Power distribution method for distributed power supply and energy storage in micro-grid

Abstract

The application discloses a micro-grid distributed power and energy storage power distribution method, and particularly relates to the technical field of micro-grid energy management, which comprises the following steps: collecting the grid-connected point state, distributed power output, load demand and energy storage SOC in real time; predicting the SOC trend curve of the future time window by using LSTM; establishing a multi-objective power distribution model with the lowest operation cost and the largest stability margin as the targets, wherein the stability margin target is associated with the SOC trend curve; calculating the dynamic weight coefficient of the operation cost and the stability target in real time by fuzzy logic according to the grid-connected point state and the SOC trend slope, solving the model to obtain the power instruction value and issuing the power instruction value for execution, triggering the pre-synchronization optimization when the grid-connected point state changes to the islanded state, adding a voltage / frequency deviation penalty term in the model, and realizing seamless switching. The application realizes prospective control and adaptive decision of power distribution, avoids overcharging and overdischarging of the energy storage, prolongs the battery life, and improves the economy and stability of the micro-grid.

Description

Technical Field

[0001] This invention relates to the field of microgrid energy management technology, and more specifically, to a method for allocating distributed power sources and energy storage power within a microgrid. Background Technology

[0002] As an effective organizational form for distributed power sources to access the distribution network, microgrids can realize the local consumption and flexible control of distributed energy, playing an important role in improving power supply reliability and promoting the utilization of renewable energy. Microgrids typically include distributed power sources such as photovoltaics and wind power, as well as energy storage systems, and can operate in grid-connected mode and islanded mode. During the operation of microgrids, how to rationally allocate the power of distributed power sources and energy storage to achieve synergistic optimization of energy balance and economic operation is a key technical issue in this field.

[0003] Existing microgrid power allocation methods mainly adopt real-time dynamic control strategies, which calculate the power command value of each unit by monitoring the current generation power, load demand and energy storage charge status.

[0004] However, such methods make decisions based solely on the current energy storage state, lacking the ability to predict future trends. When the energy storage SOC approaches the safety boundary, there is often not enough time to adjust, which can easily lead to overcharging and over-discharging problems, affecting battery life. In addition, when dealing with multi-objective optimization, the weight coefficients of the operating cost objective and the stability objective are usually preset to fixed values, which cannot be adaptively adjusted according to changes in the microgrid operating mode and the evolution of the energy storage state. It is difficult to achieve optimal decision-making under different operating conditions. In terms of mode switching, traditional methods separate the switching process from grid-connected to islanded from power distribution. Voltage and frequency surges are easily generated at the moment of switching, affecting power quality and power supply reliability.

[0005] To address the aforementioned issues, it is necessary to research a microgrid power allocation method capable of achieving forward-looking control, adaptive decision-making, and seamless switching, in order to improve the economy and stability of microgrid operation. In view of this, the present invention provides a method for power allocation of distributed power sources and energy storage within a microgrid. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for distributing distributed power sources and energy storage power within a microgrid, thereby addressing the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for distributing distributed power sources and energy storage power within a microgrid, comprising: Real-time data collection of microgrid grid connection status, distributed power output, load demand, and the state of charge (SOC) of energy storage systems; Based on the currently collected SOC and historical SOC data, the long short-term memory neural network is used to predict the SOC change trend within a preset time window in the future, and generate an SOC trend curve. The preset time window is 15 minutes to 1 hour. A multi-objective power allocation model is established with the goals of minimizing operating costs and maximizing stability margin. The stability margin objective is related to the SOC trend curve. Specifically, the stability margin objective function is constructed based on the degree of deviation between the predicted SOC value in the SOC trend curve and the preset upper and lower limits of safe SOC operation. Based on the grid connection point status and SOC trend curve, the dynamic weight coefficient pair of the operating cost target and the stability target in the multi-objective power allocation model is calculated in real time through a predefined fuzzy logic rule base. Solve the multi-objective power allocation model with dynamic weighting coefficients to obtain the power command values ​​of each distributed power source and energy storage system in the next control cycle, and issue them for execution.

[0008] Preferably, predicting the SOC change trend within a future preset time window specifically includes: inputting historical SOC sequences, historical charge and discharge power sequences, and ambient temperature data into a pre-trained long short-term memory neural network, and having the long short-term memory neural network output the predicted SOC values ​​for each time point within the future preset time window to construct an SOC trend curve; The Long Short-Term Memory (LSTM) neural network consists of an input layer, two hidden layers, and an output layer. The number of neurons in the hidden layer is 128.

[0009] Preferably, the dynamic weight coefficient pairs are calculated in real time through the fuzzy logic rule base, specifically including: converting the grid connection point state into a mode deviation signal, where the mode deviation signal is set to 0 in the grid connection state, 1 in the island state, and a continuous value between 0 and 1 when switching to the transition state. The SOC trend curve is converted into the SOC trend slope at the current moment using the difference method; The mode bias signal and SOC trend slope are used as inputs to the fuzzy logic controller. After fuzzification, fuzzy inference and defuzzification, the first weight coefficient corresponding to the operating cost target and the second weight coefficient corresponding to the stability margin target are output. The fuzzy logic controller has a preset fuzzy rule base, which includes the following rules: if the mode deviation signal is large and the SOC trend slope is negatively large, then the second weight coefficient takes the maximum value. If the mode bias signal is small and the slope of the SOC trend is small, then the first weighting coefficient should be a larger value. If the mode bias signal is moderate and the slope of the SOC trend is zero, then the two weighting coefficients are equal.

[0010] Preferably, it also includes a seamless mode switching pre-synchronization step: When an instruction to switch from grid-connected state to islanded state is detected, the pre-synchronization optimization cycle is triggered. The pre-synchronization optimization cycle is 3 to 5 control cycles before the switching action occurs. During the pre-synchronization optimization period, voltage deviation penalty terms and frequency deviation penalty terms are added to the multi-objective power allocation model; Solve the multi-objective power allocation model with added voltage deviation penalty term and frequency deviation penalty term to obtain the pre-synchronization power command value, and issue the pre-synchronization power command value for execution before the switching action occurs to reduce voltage and frequency fluctuations at the moment of switching.

[0011] Preferably, the voltage deviation penalty term is the square of the deviation between the actual voltage and the rated voltage, i.e. The frequency deviation penalty term is the square of the deviation between the actual frequency and the rated frequency, i.e. Penalty coefficient and The system's tolerance for voltage and frequency deviations is preset to a fixed value or dynamically adjusted according to the deviation range.

[0012] Preferably, the objective function of the multi-objective power allocation model is expressed as: , in, The objective function is the operating cost. Let SOC be the objective function for stability margin, which is associated with the SOC trend curve. and These are the first and second weight coefficients in the dynamic weight coefficient pair, respectively.

[0013] Preferably, the stability margin objective function The system is constructed based on the deviation of the predicted SOC value from the preset upper and lower limits of safe SOC operation in the SOC trend curve. Specifically: , in, To predict the SOC value, This is the preset upper limit for safe operation of the SOC. The SOC is the preset lower limit for safe operation. When the predicted SOC value exceeds the safe operating range, the stability margin objective function value is greater than 0, and the larger the value, the greater the deviation. When the predicted SOC value is within the safe range, the stability margin objective function value is 0.

[0014] Preferably, the multi-objective power allocation model carrying dynamic weight coefficient pairs is solved by particle swarm optimization algorithm or genetic algorithm. The population size of particle swarm optimization algorithm is set to 50 and the maximum number of iterations is 100. The crossover probability of genetic algorithm is 0.8 and the mutation probability is 0.1.

[0015] Preferably, the distributed power source includes photovoltaic power generation units and wind power generation units, and the power command value includes the active power adjustment of each distributed power source and the charging and discharging power value of the energy storage system.

[0016] Preferably, the real-time acquisition step further includes acquiring the State of Health (SOH) data of the energy storage system. The prediction process of the State of Charge (SOC) trend curve is further corrected based on the SOH data, with a correction coefficient... It shows a linear positive correlation with the SOH value, specifically: ,in and It is a preset constant, and 0 < ≤1, ≥0.

[0017] The technical effects and advantages of this invention are as follows: This invention introduces a SOC trend prediction mechanism, which elevates power allocation decisions from instantaneous control based on the current SOC to forward-looking control based on the future SOC evolution trend within a time window. Specifically, it utilizes a long short-term memory neural network to perform deep learning on historical SOC sequences, historical charge and discharge power sequences, and ambient temperature data to predict the SOC change trend within a preset time window and generate an SOC trend curve. This allows power allocation to not only be based on the current energy storage status but also to comprehensively consider the future support capacity of the energy storage. When it is predicted that the SOC will exceed the safe operating range, the system adjusts the allocation strategy in advance, effectively avoiding overcharging and over-discharging of the energy storage system, extending battery life, and providing ample adjustment margin for the stable operation of the microgrid. This invention constructs a dynamic weight adaptive adjustment mechanism based on fuzzy logic to achieve a dynamic balance between operational economy and system stability in multi-objective optimization. By converting the grid connection point state into a mode deviation signal and the SOC trend curve into the current SOC trend slope, the two are used as dual inputs of the fuzzy logic controller. After fuzzification, fuzzy inference and defuzzification, the dynamic weight coefficient pair of the operating cost target and the stability margin target is output in real time. This enables the weight coefficient to be autonomously adjusted according to the changes in the microgrid operating mode and the future state of energy storage. The power allocation model focuses on operational economy under stable grid connection conditions and prioritizes system stability under near-islanding or SOC trend deterioration conditions. This overcomes the defects of fixed weights or single adjustment basis in traditional methods. This invention proposes a pre-synchronization optimization architecture integrating switching and distribution. It incorporates the mode switching process into the power distribution model for unified optimization, achieving seamless, zero-impact switching from grid-connected to islanded mode in microgrids. When a command to switch from grid-connected to islanded mode is detected, the system triggers a pre-synchronization optimization cycle. Voltage and frequency deviation penalty terms are added to the multi-objective power distribution model. By solving the optimization model, the output of distributed power sources and energy storage systems is adjusted in advance to minimize the voltage and frequency deviations at the switching point. This architecture revolutionizes the traditional event-triggered switching into a dynamic optimization process, fundamentally suppressing electrical shocks at the moment of switching and significantly improving the power quality and reliability of microgrids. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0020] This invention provides a method for allocating distributed power sources and energy storage power within a microgrid, applicable to both grid-connected and islanded operation modes, and particularly suitable for microgrids that include photovoltaic, wind power generation units, and energy storage systems.

[0021] Step S1: Real-time monitoring of dual-scale status The system collects the grid connection status, distributed power output, load demand, and state of charge (SOC) of the energy storage system in real time. The grid connection status is used to determine whether the microgrid is in grid-connected mode or islanded mode. Distributed power output includes the actual active power of photovoltaic power generation units and wind power generation units. The load demand is the total load power within the microgrid. Based on the currently collected SOC and historical SOC data, the SOC change trend is predicted within a preset time window, generating an SOC trend curve. The preset time window is from 15 minutes to 1 hour. Specifically, the historical SOC sequence, historical charge and discharge power sequence, and ambient temperature data are input into a pre-trained long short-term memory neural network. The long short-term memory neural network outputs the predicted SOC values ​​for each time point within the preset time window, thus constructing the SOC trend curve. This curve reflects the evolution trend of the energy storage system's state of charge over a future period, providing a forward-looking basis for subsequent control. The long short-term memory neural network includes an input layer, two hidden layers, and an output layer. The number of neurons in the hidden layer is 128. The dimension of the input layer is determined according to the length of the historical sequence, and the dimension of the output layer corresponds to the number of prediction points within the preset time window.

[0022] Step S2: Construct a dynamic multi-objective power allocation model Establish a multi-objective power allocation model with the objectives of minimizing operating costs and maximizing stability margins, where the operating cost objective is... This includes the generation cost of distributed power sources, the depreciation cost of energy storage systems, and the cost of purchasing electricity from the main grid, as well as the stability margin target. It is associated with the SOC trend curve generated in step S1, specifically constructed based on the degree of deviation between the predicted SOC value in the SOC trend curve and the preset upper and lower limits of safe SOC operation. Stability margin objective function Specifically, it is constructed as follows: , in, To predict the SOC value, This is the preset upper limit for safe operation of the SOC. The SOC is the preset lower limit for safe operation. When the predicted SOC value exceeds the safe operating range, the stability margin objective function value is greater than 0, and the larger the value, the greater the deviation. When the predicted SOC value is within the safe range, the stability margin objective function value is 0.

[0023] The objective function of the multi-objective power allocation model is expressed as: , in, and These are the first and second weight coefficients in the dynamic weight coefficient pair, respectively. They are not fixed values, but are dynamically adjusted according to the real-time status.

[0024] Step S3: Adaptive Calculation of Dynamic Weight Coefficients Based on real-time monitoring of grid connection point status and SOC trend curves, dynamic weight coefficients are calculated in real-time using a fuzzy logic rule base. The specific process is as follows: First, the grid connection point status is converted into a mode deviation signal. The mode deviation signal is set to 0 in the grid connection state, 1 in the islanding state, and a continuous value between 0 and 1 in the switching transition state. For example, a ramp signal is generated according to the switching command, and the SOC trend curve is converted into the SOC trend slope at the current moment through the differential method. Then, the mode deviation signal and the SOC trend slope are used as inputs to the fuzzy logic controller. After fuzzification, the controller employs defined membership functions and fuzzy inference, and outputs the result based on a preset rule base and after defuzzification. and ; The fuzzy logic controller has a preset fuzzy rule base, which includes the following core rules: Rule 1: If the mode bias signal is large and the SOC trend slope is large and negative, then the second weighting coefficient... Significantly improved, with stability being the absolute priority; Rule 2: If the mode bias signal is small and the slope of the SOC trend is small, then the first weighting coefficient... Take the maximum value, prioritizing cost-effectiveness; Rule 3: If the mode bias signal is moderate and the slope of the SOC trend is zero, then the two weighting coefficients are equal (balancing economy and stability). Among them, fuzzy concepts such as "large", "small", "negative large", "positive small", and "medium" are quantified through predefined membership functions, and the centroid method or the maximum membership method is used to defuzzify them. In this way, the weighting coefficients can adaptively match the real-time operating conditions of the microgrid and the future state of energy storage.

[0025] Step S4: Model Solving and Command Issuance A multi-objective power allocation model with dynamic weight coefficients is solved using either particle swarm optimization (PSO) or genetic algorithm to obtain the power command values ​​for each distributed power source and energy storage system in the next control cycle. The power command values ​​include the active power adjustment of each distributed power source and the charging / discharging power of the energy storage system. The PSO algorithm uses a population size of 50 and a maximum number of iterations of 100. The genetic algorithm has a crossover probability of 0.8 and a mutation probability of 0.1. Constraints such as power balance and equipment output limits must be satisfied during the solution process. For example, power balance constraints... Distributed power generation output constraints Energy storage charging and discharging power constraints and SOC security constraints Finally, the instructions are sent to each unit for execution.

[0026] Optionally, embodiments of the present invention further include step S5, seamless mode switching pre-synchronization. When a command to switch from grid-connected to islanded state is detected, a pre-synchronization optimization cycle is triggered. This pre-synchronization optimization cycle consists of 3 to 5 control cycles prior to the switching action. During this cycle, voltage deviation penalty terms and frequency deviation penalty terms are added to the multi-objective power allocation model, meaning the objective function becomes: ,

[0027] Among them, voltage deviation penalty item It is the square of the deviation between the actual voltage and the rated voltage, i.e. Frequency deviation penalty item It is the square of the deviation between the actual frequency and the rated frequency, i.e. Penalty coefficient and The system's tolerance for voltage and frequency deviations is preset to a fixed value or dynamically adjusted according to the deviation range. Solving the model yields the pre-synchronization power command value, which is then issued and executed before the switching action occurs. This allows for advance adjustment of the output of each unit, minimizing voltage and frequency fluctuations during the switching process and achieving seamless switching.

[0028] Furthermore, to improve the accuracy of SOC trend prediction, the real-time acquisition step may also include collecting the State of Health (SOH) data of the energy storage system. During the prediction process, the prediction results of the LSTM neural network are corrected based on the SOH data, with correction coefficients... It is positively correlated with the SOH value, specifically: ,in and It is a preset constant, and 0 < ≤1, ≥0; The higher the SOH, the closer the correction coefficient is to 1, and the higher the reliability of the prediction results. When the SOH decreases, the correction coefficient decreases accordingly, and the predicted value is adjusted appropriately to reflect the aging effect.

[0029] In summary: This invention achieves a leap from current state control to future trend prediction by introducing SOC trend prediction. By dynamically adjusting the weights of multiple objectives through fuzzy logic, the control strategy adapts to the operating mode and energy storage health trend. By integrating mode switching into power distribution through pre-synchronization optimization, the switching impact is eliminated. The synergistic effect of these three factors significantly improves the economy, stability and power supply reliability of the microgrid, demonstrating outstanding substantive features and progress.

[0030] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for distributing distributed power sources and energy storage power within a microgrid, characterized in that: include: Real-time data collection of microgrid grid connection status, distributed power output, load demand, and the state of charge (SOC) of energy storage systems; Based on the currently collected SOC and historical SOC data, the long short-term memory neural network is used to predict the SOC change trend within a preset time window in the future, and generate an SOC trend curve. The preset time window is 15 minutes to 1 hour. A multi-objective power allocation model is established with the goals of minimizing operating costs and maximizing stability margin. The stability margin objective is related to the SOC trend curve. Specifically, the stability margin objective function is constructed based on the degree of deviation between the predicted SOC value in the SOC trend curve and the preset upper and lower limits of safe SOC operation. Based on the grid connection point status and SOC trend curve, the dynamic weight coefficient pair of the operating cost target and the stability target in the multi-objective power allocation model is calculated in real time through a predefined fuzzy logic rule base. Solve the multi-objective power allocation model with dynamic weighting coefficients to obtain the power command values ​​of each distributed power source and energy storage system in the next control cycle, and issue them for execution.

2. The method for distributing distributed power sources and energy storage power within a microgrid according to claim 1, characterized in that: Predicting the SOC change trend within a future preset time window specifically includes: inputting historical SOC sequences, historical charge and discharge power sequences, and ambient temperature data into a pre-trained long short-term memory neural network, and having the long short-term memory neural network output the predicted SOC values ​​at each time point within the future preset time window to construct an SOC trend curve. The Long Short-Term Memory (LSTM) neural network consists of an input layer, two hidden layers, and an output layer. The number of neurons in the hidden layer is 128.

3. The method for distributing distributed power sources and energy storage power within a microgrid according to claim 1, characterized in that: The dynamic weight coefficient pairs are calculated in real time through the fuzzy logic rule base, specifically including: converting the grid connection point status into a mode deviation signal, setting the mode deviation signal to 0 in the grid connection state, setting it to 1 in the island state, and setting it to a continuous value between 0 and 1 when switching to the transition state. The SOC trend curve is converted into the SOC trend slope at the current moment using the difference method; The mode bias signal and SOC trend slope are used as inputs to the fuzzy logic controller. After fuzzification, fuzzy inference and defuzzification, the first weight coefficient corresponding to the operating cost target and the second weight coefficient corresponding to the stability margin target are output. The fuzzy logic controller has a preset fuzzy rule base, which includes the following rules: if the mode deviation signal is large and the SOC trend slope is negatively large, then the second weight coefficient takes the maximum value. If the mode bias signal is small and the slope of the SOC trend is small, then the first weighting coefficient should be a larger value. If the mode bias signal is moderate and the slope of the SOC trend is zero, then the two weighting coefficients are equal.

4. The method for distributing distributed power sources and energy storage power within a microgrid according to claim 1, characterized in that: It also includes a seamless mode switching pre-synchronization step: When an instruction to switch from grid-connected state to islanded state is detected, the pre-synchronization optimization cycle is triggered. The pre-synchronization optimization cycle is 3 to 5 control cycles before the switching action occurs. During the pre-synchronization optimization period, voltage deviation penalty terms and frequency deviation penalty terms are added to the multi-objective power allocation model; Solve the multi-objective power allocation model with added voltage deviation penalty term and frequency deviation penalty term to obtain the pre-synchronization power command value, and issue the pre-synchronization power command value for execution before the switching action occurs to reduce voltage and frequency fluctuations at the moment of switching.

5. The method for distributing distributed power sources and energy storage power within a microgrid according to claim 4, characterized in that: The voltage deviation penalty is the square of the deviation between the actual voltage and the rated voltage, i.e. The frequency deviation penalty term is the square of the deviation between the actual frequency and the rated frequency, i.e. Penalty coefficient and The system's tolerance for voltage and frequency deviations is preset to a fixed value or dynamically adjusted according to the deviation range.

6. The method for distributing distributed power sources and energy storage power within a microgrid according to claim 1, characterized in that: The objective function of the multi-objective power allocation model is expressed as: ， in, The objective function is the operating cost. Let SOC be the objective function for stability margin, which is associated with the SOC trend curve. and These are the first and second weight coefficients in the dynamic weight coefficient pair, respectively.

7. The method for distributing distributed power sources and energy storage power within a microgrid according to claim 1, characterized in that: Stability margin objective function The system is constructed based on the deviation of the predicted SOC value from the preset upper and lower limits of safe SOC operation in the SOC trend curve. Specifically: ， in, To predict the SOC value, This is the preset upper limit for safe operation of the SOC. This is the preset lower limit for safe operation of the SOC.

8. The method for distributing distributed power sources and energy storage power within a microgrid according to claim 1, characterized in that: The multi-objective power allocation model with dynamic weight coefficients is solved using either particle swarm optimization or genetic algorithm. The population size of the particle swarm optimization algorithm is set to 50, the maximum number of iterations is 100, and the crossover probability and mutation probability of the genetic algorithm are 0.8 and 0.1 respectively.

9. The method for distributing distributed power sources and energy storage power within a microgrid according to claim 1, characterized in that: Distributed power sources include photovoltaic power generation units and wind power generation units. The power command value includes the active power adjustment of each distributed power source and the charging and discharging power value of the energy storage system.

10. The method for distributing distributed power sources and energy storage power within a microgrid according to claim 1, characterized in that: The real-time acquisition process also includes collecting the State of Health (SOH) data of the energy storage system. The prediction process for the State of Charge (SOC) trend curve further refines the prediction results based on the SOH data, with a correction coefficient... It shows a linear positive correlation with the SOH value, specifically: ,in and It is a preset constant, and 0 < ≤1, ≥0.

Images