Micro-grid and main grid energy interaction control method and system based on neural network

Abstract

The invention relates to the technical field of power grid energy control, in particular to a micro-grid and main power grid energy interaction control method and system based on a neural network, and provides the following scheme: constructing a regional power grid model comprising an upper layer model and a lower layer model; a regional power grid model is trained through a neural network model, and energy interaction strategies in different operation states are generated; dynamically adjusting an energy interaction strategy and updating the power grid model based on real-time monitoring data; a convolutional neural network is adopted to optimize the power flow of the main power grid, and energy loss is minimized; micro-grid energy management is optimized through a reinforcement learning model, and the operation cost is minimized; the main power grid and the micro power grid are subjected to collaborative optimization through a genetic algorithm, and an optimal energy interaction strategy is generated; the method can effectively improve the economical efficiency and energy efficiency of power grid operation, and is suitable for a complex and changeable power grid operation environment.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid energy control, and in particular to a method and system for interactive energy control between a microgrid and a main grid based on a neural network. Background Art

[0002] With the adjustment of the global energy structure and the development of renewable energy, the coordinated operation of microgrids and main grids has become a vital part of the power system. Microgrids usually include a variety of distributed energy sources such as photovoltaic power generation and wind power generation, as well as energy storage systems, load management systems and other components. The main grid is mainly responsible for centralized power generation and large-scale power transmission. When microgrids interact with the main grid, due to the volatility and uncertainty of distributed energy, traditional grid control methods are difficult to effectively cope with this multi-source, dynamic energy interaction scenario.

[0003] At present, the energy dispatch and distribution of power grids usually rely on preset rules and static optimization strategies, which are inefficient and delayed in responding to load fluctuations, changes in energy output, and dispatch of energy storage equipment. In addition, traditional methods cannot dynamically adapt to the complex energy interaction mode between microgrids and main grids. When the system is abnormal or the load increases suddenly, it is impossible to quickly adjust the energy distribution, resulting in increased power loss and even affecting the stability of the power grid.

[0004] For example, the Chinese patent with the authorization announcement number CN113690925B proposes a method and system for optimizing energy interaction based on microgrids, including: dividing the distribution network into several microgrids according to the power supply relationship between the load and the distributed energy units and energy storage devices; predicting the risk of renewable energy units; obtaining the operating costs of energy storage devices and non-renewable energy units respectively; constructing a cost objective function based on the risk prediction results and operating costs, combined with the energy interaction costs between microgrids, and determining the state constraints and power constraints of the microgrids during energy interaction; solving the cost objective function when the state constraints and power constraints are met at the same time, and adjusting the power of each microgrid during energy interaction according to the solution results. The power of purchasing and selling electricity from other microgrids is introduced into the cost objective function, which can reduce the power of energy interaction between microgrids and the grid during peak power consumption while reducing the operating costs of microgrids.

[0005] The above patents have the problems raised by this background technology: the energy interaction between the main grid and the microgrid mostly relies on preset scheduling strategies and lacks intelligent response to real-time monitoring data. Due to the complex and changeable operating environment of the power grid, traditional methods cannot quickly generate the optimal energy allocation plan when encountering load fluctuations or emergencies, resulting in a decrease in the operating efficiency of the power grid and may even cause energy imbalance. To solve the above problems, this application designs a method and system for controlling the energy interaction between the microgrid and the main grid based on a neural network. Summary of the invention

[0006] The technical problem to be solved by the present invention is to address the deficiencies of the prior art, and provide a method and system for controlling the energy interaction between a microgrid and a main grid based on a neural network, construct a regional grid model including upper and lower models; train the regional grid model through a neural network model to generate energy interaction strategies under different operating states; dynamically adjust the energy interaction strategy and update the grid model based on real-time monitoring data; use a convolutional neural network to optimize the main grid flow and minimize energy loss; optimize microgrid energy management through a reinforcement learning model to minimize operating costs; and generate the optimal energy interaction strategy by collaboratively optimizing the main grid and the microgrid through a genetic algorithm. The present invention can effectively improve the economy and energy efficiency of grid operation, and is suitable for complex and changeable grid operation environments.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for interactively controlling energy between a microgrid and a main grid based on a neural network, the method comprising:

[0009] Constructing a regional power grid model based on the information of the microgrid and the main grid in the region, wherein the regional power grid model includes a lower-layer microgrid model and an upper-layer main grid model;

[0010] Constructing a neural network model, training the regional power grid model through the neural network model, and obtaining power grid energy interaction strategies under different operating conditions;

[0011] Acquire real-time monitoring data of the microgrid and the main grid, adjust the grid energy interaction strategy according to the real-time monitoring data, and update the regional grid model.

[0012] The constructing of the regional power grid model comprises:

[0013] Collect multi-source data related to microgrids and main grids in the region;

[0014] Constructing an upper main grid model according to the topological structure of the main grid, wherein the upper main grid model includes power generation, transmission and dispatching functions of the main grid;

[0015] Constructing a lower-layer microgrid model according to the multi-source information of each microgrid, wherein the lower-layer microgrid model includes a sub-model of a single microgrid, and the sub-model includes a microgrid energy storage device, a local area load, and a microgrid power generation device;

[0016] The upper main grid model also includes a distribution network control center, which is used to interact with the lower microgrid model, obtain lower-level demands and lower-level information, and distribute energy with the main grid according to the lower-level demands and lower-level information.

[0017] The neural network model comprises:

[0018] The main grid optimization layer is used to optimize the energy allocation scheme of the main grid according to the upper main grid model, under the constraint condition of satisfying the main grid power flow balance, with the goal of achieving the main grid optimal power flow, and output the minimum energy loss scheme;

[0019] The microgrid optimization layer is used to optimize the energy allocation scheme of the microgrid and output the energy optimization management scheme based on the lower-layer microgrid model and under the condition that the electrical boundary conditions of the microgrid are met, with the goal of minimizing the operating cost of the microgrid;

[0020] The double-layer optimization layer is used to process the coordinated optimization of the main grid and the microgrid, update the fitness of the energy minimum loss scheme and the energy optimization management scheme through a genetic algorithm, and output a grid energy interaction strategy.

[0021] The optimization of the energy distribution scheme of the main power grid includes:

[0022] Processing multi-source data of the main power grid to extract an input parameter sequence, wherein the multi-source data includes node voltage, branch power flow, distributed power output and load demand;

[0023] Generate a power flow allocation scheme through a convolutional neural network, set the objective function of the main power grid optimization to minimize the active power loss in the main power grid power flow, and calculate whether any node in the main power grid satisfies the power flow balance equation according to the power flow allocation scheme;

[0024] If any node satisfies the power distribution equation, the active power of each node is counted and the energy minimum loss plan is output. If there are nodes that do not satisfy the power distribution equation, the nodes that do not satisfy the power distribution equation are further analyzed to identify the key parameters that cause imbalance. The key parameters are adjusted through the back propagation mechanism to generate a new power distribution plan and make a new judgment.

[0025] The optimization of the energy distribution scheme of the microgrid includes:

[0026] According to the lower-layer microgrid model, the renewable energy generation power, user load demand and energy storage status in the microgrid are collected to generate a multi-dimensional input sequence;

[0027] Constructing a reinforcement learning model, training the multidimensional input sequence through the reinforcement learning model, and outputting an energy allocation scheme for the microgrid, wherein the reinforcement learning model includes a gated recurrent unit and an A3C algorithm;

[0028] Determine whether the generated energy allocation scheme meets the electrical boundary conditions of the microgrid, wherein the electrical boundary conditions include the charging and discharging power limit of the energy storage device, the load demand satisfaction and the power transaction limit with the main grid;

[0029] If the electrical boundary conditions are met, the energy allocation plan is output as an energy optimization management plan. If the electrical boundary conditions are not met, the network weights of the reinforcement learning model are updated through a back-propagation mechanism to generate a new energy allocation plan until the electrical boundary conditions are met.

[0030] The double-layer optimization layer comprises:

[0031] The energy minimum loss scheme of the upper main grid and the energy optimization management scheme of the lower microgrid are jointly optimized through genetic algorithms;

[0032] In each iteration of the genetic algorithm, the energy optimization management scheme of the microgrid is first optimized through the lower-level microgrid model, and then the results are fed back to the upper-level main grid model for power flow optimization. Through the collaboration of the two layers, the energy interaction strategy between the main grid and the microgrid is output;

[0033] When the genetic algorithm converges, the optimal energy interaction strategy between the main grid and the microgrid is output.

[0034] The genetic algorithm comprises:

[0035] According to the energy interaction relationship between the energy minimum loss scheme of the upper main grid and the energy optimization management scheme of the lower microgrid, the scheme coupling degree is calculated, and an initial solution is generated according to the scheme coupling degree;

[0036] A fitness function is defined in the genetic algorithm, which is used to minimize the energy loss and operating cost of the main grid and the microgrid at the same time. The calculation formula of the initial solution fitness is:

[0037] ;

[0038] in, represents the initial solution fitness, and represents the weight coefficient, s represents a single node in the main grid, and n represents the total number of nodes in the main grid. represents the energy loss of the sth node of the main grid in the initial solution, represents the energy interaction coefficient of the main grid and microgrid scheme, represents the node scheduling time of the sth node in the main grid, d represents a single microgrid, m represents the total number of microgrids, represents the operating cost of the dth microgrid in the initial solution, Represents the state of the energy storage device of the dth microgrid. If the energy storage device inventory is less than the energy storage threshold, =0, if the storage capacity of the energy storage device is greater than or equal to the energy storage threshold, =1, Represents a constant greater than 0;

[0039] The fitness of the initial solution is iteratively optimized through the selection, crossover and mutation operations in the genetic algorithm until the maximum number of iterations is reached. After each iteration, the optimization degree of the current energy interaction strategy is calculated. If the energy interaction strategy meets the fitness threshold, the final power grid energy interaction strategy is output, including the energy minimum loss plan of the main grid and the energy optimization management plan of the microgrid.

[0040] The adjusting the grid energy interaction strategy includes:

[0041] The monitoring data are collected by sensors deployed on each node of the main grid and the microgrid, wherein the monitoring data includes the main grid monitoring data and the microgrid monitoring data;

[0042] The monitoring data is transmitted to the distribution network control center, and the distribution network control center calculates the deviation between the current operating state and the power grid energy interaction strategy according to the input voltage, current and power data. If the deviation is greater than or equal to the deviation threshold, it indicates that there is a deviation in the current operating state;

[0043] If the grid energy interaction strategy deviates, the target parameters of the regional grid model are reset, including the target voltage of the node, load distribution, charging and discharging power of the energy storage device, and the power exchange between the main grid and the microgrid;

[0044] According to the target parameters of the regional power grid model, the monitoring data is trained through the neural network model to adjust the power grid energy interaction strategy.

[0045] A microgrid and main grid energy interaction control system based on a neural network, the system comprising a regional grid modeling module, an energy interaction distribution module and an energy interaction scheduling module;

[0046] The regional power grid modeling module is used to construct a regional power grid model according to the information of the microgrid and the main grid in the region, and the regional power grid model includes a lower microgrid model and an upper main grid model;

[0047] The energy interaction distribution module is configured with an energy optimization strategy, which is used to optimize the energy interaction of the regional power grid through a neural network model to generate an energy interaction strategy;

[0048] The energy interaction scheduling module is used to be responsible for the global scheduling of power grid operation, monitor the power grid operation status and control the execution of energy interaction.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] 1. The present invention achieves the coordinated optimization of the main grid and the microgrid by constructing a neural network-based energy interaction control system between the microgrid and the main grid, using the regional grid model and reinforcement learning algorithm. By real-time monitoring and adjusting the grid energy distribution, it ensures that the energy interaction between the main grid and the microgrid is always in the optimal state under different operating conditions, effectively reducing the grid energy loss and minimizing the operating cost;

[0051] 2. The present invention not only optimizes the energy loss of the main grid through a double-layer optimization structure, but also introduces an energy optimization management scheme for the microgrid. The genetic algorithm is used to jointly optimize the energy interaction strategies of the main grid and the microgrid to ensure that the two work together. This double-layer optimization mechanism effectively improves the stability and efficiency of the overall operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0053] Figure 1 This is a flow chart of a method for controlling energy interaction between a microgrid and a main grid based on a neural network according to Embodiment 1 of the present invention;

[0054] Figure 2 This is a schematic diagram of the regional power grid model structure of Example 1 of the present invention;

[0055] Figure 3 This is a diagram of the energy distribution network structure of Example 1 of the present invention;

[0056] Figure 4 This is a schematic diagram of the working principle of the main power grid optimization layer in Example 1 of the present invention;

[0057] Figure 5 This is a schematic diagram of the working principle of the microgrid optimization layer in Example 1 of the present invention;

[0058] Figure 6 This is a module diagram of a microgrid and main grid energy interaction control system based on a neural network according to Example 2 of the present invention. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0060] Example 1

[0061] See also Figure 1 The present invention provides an embodiment: a method for controlling energy interaction between a microgrid and a main grid based on a neural network, wherein the specific steps of the method are as follows:

[0062] S1: Build a regional power grid model based on the historical operation information of the microgrid and the main grid;

[0063] S2: constructing an energy distribution network, taking the regional power grid model as an input parameter of the energy distribution network, training the input parameter through the energy distribution network, and outputting a power grid energy interaction strategy;

[0064] S3: Obtain monitoring data of the microgrid and main grid in real time, including current, voltage, load, and renewable energy generation, and evaluate the current operating status of the grid;

[0065] S4: adjusting the power grid energy interaction strategy according to the current operating state, and updating the regional power grid model;

[0066] See also Figure 2, a schematic diagram of the structure of a regional power grid model according to an embodiment of the present invention, based on the hierarchical control structure of the regional power grid, a distributed control method is adopted to ensure efficient energy distribution, the lower microgrid adjusts the local power supply and demand balance through energy storage equipment, and the upper main grid dynamically adjusts the operating status of the transmission lines and generator sets to ensure sufficient and efficient power supply in the region, the regional power grid model includes an upper main grid model and a lower microgrid model, the upper main grid model includes distributed power source information, regional total load information, a main grid and a distribution network control center, the distributed power source information is used to monitor and manage energy information obtained from distributed power sources, such as wind power, solar energy, etc., which will be transmitted to the main grid, the main grid is used to coordinate the power supply of the entire region, and aggregate the power input from distributed power sources and other energy sources. Through the main grid, power can be transmitted in a wider area, the regional total load information is used to receive and process the total load demand of the entire region, ensure the balance between power supply and load demand, and transmit it to the main grid, the distribution network control center is the hub connecting the upper main grid and the lower microgrid. It is responsible for coordinating the operating status of each microgrid, and dynamically adjusting the power generation and dispatching strategy of the main grid according to the load demand and energy storage situation of the microgrid. The lower microgrid model in the figure includes microgrid 1, microgrid 2 and microgrid 3. Each microgrid operates independently but keeps connected to the main grid at the same time. Each microgrid includes energy collection information, energy storage device information and regional load information. The energy collection information is used to monitor the energy flow of the microgrid to ensure that its power output matches the demand and determine the corresponding power energy source of the microgrid. The energy storage device information is used to manage the energy storage equipment in the microgrid, such as battery packs, to ensure that sufficient energy support can be provided during peak demand, or to store excess electric energy during low demand. The regional load information contains the power load demand of specific areas in the microgrid, such as residential, industrial or commercial areas.

[0067] The specific steps of S1 are as follows:

[0068] S1.1: Multi-source sensors collect historical multi-dimensional data of the microgrid and the main grid, wherein the historical multi-dimensional data includes load data, power generation data, energy storage device status data, meteorological data, and topological structure data;

[0069] To ensure the integrity and accuracy of the data, a distributed data processing framework is used to distribute the data on different edge nodes for local processing to reduce latency.

[0070] S1.2: The upper main grid model is constructed according to the actual topological structure of the main grid, including power generation, transmission and dispatching functions. Each function interacts with the lower microgrid model through the distribution network control center to obtain load demand and power generation capacity information in real time. The transmission lines of the main grid are dynamically adjusted according to the current load and power generation conditions to avoid overload or underload.

[0071] The distribution network control center is the core of the interaction between the upper main grid and the lower microgrid. It monitors the operating status of each microgrid in real time, and distributes and dispatches energy with the main grid through load information, power generation information, energy storage status and other data. The dispatch center also dynamically adjusts the power generation and transmission strategies of the upper main grid based on the needs of the lower layer to ensure energy supply balance, improve the overall energy utilization efficiency of the power grid, reduce energy loss, and optimize the configuration of power resources.

[0072] S1.3: Each microgrid independently constructs a sub-model, including energy storage equipment, regional loads, and power generation devices. The sub-model not only considers the balance of power supply and demand of each microgrid, but also predicts future energy demand through load prediction algorithms. The energy storage equipment in the sub-model decides whether to release or store power based on real-time load demand and power generation capacity. The sub-models are connected in series through the geographical location of the microgrid to form a lower-level microgrid model.

[0073] See also Figure 3 , a structural diagram of an energy distribution network according to an embodiment of the present invention, wherein the energy distribution network comprises:

[0074] The main grid optimization layer is used to optimize the energy allocation scheme of the main grid according to the upper main grid model, under the constraint condition of satisfying the main grid power flow balance, with the goal of achieving the main grid optimal power flow, and output the minimum energy loss scheme;

[0075] The microgrid optimization layer is used to optimize the energy allocation scheme of the microgrid and output the energy optimization management scheme based on the lower-layer microgrid model and under the condition that the electrical boundary conditions of the microgrid are met, with the goal of minimizing the operating cost of the microgrid;

[0076] The double-layer optimization layer is used to process the coordinated optimization of the main grid and the microgrid, update the fitness of the energy minimum loss scheme and the energy optimization management scheme through a genetic algorithm, and output a grid energy interaction strategy;

[0077] See also Figure 4 , a schematic diagram of the working principle of the main power grid optimization layer according to an embodiment of the present invention, wherein the main power grid optimization layer outputs a minimum energy loss solution, including:

[0078] First, the multi-source data of the main power grid is collected and processed, including node voltage, branch power flow, distributed power output and load demand. Through the data preprocessing module, abnormal data and noise are eliminated to ensure the validity and accuracy of the input data. After data processing, all input parameters are sorted into sequence inputs. Each feature dimension represents different power grid operation status parameters, such as node voltage changes, branch power flow direction, power output of generator sets, etc., as input features;

[0079] Data standardization technology is used to normalize various types of data to ensure that the data scales of different data sources are consistent. At the same time, key parameters (such as phase difference between nodes, voltage amplitude, etc.) are extracted through feature engineering to reduce the impact of irrelevant features on calculations;

[0080] Based on the generated input features, the local features between nodes are calculated through convolutional neural networks. The power flow allocation scheme is generated by aggregating the local features across nodes. The parameters in the network structure are adjusted through the back-propagation mechanism to ensure that the model can accurately predict the energy allocation scheme.

[0081] Receive a preprocessed multi-dimensional input feature sequence. Through a series of convolution operations, CNN can capture the spatial correlation between nodes. For example, the nonlinear relationship between voltage fluctuations and power flow can be extracted through the sliding window operation of convolution filters. Each layer of convolution generates a set of feature maps that can reflect the power flow transmission rules between different nodes;

[0082] During the convolution process, the ReLU activation function is used to model the nonlinear relationship, so that the network can capture the complex functional relationship in the power flow balance equation;

[0083] After multiple layers of convolution and pooling operations, the network finally generates a preliminary prediction result of the power flow distribution plan, which is further adjusted by weight parameters to output a power flow distribution plan that satisfies the power flow balance equation.

[0084] According to the generated power flow distribution plan, the convolutional neural network will verify each node through the power flow balance equation. The core of the power flow balance equation is to ensure that the input and output power of each node are equal. The calculation formula of the power flow balance equation is as follows:

[0085] ;

[0086] in, represents the input power of node i in the main grid, represents the node voltage of node i, n represents the total number of nodes in the main grid, j represents the jth node in the main grid, and the jth node is different from node i. represents the power loss of node j, represents the penalty coefficient, represents the node voltage of node j, represents the real part of the admittance of nodes i and j, represents the imaginary part of the admittance of nodes i and j, represents the phase difference between node i and node j;

[0087] Through the node verification mechanism, the network can identify nodes that do not meet the power distribution equation, providing a basis for subsequent adjustments. During the verification process, the calculation module in the network updates the node power in real time to ensure power balance;

[0088] If some nodes are found to not satisfy the power distribution equation during the verification of the power balance equation, the convolutional neural network will automatically enter the adaptive adjustment mode. The network first identifies the key parameters that cause imbalance, such as low node voltage or excessive power loss. On this basis, the network weight parameters are adjusted through the back-propagation mechanism. The network will re-learn the power relationship between nodes and generate a new power distribution plan. The network adjustment process is carried out through the reinforcement learning mechanism, gradually optimizing the power transmission path. The back-propagation mechanism and reinforcement learning technology are used to enable the network to adaptively adjust the power distribution plan, and finally achieve the optimal result that all nodes satisfy the power distribution equation.

[0089] When all nodes satisfy the power flow balance equation, the energy loss is calculated based on the active power and reactive power of each node. The power transmission loss of each node is counted, and the minimum energy loss solution is finally output. By dynamically updating the power allocation of each node, the network can output a power flow allocation solution that meets the minimum loss requirements;

[0090] See also Figure 5 , a schematic diagram of the working principle of the microgrid optimization layer according to an embodiment of the present invention, wherein the working process of the microgrid optimization layer includes:

[0091] According to the lower-level microgrid model, the power generation of renewable energy (such as photovoltaic and wind power), user load demand, and the current status of energy storage devices (including charging and discharging rates) in the microgrid are collected and integrated into time series input;

[0092] Construct a reinforcement learning model based on the gated recurrent unit (GRU) and the asynchronous advantage actor-critic algorithm (A3C). Specifically, the GRU is used to process the temporal correlation in the input data, and the A3C algorithm is used to generate the energy allocation strategy.

[0093] The input data in the microgrid is time series data, including photovoltaic power generation, wind power generation, load demand, energy storage equipment status and transaction power with the main grid. The input data has many feature dimensions and strong time correlation, which requires the use of GRU network to capture the time dependency of the data. GRU is used to process the input time series data, capture the dynamic characteristics of energy distribution in the microgrid over time, and output the prediction results of load demand and energy supply for multiple time steps in the future. The prediction results will be input into the actor network and critic network of the A3C algorithm to further generate energy allocation strategies;

[0094] A3C includes an actor network and a critic network. The actor network is responsible for selecting actions, such as adjusting the charging and discharging power of energy storage equipment or the power traded with the main grid. Action selection is implemented through a policy function. The probability distribution of microgrid evolution is generated according to the current state and prediction results. The corresponding action is selected according to the probability distribution of microgrid evolution. The critic network calculates the value function of the action generated by the actor network through the value function to estimate its long-term benefits, and optimizes the actions and long-term benefits according to the network weights, and outputs the energy allocation plan of the microgrid.

[0095] The goal of reinforcement learning is to make decisions on the energy allocation of microgrids through intelligent agents based on the objective function to minimize operating costs. The core of GRU is to control the update method of hidden states through update gates and reset gates, so as to effectively capture the long-term dependency characteristics in the input data. At the same time, the A3C algorithm calculates the advantage of the current strategy compared to the benchmark strategy through the advantage function, making the strategy update more directional and effective. Through the asynchronous update of multiple intelligent agents, the global optimal strategy can be quickly found and the convergence speed can be significantly improved;

[0096] After generating the energy allocation plan, the network will automatically determine whether the plan meets the electrical boundary conditions of the microgrid. The specific electrical boundary conditions include:

[0097] The charging and discharging power limit of the energy storage device must be within its physical limit;

[0098] Load demand satisfaction: it must be ensured that the user's load demand is met, and the cost of unsatisfied load (such as activating backup generators) will be included in the total cost;

[0099] The power transaction limit with the main grid is that the power interaction with the main grid does not exceed the maximum allowed range;

[0100] By judging the electrical boundary conditions, it is ensured that the generated energy distribution scheme is feasible in actual operation and meets the physical and economic requirements of the microgrid;

[0101] If the generated energy allocation scheme does not meet the electrical boundary conditions, the reinforcement learning model will update the network weights through the back-propagation mechanism. This process regenerates a new energy allocation scheme by continuously adjusting the network parameters until all conditions are met;

[0102] The double-layer optimization layer uses a genetic algorithm to collaboratively optimize the energy minimum loss scheme of the upper main grid and the energy optimization management scheme of the lower microgrid. First, the genetic algorithm is used to evaluate the initial solutions in the main grid and the microgrid, and a fitness function is defined. Through multiple iterations, the global optimality of the energy interaction between the main grid and the microgrid is finally achieved. The double-layer optimization layer includes:

[0103] The energy minimum loss scheme of the upper main grid and the energy optimization management scheme of the lower microgrid are jointly optimized through genetic algorithms;

[0104] Through the two-layer optimization structure, the genetic algorithm can coordinately optimize the energy distribution between the upper and lower power grids, ensuring that the coupling relationship between the two power grids is fully utilized. The use of fitness function can not only minimize energy loss, but also take into account the operating cost of the microgrid;

[0105] In each iteration of the genetic algorithm, the energy optimization management scheme of the microgrid is first optimized through the lower-level microgrid model, and then the results are fed back to the upper-level main grid model for power flow optimization. Through the collaboration of the two layers, the energy interaction strategy between the main grid and the microgrid is output;

[0106] When the genetic algorithm converges, the optimal energy interaction strategy between the main grid and the microgrid is output;

[0107] The genetic algorithm comprises:

[0108] According to the energy interaction relationship between the energy minimum loss scheme of the upper main grid and the energy optimization management scheme of the lower microgrid, the scheme coupling degree is calculated, and the scheme coupling degree reflects the mutual influence between the main grid and the microgrid, and an initial solution is generated according to the scheme coupling degree;

[0109] A fitness function is defined in the genetic algorithm, which is used to minimize the energy loss and operating cost of the main grid and the microgrid at the same time. The calculation formula of the initial solution fitness is:

[0110] ;

[0111] in, represents the initial solution fitness, and represents the weight coefficient, s represents a single node in the main grid, and n represents the total number of nodes in the main grid. represents the energy loss of the sth node of the main grid in the initial solution, represents the energy interaction coefficient of the main grid and microgrid scheme, represents the node scheduling time of the sth node in the main grid, d represents a single microgrid, m represents the total number of microgrids, represents the operating cost of the dth microgrid in the initial solution, Represents the state of the energy storage device of the dth microgrid. If the energy storage device inventory is less than the energy storage threshold, =0, if the storage capacity of the energy storage device is greater than or equal to the energy storage threshold, =1, Represents a constant greater than 0;

[0112] The fitness of the initial solution is iteratively optimized through the selection, crossover and mutation operations in the genetic algorithm until the maximum number of iterations is reached. After each iteration, the optimization degree of the current energy interaction strategy is calculated. If the energy interaction strategy meets the fitness threshold, the final grid energy interaction strategy is output, including the energy minimum loss plan of the main grid and the energy optimization management plan of the microgrid. Each iteration in the genetic algorithm calculates the fitness according to the new energy interaction strategy. If the fitness of the strategy meets the predetermined threshold, the final grid energy interaction strategy is output. Each iteration ensures that the solution gradually tends to the global optimum by evaluating the power flow, loss and energy storage status in the energy interaction;

[0113] Selection operation: select individuals with high fitness to enter the next round to ensure that better solutions can be passed on to future generations;

[0114] Crossover operation: Two solutions with high fitness generate a new solution through crossover operation. The crossover operation mainly combines the energy allocation scheme of the main grid with the energy storage scheduling strategy of the microgrid to form a new grid energy interaction strategy.

[0115] Mutation operation: Introducing small-scale mutations in each generation of solutions to ensure the diversity of solutions and prevent the algorithm from falling into local optimality. Mutation operations usually fine-tune the energy allocation or energy storage scheduling of a single node.

[0116] The specific steps of S4 are as follows:

[0117] S4.1: Collect monitoring data through sensors deployed on each node of the main grid and the microgrid, where the monitoring data includes main grid monitoring data and microgrid monitoring data;

[0118] Collect the voltage, current and power of the main grid nodes to ensure that the nodes are in normal operation, collect the renewable energy power generation of the microgrid, the charging and discharging status of the energy storage equipment and the load demand of the microgrid, and the sensors use low-latency communication protocols (such as 5G or LoRa) to transmit the monitoring data to the distribution network control center to ensure the stability and accuracy of real-time data transmission;

[0119] S4.2: The monitoring data is transmitted to the distribution network control center, and the distribution network control center calculates the deviation between the current operating state and the grid energy interaction strategy according to the input voltage, current and power data. If the deviation is greater than or equal to the deviation threshold, it indicates that there is a deviation in the current operating state;

[0120] S4.3: If the grid energy interaction strategy is deviated from, the target parameters of the regional grid model are reset, including the target voltage of the node, load distribution, charging and discharging power of the energy storage device, and the power exchange amount between the main grid and the microgrid;

[0121] S4.4: According to the target parameters of the regional power grid model, the monitoring data is trained by the neural network model to adjust the power grid energy interaction strategy;

[0122] The neural network achieves real-time optimization in complex dynamic environments in the power grid through multi-level data processing and parameter adjustment. The quantitative calculation of the deviation provides a standard for judging the operating status of the invention, and the adaptive adjustment of the neural network ensures that the power grid can still operate stably under complex conditions.

[0123] Example 2

[0124] See also Figure 6 , the present invention provides an embodiment: a microgrid and main grid energy interaction control system based on a neural network, the system comprising a regional grid modeling module, an energy interaction allocation module and an energy interaction scheduling module;

[0125] The regional power grid modeling module is used to construct a regional power grid model according to the information of the microgrid and the main grid in the region, and the regional power grid model includes a lower microgrid model and an upper main grid model;

[0126] The energy interaction distribution module is configured with an energy optimization strategy, which is used to optimize the energy interaction of the regional power grid through a neural network model to generate an energy interaction strategy;

[0127] The energy interaction scheduling module is used to be responsible for the global scheduling of power grid operation, monitor the power grid operation status and control the execution of energy interaction;

[0128] The regional power grid modeling module comprises:

[0129] The data acquisition unit is used to collect microgrid and main grid data in the area, including power generation power, load demand, energy storage status, etc.

[0130] The data processing unit is used to pre-process the collected multi-source data, such as data cleaning and normalization, to ensure data consistency.

[0131] A power grid model building unit is used to build a lower-layer microgrid model and an upper-layer main grid model based on the preprocessed data, and finally build a whole regional power grid model;

[0132] The energy interactive distribution module comprises:

[0133] The neural network optimization unit is used to optimize the data of the regional power grid through a neural network and generate an energy interaction strategy.

[0134] The strategy generation unit is used to generate an energy distribution plan for the main grid and the microgrid according to the output results of the neural network.

[0135] A cost calculation unit, used to calculate the total cost of the energy allocation scheme and further optimize the energy interaction strategy based on the calculation results;

[0136] The energy interaction scheduling module includes:

[0137] Global dispatch unit: responsible for global dispatch and implementation according to the energy interaction strategy between the main grid and the microgrid.

[0138] State monitoring unit: monitors the operation status of the power grid in real time and detects whether the operation deviates from the energy interaction strategy.

[0139] Feedback adjustment unit: adjusts strategies based on monitoring data to ensure that the power grid operates in the best condition.

[0140] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for interactive energy control between microgrid and main grid based on neural network, characterized in that: The method comprises: Constructing a regional power grid model based on the information of the microgrid and the main grid in the region, wherein the regional power grid model includes a lower-layer microgrid model and an upper-layer main grid model; Constructing a neural network model, training the regional power grid model through the neural network model, and obtaining power grid energy interaction strategies under different operating conditions; Acquire real-time monitoring data of the microgrid and the main grid, adjust the grid energy interaction strategy according to the real-time monitoring data, and update the regional grid model; The neural network model comprises: The main grid optimization layer is used to optimize the energy allocation scheme of the main grid according to the upper main grid model, under the constraint condition of satisfying the main grid power flow balance, with the goal of achieving the main grid optimal power flow, and output the minimum energy loss scheme; The microgrid optimization layer is used to optimize the energy allocation scheme of the microgrid and output the energy optimization management scheme based on the lower-layer microgrid model and under the condition that the electrical boundary conditions of the microgrid are met, with the goal of minimizing the operating cost of the microgrid; The double-layer optimization layer is used to process the coordinated optimization of the main grid and the microgrid, update the fitness of the energy minimum loss scheme and the energy optimization management scheme through a genetic algorithm, and output a grid energy interaction strategy.

2. The method for energy interaction control between microgrid and main grid based on neural network according to claim 1 is characterized in that: The constructing of the regional power grid model comprises: Collect multi-source data related to microgrids and main grids in the region; Constructing an upper main grid model according to the topological structure of the main grid, wherein the upper main grid model includes power generation, transmission and dispatching functions of the main grid; Constructing a lower-layer microgrid model according to the multi-source information of each microgrid, wherein the lower-layer microgrid model includes a sub-model of a single microgrid, and the sub-model includes a microgrid energy storage device, a local area load, and a microgrid power generation device; The upper main grid model also includes a distribution network control center, which is used to interact with the lower microgrid model, obtain lower-level demands and lower-level information, and distribute energy with the main grid according to the lower-level demands and lower-level information.

3. The method for energy interaction control between microgrid and main grid based on neural network according to claim 1 is characterized in that: The optimization of the energy distribution scheme of the main power grid includes: Processing multi-source data of the main power grid to extract an input parameter sequence, wherein the multi-source data includes node voltage, branch power flow, distributed power output and load demand; Generate a power flow allocation scheme through a convolutional neural network, set the objective function of the main power grid optimization to minimize the active power loss in the main power grid power flow, and calculate whether any node in the main power grid satisfies the power flow balance equation according to the power flow allocation scheme; If any node satisfies the power distribution equation, the active power of each node is counted and the energy minimum loss plan is output. If there are nodes that do not satisfy the power distribution equation, the nodes that do not satisfy the power distribution equation are further analyzed to identify the key parameters that cause imbalance. The key parameters are adjusted through the back propagation mechanism to generate a new power distribution plan and make a new judgment.

4. The method for energy interaction control between microgrid and main grid based on neural network according to claim 1 is characterized in that: The optimization of the energy distribution scheme of the microgrid includes: According to the lower-layer microgrid model, the renewable energy generation power, user load demand and energy storage status in the microgrid are collected to generate a multi-dimensional input sequence; Constructing a reinforcement learning model, training the multidimensional input sequence through the reinforcement learning model, and outputting an energy allocation scheme for the microgrid, wherein the reinforcement learning model includes a gated recurrent unit and an A3C algorithm; Determine whether the generated energy allocation scheme meets the electrical boundary conditions of the microgrid, wherein the electrical boundary conditions include the charging and discharging power limit of the energy storage device, the load demand satisfaction and the power transaction limit with the main grid; If the electrical boundary conditions are met, the energy allocation plan is output as an energy optimization management plan. If the electrical boundary conditions are not met, the network weights of the reinforcement learning model are updated through a back-propagation mechanism to generate a new energy allocation plan until the electrical boundary conditions are met.

5. The method for energy interaction control between microgrid and main grid based on neural network according to claim 1 is characterized in that: The double-layer optimization layer comprises: The energy minimum loss scheme of the upper main grid and the energy optimization management scheme of the lower microgrid are jointly optimized through genetic algorithms; In each iteration of the genetic algorithm, the energy optimization management scheme of the microgrid is first optimized through the lower-level microgrid model, and then the results are fed back to the upper-level main grid model for power flow optimization. Through the collaboration of the two layers, the energy interaction strategy between the main grid and the microgrid is output; When the genetic algorithm converges, the optimal energy interaction strategy between the main grid and the microgrid is output.

6. The method for controlling energy interaction between a microgrid and a main grid based on a neural network according to claim 5 is characterized in that: The genetic algorithm comprises: According to the energy interaction relationship between the energy minimum loss scheme of the upper main grid and the energy optimization management scheme of the lower microgrid, the scheme coupling degree is calculated, and an initial solution is generated according to the scheme coupling degree; A fitness function is defined in the genetic algorithm, which is used to minimize the energy loss and operating cost of the main grid and the microgrid at the same time. The calculation formula of the initial solution fitness is: ， in, represents the initial solution fitness, and represents the weight coefficient, s represents a single node in the main grid, and n represents the total number of nodes in the main grid. represents the energy loss of the sth node of the main grid in the initial solution, represents the energy interaction coefficient of the main grid and microgrid scheme, represents the node scheduling time of the sth node in the main grid, d represents a single microgrid, m represents the total number of microgrids, represents the operating cost of the dth microgrid in the initial solution, Represents the state of the energy storage device of the dth microgrid. If the energy storage device inventory is less than the energy storage threshold, =0, if the storage capacity of the energy storage device is greater than or equal to the energy storage threshold, =1, Represents a constant greater than 0; The fitness of the initial solution is iteratively optimized through the selection, crossover and mutation operations in the genetic algorithm until the maximum number of iterations is reached. After each iteration, the optimization degree of the current energy interaction strategy is calculated. If the energy interaction strategy meets the fitness threshold, the final power grid energy interaction strategy is output, including the energy minimum loss plan of the main grid and the energy optimization management plan of the microgrid.

7. The method for controlling energy interaction between a microgrid and a main grid based on a neural network according to claim 2 is characterized in that: The adjusting the grid energy interaction strategy includes: The monitoring data are collected by sensors deployed on each node of the main grid and the microgrid, wherein the monitoring data includes the main grid monitoring data and the microgrid monitoring data; The monitoring data is transmitted to the distribution network control center, and the distribution network control center calculates the deviation between the current operating state and the power grid energy interaction strategy according to the input voltage, current and power data. If the deviation is greater than or equal to the deviation threshold, it indicates that there is a deviation in the current operating state; If the grid energy interaction strategy deviates, the target parameters of the regional grid model are reset, including the target voltage of the node, load distribution, charging and discharging power of the energy storage device, and the power exchange between the main grid and the microgrid; According to the target parameters of the regional power grid model, the monitoring data is trained through the neural network model to adjust the power grid energy interaction strategy.

8. A microgrid and main grid energy interaction control system based on a neural network, used to implement the microgrid and main grid energy interaction control method based on a neural network as described in any one of claims 1 to 7, characterized in that: The system includes a regional power grid modeling module, an energy interactive distribution module and an energy interactive scheduling module; The regional power grid modeling module is used to construct a regional power grid model according to the information of the microgrid and the main grid in the region, and the regional power grid model includes a lower microgrid model and an upper main grid model; The energy interaction distribution module is configured with an energy optimization strategy, which is used to optimize the energy interaction of the regional power grid through a neural network model to generate an energy interaction strategy; The energy interaction scheduling module is used to be responsible for the global scheduling of power grid operation, monitor the power grid operation status and control the execution of energy interaction.

Images