Segmented energy storage direct-hanging confluence method based on photovoltaic direct-current networking

Abstract

The invention discloses a segmented energy storage direct-hanging convergence method based on photovoltaic direct-current networking, and relates to the technical field of segmented energy storage of photovoltaic direct-current networking, and the method comprises the steps: constructing a direct-current networking system, collecting the operation data of the direct-current networking, and constructing a data set; based on the data set, dynamically adjusting the output power of a plurality of energy storage modules in real time through an adaptive power distribution algorithm, and balancing each energy storage module according to the use frequency and depth; and constructing a long-term optimization strategy of the direct-current networking system based on real-time adjustment and control of the direct-current networking system, and continuously optimizing system configuration and algorithm performance. By combining direct current networking, segmented energy storage and an artificial intelligence algorithm, the electric energy conversion efficiency is remarkably improved, the equipment loss is reduced, load variable frequency driving is realized, the system structure is simplified, and the fault rate and the equipment cost are reduced. And meanwhile, energy storage utilization is optimized, the response speed and stability of the system are improved, and access of renewable energy sources is better adapted.

Description

Technical Field

[0001] The invention relates to the technical field of photovoltaic direct current networking segmented energy storage, and in particular to a segmented energy storage direct-hanging confluence method based on photovoltaic direct current networking. Background Art

[0002] At present, the typical design of the power supply system of large thermal power units is to connect a step-down transformer at the power output of the power generation unit, which can be divided into two stages of transformation. The first stage is to step down the 20kV at the generator output to 6kV, and the second stage is to step down the 6kV to 400V. Both use industrial frequency transformers and involve AC power. With the continuous increase in the capacity of the unit, the capacity of the power supply system is also increasing. The capacity of the high-voltage power transformer is getting larger, and the short-circuit current of the power supply system is very large, which puts higher requirements on the short-circuit breaking capacity of the high-voltage switch. In order to meet the requirements of short-circuit thermal stability, the high-voltage cable of the power supply system needs to select thicker cables, which increases the investment cost.

[0003] Traditional high-voltage power plant transformers use industrial frequency transformers, which are huge in size. In order to obtain a larger capacity of the power plant system, they are often remodeled for capacity expansion. The cost of the transformation is huge, the construction period is long, and the normal power generation of the power plant is delayed. The use of DC networking in the power plant system can effectively solve some bottleneck problems in the development of traditional AC power plant systems. Compared with traditional AC power plant systems, DC networking has many advantages: DC networking has a larger power supply capacity; DC networking does not need to consider phase angle and frequency, and can realize asynchronous system interconnection. The power plant load is connected to the DC power plant system through a rectifier, which can improve the efficiency of power conversion and reduce equipment loss, realize load variable frequency drive, simplify the internal circuit of the load, and reduce the failure rate and equipment cost. Summary of the invention

[0004] In view of the problems existing in the existing segmented energy storage direct-hanging confluence method based on photovoltaic direct current networking, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is how to improve the efficiency of electric energy conversion and reduce equipment loss, realize load variable frequency drive, simplify the internal circuit of the load and reduce the failure rate and equipment cost.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, an embodiment of the present invention provides a segmented energy storage direct-hanging convergence method based on photovoltaic DC networking, which includes constructing a DC networking system, collecting DC networking operation data, and constructing a data set; based on the data set, dynamically adjusting the output power of multiple energy storage modules in real time through an adaptive power allocation algorithm, and balancing each energy storage module according to the frequency and depth of use; based on real-time regulation and control of the DC networking system, constructing a long-term optimization strategy for the DC networking system, and continuously optimizing the system configuration and algorithm performance.

[0007] As a preferred solution of the segmented energy storage direct-connection convergence method based on photovoltaic DC networking described in the present invention, wherein: the DC networking system is constructed, including a thermal power grid-connected energy storage T-type access unit and an energy storage convergence access unit; the thermal power grid-connected energy storage T-type access unit is used to connect the energy storage system to the generator outlet through a convergence energy storage bus; the energy storage convergence access unit is used to boost the energy storage and directly connect it to the generator outlet, and the energy storage system directly responds to the frequency modulation command; the energy storage convergence access unit connects multiple energy storage modules through the energy storage bus, and the energy storage modules include energy storage grid-connected switches, energy storage transformers, energy storage AC cabinets and energy storage battery modules.

[0008] As a preferred solution of the segmented energy storage direct-hanging convergence method based on photovoltaic DC networking described in the present invention, wherein: the DC networking operation data is collected, including system power data, DC bus data, energy storage module data, generator data, load and renewable energy data; a DC bus power balance equation is constructed based on the DC bus data; the DC bus power balance equation is specifically that at any time t, the sum of the output power of the energy storage module and the sum of the output power of the renewable energy is equal to the power output of the converter to the AC side, so as to maintain the power balance of the DC bus.

[0009] Constructing a DC bus voltage dynamic equation based on the DC bus power balance equation includes the following steps: Determine the DC bus equivalent capacitance; obtain the DC bus voltage at the current moment; obtain the DC bus net power according to the DC bus power balance equation; combine the DC bus equivalent capacitance, the DC bus voltage and the DC bus net power into a differential equation to obtain the rate of change of the DC bus voltage over time; integrate the rate of change of the DC bus voltage over time to obtain the change of the DC bus voltage over time.

[0010] Constructing an AC side power balance equation based on the DC bus voltage dynamic equation includes the following steps: obtaining the power output by the converter to the AC side; measuring the change in generator output power at the same time; recording the change in load demand at the same time; and obtaining the AC side power balance deviation according to the power output by the converter to the AC side, the change in generator output power and the change in load demand.

[0011] A system frequency dynamic equation is constructed based on the AC side power balance equation, including the following steps: obtaining a change in mechanical power input; calculating a difference between the change in mechanical power input and the AC side power balance deviation; and determining a rate at which the system frequency changes over time based on a system inertia constant and the difference between the change in mechanical power input and the AC side power balance deviation.

[0012] The SOC dynamic equation of the energy storage module is constructed based on the frequency dynamic equation, including the following steps: obtaining the charge and discharge efficiency of the ith energy storage module; obtaining the rated capacity of the ith energy storage module; obtaining the output power of the ith energy storage module; and calculating the rate at which the state of charge of the ith energy storage module changes with time using the output power of the ith energy storage module, the rated capacity of the ith energy storage module and the charge and discharge efficiency of the ith energy storage module.

[0013] As a preferred solution of the segmented energy storage direct-connection confluence method based on photovoltaic DC networking described in the present invention, wherein: the adaptive power allocation algorithm adopts a deep reinforcement learning framework, including state space, action space, reward function, strategy function and value function; the state space defines the current state of the system; the action space defines the action taken by the algorithm; the reward function defines the control target and guides the algorithm to learn the optimal strategy; a deep reinforcement learning model is constructed based on the strategy function and value function, wherein: a strategy network is used to select actions and parse action vectors, and the energy storage module executes action instructions to update the output power of the energy storage module, update the output power of the generator, and calculate the converter power; the control instructions are sent to each device; the strategy network is expressed as: ; in, Represents a parameterized strategy, given the state s Next select action a probability; represents the policy network parameters; represents the action vector; represents the state vector; The value function is expressed as: ; in, Indicates the current state. represents the expected value under strategy π, represents the discount factor, 0 < ≤ 1; represents the time step; Indicated in t + k Rewards at all times.

[0014] As a preferred solution of the segmented energy storage direct-hanging confluence method based on photovoltaic DC networking described in the present invention, wherein: based on the deep reinforcement learning model, the A2C algorithm is used for learning, and the policy gradient is updated, which is expressed as: ; in, Express Find the gradient, represents the objective function; represents the advantage function; the advantage function is expressed as: ; in, Indicates immediate reward, Indicates the next state The value function of Represents the next state; and then updates the value function to: ; in, represents the learning rate, 0 < ≤ 1.

[0015] As a preferred solution of the segmented energy storage direct-hanging convergence method based on photovoltaic DC networking described in the present invention, wherein: based on the adaptive power allocation algorithm, the DC networking system is regulated and controlled, including: collecting relevant data, constructing the state vector, and further obtaining the average SOC of the energy storage module; selecting actions based on the strategy network, parsing actions, and then executing actions, and sending control instructions to each device; observing the system response within a control cycle, re-collecting the system state, and calculating the reward based on the reward function; based on the A2C algorithm, the deep reinforcement learning model is learned and updated, and the DC networking system is cyclically controlled.

[0016] As a preferred solution of the segmented energy storage direct-hanging confluence method based on photovoltaic DC networking described in the present invention, the long-term optimization strategy includes energy storage module life management, system configuration optimization and algorithm continuous improvement; the energy storage module life management records the charge and discharge conditions of each energy storage module, calculates the depth of each charge and discharge, counts the number of cycles at different depths, and estimates the remaining life, which is expressed as: ; in, Indicates The estimated remaining life of each energy storage module, represents the initial life expectancy, Indicates Secondary discharge depth, Indicated in The rated number of cycles under depth; adjusting the reference state of charge of each energy storage module based on the estimated remaining life, and formulating a preventive maintenance plan; the system configuration optimization is used to minimize the total cost and meet the system performance requirements, including: by collecting and analyzing the operating data of different types of energy storage modules, evaluating the performance of various energy storage technologies in different application scenarios, among which the factors considered include: efficiency, response speed, cycle life and self-discharge rate; The optimization of energy storage module combination is expressed as: ; in, represents the total system cost, Indicates The capital cost of energy storage Indicates The operation and maintenance cost of the energy storage system; continuously improve the algorithm by collecting long-term operation data, including the operating status, environmental conditions and load conditions of various components, and constructing an offline training data set, including input features and corresponding system performance indicators; optimize key parameters, including learning rate , discount factor and reward weights; use the optimized network structure and hyperparameters to retrain the model on the offline dataset, and use the cross-validation method to evaluate the model performance.

[0017] In the second aspect, an embodiment of the present invention provides a segmented energy storage direct-hung convergence system based on photovoltaic DC networking, which includes: a construction module, which is used to construct a DC networking system, collect DC networking operation data, and construct a data set; a control module, which is used to dynamically adjust the output power of multiple energy storage modules in real time according to the data set through an adaptive power allocation algorithm, and balance each energy storage module according to the frequency and depth of use; an optimization module, which is used to construct a long-term optimization strategy for the DC networking system based on real-time regulation and control of the DC networking system, and continuously optimize the system configuration and algorithm performance.

[0018] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, any step of the above-mentioned segmented energy storage direct-hanging convergence method based on photovoltaic DC networking is implemented.

[0019] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned segmented energy storage direct-hanging confluence method based on photovoltaic DC networking is implemented.

[0020] The beneficial effect of the present invention is that the present invention significantly improves the efficiency of power conversion, reduces equipment loss, realizes load frequency conversion drive, simplifies the system structure, and reduces the failure rate and equipment cost by combining DC networking, segmented energy storage and artificial intelligence algorithms. At the same time, it optimizes the utilization of energy storage, improves the response speed and stability of the system, and better adapts to the access of renewable energy. Through intelligent management and long-term optimization strategies, not only can short-term performance improvements be achieved, but also continuous improvement of the system in long-term operation can be ensured, thereby bringing comprehensive beneficial effects in multiple aspects such as efficiency, cost, reliability and environmental adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0022] Figure 1 It is a flow chart of the segmented energy storage direct-connection method based on photovoltaic DC networking; Figure 2 This is a diagram of a T-type segmented system for direct-mounted energy storage confluence.

[0023] In the figure: 1. Thermal power grid-connected energy storage T-type access unit; 2. Energy storage convergence access unit; 1-1. #1 generator; 1-2. #1 main transformer; 1-3. 30kV power grid system; 1-4. #1GCB switch; 1-5. #2GCB switch; 1-6. #2 generator; 1-7. #2 main transformer; 2-1. Energy storage centralized step-up transformer (10kV-20kV); 2-2. 10kV energy storage busbar; 2-3. #1 energy storage grid-connected switch; 2-4. #1 energy storage transformer; 2-5. #1 energy storage converter cabinet; 2-6. #1 energy storage battery module; 2-7. #n energy storage grid-connected switch; 2-8. #n energy storage transformer; 2-9. #n energy storage converter cabinet; 2-10. #n energy storage battery module. DETAILED DESCRIPTION

[0024] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0025] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0026] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0027] Example 1, reference Figure 1 and Figure 2, which is the first embodiment of the present invention, and provides a segmented energy storage direct-hanging confluence method based on photovoltaic DC networking, including: S1: Build a DC networking system, collect DC networking operation data, and build a data set.

[0028] like Figure 2 As shown, the DC networking system includes a thermal power grid-connected energy storage T-type access unit 1 and an energy storage converging access unit 2, wherein: The generator in the thermal power grid-connected energy storage T-type access unit 1 is connected to the 330kV power grid through a step-up transformer, and the energy storage system is connected to the generator outlet using a converging energy storage bus. A large current breaking GCB device is used on the converging energy storage bus. Only one converging energy storage bus is needed to achieve the purpose of energy storage access to two thermal power units. The energy storage convergence access unit is used to boost the energy storage and directly connect it to the generator outlet. The energy storage system directly responds to the frequency regulation command.

[0029] The reference numerals are further explained, including: 1—thermal power grid-connected energy storage T-type access unit; 2—energy storage convergence access unit; 1-1—#1 generator; 1-2—#1 main transformer; 1-3—330kV power grid system; 1-4—#1GCB switch; 1-5—#2GCB switch; 1-6—#2 generator; 1-7—#2 main transformer; 2-1—energy storage centralized step-up transformer (10kV-20kV); 2-2—10kV energy storage busbar; 2-3—#1 energy storage grid-connected switch; 2-4—#1 energy storage transformer; 2-5—#1 energy storage converter cabinet; 2-6—#1 energy storage battery module; 2-7—#n energy storage grid-connected switch; 2-8—#n energy storage transformer; 2-9—#n energy storage converter cabinet; 2-10#n energy storage battery module.

[0030] Collect DC network operation data, including system power data, DC bus data, energy storage module data, generator data, load and renewable energy data; Based on the collected data, a data set is constructed to further build a DC networking system model, including: The DC bus power balance equation is constructed based on the DC bus data. Specifically, at any time t, the sum of the output power of the energy storage module and the output power of the renewable energy is equal to the power output of the converter to the AC side to maintain the power balance of the DC bus. The formula is expressed as follows: ; in, represents the DC bus net power, Indicates The output power of each energy storage module is positive for charging and negative for discharging. is the renewable energy output power, Indicates the power output by the converter to the AC side.

[0031] The DC bus voltage dynamic equation is constructed based on the DC bus power balance equation, including the following steps: determining the DC bus equivalent capacitance; obtaining the DC bus voltage at the current moment; obtaining the DC bus net power according to the DC bus power balance equation; combining the DC bus equivalent capacitance, DC bus voltage and DC bus net power into a differential equation to obtain the rate of change of the DC bus voltage over time. Integrate the rate of change of the DC bus voltage over time to obtain the change of the DC bus voltage over time. The specific formula is expressed as: ; in, is the DC bus equivalent capacitance, Indicates the DC bus voltage.

[0032] The AC side power balance equation is constructed based on the DC bus voltage dynamic equation, including the following steps: obtaining the power output by the converter to the AC side; measuring the change in the generator output power at the same time; recording the change in the load demand at the same time; and obtaining the AC side power balance deviation based on the power output by the converter to the AC side, the change in the generator output power and the change in the load demand. The specific formula is: ; in, Indicates the AC side power balance deviation, Indicates the change in generator output power, Indicates the load demand change.

[0033] The system frequency dynamic equation is constructed based on the AC side power balance equation, including the following steps: obtaining the mechanical power input change; calculating the difference between the mechanical power input change and the AC side power balance deviation; determining the rate at which the system frequency changes over time based on the system inertia constant, the difference between the mechanical power input change and the AC side power balance deviation. The specific formula is: ; in, represents the system frequency, H represents the system inertia constant, Indicates the change in mechanical power input.

[0034] The SOC dynamic equation of the energy storage module is constructed based on the frequency dynamic equation, including the following steps: obtaining the charge and discharge efficiency of the i-th energy storage module; obtaining the rated capacity of the i-th energy storage module; obtaining the output power of the i-th energy storage module; using the output power of the i-th energy storage module, the rated capacity of the i-th energy storage module and the charge and discharge efficiency of the i-th energy storage module, calculate the rate at which the state of charge of the i-th energy storage module changes over time. The specific formula is expressed as: ; in, Indicates The charge state of each energy storage module, Indicates The charging and discharging efficiency of each energy storage module is Indicates The rated capacity of each energy storage module.

[0035] S2: Based on the data set, an adaptive power allocation algorithm is constructed to dynamically adjust the output power of the energy storage system and the generator in real time, and balance each energy storage module based on the frequency and depth of use.

[0036] The adaptive power allocation algorithm is constructed using a deep reinforcement learning framework, which consists of five components: state space, action space, reward function, strategy function, and value function. Initialize relevant data parameters and initialize strategy network parameters and value functions; Collect relevant operating data, including system frequency , DC bus voltage , Energy storage module data , , generator data , load and renewable energy data; Construct a state vector, defining the current state of the system, It is expressed as: ; in, Indicates the system frequency deviation, Indicates the charging status list of all energy storage modules. Indicates the output power list of the energy storage module.

[0037] The action space defines the actions taken by the algorithm, then the action vector It is expressed as: ;

[0038] Calculating system deviation : ; Calculate DC bus voltage deviation : ; Calculate the average energy storage module SOC : ; Where N is the number of energy storage modules, represents the nominal frequency, Indicates The ideal charging state of each energy storage module is Indicates the average charge state of all energy storage modules.

[0039] Use the strategy network to select actions and parse action vectors. Execute action instructions through the energy storage module to update the output power of the energy storage module, update the output power of the generator, calculate the converter power, and send control instructions to each device.

[0040] Among them, the policy network is expressed as: ; in, represents the action vector; represents the state vector; Observe the system response within a control cycle, re-collect the system state, and calculate the reward based on the reward function, which is expressed as: ; ; ; ; ; ; in, represents the frequency deviation reward, Indicates the energy storage system status reward, represents the generator efficiency bonus, represents the DC bus voltage bonus, represents the energy storage module coordination reward, represents the nominal frequency, Indicates The ideal charging state of each energy storage module is represents the ideal output power of the generator, Indicates the DC bus reference voltage, Indicates the average state of charge of all energy storage modules; , , , Represents the weight coefficient.

[0041] The deep reinforcement learning model is constructed by using the policy function and the value function. The definition of is expressed as: ; in, represents the policy network parameters, Indicates the current state. represents the expected value under strategy π, represents the discount factor, 0 < ≤ 1; represents the time step; Indicated in t + k Rewards at all times.

[0042] Based on the deep reinforcement learning model, the A2C algorithm is used for learning and the policy gradient is updated, which is expressed as: ; in, Indicates the parameters Find the gradient, represents the objective function, that is, the expected cumulative reward, Represents a parameterized strategy, given the state s Next select action a probability; represents the advantage function.

[0043] The advantage function is expressed as: ; in, Indicates immediate reward, Indicates the next state The value function of Indicates the next state; Then the updated value function is: ; in, represents the learning rate, 0 < ≤ 1.

[0044] Multiple cycles are performed to continuously update the value function and thus update the output power of the energy storage module and the generator output power to ensure that the power scheduling reaches the current optimal level.

[0045] It should be noted that the formula is the definition of the value function, which means that from the state sAt the beginning, the expected sum of all future discounted rewards obtained by executing the policy π. This is a theoretical definition, not an actual update rule. In actual implementation, the update rule of TD learning is used to gradually approach this theoretical value. This is why this embodiment needs to use the form of incremental update: .

[0046] S3: Based on the real-time regulation and control of the DC networking system, a long-term optimization strategy for the DC networking system is constructed to continuously optimize the system configuration and algorithm performance.

[0047] Long-term optimization strategies, including energy storage module life management, system configuration optimization, and continuous algorithm improvement; The energy storage module life management records the charge and discharge conditions of each energy storage module, calculates the depth of each charge and discharge, counts the number of cycles at different depths, and estimates the remaining life, which is expressed as: ; in, Indicates The estimated remaining life of each energy storage module, represents the initial life expectancy, Indicates Secondary discharge depth, Indicated in Rated number of cycles at depth.

[0048] Adjust the reference state of charge of each energy storage module based on the estimated remaining life to develop a preventive maintenance plan; System configuration optimization is used to minimize total cost and meet system performance requirements, including: By collecting and analyzing the operating data of different types of energy storage modules, the performance of various energy storage technologies in different application scenarios is evaluated, including factors such as efficiency, response speed, cycle life and self-discharge rate; The optimization of energy storage module combination is expressed as: ; in, represents the total system cost, Indicates The capital cost of energy storage Indicates The operation and maintenance costs of this type of energy storage; Continuously improve the algorithm by collecting long-term operation data, including the operating status of various components, environmental conditions and load conditions, and constructing offline training data sets, including input features and corresponding system performance indicators; optimize key parameters, including but not limited to learning rate , discount factor and reward weights; The input characteristics include: system frequency deviation, state of charge (SOC) of each energy storage module, output power of the energy storage module, DC bus voltage, generator output power, load demand, renewable energy output power and environmental conditions.

[0049] System performance indicators include: frequency stability, voltage stability, energy storage utilization efficiency, generator operation efficiency, overall system energy loss, renewable energy consumption rate, energy storage module life expectancy and response time.

[0050] Using the optimized network structure and hyperparameters, the model is retrained on the offline dataset, and the model performance is evaluated using the cross-validation method.

[0051] If the performance of different folds in cross-validation varies greatly, it indicates that the data distribution is uneven or the model is unstable. The data distribution should be checked to ensure that each fold represents the characteristics of the overall data. The number of cross-validation folds should be increased to obtain more stable estimates, and stratified sampling methods should be used to ensure that each fold has a similar data distribution. If it is time series data, use the time series cross-validation method.

[0052] If the model performs well on the validation set but performs poorly on new data, the model has a generalization problem and the diversity of the training data should be increased, stronger regularization techniques should be used, and a simpler model structure should be used to have better generalization ability.

[0053] By systematically analyzing these possible situations and taking corresponding measures, the model can be continuously improved to improve its performance and reliability. This process is usually iterative and requires repeated attempts and adjustments until a satisfactory result is achieved.

[0054] In summary, the present invention significantly improves the efficiency of power conversion, reduces equipment loss, realizes load frequency conversion drive, simplifies the system structure, and reduces the failure rate and equipment cost by combining DC networking, segmented energy storage and artificial intelligence algorithms. At the same time, it optimizes the utilization of energy storage, improves the response speed and stability of the system, and better adapts to the access of renewable energy. Through intelligent management and long-term optimization strategies, not only can short-term performance improvements be achieved, but also continuous improvement of the system in long-term operation can be ensured, thereby bringing comprehensive beneficial effects in multiple aspects such as efficiency, cost, reliability and environmental adaptability.

[0055] Example 2, reference Figure 1On the basis of the first embodiment, this embodiment further provides a segmented energy storage direct-hung confluence system based on photovoltaic DC networking, including: a construction module, used to construct a DC networking system, collect DC networking operation data, and construct a data set; a control module, used to dynamically adjust the output power of multiple energy storage modules in real time according to the data set through an adaptive power allocation algorithm, and balance each energy storage module according to the frequency and depth of use; an optimization module, used to construct a long-term optimization strategy for the DC networking system based on real-time regulation and control of the DC networking system, and continuously optimize the system configuration and algorithm performance.

[0056] This embodiment also provides a computer device, which is suitable for the case of a segmented energy storage direct-hook convergence method based on photovoltaic DC networking, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the segmented energy storage direct-hook convergence method based on photovoltaic DC networking as proposed in the above embodiment.

[0057] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0058] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for implementing the segmented energy storage direct-connection confluence method based on photovoltaic DC networking as proposed in the above embodiment is implemented.

[0059] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0060] Embodiment 3, based on the previous two embodiments, this embodiment provides a segmented energy storage direct-hanging convergence method based on photovoltaic DC networking. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0061] The present invention is compared with the traditional model predictive control (MPC) method. The experiment is based on a simplified power system model, which includes multiple power generation and energy storage units such as wind power, photovoltaic power, thermal power and energy storage.

[0062] The experimental time was set to 7 days (168 hours) and the time step was 1 hour.

[0063] The system is controlled using the deep reinforcement learning method (DRL) of the present invention and the traditional MPC method respectively.

[0064] The DRL method uses optimized network structure and hyperparameters, including: Network structure: 3 layers of LSTM + 2 layers of fully connected layers; Learning rate: 0.001, using Adam optimizer; Discount factor: 0.99; Experience replay buffer size: 10000; Target network update frequency: every 100 steps; The MPC method uses a rolling horizon of 24 hours and a prediction step of 1 hour.

[0065] The experimental results are shown in Table 1: Table 1 Comparative test performance difference results , Table 1 clearly shows that the multi-objective optimization control method for power systems based on deep reinforcement learning proposed in the present invention is superior to the traditional MPC method in all key indicators. The DRL method has achieved remarkable results in reducing operating costs, improving the utilization rate of renewable energy, and enhancing system stability. It is particularly noteworthy that the DRL method performed well in decision-making speed and the ability to cope with emergencies, which increased by 97.8% and 57.1% respectively. These results fully demonstrate the superiority of the method of the present invention in dealing with complex modern power systems, especially when a high proportion of renewable energy is connected. This method can not only better balance multiple optimization objectives, but also has higher computational efficiency and adaptability, providing a new and effective way for the intelligent control of power systems.

[0066] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A segmented energy storage direct-hanging confluence method based on photovoltaic DC networking, characterized by: include, Build a DC networking system, collect DC networking operation data, and build a data set; Based on the data set, the output power of multiple energy storage modules is dynamically adjusted in real time through an adaptive power allocation algorithm, and each energy storage module is balanced according to the frequency and depth of use; Based on the real-time regulation and control of the DC networking system, a long-term optimization strategy for the DC networking system is constructed to continuously optimize the system configuration and algorithm performance.

2. The segmented energy storage direct-hanging converging method based on photovoltaic DC networking according to claim 1 is characterized in that: Constructing the DC networking system, including a thermal power grid-connected energy storage T-type access unit and an energy storage converging access unit; The thermal power grid-connected energy storage T-type access unit is used to connect the energy storage system to the generator outlet through the energy storage busbar; The energy storage confluence access unit is used to boost the energy storage and directly access the generator outlet, and the energy storage system directly responds to the frequency modulation command; The energy storage convergence access unit is connected to a plurality of energy storage modules via an energy storage bus, and the energy storage modules include an energy storage grid-connected switch, an energy storage transformer, an energy storage AC cabinet and an energy storage battery module.

3. The segmented energy storage direct-hanging confluence method based on photovoltaic DC networking according to claim 2 is characterized in that: Collecting the DC network operation data, including system power data, DC bus data, energy storage module data, generator data, load and renewable energy data; Constructing a DC bus power balance equation based on the DC bus data; the DC bus power balance equation is specifically that at any time t, the sum of the output power of the energy storage module and the output power of the renewable energy is equal to the power output of the converter to the AC side, so as to maintain the power balance of the DC bus; Constructing a DC bus voltage dynamic equation based on the DC bus power balance equation includes the following steps: Determine the DC bus equivalent capacitance; Get the DC bus voltage at the current moment; Obtaining the DC bus net power according to the DC bus power balance equation; The DC bus equivalent capacitance, the DC bus voltage and the DC bus net power are combined into a differential equation to obtain a rate of change of the DC bus voltage over time; Integrating the rate of change of the DC bus voltage over time to obtain a change of the DC bus voltage over time; Constructing an AC side power balance equation based on the DC bus voltage dynamic equation includes the following steps: Obtaining the power outputted by the converter to the AC side; Determine the change in generator output power at the same time; Record the load demand changes at the same time; Obtaining an AC side power balance deviation according to the power output by the converter to the AC side, a change in the generator output power, and a change in the load demand; Constructing a system frequency dynamic equation based on the AC side power balance equation includes the following steps: Obtain the change in mechanical power input; Calculating the difference between the mechanical power input variation and the AC side power balance deviation; Determining the rate at which the system frequency changes over time according to the system inertia constant, the difference between the mechanical power input change and the AC side power balance deviation; The SOC dynamic equation of the energy storage module is constructed based on the frequency dynamic equation, including the following steps: Obtain the charging and discharging efficiency of the i-th energy storage module; Obtain the rated capacity of the i-th energy storage module; Obtain the output power of the i-th energy storage module; The rate at which the state of charge of the i-th energy storage module changes with time is calculated using the output power of the i-th energy storage module, the rated capacity of the i-th energy storage module, and the charge and discharge efficiency of the i-th energy storage module.

4. The segmented energy storage direct-hanging converging method based on photovoltaic DC networking according to claim 3 is characterized in that: The adaptive power allocation algorithm adopts a deep reinforcement learning framework, including a state space, an action space, a reward function, a strategy function, and a value function; The state space defines the current state of the system; The action space defines the actions taken by the algorithm; The reward function defines the control objective and guides the algorithm to learn the optimal strategy; A deep reinforcement learning model is constructed based on the policy function and the value function, wherein: Use the strategy network to select actions and parse the action vector. The energy storage module executes the action instruction to update the output power of the energy storage module, update the output power of the generator, and calculate the converter power. The control instruction is sent to each device. The strategy network is expressed as: ； in, Represents a parameterized strategy, given the state s Next select action a probability; represents the policy network parameters; represents the action vector; represents the state vector; The definition of the value function is expressed as: ； in, Indicates the current state. represents the expected value under strategy π, represents the discount factor, 0 < ≤ 1; represents the time step; Indicated in t + k Rewards at all times.

5. The segmented energy storage direct-hanging converging method based on photovoltaic DC networking according to claim 4 is characterized in that: Based on the deep reinforcement learning model, the A2C algorithm is used for learning and the policy gradient is updated, which is expressed as: ； in, Express Find the gradient, represents the objective function; represents the advantage function; The advantage function is expressed as: ； in, Indicates immediate reward, Indicates the next state The value function of Indicates the next state; Then the updated value function is: ； in, Represents the learning rate, 0 < ≤ 1.

6. The segmented energy storage direct-hanging converging method based on photovoltaic DC networking according to claim 5 is characterized in that: The DC networking system is regulated and controlled based on the adaptive power allocation algorithm, including: Collect relevant data, construct the state vector, and further obtain the average SOC of the energy storage module; Based on the policy network, the action is selected, analyzed, and then executed, and control instructions are sent to each device; Observe the system response within a control cycle, re-collect the system state, and calculate the reward based on the reward function; Based on the A2C algorithm, the deep reinforcement learning model is learned and updated to cyclically control the DC networking system.

7. The segmented energy storage direct-hanging converging method based on photovoltaic DC networking according to claim 6 is characterized in that: The long-term optimization strategy includes energy storage module life management, system configuration optimization and continuous algorithm improvement; The energy storage module life management records the charge and discharge conditions of each energy storage module, calculates the depth of each charge and discharge, counts the number of cycles at different depths, and estimates the remaining life, which is expressed as: ； in, Indicates The estimated remaining life of each energy storage module, represents the initial life expectancy, Indicates Secondary discharge depth, Indicated in Rated number of cycles at depth; Adjust the reference state of charge of each energy storage module based on the estimated remaining life to develop a preventive maintenance plan; The system configuration is optimized to minimize total cost and meet system performance requirements, including: By collecting and analyzing the operating data of different types of energy storage modules, the performance of various energy storage technologies in different application scenarios is evaluated, including factors such as efficiency, response speed, cycle life and self-discharge rate; The optimization of energy storage module combination is expressed as: ； in, represents the total system cost, Indicates The capital cost of energy storage Indicates The operation and maintenance costs of this type of energy storage; Continuously improve the algorithm by collecting long-term operation data, including the operating status of various components, environmental conditions and load conditions, constructing offline training data sets, including input features and corresponding system performance indicators; optimizing key parameters, including learning rate , discount factor and reward weights; Using the optimized network structure and hyperparameters, the model is retrained on the offline dataset, and the model performance is evaluated using the cross-validation method.

8. A segmented energy storage direct-hanging bus system based on photovoltaic DC networking, based on the segmented energy storage direct-hanging bus method based on photovoltaic DC networking according to any one of claims 1 to 7, characterized in that: include, A construction module is used to construct a DC networking system, collect DC networking operation data, and construct a data set; The control module is used to dynamically adjust the output power of multiple energy storage modules in real time according to the data set through an adaptive power allocation algorithm, and balance each energy storage module according to the frequency and depth of use; The optimization module is used to build a long-term optimization strategy for the DC networking system based on the real-time regulation and control of the DC networking system, and continuously optimize the system configuration and algorithm performance.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the segmented energy storage direct-hanging confluence method based on photovoltaic DC networking are implemented as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the segmented energy storage direct-hanging confluence method based on photovoltaic DC networking as described in any one of claims 1 to 7 are implemented.

Images