trained for 15,000 rounds, with each round consisting of 24 steps. Five parallel environments were used to accelerate the sampling process. Since generalized advantage estimation was used, the entire trajectory was required to calculate the advantage; therefore, the replay buffer size was equal to the cycle length, and the empirical replay area size was equal to the cycle length of 24. The hyperparameter discount factor and decay parameter used in generalized advantage estimation were chosen to be 0.95 and 0.99, respectively. Based on the IEEE 33-node distribution network system, a hierarchical communication multi-agent constrained policy optimization algorithm was trained. The training rounds and reward curves are shown below. Figure 5 As shown. Figure 5The reward curves for the joint scheduling training process of the electric heavy-duty truck aggregation and the wind turbine aggregation are shown. In the initial learning phase, the rewards are randomly selected as the agents accumulate experience through random exploration. As the number of training rounds increases, the learning process continuously accumulates experience, and the reward value increases, reaching a point of near-stabilization at 15,000 rounds, with the reward value remaining stable at approximately -23,744,608. The agents have learned the optimal actions for the electric heavy-duty truck aggregation and the wind turbine aggregation. The results show that the hierarchical communication multi-agent constrained policy optimization algorithm proposed in this embodiment successfully learns a policy that maximizes rewards. To test the robustness of the algorithm, the policy is tested under both maximum and normal wind speed conditions.

[0083] Depend on Figure 6 It can be seen that the wind speed fluctuates significantly throughout the day, exhibiting a clear non-periodic variation with substantial fluctuations between peaks and troughs. Furthermore, the maximum wind speed of the day, 7.13 m / s at 4 AM, is lower than the rated wind speed of 13 m / s, meaning the wind turbines cannot operate at their rated power, resulting in relatively low power generation from the wind farm. At this time, the main source of renewable energy is photovoltaic power. The wind farm output and the output of different wind turbine clusters within the standard day are as follows: Figure 7 As shown in the figure. It is understandable that the wind turbine output is positively correlated with wind speed. The output of the wind turbine cluster fluctuates significantly, and if left uncontrolled, it will cause a major impact on the power distribution network. Taking a wind speed of 7.13 m / s at point 4 as an example, the distribution of wind turbine locations and the flow field in the four wind turbine clusters are shown below. Figure 8 , Figure 9 , Figure 10 and Figure 11 As shown in the diagram. By adjusting yaw angle, tilt angle, and axial induction factor, the wind turbines in the wind farm reduce the wake effect between turbines, maximizing the wind farm's power output and reducing the net load on the distribution network. The output power of turbine cluster 1 is 157.29 kW, an increase of 51.07% compared to the uncontrolled wind farm's 104.12 kW; the output power of turbine cluster 2 is 148.4 kW, an increase of 30.55% compared to the uncontrolled wind farm's 104.12 kW; the output power of turbine cluster 3 is 119.85 kW, an increase of 40.35% compared to the uncontrolled wind farm's 85.39 kW; and the output power of turbine cluster 4 is 114.056 kW, an increase of 39.62% compared to the uncontrolled wind farm's 81.69 kW. The power distribution of the distribution network under normal wind speed conditions is shown below. Figure 12 As shown.

[0084] according to Figure 12It can be seen that the electric heavy-duty truck aggregation, acting as an energy storage station, performs charging and discharging actions based on load and new energy output. Under the condition of meeting battery energy constraints, the optimized curve is smoother and better than the uncontrolled total load curve. The hierarchical communication multi-agent constrained strategy optimization algorithm schedules the electric heavy-duty truck load by allocating the aggregator's charging demand to load valleys and discharging according to the load peak, achieving peak shaving and valley filling. During scheduling, the four electric heavy-duty truck aggregators in the hierarchical communication multi-agent constrained strategy optimization algorithm provide charging and discharging actions based on load and electric heavy-duty truck battery status. The electric heavy-duty truck aggregation discharges during peak hours (18-23) and charges during the remaining time of scheduling. After obtaining the electric heavy-duty truck aggregation load optimized by the hierarchical communication multi-agent constrained strategy optimization algorithm, the power flow is calculated to obtain network loss and voltage offset. The electric heavy-duty truck load under disordered discharge conditions is connected to the corresponding node, and the power flow is calculated to obtain network loss and voltage offset, serving as a comparison with the optimized situation after the hierarchical communication multi-agent constrained strategy optimization algorithm. Under normal wind speed conditions, the total daily active power loss before distribution network optimization was 1.047 MW, and the reactive power loss was 0.58 MVar. After dispatching, the total daily active power loss was 1 MW, a decrease of 4.5% compared to before optimization, and the total daily reactive power loss was 0.57 MVar, a decrease of 1.7% compared to before optimization.

[0085] On a standard day with the highest wind speed, the relationship between wind speed and time is as follows: Figure 13 As shown, from 3 PM to 6 PM, the wind speed is 13 m / s higher than the rated wind speed, and the fan can operate at its maximum power.

[0086] At maximum wind speed, the wind farm output and the output of different wind turbine clusters are as follows: Figure 4 As shown. By Figure 14 It can be seen that the wind speed fluctuations between peaks and valleys are greater than those under normal conditions. Furthermore, when the maximum wind speed of the day is 3 PM, exceeding the rated wind speed by 13 m / s, the wind turbines can operate at their rated power, resulting in higher power generation from the wind farm. At this time, the main source of renewable energy output is from the wind farm. The wind farm output and the output of different wind turbine clusters at maximum wind speed are shown below. Figure 13 As shown, the output of the wind turbine cluster fluctuates significantly, reaching full power operation between 16 and 18 pm. Without control, this will have a greater impact on the power distribution network, requiring a more robust dispatching strategy.

[0087] Taking a wind speed of 10.01 m / s at four points as an example, the distribution of the wind turbine locations and the flow field of the four wind turbine clusters are as follows: Figure 15 , Figure 16 , Figure 17 and Figure 18As shown, the wind turbines in the wind farm achieve maximum power by adjusting yaw angle, tilt angle, and axial sensing factor, reducing wake effects between turbines and increasing the overall power generation of the wind farm. The output power of turbine cluster 1 is 462.83kW, an increase of 42.9% compared to the uncontrolled wind farm's 323.88kW; the output power of turbine cluster 2 is 438.59kW, an increase of 35.42% compared to the uncontrolled wind farm's 323.88kW; the output power of turbine cluster 3 is 354.79kW, an increase of 33.89% compared to the uncontrolled wind farm's 264.99kW; and the output power of turbine cluster 4 is 339.40kW, an increase of 33.35% compared to the uncontrolled wind farm's 254.51kW.

[0088] Under maximum wind speed conditions, the control of the power distribution network load by the electric heavy-duty truck cluster is as follows: Figure 19 As shown, the load curve shows a significant increase between 9 and 18 hours, due to the overall increase in wind and solar power output. During this time, the electric heavy-duty truck aggregators charge, absorbing more renewable energy generation. The hierarchical communication multi-agent constrained strategy optimization algorithm schedules the electric heavy-duty truck load by allocating the aggregators' charging needs to off-peak hours and discharging according to load peaks, thus achieving peak shaving and valley filling. During scheduling, the four electric heavy-duty truck aggregators in the hierarchical communication multi-agent constrained strategy optimization algorithm provide charging and discharging actions based on the load and the electric heavy-duty truck battery status. The electric heavy-duty truck aggregators discharge during peak hours (18-23 hours) and charge during the remaining time of scheduling, reducing the load pressure on the distribution network and the load impact brought by the surge in renewable energy generation.

[0089] Under maximum wind speed conditions, the total daily active power loss before distribution network optimization was 1.189 MW, and the total daily reactive power loss was 0.74 MVar. After dispatching, the total daily active power loss was 1.103 MW, a decrease of 7.2% compared to before optimization, and the total daily reactive power loss was 0.68 MVar, a decrease of 8.1% compared to before optimization.

[0090] Under normal and maximum wind speeds, the uncontrolled and optimized load curves of the distribution network are as follows: Figure 20 As shown. According to Figure 20 The medium load curve shows the optimized load curve (i.e. Figure 20 The stability of the load (as shown by the solid line in the image) is significantly improved because the electric heavy-duty truck aggregate acts as an energy storage system, promoting peak shaving and valley filling of the load. At maximum wind speed, the load peaks between 9:00 and 18:00 because the output of new energy sources is high at this time, and the load follows the power generation side. The electric heavy-duty truck load can absorb wind power and photovoltaic power generation.

[0091] Based on time-of-use pricing parameters and grid connection prices, evaluation indicators for different strategies are calculated as shown in Table 1. All values ​​in the table are in yuan. The evaluation indicators include the total operating cost of the distribution network, network loss costs, electric heavy-duty truck subsidy costs, electric heavy-duty truck electricity sales revenue, and electric heavy-duty truck charging costs.

[0092] Table 1 Comparison of key indicators under four scenarios

[0093] As shown in Table 1, under normal wind speeds, compared to the unoptimized strategy, the optimized strategy, although increasing the subsidy cost for electric heavy-duty trucks, reduced distribution network losses by 4.49% and achieved a smaller grid load variance, resulting in a 16.49% reduction in distribution network operating costs. Furthermore, under maximum wind speeds, compared to the unoptimized strategy, the optimized strategy, achieved through the wind farm-electric heavy-duty truck aggregation, reduced distribution network operating costs by 9.97% and network losses by 7.23% while still ensuring full absorption of new energy.

[0094] However, as wind speed changes, the charging and discharging strategies of electric heavy-duty trucks will be adjusted according to the distribution network status, and the subsidy costs and charging fees for electric heavy-duty trucks will also be adjusted accordingly. Under maximum wind speed conditions, the optimized dispatching scheme can significantly reduce the subsidy costs for electric heavy-duty trucks, reduce charging costs for electric heavy-duty truck users by 14.11%, while increasing the revenue from electricity sales by electric heavy-duty trucks, improving user satisfaction, and reducing distribution network load fluctuations and economic costs.

[0095] Based on the same general inventive concept, this invention also protects a joint optimization scheduling system for wind farm electric heavy-duty truck clusters. The joint optimization scheduling system for wind farm electric heavy-duty truck clusters provided by this invention will be described below. The joint optimization scheduling system for wind farm electric heavy-duty truck clusters described below can be referred to in correspondence with the joint optimization scheduling method for wind farm electric heavy-duty truck clusters described above.

[0096] like Figure 21 As shown, the wind farm electric heavy-duty truck cluster joint optimization scheduling system provided in this embodiment of the invention specifically includes: Modeling module 210 is used to delineate the wind turbine aggregate and the electric heavy truck aggregate, and to construct the wind farm power model and the electric heavy truck charging and discharging model respectively.

[0097] The configuration module 220 is used to construct an objective function and set objective constraints with the goals of minimizing the operating cost of the distribution network and maximizing the absorption of new energy.

[0098] The decision module 230 is used to guide the wind turbine cluster and the electric heavy truck cluster to exchange information through a hierarchical communication topology based on the wind farm power model, the electric heavy truck charging and discharging model, the objective function and the objective constraints, so as to obtain the scheduling decision data of the wind farm electric heavy truck cluster.

[0099] Regarding the systems in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments of the relevant methods, and will not be elaborated further here.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A joint optimization scheduling method for electric heavy-duty truck clusters in wind farms, characterized in that, include: Delineate the wind turbine cluster and the electric heavy truck cluster, and construct wind farm power model and electric heavy truck charging and discharging model respectively; With the goals of minimizing the operating cost of the distribution network and maximizing the absorption of new energy sources, an objective function is constructed and objective constraints are set. Based on the wind farm power model, the electric heavy truck charging and discharging model, the objective function, and the objective constraints, the wind turbine cluster and the electric heavy truck cluster are guided to interact through a hierarchical communication topology to obtain scheduling decision data for the wind farm electric heavy truck cluster.

2. The joint optimization scheduling method for wind farm electric heavy-duty truck clusters according to claim 1, characterized in that, Define the wind turbine cluster and the electric heavy-duty truck cluster, including: Based on the spatial distribution of wind turbines and the wake effect, the wind farm is divided into multiple wind turbine clusters, each of which contains multiple wind turbines; Based on the user's travel chain, the electric heavy-duty truck cluster is used as a mobile energy storage unit to form an electric heavy-duty truck aggregation.

3. The joint optimization scheduling method for wind farm electric heavy-duty truck clusters according to claim 1, characterized in that, Constructing a wind farm power model includes: Determine the steady-state output power of the wake region based on the factors affecting the power of a single wind turbine; The steady-state output power corresponding to each wind turbine aggregate is superimposed to obtain the aggregate output power; The output power of all aggregates is superimposed to calculate the wind field output power, thus obtaining the wind field power model.

4. The joint optimization scheduling method for wind farm electric heavy-duty truck clusters according to claim 1, characterized in that, Constructing a charging and discharging model for electric heavy-duty trucks, including: Based on battery loss factors, the subsidy cost coefficient is calculated using battery cost, battery capacity, and depth of discharge, and the charge / discharge power boundary values ​​are determined. Based on the subsidy cost coefficient and the charging and discharging power boundary value, the energy interaction relationship between the electric heavy-duty truck aggregate and the external system is established, and the charging and discharging model of the electric heavy-duty truck is obtained.

5. The joint optimization scheduling method for wind farm electric heavy-duty truck clusters according to claim 1, characterized in that, With the objectives of minimizing distribution network operating costs and maximizing renewable energy consumption, an objective function is constructed, including: The operating cost of the distribution network is obtained by summing the operating and maintenance costs of new energy power generation, the network loss costs of distribution network lines, the cost of purchasing electricity, and the cost of dispatching subsidies for electric heavy trucks. The difference between the sum and the revenue from selling electricity is then calculated. The minimum value of the operating cost of the distribution network is then obtained, and the first sub-function is established. Determine the net load of the distribution network, find the minimum value of the net load of the distribution network, and establish the second sub-function; The first sub-function and the second sub-function are used as the target function.

6. The joint optimization scheduling method for wind farm electric heavy-duty truck clusters according to claim 1, characterized in that, Define the target constraints, including: Based on the power purchase range, the upper and lower limits of the power seller's ramp-up, and the reserve capacity constraints, establish the power purchase constraint sub-conditions for the distribution network; Based on the premise that the output of new energy sources does not exceed the maximum limit, establish output-reserve-ramp constraint sub-conditions; Based on the power balance equation and the fact that voltage, active power, and reactive power all meet the safe range, nodal power and voltage constraint subconditions are established. Based on the premise that the output of new energy sources does not exceed the real-time maximum possible value, establish a constraint sub-condition for the output of new energy sources; The power purchase constraint sub-condition of the distribution network, the output-reserve-ramp constraint sub-condition, the node power and voltage constraint sub-condition, and the new energy output constraint sub-condition are taken as target constraint conditions.

7. The joint optimization scheduling method for wind farm electric heavy-duty truck clusters according to claim 1, characterized in that, The hierarchical communication topology is established through the following process: Lower-level nodes transmit local information to higher-level nodes, which then aggregate the information to form aggregate perception, serving as the aggregate topology within the group. Information is exchanged between high-level nodes of adjacent aggregates as a shared topology between groups; Higher-level nodes feed back global information to lower-level nodes, serving as a shared topology within the group; The intra-group aggregation topology, the inter-group shared topology, and the intra-group shared topology are used as a hierarchical communication topology structure.

8. The joint optimization scheduling method for wind farm electric heavy-duty truck clusters according to claim 1, characterized in that, The method further includes: In a pre-defined simulation environment, the wind turbine assembly and the electric heavy truck assembly are controlled to execute the scheduling decision data to obtain simulation data; Based on the simulation data, the scheduling decision data is optimized and updated.

9. The joint optimization scheduling method for wind farm electric heavy-duty truck clusters according to claim 8, characterized in that, Based on the simulation data, the scheduling decision data is optimized and updated, including: Based on the simulation data, the output improvement index of the wind turbine aggregate under normal wind speed and maximum wind speed scenarios is determined, and based on the output improvement index, the adjustment range and step size of yaw angle, tilt angle and axial sensing factor in the scheduling decision data are iteratively optimized. Based on the power distribution simulation results in the simulation data, the magnitude of the reduction in network loss and the change in user costs are determined, and the charging and discharging power boundaries and time windows in the scheduling decision data are adjusted according to the magnitude of the reduction in network loss and the change in user costs.

10. A joint optimization scheduling system for electric heavy-duty truck clusters in wind farms, characterized in that, include: The modeling module is used to delineate the wind turbine aggregate and the electric heavy truck aggregate, and to build the wind farm power model and the electric heavy truck charging and discharging model respectively. The module is designed to construct an objective function and set objective constraints with the goals of minimizing the operating cost of the distribution network and maximizing the absorption of new energy sources. The decision module is used to guide the wind turbine cluster and the electric heavy truck cluster to interact through a hierarchical communication topology based on the wind farm power model, the electric heavy truck charging and discharging model, the objective function, and the objective constraints, so as to obtain the scheduling decision data of the wind farm electric heavy truck cluster.

Images