Distributed micro-grid system and scheduling control method thereof

Abstract

The invention belongs to the technical field of micro-grids, and particularly relates to a distributed micro-grid system and a dispatching control method thereof.The micro-grid system comprises a wind power generation device, a photovoltaic power generation device, a hydroelectric power generation device, a tidal energy power generation device and a diesel engine power generation device; the energy storage equipment comprises a storage battery energy storage device, a super capacitor energy storage device, a flywheel energy storage device and a water electrolysis hydrogen production and fuel cell energy storage device; the water electrolysis hydrogen production and fuel cell energy storage device comprises a water electrolysis device, a hydrogen storage tank and a fuel cell; the water electrolysis device is connected with the alternating current bus through an electrolysis control switch and conveys prepared hydrogen to the hydrogen storage tank for storage, and the hydrogen storage tank is connected to the fuel cell through a hydrogen control valve. A micro-grid energy optimization scheduling problem is described as a Markov decision process, and a WGAIL-based energy optimization scheduling solving algorithm is adopted, so that the stability of a training process is improved, a better strategy can be obtained, and the energy optimization scheduling of the micro-grid can be realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of microgrids, and in particular relates to a distributed microgrid system and a dispatching and control method thereof. Background Art

[0002] The power grid is an important lifeline for modern industrial production and residents' lives. The power grid is composed of power generation systems, substation systems, transmission systems, distribution systems and power consumption systems. Traditional power generation systems mainly rely on thermal power generation, which consumes huge amounts of coal energy and has serious carbon emission pollution. In addition, large centralized power grid systems are generally used. Once a failure occurs, it will have a great impact on users.

[0003] In recent years, with the development of clean new energy power generation technology, and in order to solve the current shortage of traditional fossil energy and environmental pollution problems, the proportion of renewable energy in the power grid has increased significantly, and various power generation methods such as wind power, hydropower, and photovoltaic power generation have developed comprehensively, promoting the rapid development of distributed microgrids. Microgrids, also known as microgrids, generally refer to small power generation and distribution systems composed of various forms of distributed power sources, energy storage devices, energy conversion devices, loads, monitoring and protection devices, etc.

[0004] However, since renewable energy sources such as hydropower, wind power, and photovoltaic power generation are greatly affected by natural conditions, the scope of application of microgrids is still relatively limited. In addition, with the increasing number of DC loads, a situation of coexistence of AC and DC equipment has been formed. Currently, microgrids are generally affected by uncertain factors such as volatility and intermittency, making large-scale renewable energy grid connection and consumption face huge challenges. In addition, the energy storage system is an important component and key supporting technology in the microgrid, but the current microgrid mainly relies on a single battery energy storage system, and the energy storage capacity is very limited. The energy storage system urgently needs to be further improved and perfected.

[0005] For the problem of energy optimization and scheduling of microgrids with uncertainty, the more mature methods currently include heuristic algorithms, robust optimization algorithms and model predictive control algorithms. However, each of the three algorithms has certain advantages and disadvantages. Since the large-scale grid connection of new energy sources has aggravated the uncertainty of the source-load side, it is difficult for many actual microgrid systems to obtain an accurate prediction model. Summary of the invention

[0006] In view of the above situation, the present invention provides a distributed microgrid system, which further expands the power generation types and energy storage systems of the distributed microgrid. At the same time, the scheduling control method provided in conjunction can better realize the energy scheduling optimization of the microgrid and reduce the operating cost of the microgrid.

[0007] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0008] A distributed microgrid system comprises a main grid and a distributed microgrid system. The microgrid system and the main grid can be switched on and off to achieve switching control between a grid-connected operation mode and an island operation mode.

[0009] The microgrid system includes a renewable energy power generation unit, a traditional fossil energy power generation unit, an energy storage device, a user load and an energy management controller.

[0010] The renewable energy power generation unit includes a wind power generation device, a photovoltaic power generation device, a hydropower generation device and a tidal power generation device.

[0011] Furthermore, in order for the microgrid to operate stably, the wind power generation device in the microgrid system must meet The constraints of is the maximum output power of the fan (kW);

[0012] The photovoltaic power generation device must meet the The constraints of is the maximum permissible photovoltaic output power (kW);

[0013] The traditional fossil energy power generation unit is mainly a diesel generator. In order to reduce the microgrid's dependence on the main grid, when the renewable energy power generation unit is insufficient, a diesel generator can be used to supplement the power, and its operating constraints are:

[0014]

[0015] in, and are the minimum and maximum output power of the diesel generator respectively;

[0016] The objective function of the microgrid system provided by the present invention is:

[0017]

[0018] Among them, C wt (t), C MG (t) and are the interaction costs of wind power generation, photovoltaic power generation, main power grid and the total operating cost of the mth diesel generator at time t, respectively. The expressions are as follows:

[0019]

[0020] Among them, c wt is the unit price of wind power per kilowatt (yuan / kW), c MG (t) is the real-time electricity price of the main grid at time t, a m 、b m 、cm are the coefficients of the generator respectively.

[0021] The energy storage equipment includes a battery energy storage device, a supercapacitor energy storage device, a flywheel energy storage device, a water electrolysis hydrogen production device and a fuel cell energy storage device.

[0022] The energy storage equipment constraints are:

[0023]

[0024] Among them, SOC t-1 is the state of charge of the energy storage battery in the previous period, and the sampled binary number δ t Represents the state of the battery at time t. When δ t =1 is charging state, δ t =0 is the discharge state; μ c and ρ disc are the charging efficiency and discharging efficiency, respectively. is the maximum capacity of the energy storage battery, P t bat ,ch- ,dis- is the charging and discharging power of the energy storage at time t; in order to avoid overcharging and discharging, it is also necessary to add constraints to the state of charge:

[0025] SOC min ≤SOC t ≤SOC max

[0026] Among them, SOC min and SOC max They are the minimum and maximum state of charge of the energy storage battery respectively.

[0027] The microgrid system is provided with an AC bus, which is preferably an AC380V AC bus. The renewable energy power generation unit and the traditional fossil energy power generation unit are each connected to the AC bus through a power transmission control module.

[0028] The AC bus is connected to the battery energy storage device, the supercapacitor energy storage device, the flywheel energy storage device and the fuel cell respectively through the energy storage control module.

[0029] The water electrolysis hydrogen production and fuel cell energy storage device includes a water electrolysis device, a hydrogen storage tank and a fuel cell; the water electrolysis device is connected to the AC bus through an electrolysis control switch, and the produced hydrogen is transported to the hydrogen storage tank for storage, and the hydrogen storage tank is then connected to the fuel cell through a hydrogen control valve.

[0030] Furthermore, the hydrogen control valve is an electric regulating valve, which is connected to the energy management controller via a control cable and is used to achieve regulation and control of hydrogen delivery from the hydrogen storage tank to the fuel cell.

[0031] Preferably, the water electrolysis hydrogen production and fuel cell energy storage device also includes an oxygen storage tank, which is connected to the fuel cell via an oxygen control valve. The oxygen control valve is an electric regulating valve, which is connected to the energy management controller via a control cable to achieve regulation and control of oxygen delivery from the oxygen storage tank to the fuel cell.

[0032] Furthermore, the AC busbar of the microgrid system is connected to the main grid via a grid-connected line, and a switching control switch and a transformer are provided on the grid-connected line.

[0033] The interaction power constraint between the microgrid system and the main grid is:

[0034]

[0035] in, They are respectively the lower and upper limits of the output power for interaction between the system and the main grid.

[0036] Furthermore, the AC bus of the microgrid system is connected to the user load through a power supply line, a power supply control switch is provided on the power supply line, the user load includes an AC load and a DC load, and the DC load is connected to the power supply line through a power supply control module.

[0037] Furthermore, the power transmission control module is a DC to AC inverter, the power supply control module is an AC to DC rectifier, and the energy storage control module is an AC to DC converter that can convert both AC to DC and DC to AC.

[0038] Furthermore, the energy management controller is connected to each power transmission control module, energy storage control module, power supply control module and electrolysis control switch, power supply control switch and switching control switch through a control cable.

[0039] The present invention also provides a dispatching control method based on the above-mentioned distributed microgrid system, and the specific steps are as follows:

[0040] S1. The energy optimization scheduling problem of the microgrid system is modeled as a Markov decision process. The optimal energy scheduling strategy is obtained through continuous interactive learning between the intelligent agent and the environment, and the optimization scheduling problem is transformed with the goal of minimizing economic costs and balancing energy supply and demand.

[0041] The Markov decision process consists of a four-tuple {S, A, R, P}, where S is the state space, A is the action space, R is the reward function, and P is the state transition probability;

[0042] S2. The energy optimization scheduling solution algorithm based on WGAIL is combined with the generative adversarial imitation learning method of Wasserstein distance and PPO algorithm. Its structure consists of expert strategy data, generator and discriminator. The strategy network of the intelligent agent is updated through the reinforcement learning algorithm. After the feedback iterative update of the discriminator, the optimal decision of the energy scheduling problem in the LDR scenario is finally obtained.

[0043] The expert strategy data acquisition uses a proximal strategy optimization algorithm based on the truncated objective function method to train the expert strategy model in a microgrid environment and obtain expert sample data; the generator is composed of a complete reinforcement learning framework, including an agent and an environment; the discriminator is a neural network composed of a fully connected layer, and the state-action expert sample data {s E ,a E} and state-action generation sample data {s G ,a G}, and the output is obtained after passing through the neural network;

[0044] S3. Finally, through the analysis of power data examples, the convergence stability of the energy optimization scheduling solution algorithm based on WGAIL in step S2 and its effectiveness in dealing with the microgrid energy optimization scheduling problem are verified to achieve the optimal scheduling control and energy coordination management of the distributed power sources in the microgrid.

[0045] The present invention also includes other components that enable it to be used normally, which are all conventional means in the field. In addition, devices or components not limited in the present invention, such as: wind power generation devices, photovoltaic power generation devices, hydropower generation devices, tidal power generation devices, diesel engine power generation devices, battery energy storage devices, supercapacitor energy storage devices, flywheel energy storage devices, and energy management controllers, etc., all adopt the existing technology in the field.

[0046] The beneficial effects of the present invention are as follows:

[0047] The distributed microgrid system provided by the present invention further expands the types of power generation of distributed microgrids on the basis of the common types of power generation devices in existing distributed microgrid systems, such as adding relatively advanced tidal power generation devices, and further improves the energy storage system of the microgrid, such as adding flywheel energy storage devices and electrolytic hydrogen energy storage devices. When the traditional battery energy storage device and the supercapacitor energy storage device are charged to saturation, the excess power generation energy of the distributed power source can be further converted into the kinetic energy of the flywheel for storage, and converted into hydrogen and oxygen chemical energy for storage through electrolysis of water; it can even switch to the main grid grid-connected operation mode while ensuring the power consumption of user loads within the microgrid system, exchange energy with the main grid, and upload the power generation to the main grid.

[0048] In addition, the present invention also provides a scheduling and control method that can be better applied to the distributed microgrid system, describes the microgrid energy optimization scheduling problem as a Markov decision process, and uses the WGAIL energy optimization scheduling solution algorithm combined with the generative adversarial imitation learning method of the Wasserstein distance and the PPO algorithm to improve the stability of the training process and obtain a better strategy; the convergence effect during the training process is significantly improved compared with the GAIL algorithm and the PPO algorithm; in terms of economic scheduling, it can reduce the operating cost of the microgrid system; in terms of algorithm running time, it is shorter than the other two methods, and can achieve energy optimization scheduling of the microgrid. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a schematic diagram of the structure of the distributed microgrid system in Example 1.

[0050] Figure 2 Schematic diagram of the WGAIL algorithm structure of the scheduling control method in Example 2. DETAILED DESCRIPTION

[0051] The technical solution of the present invention will be clearly and completely described below in conjunction with specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments.

[0052] Example 1

[0053] like Figure 1 As shown, a distributed microgrid system includes a main grid 22 and a distributed microgrid system. The microgrid system can be switched on and off with the main grid to achieve switching control between a grid-connected operation mode and an island operation mode.

[0054] The microgrid system includes a renewable energy power generation unit, a traditional fossil energy power generation unit, an energy storage device, a user load and an energy management controller 20.

[0055] The renewable energy power generation unit includes a wind power generation device 1 , a photovoltaic power generation device 2 , a hydropower generation device 3 and a tidal power generation device 4 .

[0056] In order for the microgrid to operate stably, the wind power generation device in the microgrid system must meet the following requirements: The constraints of is the maximum output power of the fan (kW);

[0057] The photovoltaic power generation device must meet the The constraints of is the maximum permissible photovoltaic output power (kW);

[0058] The traditional fossil energy power generation unit is mainly a diesel generator 5; in order to reduce the dependence of the microgrid on the main grid, when the output of the renewable energy power generation unit is insufficient, the diesel generator can be used to supplement the power, and its operating constraints are:

[0059]

[0060] in, and are the minimum and maximum output power of the diesel generator respectively;

[0061] The objective function of the microgrid system provided by the present invention is:

[0062]

[0063] Among them, C wt (t), C MG (t) and are the interaction costs of wind power generation, photovoltaic power generation, main power grid and the total operating cost of the mth diesel generator at time t, respectively. The expressions are as follows:

[0064]

[0065] Among them, c wt is the unit price of wind power per kilowatt (yuan / kW), c MG (t) is the real-time electricity price of the main grid at time t, a m 、b m 、c m are the coefficients of the generator respectively.

[0066] The energy storage equipment includes a battery energy storage device 6, a supercapacitor energy storage device 7, a flywheel energy storage device 8, and a water electrolysis hydrogen production and fuel cell energy storage device.

[0067] The energy storage equipment constraints are:

[0068]

[0069] Among them, SOC t-1 is the state of charge of the energy storage battery in the previous period, and the sampled binary number δ t Represents the state of the battery at time t. When δ t =1 is charging state, δ t =0 is the discharge state; μ c and ρ disc are the charging efficiency and discharging efficiency, respectively. is the maximum capacity of the energy storage battery, P t bat ,ch- ,dis-is the charging and discharging power of the energy storage at time t; in order to avoid overcharging and discharging, it is also necessary to add constraints to the state of charge:

[0070] SOC min ≤SOC t ≤SOC max

[0071] Among them, SOC min and SOC max They are the minimum and maximum state of charge of the energy storage battery respectively.

[0072] The microgrid system is provided with an AC380V AC bus, a renewable energy power generation unit and a traditional fossil energy power generation unit, each of which is connected to the AC bus through a power transmission control module.

[0073] The AC bus is connected to the battery energy storage device, the supercapacitor energy storage device, the flywheel energy storage device and the fuel cell respectively through the energy storage control module.

[0074] The water electrolysis hydrogen production and fuel cell energy storage device comprises a water electrolysis device 9, a hydrogen storage tank 10 and a fuel cell 11; the water electrolysis device is connected to the AC busbar via an electrolysis control switch 15, and the produced hydrogen is transported to the hydrogen storage tank for storage, and the hydrogen storage tank is then connected to the fuel cell via a hydrogen control valve 12. The hydrogen control valve is an electric regulating valve, which is connected to an energy management controller via a control cable, and is used to realize the regulation and control of hydrogen transport from the hydrogen storage tank to the fuel cell.

[0075] The water electrolysis hydrogen production and fuel cell energy storage device also includes an oxygen storage tank 13, which is connected to the fuel cell via an oxygen control valve 14. The oxygen control valve is an electric regulating valve, which is connected to the energy management controller via a control cable to achieve regulation and control of oxygen delivery from the oxygen storage tank to the fuel cell.

[0076] The AC busbar of the microgrid system is connected to the main grid via a grid-connected line, and a switching control switch 16 and a transformer 21 are provided on the grid-connected line.

[0077] The interaction power constraint between the microgrid system and the main grid is:

[0078]

[0079] in, They are respectively the lower and upper limits of the output power for interaction between the system and the main grid.

[0080] The AC busbar of the microgrid system is connected to the user load via a power supply line, on which a power supply control switch 17 is provided. The user load includes an AC load 18 and a DC load 19, and the DC load is connected to the power supply line via a power supply control module.

[0081] The power transmission control module is a DC to AC inverter, the power supply control module is an AC to DC rectifier, and the energy storage control module is an AC to DC converter that can convert both AC to DC and DC to AC.

[0082] The energy management controller is connected to each power transmission control module, energy storage control module, power supply control module and electrolysis control switch, power supply control switch and switching control switch through a control cable.

[0083] Example 2

[0084] like Figure 2 As shown, based on the distributed microgrid system provided in Example 1, the dispatching control method corresponding to the distributed microgrid system has the following specific steps:

[0085] S1. The energy optimization scheduling problem of the microgrid system is modeled as a Markov decision process. The optimal energy scheduling strategy is obtained through continuous interactive learning between the intelligent agent and the environment, and the optimization scheduling problem is transformed with the goal of minimizing economic costs and balancing energy supply and demand.

[0086] The Markov decision process consists of a four-tuple {S, A, R, P}, where S is the state space, A is the action space, R is the reward function, and P is the state transition probability;

[0087] The state space S is defined at time t according to the microgrid environment and is described as:

[0088]

[0089] Among them, P t pv is the photovoltaic output power at time t, P t load is the load demand power at time t, is the state of charge of the energy storage battery at time t;

[0090] The agent actions defined in action space A are to satisfy the system energy balance equation and to achieve real-time demand in load demand response scenarios by controlling the actions of wind power, diesel generators, energy storage batteries and the main power grid. The action space is defined as:

[0091]

[0092] Among them, Pt wt , and P t MG They represent the output power of wind power, diesel generator and main grid at time t, respectively, P t bat,ch-,dis- is the charging and discharging power of the energy storage battery at time t;

[0093] The purpose of energy scheduling based on LDR is to minimize electricity costs and achieve power balance between supply and demand. The reward function is constructed based on the cost of equipment power generation and the cost function of the power balance condition. The reward function at time t is defined as:

[0094]

[0095] The electricity cost at time t is defined as C(t) = C wt (t)+C MG (t)+C DG (t), is the cost function of the power balance condition not being met at time t, n1 and n2 are the weight coefficients of the two respectively; The specific description is:

[0096]

[0097] in, is the energy supply and demand relationship variable, according to χ t The positive and negative select the charging and discharging actions of the energy storage battery, n 21 and n 22 They are the energy balance reward coefficients when the system energy supply exceeds demand and when the system energy supply falls short of demand after energy storage regulation;

[0098] The expression of state transition probability P is:

[0099]

[0100] It means that the system takes action a from the state at time t k After that, the probability of transitioning to the state at time t+1.

[0101] S2. The energy optimization scheduling solution algorithm based on WGAIL is combined with the generative adversarial imitation learning method of Wasserstein distance and PPO algorithm. Its structure consists of expert strategy data, generator and discriminator. The strategy network of the intelligent agent is updated through the reinforcement learning algorithm. After the feedback iterative update of the discriminator, the optimal decision of the energy scheduling problem in the LDR scenario is finally obtained.

[0102] The expert strategy data acquisition uses a proximal strategy optimization algorithm based on the truncated objective function method to train the expert strategy model in a microgrid environment and obtain expert sample data; the generator is composed of a complete reinforcement learning framework, including an agent and an environment; the discriminator is a neural network composed of a fully connected layer, and the state-action expert sample data {s E ,a E} and state-action generation sample data {s G ,a G}, and the output is obtained after passing through the neural network;

[0103] Among them, expert experience collection imitation learning requires fitting of expert data sets, and the quality of expert data determines the success of learning; the final decision model is obtained by training in a microgrid environment through a proximal strategy optimization algorithm. The trained model is an expert strategy model, and state-action expert data samples are collected through the expert strategy model;

[0104] The generator structure is composed of a complete reinforcement learning framework. The reinforcement learning algorithm adopts a proximal policy optimization algorithm, which contains two neural networks, namely the policy network Actor and the value network Critic. The input of the policy network Actor is the state of the environment, and the output is the probability of each action. The value network Critic evaluates the current state and outputs the evaluation value. In general, the policy network and the value network have a common feature extraction layer.

[0105] Its state space is 3-dimensional and its action space is 4-dimensional, and the neural network is constructed based on this setting; specifically, the number of neurons in the input layer of the policy network Actor is the dimension of the state space, that is, the input layer has 3 neurons;

[0106] The hidden layer uses a 10-layer multilayer perceptron neural network, with 256 neurons in each layer. The activation function of the neurons in the first 8 layers is Tanh, which aims to limit the output of the neurons to (-1, 1), while the activation function of the neurons in the 9th and 10th layers is Identity, which is an identity function and has no effect on the output.

[0107] The number of neurons in the output layer is the dimension of the action space, and the output layer has 4 neurons. The value network Critic is used to evaluate the quality of the action, and its input layer contains 4 neurons.

[0108] There are 7 hidden layers, each containing 256 neurons; the activation function of the first 5 layers is Tanh, and the activation function of the 6th and 7th layers is Identity; the output layer has only 1 neuron;

[0109] The discriminator is a neural network composed of fully connected layers, including input layer, hidden layer and output layer, which is used to identify the difference between real data and generated data;

[0110] The input of the input layer is a state-action pair, so the number of neurons in the input layer is the sum of the dimensions of the state space and the action space, a total of 7 neurons;

[0111] There are 5 hidden layers, each with 100 neurons, and the activation function is Tanh; the output layer has only one neuron; in the discriminator network, the distance between the generated strategy and the expert strategy is measured using the Wasserstein distance:

[0112]

[0113] For the Wasserstein distance that satisfies the Lipschitz constraint, a gradient penalty method is proposed to increase the constraints on the discriminator network gradient, which is described as:

[0114]

[0115] In the formula, λ is taken as 10;

[0116] The objective function of the discriminator network is:

[0117]

[0118] Since most of the reward values ​​in reinforcement learning are negative, the reward function in the discriminator network is described as:

[0119] r D (s,a)=exp(D(s,a))

[0120] The reward value calculated for the generated data is fed back to the generator and stored for subsequent calculation of the advantage function.

[0121] S3. Finally, through the analysis of power data examples, the convergence stability of the energy optimization scheduling solution algorithm based on WGAIL in step S2 and its effectiveness in dealing with the microgrid energy optimization scheduling problem are verified to achieve the optimal scheduling control and energy coordination management of the distributed power sources in the microgrid.

[0122] The technical solution of the present invention is not limited to the above-mentioned specific embodiments. Without departing from the scope and spirit of the described embodiments, many modifications and changes are obvious to ordinary technicians in this technical field. Any technical deformation made within the spirit and principles of the present invention falls within the protection scope of the present invention.

Claims

1. A distributed microgrid system, characterized in that: It includes a main grid and a distributed microgrid system. The microgrid system can be switched on and off with the main grid to realize switching control between the grid-connected operation mode and the island operation mode. The microgrid system includes a renewable energy power generation unit, a traditional fossil energy power generation unit, an energy storage device, a user load and an energy management controller. The renewable energy power generation unit includes a wind power generation device, a photovoltaic power generation device, a hydropower generation device and a tidal power generation device. The traditional fossil energy power generation unit includes a diesel engine power generation device. The energy storage device includes a battery energy storage device, a supercapacitor energy storage device, a flywheel energy storage device, an electrolytic Water hydrogen production and fuel cell energy storage device; the water electrolysis hydrogen production and fuel cell energy storage device includes a water electrolysis device, a hydrogen storage tank and a fuel cell; an AC bus is provided in the microgrid system, and the renewable energy power generation unit and the traditional fossil energy power generation unit are respectively connected to the AC bus through a power transmission control module; the AC bus is respectively connected to the battery energy storage device, the supercapacitor energy storage device, the flywheel energy storage device and the fuel cell through the energy storage control module, the water electrolysis device is connected to the AC bus through an electrolysis control switch, and the produced hydrogen is transported to the hydrogen storage tank for storage, and the hydrogen storage tank is then connected to the fuel cell through a hydrogen control valve.

2. The distributed microgrid system according to claim 1, characterized in that: The AC busbar of the microgrid system is connected to the main grid via a grid-connected line, and a switching control switch and a transformer are provided on the grid-connected line.

3. The distributed microgrid system according to claim 2, characterized in that: The AC bus of the microgrid system is connected to the user load through a power supply line, and a power supply control switch is provided on the power supply line. The user load includes an AC load and a DC load, and the DC load is connected to the power supply line through a power supply control module.

4. The distributed microgrid system according to claim 3, characterized in that: The power transmission control module is a DC to AC inverter, the power supply control module is an AC to DC rectifier, and the energy storage control module is an AC to DC converter that can convert both AC to DC and DC to AC.

5. The distributed microgrid system according to claim 4, characterized in that: The AC busbar of the microgrid system is an AC380V AC busbar.

6. The distributed microgrid system according to claim 5, characterized in that: The energy management controller is connected to each power transmission control module, energy storage control module, power supply control module and electrolysis control switch, power supply control switch and switching control switch through a control cable.

7. The distributed microgrid system according to claim 6, characterized in that: The hydrogen control valve is an electric regulating valve, which is connected to the energy management controller through a control cable and is used to achieve regulation and control of hydrogen delivery from the hydrogen storage tank to the fuel cell.

8. The distributed microgrid system according to claim 7, characterized in that: The water electrolysis hydrogen production and fuel cell energy storage device also includes an oxygen storage tank, which is connected to the fuel cell via an oxygen control valve. The oxygen control valve is an electric regulating valve, which is connected to the energy management controller via a control cable to achieve regulation and control of oxygen delivery from the oxygen storage tank to the fuel cell.

9. A dispatching control method for a distributed microgrid system according to any one of claims 1 to 8, characterized in that: Here are the steps: S1. The energy optimization scheduling problem of the microgrid system is modeled as a Markov decision process. The optimal energy scheduling strategy is obtained through continuous interactive learning between the intelligent agent and the environment, and the optimization scheduling problem is transformed with the goal of minimizing economic costs and balancing energy supply and demand. The Markov decision process consists of a four-tuple {S, A, R, P}, where S is the state space, A is the action space, R is the reward function, and P is the state transition probability; S2. The energy optimization scheduling solution algorithm based on WGAIL is combined with the generative adversarial imitation learning method of Wasserstein distance and PPO algorithm. Its structure consists of expert strategy data, generator and discriminator. The strategy network of the intelligent agent is updated through the reinforcement learning algorithm. After the feedback iterative update of the discriminator, the optimal decision of the energy scheduling problem in the LDR scenario is finally obtained. The expert strategy data acquisition uses a proximal strategy optimization algorithm based on the truncated objective function method to train the expert strategy model in a microgrid environment and obtain expert sample data; the generator is composed of a complete reinforcement learning framework, including an agent and an environment; the discriminator is a neural network composed of a fully connected layer, and the state-action expert sample data {s E ,a E } and state-action generation sample data {s G ,a G }, and the output is obtained after passing through the neural network; S3. Through the analysis of power data examples, the convergence stability of the energy optimization scheduling solution algorithm based on WGAIL in step S2 and its effectiveness in dealing with the microgrid energy optimization scheduling problem are verified to achieve the optimal scheduling control and energy coordination management of the microgrid distributed power supply.

Images