Microgrid energy supply wave balancing method and system

Abstract

The present invention discloses a microgrid energy supply wave balancing method and system, which relates to the technical field of power grid heat supply balancing technology. The method comprises the following steps: deploying high-precision sensors to monitor the microgrid's energy supply, load demand, load output power, voltage, and frequency in real time; constructing an energy fluctuation depth model based on real-time and historical data, and outputting the energy fluctuation trend within a preset time; setting an adaptive energy regulation controller to automatically adjust the output power and operating status information of each power generation unit based on the prediction results through a power control algorithm; using the adjustment results of the power control algorithm and the current power grid state as input information through an energy balance control strategy to solve the energy balance parameter target and feed it back to the adaptive energy regulation controller; and dynamically adjusting the load distribution based on the final output result of the adaptive energy regulation controller. The present invention solves the problem that traditional heating systems cannot respond in real time, resulting in delayed balance regulation.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid heat supply balancing, and in particular to a microgrid energy supply wave balancing method and system. Background Art

[0002] Microgrids are usually composed of distributed energy resources (such as solar energy, wind energy, etc.) and energy storage systems (such as battery packs), aiming to achieve flexible and efficient use of energy.

[0003] In microgrids, the problem of unstable energy supply is becoming increasingly prominent due to the volatility and intermittency of renewable energy sources (such as solar and wind energy). This instability not only causes voltage fluctuations and frequency drift, but also affects the overall reliability and energy efficiency of the power system. Existing energy balancing methods usually rely on energy storage systems or backup power sources for regulation, but these methods have limited response speed and regulation accuracy when dealing with frequent and severe energy fluctuations. Therefore, it is necessary to develop a dynamic energy balancing control method to achieve smooth energy supply and ensure the efficient and stable operation of microgrids.

[0004] For example, the existing Chinese patent with publication number CN113572197A discloses a comprehensive self-consistent energy microgrid configuration method and energy control method based on hydrogen energy storage. The microgrid's intelligent management and control system collects the microgrid load's electricity, heat, and cooling energy demand in real time, as well as the real-time pressure value of hydrogen in the hydrogen storage tank in the microgrid, and relies on the grid-connected and grid-off discrimination module to identify the relationship between the microgrid and the grid. When the microgrid is connected to the grid or the microgrid is operating independently from the grid, the microgrid's energy control is based on the real-time pressure value of hydrogen in the hydrogen storage tank to characterize the energy storage capacity. Based on the real-time hydrogen pressure value and the set hydrogen pressure boundary threshold judgment result, and combined with whether the wind and photovoltaic power in the microgrid are output, and the microgrid load cooling and heating supply as constraints, the corresponding energy control strategy is activated. Based on the wind and solar power forecast results and the typical daily load curve, the invention adjusts the command deviation caused by wind and solar power randomness and load fluctuations in real time according to the output of each unit within the day, ensuring the real-time balance of system energy and power.

[0005] However, this design relies on wind and solar power forecasts to adjust the output of each unit in real time to mitigate command deviations caused by wind and solar power randomness and load fluctuations. However, the accuracy of wind and solar power forecasts can be affected by various factors, such as weather changes and equipment failures. This can lead to deviations between the forecast and actual values, making it impossible to maximize energy utilization efficiency while ensuring rapid response and high-precision regulation, thereby affecting the stability and reliability of the system. To this end, the present invention provides a method and system for balancing energy supply waves in a microgrid. Summary of the Invention

[0006] The object of the present invention is to provide a microgrid energy supply wave balancing method and system to solve the existing problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a microgrid energy supply wave balancing method, comprising the following steps:

[0008] S1. Deploy high-precision sensors to monitor the microgrid's energy supply, load demand, load output power, voltage, and frequency in real time. Build a deep energy fluctuation model based on real-time and historical data to output energy fluctuation trends within a preset timeframe.

[0009] S2. Setting an adaptive energy regulation controller to automatically adjust the output power and working status information of each power generation unit based on the prediction results through a power control algorithm;

[0010] S3. Using the energy balance control strategy, the adjustment result of the power control algorithm and the current grid state are used as input information to solve the energy balance parameter target and feed it back to the adaptive energy regulation controller;

[0011] S4. Dynamically adjust load distribution according to the final output result of the adaptive energy regulation controller.

[0012] A further improvement of the present invention is that the specific construction process of the energy fluctuation depth model includes:

[0013] S11, install current sensor and voltage sensor at the load point, obtain load demand at the same time, and set sampling period t to obtain input data x={x P ,x I ,x U}, where x P represents the load demand dataset, x I represents the current data set, x U represents the voltage dataset;

[0014] S12, create a special time node data set, for x in the input data P Add an additional tag hz to the input sample at the special time node to obtain the tag load requirement dataset x P _hz, add to the input data to get the input sequence inx = {x P ,x I ,x U ,x P _hz};

[0015] S13. Create an energy encoder to map the input sequence to a high-dimensional potential representation y i =f θ (inx), f θ() represents a high-dimensional mapping function, i∈{1,2,…,Tc}, Tc represents the total number of sampling times with the sampling period t as the step length, corresponding to time T; the feature y i Mapped to the hidden dimension, and the output of the hidden dimension is fed to the energy encoder to obtain x U and x I Append to the feature vector to obtain the first layer input feature λ1; pass λ1 through the energy encoder to obtain the first load feature output vector y1;

[0016] S14, convert the first load characteristic output vector again to obtain Obtain the energy fluctuation depth model , where W n Represents a mapping matrix, y T+n represents the predicted load demand at step n with a step length of t.

[0017] A further improvement of the present invention is that the energy encoder includes an N-layer convolutional structure, and a multi-head attention mechanism is applied to capture the dependency between the load demand, current and voltage in the input sequence inx, which is expressed as λ χ =MHA(||λ χ-1 ||2)+λ χ-1 , MHA() represents the multi-head attention block, which uses a feedforward neural network to perform a nonlinear transformation on the result of the attention mechanism, expressed as MLP() represents a feedforward neural network, χ∈{2,3,…,N+1}, λ χ represents the input features of the χth layer, represents the output features of the χth layer.

[0018] A further improvement of the present invention is that the adaptive energy regulation controller is equipped with a power control algorithm and an energy balance control strategy at the same time. First, the power control algorithm is used to quickly respond to energy fluctuations and perform rough regulation of voltage and power, and then the energy balance control strategy is used to perform fine regulation of voltage and power.

[0019] The present invention is further improved in that the power control algorithm specifically comprises the following steps: setting a target of minimizing the mismatch between supply and demand, and The variance of Indicates the first step after the current time T with a step length of t The output power of the actual load demand in each time step is: Indicates the first step after the current time T with a step length of t The actual load demand and set the output power constraint of the load demand Indicates the minimum output power required by the set load. Indicates the maximum output power required by the set load;

[0020] Output power, load demand, and voltage based on the current load demand and current and the predicted load demand Construct an optimization model for the next n steps, and obtain the time step based on the solution of the optimization model The optimal output power.

[0021] A further improvement of the present invention is that the optimization model of the future n-th step is expressed as:

[0022]

[0023] Among them, K sup_(1:n-1) It represents the power sequence of the power generation unit from the first step with a step length of t after time T to the n-1th step, α1, α2, α3 and α4 represent the weight coefficients, Indicates the set target output power.

[0024] A further improvement of the present invention is that the energy balance control strategy specifically comprises the following steps:

[0025] S31, initialize the particle state at the moment γ = 0, and set the initial weight for each particle man represents the total number of particles, and each particle consists of two instructions to increase or decrease the output power;

[0026] S32, extract the predicted Hourly load demand data and And initialize the fading factor σ;

[0027] S33, design particle performance fitness function FF;

[0028] S34. Calculate correction items Add the new correction term to the predicted energy allocation strategy to obtain the updated particle position;

[0029] S35. Update particle weights based on the proportion of particle performance fitness functions;

[0030] S36, collecting the cumulative weight distribution of each particle, extracting particles with unchanged weights for three consecutive times, and resampling each extracted particle;

[0031] S37, repeat S33 to S36 until the in Indicates the set energy supply and demand balance error threshold, and outputs the output power at this time to S4.

[0032] A further improvement of the present invention is that the particle performance fitness function FF is expressed as:

[0033]

[0034] in, Indicates the degree of voltage fluctuation. Indicates the degree of current fluctuation, which is obtained through standard deviation. Represents the weight coefficient in the particle performance fitness function. When it is less than the set demand threshold, when When it is greater than or equal to the set demand threshold,

[0035] The present invention is further improved in that the specific process of resampling each extracted particle includes: initializing a cumulative weight array whose length is the same as the number of particles and initializing all elements to 0; traversing the particle set and accumulating the weight of each particle to the corresponding position and subsequent positions of the cumulative weight array; initializing a new particle set whose size is the same as the original particle set; generating a random number uniformly distributed in the interval [0,1]; starting from the starting position of the cumulative weight array, gradually accumulating the weight with compensation j until the cumulative sum is greater than or equal to the random number; at this time, the particle corresponding to the stopping position is the selected particle; copying the selected particle to the new particle set; repeating the above process until the new particle set is filled.

[0036] In another aspect, the present invention provides a microgrid energy supply wave balancing system, comprising:

[0037] The real-time energy fluctuation monitoring module deploys high-precision sensors to monitor the microgrid's energy supply, load demand, load output power, voltage, and frequency in real time. It builds a deep energy fluctuation model based on real-time and historical data and outputs energy fluctuation trends within a preset timeframe.

[0038] The real-time energy fluctuation monitoring module includes a special time node data set creation unit and an energy encoder creation unit. The special time node data set creation unit is used to create x in the input data. P An additional tag is added to the input sample at the special time node to obtain a labeled load demand dataset; the energy encoder creation unit is used to map the input sequence to the high dimension and the hidden dimension to obtain a first load feature output vector as the input of the energy fluctuation deep model;

[0039] The dynamic energy balance control module is equipped with an adaptive energy regulation controller, which automatically adjusts the output power and working status information of each power generation unit based on the prediction results through the power control algorithm;

[0040] The energy balance optimization module uses the adjustment results of the power control algorithm and the current grid state as input information through the energy balance control strategy to solve the energy balance parameter target and feed it back to the adaptive energy regulation controller;

[0041] The dynamic load balancing module dynamically adjusts the load distribution according to the final output result of the adaptive energy regulation controller.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. The present invention first uses an adaptive energy regulation controller equipped with a power control algorithm and an energy balance control strategy to automatically adjust the output power and operating status information of each power generation unit based on the prediction results and the current grid status. The power control algorithm can quickly respond to energy fluctuations and perform coarse voltage and power adjustments to ensure stable operation of the grid. The energy balance control strategy further performs fine voltage and power adjustments to optimize energy distribution and improve energy utilization efficiency.

[0044] 2. Based on the output of the adaptive energy regulation controller, the system can dynamically adjust load distribution to ensure that each load receives a reasonable energy supply. This dynamic adjustment can be optimized in real time based on actual demand and grid status, improving the flexibility and responsiveness of the microgrid.

[0045] 3. The entire microgrid energy supply wave balancing method is intelligent and automated, reducing manual intervention and decision-making costs. The system can automatically learn, predict and optimize energy distribution strategies, improving the efficiency and accuracy of energy management.

[0046] 4. Through real-time monitoring, prediction and adaptive adjustment, the system can ensure the supply and demand balance and stable operation of the microgrid; this balance and stability helps to improve energy utilization efficiency, reduce energy waste, and enhance the reliability and resilience of the microgrid. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a flow chart of the microgrid energy supply wave balancing method of the present invention;

[0048] Figure 2 This is a flow chart of the power control algorithm in the microgrid energy supply wave balancing method of the present invention;

[0049] Figure 3 This is a framework diagram of the microgrid energy supply wave balancing system of the present invention. DETAILED DESCRIPTION

[0050] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0051] The term "and / or" simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, or B exists alone. Additionally, the character " / " generally indicates an "or" relationship between the related objects.

[0052] Example 1

[0053] Figure 1 The flowchart of the microgrid energy supply wave balancing method disclosed in this embodiment is shown, and the steps are as follows:

[0054] S1. Deploy high-precision sensors to monitor the microgrid's load demand, load output power, voltage, and frequency in real time. Build a deep energy fluctuation model based on real-time and historical data to output energy fluctuation trends within a preset timeframe.

[0055] The specific construction process of the energy fluctuation depth model includes:

[0056] S11, install current sensor and voltage sensor at the load point, obtain load demand at the same time, and set sampling period t to obtain input data x={x P ,x I ,x U}, where x P represents the load demand dataset, x I represents the current data set, x U represents the voltage dataset;

[0057] S12, create a special time node data set, for x in the input data P Add an additional tag hz to the input sample at the special time node to obtain the tag load requirement dataset x P _hz, add to the input data to get the input sequence inx = {x P ,x I ,x U ,x P _hz};

[0058] If a significant event (such as a load peak) is known to occur at time t, this embodiment can insert a marker hz at this location. During model processing, the marker hz is linearly projected to hz' and then fed into the Transformer along with the other linearly projected data. In the Transformer's multi-head attention mechanism, the marker hz helps the model better focus on the associated contextual information, thereby improving prediction accuracy.

[0059] Special time node data sets are determined by staff based on experience or experimental data.

[0060] Doing so allows the model to better capture information at key time points when processing complex time series data and make full use of this information when making predictions.

[0061] S13. Create an energy encoder to map the input sequence to a high-dimensional potential representation y i =f θ (inx), f θ () represents a high-dimensional mapping function, i∈{1,2,…,Tc}, Tc represents the total number of sampling times with the sampling period t as the step length, corresponding to time T; the feature y i Mapped to the hidden dimension W Trans , and feed the output of the hidden dimension to the energy encoder to obtain x U and x I Append to the feature vector to obtain the first layer input feature λ1; pass λ1 through the energy encoder to obtain the first load feature output vector y1;

[0062] The energy encoder consists of an N-layer convolutional structure and uses a multi-head attention mechanism to capture the dependencies between load demand, current, and voltage in the input sequence inx, denoted as λ χ =MHA(||λ χ-1 ||2)+λ χ-1 , MHA() represents the multi-head attention block, which uses a feedforward neural network to perform a nonlinear transformation on the result of the attention mechanism, expressed as MLP() represents a feedforward neural network, χ∈{2,3,…,N+1}, λ χ represents the input features of the χth layer, represents the output features of the χth layer.

[0063] Since the amount of data required for energy fluctuations increases with the number of users, there are high requirements for model scalability. The encoder part can process the entire sequence in parallel instead of step by step like the recurrent neural network (RNN), which greatly improves the computational efficiency and scalability.

[0064] Training is performed using a backpropagation algorithm that calculates the gradient of the loss function with respect to each weight and uses these gradients to update the weights in order to minimize the difference between the predicted and actual outputs, thereby optimizing model performance.

[0065] S14, convert the first load characteristic output vector again to obtain In order to predict the probability distribution of the occurrence of potential features at future moments, this embodiment uses the following formula to maintain the probability distribution of the potential features at future moments and The energy fluctuation depth model is obtained Among them, W n Represents a mapping matrix, y T+n represents the predicted load demand at step n with a step length of t.

[0066] S2. An adaptive energy regulation controller is set up to automatically adjust the output power and working status information of each power generation unit based on the prediction results through a power control algorithm; it is used to make decisions in a short time; the adaptive energy regulation controller is equipped with a power control algorithm and an energy balance control strategy at the same time, firstly quickly responding to energy fluctuations through the power control algorithm to perform rough voltage and power adjustments, and then finely adjusting the voltage and power through the energy balance control strategy.

[0067] The specific steps of the power control algorithm include: setting the goal of minimizing the mismatch between supply and demand, and The variance of Indicates the first step after the current time T with a step length of t The output power of the actual load demand in each time step is: Indicates the first step after the current time T with a step length of t The actual load demand and set the output power constraint of the load demand Indicates the minimum output power required by the set load. Indicates the maximum output power required by the set load;

[0068] Output power, load demand, and voltage based on the current load demand and current and the predicted load demand Construct an optimization model for the next n steps, and obtain the current time step based on the solution of the optimization model The optimal output power.

[0069] The optimization model for the next n-th step is expressed as:

[0070]

[0071] Among them, K sup_(1:n-1) It represents the power sequence of the power generation unit from the first step with a step length of t after time T to the n-1th step, α1, α2, α3 and α4 represent the weight coefficients, Indicates the set target output power.

[0072] S3. Using the energy balance control strategy, the adjustment result of the power control algorithm and the current grid state are used as input information to solve the energy balance parameter target and feed it back to the adaptive energy regulation controller;

[0073] S4. Dynamically adjust load distribution according to the final output result of the adaptive energy regulation controller.

[0074] Example 2

[0075] Figure 2 The power control algorithm flow chart of the microgrid energy supply wave balancing method disclosed in this embodiment is shown. The adjustment results of the power control algorithm and the current grid state are used as input information to solve the energy balance parameter target and feed it back to the adaptive energy regulation controller. The steps are as follows:

[0076] S31, initialize the particle state at the moment γ = 0, and set the initial weight for each particle man represents the total number of particles, and each particle consists of two instructions to increase or decrease the output power;

[0077] Conventional particle filter algorithms have poor tracking capabilities when facing sudden changes and cannot efficiently solve the problem to be solved by this embodiment, namely:

[0078] Once the power control algorithm generates a regulation result, it must be executed promptly. This typically involves sending control instructions to individual generating units to adjust their output power and operating status. These instructions might increase or decrease output power, or start or stop certain generating units. During execution, the accuracy and timeliness of these instructions must be ensured to avoid unnecessary impact or fluctuations on the power grid.

[0079] After implementing the adjustments, the system needs to monitor the grid's status in real time, including the stability of energy supply, the degree to which load demands are met, and the stability of voltage and frequency. If any issues or deviations are detected, the system needs to provide immediate feedback and make adjustments. This real-time monitoring and feedback mechanism is key to ensuring the continued effectiveness of the power control algorithm.

[0080] As time passes and grid conditions change, power control algorithms may require continuous optimization and adjustment. This includes updating prediction models based on new data and adjusting algorithm parameters to improve accuracy and efficiency. Continuous optimization and adjustment ensure that power control algorithms always adapt to the grid's actual needs and perform optimally.

[0081] Therefore, this embodiment, by combining strong tracking filtering, effectively generates a targeted correction term during the update process of the real-time measurement data fusion prediction error. This correction term is used to optimize the particle set in the PF algorithm, allowing particles to quickly move to areas with higher likelihood, thereby effectively curbing the phenomenon of particle degradation. The specific steps include:

[0082] S32, extract the predicted Time-of-day load demand dataset and And initialize the fading factor σ to adjust the weights of historical data and real-time observation data in the update process;

[0083] S33, design particle performance fitness function FF;

[0084] S34. Calculate correction items Add the new correction term to the predicted energy allocation strategy to obtain the updated particle position;

[0085] S35. Update particle weights based on the proportion of the particle performance fitness function; (the proportion of the fitness of the mth particle to the total fitness of man particles represents the new particle weight)

[0086] S36, collecting the cumulative weight distribution of each particle, extracting particles with unchanged weights for three consecutive times, and resampling each extracted particle;

[0087] S37, repeat S33 to S36 until the in Indicates the set energy supply and demand balance error threshold, and outputs the output power at this time to S4.

[0088] The particle performance fitness function FF is expressed as:

[0089]

[0090] in, Indicates the degree of voltage fluctuation. Indicates the degree of current fluctuation, which is obtained through standard deviation. Represents the weight coefficient in the particle performance fitness function. When it is less than the set demand threshold, when When it is greater than or equal to the set demand threshold, This prioritizes efficiency at low loads and stability at high loads.

[0091] The specific process of resampling each extracted particle includes: initializing a cumulative weight array whose length is the same as the number of particles and initializing all elements to 0; traversing the particle set, accumulating the weight of each particle to the corresponding position and subsequent positions of the cumulative weight array, for example, the weight of the mth particle will be added to the mth element and all subsequent elements of the cumulative weight array; initializing a new particle set whose size is the same as the original particle set; generating a random number uniformly distributed in the interval [0,1]; starting from the starting position of the cumulative weight array, gradually accumulating the weight with compensation j until the cumulative sum is greater than or equal to the random number; at this time, the particle corresponding to the stopping position is the selected particle; copying the selected particle to the new particle set; repeating the above process until the new particle set is filled.

[0092] The threshold and weight can be set by default according to the present invention, or can be set by the operator.

[0093] Example 3

[0094] Figure 3 The framework diagram of the microgrid energy supply wave balancing system of the present invention is shown. Based on the same inventive concept as that of Example 1 and Example 2, the present invention provides a microgrid energy supply wave balancing system, including:

[0095] The real-time energy fluctuation monitoring module deploys high-precision sensors to monitor the microgrid's energy supply, load demand, load output power, voltage, and frequency in real time. It builds a deep energy fluctuation model based on real-time and historical data and outputs energy fluctuation trends within a preset timeframe.

[0096] The dynamic energy balance control module is equipped with an adaptive energy regulation controller, which automatically adjusts the output power and working status information of each power generation unit based on the prediction results through the power control algorithm;

[0097] The energy balance optimization module uses the adjustment results of the power control algorithm and the current grid state as input information through the energy balance control strategy to solve the energy balance parameter target and feed it back to the adaptive energy regulation controller;

[0098] The dynamic load balancing module dynamically adjusts the load distribution according to the final output result of the adaptive energy regulation controller.

[0099] The real-time energy fluctuation monitoring module includes a special time node data set creation unit and an energy encoder creation unit;

[0100] The special time node data set creation unit is used to create x P Add an additional tag to the input sample at the special time node to obtain a tag load demand dataset;

[0101] The energy encoder creation unit is used to map the input sequence to the high dimension and the hidden dimension to obtain the first load feature output vector as the input of the energy fluctuation deep model.

[0102] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0103] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0104] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0105] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0106] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which are all protected by the present invention.

Claims

1. A microgrid energy supply wave balancing method, characterized by: The following steps are involved: S1. Deploy high-precision sensors to monitor the microgrid's energy supply, load demand, load output power, voltage, and frequency in real time. Build a deep energy fluctuation model based on real-time and historical data to output energy fluctuation trends within a preset timeframe. S2. Setting an adaptive energy regulation controller to automatically adjust the output power and working status information of each power generation unit based on the prediction results through a power control algorithm; S3. Using the energy balance control strategy, the adjustment result of the power control algorithm and the current grid state are used as input information to solve the energy balance parameter target and feed it back to the adaptive energy regulation controller; S4. Dynamically adjust the load distribution according to the final output result of the adaptive energy regulation controller. The specific construction process of the energy fluctuation depth model includes: S11, install current sensor and voltage sensor at the load point, obtain load demand at the same time, and set sampling period t to obtain input data x={x P ,x I ,x U }, where x P represents the load demand dataset, x I represents the current data set, x U represents the voltage dataset; S12, create a special time node data set, for x in the input data P Add an additional tag hz to the input sample at the special time node to obtain the tag load requirement dataset x P _hz, add to the input data to get the input sequence inx = {x P ,x I ,x U ,x P _hz}; S13. Create an energy encoder to map the input sequence to a high-dimensional potential representation y i =f θ (inx), f θ () represents a high-dimensional mapping function, i∈{1,2,…,Tc}, Tc represents the total number of sampling times with the sampling period t as the step length, corresponding to time T; the feature y i Mapped to the hidden dimension, and the output of the hidden dimension is fed to the energy encoder to obtain x U and x I Append to the feature vector to obtain the first layer input feature λ1; pass λ1 through the energy encoder to obtain the first load feature output vector y1; S14, convert the first load characteristic output vector again to obtain Obtain the energy fluctuation depth model Among them, W n Represents a mapping matrix, y T+n represents the predicted load demand at step n with a step length of t.

2. The microgrid energy supply wave balancing method according to claim 1, characterized in that: The energy encoder consists of an N-layer convolutional structure and uses a multi-head attention mechanism to capture the dependencies between load demand, current, and voltage in the input sequence inx, denoted as λ χ =MHA(||λ χ-1 ||2)+λ χ-1 , MHA() represents the multi-head attention block, which uses a feedforward neural network to perform a nonlinear transformation on the result of the attention mechanism, expressed as MLP() represents a feedforward neural network, χ∈{2,3,…,N+1}, λ χ represents the input features of the χth layer, represents the output features of the χth layer.

3. The microgrid energy supply wave balancing method according to claim 2, characterized in that: The adaptive energy regulation controller is equipped with a power control algorithm and an energy balance control strategy. First, the power control algorithm is used to quickly respond to energy fluctuations and perform rough voltage and power regulation, and then the energy balance control strategy is used to perform fine voltage and power regulation.

4. The microgrid energy supply wave balancing method according to claim 3, characterized in that: The specific steps of the power control algorithm include: setting the goal of minimizing the mismatch between supply and demand, and The variance of Indicates the first step after the current time T with a step length of t The output power of the actual load demand in each time step is: Indicates the first step after the current time T with a step length of t The actual load demand and set the output power constraint of the load demand Indicates the minimum output power required by the set load. Indicates the maximum output power required by the set load; Output power, load demand, and voltage based on the current load demand and current and the predicted load demand Construct an optimization model for the next n steps, and obtain the time step based on the solution of the optimization model The optimal output power.

5. The microgrid energy supply wave balancing method according to claim 4, characterized in that: The optimization model for the next n-th step is expressed as: Among them, K sup_(1:n-1) It represents the power sequence of the power generation unit from the first step with a step length of t after time T to the n-1th step, α1, α2, α3 and α4 represent the weight coefficients, Indicates the set target output power.

6. The microgrid energy supply wave balancing method according to claim 3, characterized in that: The specific steps of the energy balance control strategy include: S31, initialize the particle state at the moment γ = 0, and set the initial weight for each particle man represents the total number of particles, and each particle consists of two instructions to increase or decrease the output power; S32, extract the predicted Hourly load demand data and And initialize the fading factor σ; S33, design particle performance fitness function FF; S34. Calculate correction items Add the new correction term to the predicted energy allocation strategy to obtain the updated particle position; S35. Update particle weights based on the proportion of particle performance fitness functions; S36, collecting the cumulative weight distribution of each particle, extracting particles with unchanged weights for three consecutive times, and resampling each extracted particle; S37, repeat S33 to S36 until the in Indicates the set energy supply and demand balance error threshold, and outputs the output power at this time to S4.

7. The microgrid energy supply wave balancing method according to claim 6, characterized in that: The particle performance fitness function FF is expressed as: in, Indicates the degree of voltage fluctuation. Indicates the degree of current fluctuation, which is obtained through standard deviation. Represents the weight coefficient in the particle performance fitness function. When it is less than the set demand threshold, when When it is greater than or equal to the set demand threshold, 8. The microgrid energy supply wave balancing method according to claim 6, characterized in that: The specific process of resampling each extracted particle includes: initializing a cumulative weight array whose length is the same as the number of particles and initializing all elements to 0; traversing the particle set, accumulating the weight of each particle to the corresponding position and subsequent positions of the cumulative weight array; initializing a new particle set whose size is the same as the original particle set; generating a random number uniformly distributed in the interval [0,1]; starting from the starting position of the cumulative weight array, gradually accumulating the weight with compensation j until the cumulative sum is greater than or equal to the random number; at this time, the particle corresponding to the stopping position is the selected particle; copying the selected particle to the new particle set; repeating the above process until the new particle set is filled.

9. A microgrid energy supply wave balancing system, configured to execute the microgrid energy supply wave balancing method according to any one of claims 1 to 8, characterized in that: include: Real-time energy fluctuation monitoring module, deploying high-precision sensors to monitor the microgrid's energy supply, load demand, load output power, voltage, and frequency in real time; Build an energy fluctuation depth model through real-time data and historical data, and output the energy fluctuation trend within a preset time; The real-time energy fluctuation monitoring module includes a special time node data set creation unit and an energy encoder creation unit. The special time node data set creation unit is used to create x in the input data. P Add an additional tag to the input sample at the special time node to obtain a tag load demand dataset; The energy encoder creation unit is used to map the input sequence onto the high dimension and the hidden dimension to obtain a first load feature output vector as an input to the energy fluctuation deep model; The dynamic energy balance control module is equipped with an adaptive energy regulation controller, which automatically adjusts the output power and working status information of each power generation unit based on the prediction results through the power control algorithm; The energy balance optimization module uses the adjustment results of the power control algorithm and the current grid state as input information through the energy balance control strategy to solve the energy balance parameter target and feed it back to the adaptive energy regulation controller; The dynamic load balancing module dynamically adjusts the load distribution according to the final output result of the adaptive energy regulation controller.

Images