Energy cluster and energy storage collaborative virtual power plant control method, system and device

Abstract

The present application relates to the technical field of virtual power plant control, and provides a kind of based on energy cluster and energy storage collaborative virtual power plant control method, system and device.The method comprises: collecting the real-time data of energy equipment, and the real-time data is preprocessed to obtain preprocessed real-time data;According to the preprocessed real-time data, use LSTM attention map neural network to obtain prediction result;According to the prediction result, the configuration and collaborative optimization of multivariate energy storage system are carried out to obtain the charge-discharge power reference value;According to the charge-discharge power reference value, the generator set in virtual power plant is adjusted.The present application realizes the efficient and stable operation of virtual power plant by constructing the optimization control method of distributed energy cluster and multivariate energy storage collaboration, improves energy utilization and support ability to power grid.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of virtual power plant control, and provides a virtual power plant control method, system and device based on energy cluster and energy storage cooperation. BACKGROUND

[0002] In the existing distributed energy individual control mode, solar photovoltaic power stations, wind power plants and the like are each for war, lacking effective cooperation mechanisms. When sunlight is suddenly blocked by clouds, the photovoltaic power station power generation power drops sharply, and at this time, if the wind power plant cannot timely supplement the power gap due to low wind speed, and there is no other energy or energy storage system around for adjustment, it will lead to unstable power supply in the local area. When the light is sufficient and the wind is strong, each energy overgenerates, but cannot cooperate with each other to reasonably store or transport the excess power, ultimately causing a large amount of energy to be wasted.

[0003] A single energy storage control mode greatly limits the performance of the energy storage system. Taking lithium batteries as an example, in some industrial production high-power power consumption scenarios, the energy storage system needs to quickly release a large amount of power. Due to the characteristics of lithium batteries, it is difficult to maintain good performance in the process of high-power fast charging and discharging, and problems such as unstable voltage and shortened service life may occur. Although supercapacitors can quickly respond to high-power demand, their energy storage capacity is limited and cannot meet long-term and large-scale energy storage needs. This makes the energy storage system appear to be inadequate when facing complex and diverse energy fluctuations.

[0004] The traditional centralized control architecture has obvious disadvantages in the application of virtual power plants. Because distributed energy and energy storage devices are widely distributed, a large amount of data needs to be transmitted to the central controller for processing. Information inevitably delays in the transmission process, and the central controller also needs time to process data. When the power grid load suddenly increases, or the distributed energy power generation power changes greatly due to weather and other factors, the central controller has difficulty in making accurate and effective control instructions in time, and cannot quickly adjust the distribution of energy and the working state of the energy storage system, which seriously affects the response ability of the virtual power plant to the dynamic changes of the power grid. SUMMARY

[0005] The present application aims to at least solve one of the technical problems in the related art. To this end, the present application provides a virtual power plant control method, system and device based on energy cluster and energy storage cooperation, which effectively solves the problem of distributed energy being in charge of its own affairs and lacking cooperation, avoids energy waste, guarantees power supply stability and reliability, and improves energy utilization efficiency.

[0006] The present application provides a virtual power plant control method based on energy cluster and energy storage cooperation, comprising:

[0007] S1: collect real-time data of energy equipment, and pre-process the real-time data to obtain pre-processed real-time data;

[0008] S2: using an LSTM attention map neural network to predict according to the pre-processed real-time data to obtain a prediction result;

[0009] S3: configuring and cooperatively optimizing a multi-element energy storage system according to the prediction result to obtain a charge-discharge power reference value;

[0010] S4: adjusting a generator set in a virtual power plant according to the charge-discharge power reference value.

[0011] According to the energy cluster and energy storage collaborative virtual power plant control method provided by the application, the real-time data includes power generation power, equipment state and environmental parameters.

[0012] According to the energy cluster and energy storage collaborative virtual power plant control method provided by the application, the pre-processing method comprises:

[0013] S11: sorting the collected real-time data in time sequence to obtain original real-time data;

[0014] S12: screening out original real-time data with a length less than a length threshold to obtain screened original real-time data;

[0015] S13: filling in missing values of the screened original real-time data to obtain the pre-processed real-time data.

[0016] According to the energy cluster and energy storage collaborative virtual power plant control method provided by the application, step S2 comprises:

[0017] S21: optimizing the pre-processed real-time data according to a dynamic gated convolution to obtain optimized real-time data;

[0018] S22: using an LSTM algorithm to predict the pre-processed real-time data to obtain a hidden state;

[0019] S23: dynamically allocating and optimizing the hidden state using a graph neural network and an attention mechanism to obtain a prediction result.

[0020] According to the energy cluster and energy storage collaborative virtual power plant control method provided by the application, step S21 comprises:

[0021] S211: using a dynamic gated convolution to optimize a prediction input:

[0022]

[0023] wherein, is a final convolution kernel, is a first activation function, is a multi-layer perception function, is a past pre-processed real-time data of a time instant, is a current time instant, is a backtracking time step, is a window start time, is an initial convolution kernel;

[0024] S212: calculating a learnable dilation rate :

[0025]

[0026] wherein, is a rounding function, is a dilation rate learning parameter, is a prediction scaling factor function, represents a pre-processed real-time data of a current time instant, is a maximum dilation rate;

[0027] S213: calculating a channel gate:

[0028]

[0029] wherein, is a channel gate of an energy device, is a serial number of the energy device, , is a total number of energy devices, is a normalization weight parameter, is a global average pooling function;

[0030] S214: fusing an LSTM input:

[0031]

[0032]

[0033] wherein, is an LSTM additional value, is a one-dimensional convolution function, is a channel gate of an , is a channel gate of an , is an optimized real-time data.

[0034] According to the energy cluster and energy storage collaborative virtual power plant control method provided by the application, step S23 comprises:

[0035] S231: using a graph neural network to optimize the hidden state:

[0036]

[0037] wherein, is the updated power station hidden state, is a second activation function, is a neighbor power station of the power station , is a set of neighbor power stations of the power station , is the degree of the power station , is the degree of the power station ; is a learnable matrix; is the original hidden state of the neighbor power station at time t;

[0038] S232: calculating the importance score of each time step:

[0039]

[0040] wherein, is the importance score of the current time, is an attention initial weight matrix, denotes matrix transposition, is a hyperbolic tangent activation function, is a trainable weight, is a bias, is the updated hidden state;

[0041] S233: calculating the context vector :

[0042]

[0043]

[0044] wherein, is a context weight, is the importance score of the neighbor power station;

[0045] S234: combining the context vector and the updated hidden state to obtain a prediction result:

[0046]

[0047] wherein, is a prediction result, is an output trainable weight, is an output bias.

[0048] According to the energy cluster and energy storage collaborative virtual power plant control method provided by the application, step S3 comprises:

[0049] S31: establishing different SOC dynamic equations according to different energy storage technologies;

[0050] S32: performing MPC-RL modeling on the power grid system to obtain an MPC-RL algorithm;

[0051] S33: inputting the SOC dynamic equation and the prediction result into the MPC-RL algorithm to obtain a charging and discharging power reference value.

[0052] The application also provides an energy cluster and energy storage collaborative virtual power plant control system, comprising:

[0053] A data preprocessing module: collecting real-time data of energy equipment and preprocessing the real-time data to obtain preprocessed real-time data;

[0054] A data prediction module: using an LSTM attention map neural network to predict a prediction result according to the preprocessed real-time data;

[0055] A data collaborative optimization module: configuring and collaboratively optimizing a multi-element energy storage system according to the prediction result to obtain a charging and discharging power reference value;

[0056] A generator set adjustment module: adjusting the generator set in the virtual power plant according to the charging and discharging power reference value.

[0057] The application also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the energy cluster and energy storage collaborative virtual power plant control method according to any one of the above embodiments when executing the program.

[0058] The one or more technical solutions in the embodiments of the application have at least one of the following technical effects:

[0059] The energy cluster and energy storage collaborative virtual power plant control method, system and device provided by the application can realize real-time collection and accurate prediction of distributed energy data by using the Internet of Things technology and the LSTM model and adding an attention mechanism in the model, and can construct a global optimization model to realize reasonable allocation of energy. The application effectively solves the problem of distributed energy being managed separately and lacking collaboration, avoids energy waste, guarantees power supply stability and reliability, and improves energy utilization efficiency.

[0060] Precise mathematical models of different energy storage technologies are established, and a hierarchical control strategy is adopted. The upper-level MPC-RL algorithm performs global optimization, while the lower-level energy storage devices implement precise local control. This approach takes into account grid power fluctuations, the stability of energy storage charging and discharging, and battery health, thereby improving the adaptability of the energy storage system under different time scales and power demands, and fully leveraging the characteristics of different energy storage technologies.

[0061] The optimization control aims to minimize the overall operating cost and maximize the energy utilization rate of the virtual power plant. By rationally arranging the generation and discharging of distributed energy generation and energy storage systems, energy loss and operating costs are reduced, thereby improving the economic efficiency of the virtual power plant.

[0062] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0064] Figure 1 This is a flowchart illustrating the virtual power plant control method based on energy cluster and energy storage collaboration provided by the present invention.

[0065] Figure 2 This is a structural block diagram of the virtual power plant control device based on energy cluster and energy storage collaboration provided by the present invention.

[0066] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0067] Figure label:

[0068] 101. Data preprocessing module; 102. Data prediction module; 103. Data collaborative optimization module; 104. Generator set regulation module; 810. Processor; 820. Communication interface; 830. Memory; 840. Communication bus. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. The following embodiments are used to illustrate this invention but cannot be used to limit the scope of this invention.

[0070] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0071] The following is combined Figures 1 to 3 This invention is described.

[0072] Example

[0073] like Figure 1 As shown, Figure 1 The flowchart illustrates a virtual power plant control method based on energy clusters and energy storage collaboration, including the following steps:

[0074] S1: Collect real-time data from energy equipment and preprocess the real-time data to obtain preprocessed real-time data;

[0075] S2: Based on the preprocessed real-time data, an LSTM attention graph neural network is used to make predictions to obtain prediction results;

[0076] S3: Based on the prediction results, perform multi-element energy storage system configuration and collaborative optimization to obtain a reference value for charging and discharging power;

[0077] S4: Adjust the generator sets in the virtual power plant according to the charging and discharging power reference value.

[0078] Specifically, real-time data includes power generation, equipment status, and environmental parameters. Using IoT technology, numerous sensors are deployed on distributed energy devices (such as photovoltaic panels and wind turbines) to collect data on power generation, equipment status, and environmental parameters (light intensity, wind speed, temperature, etc.) in real time. This data is then transmitted to a central control platform via wireless communication networks (such as ZigBee and LoRa).

[0079] Specifically, preprocessing methods include:

[0080] S11: Sort the collected real-time data in chronological order to obtain the raw real-time data;

[0081] S12: Filter out the raw real-time data whose length is less than the length threshold to obtain the filtered raw real-time data;

[0082] S13: Fill in missing values ​​in the filtered raw real-time data to obtain the preprocessed real-time data.

[0083] Specifically, step S2 includes:

[0084] S21: Optimize the preprocessed real-time data input by dynamically gated convolution to obtain optimized real-time data.

[0085] Step S21 includes:

[0086] S211: Optimizes prediction input using dynamic gated convolution:

[0087]

[0088] in, For the final convolution kernel, As the first activation function, For multilayer perceptron functions, For the past Preprocessed real-time data at each moment, At this moment, Rewind time step, Window start time;

[0089] S212: Calculate the learnable expansion rate :

[0090]

[0091] in, For the floor function, For the learning parameters of the expansion rate, To predict the scaling factor function, This represents the preprocessed real-time data at the current moment. The maximum expansion rate;

[0092] S213: Calculate channel gating:

[0093]

[0094] in, For the first Access control for individual energy devices For the ordinal number of the energy equipment, , The total number of energy devices, For normalized weight parameters, This is the global average pooling function;

[0095] S214: Fusion LSTM input:

[0096]

[0097]

[0098] in, Adding value to LSTM It is a one-dimensional convolution function. For the first Energy equipment , For the first Energy equipment , This is the optimized real-time data.

[0099] S22: The hidden state is obtained by using the LSTM algorithm to predict and preprocess the real-time data. Step S22 includes:

[0100] S221: Decide which information should be added to current memory:

[0101]

[0102] In the formula, The input gate state determines which information should be added to the current memory. The sigmoid activation function is used to compress the input to between 0 and 1, controlling the information flow. This is the concatenation of the hidden state from the previous moment and the current input. For input gate paranoia, This is the input gate weight matrix.

[0103] S222: Decide which states should be discarded from old memories:

[0104]

[0105] In the formula In the forget gate state, Here is the forget gate weight matrix. This is the forget gate bias term.

[0106] S223: Determine which information should be output to the next time step:

[0107]

[0108] In the formula Output gate state This is the output gate weight matrix. This is the output gate bias term.

[0109] S224: Calculate the hidden state:

[0110]

[0111]

[0112] For convolution operations, This is the hidden state from the previous moment. Candidate memory stake matrix For candidate memory bias terms, This is a digit-wise multiplication.

[0113] S23: Dynamically allocate and optimize the hidden states using a graph neural network and an attention mechanism to obtain the prediction result. Step S23 includes:

[0114] S231: Optimize the hidden state using a graph neural network:

[0115]

[0116] in, For the updated power station The hidden state, For the second activation function, For power station The neighboring power station, For power station The collection of neighboring power stations, For power station The degree, For power station The degree; It is a learnable matrix; For neighboring power stations The original hidden state at time t;

[0117] S232: Calculate the importance score for each time step:

[0118]

[0119] in, Score the importance of the current moment. The initial weight matrix for attention. Indicates matrix transpose. The hyperbolic tangent activation function is used. For trainable weights, This is the bias value. This is the updated hidden state;

[0120] S233: Calculate the context vector :

[0121]

[0122]

[0123] in, For context weights, Score the importance of neighboring power stations;

[0124] S234: Combine the context vector with the updated hidden state to obtain the prediction result:

[0125]

[0126] in, For the predicted results, To output trainable weights, This is the output bias.

[0127] Example input into model calculation:

[0128] Task: Predict the photovoltaic power generation at 12 noon tomorrow.

[0129] Input data: Light intensity over the past 3 hours (one data point per hour):

[0130] (Normalized values, 0 represents no light, 1 represents the strongest light), where, To illustrate the input data set, This is the first set of schematic data. This is the second set of sample data. This is the third set of schematic data.

[0131] LSTM hidden states (assuming they have been calculated through forward propagation, representing the trend of light intensity changes and device status)

[0132]

[0133] This is the first indicated hidden state. This is the second indicated hidden state. This indicates the third hidden state.

[0134] Step 1: Calculate attention score

[0135] Concatenating input and hidden states:

[0136]

[0137] Calculate the score:

[0138]

[0139] in, The first indicator of importance is scored. The second indicator of importance is scored. The third indicator is given a score based on its importance.

[0140] Step 2: Normalize attention weights

[0141] ;

[0142] ;

[0143] in, The first example illustrates the context weight. The second illustration shows the context weight. This is the third example of context weight.

[0144] Step 3: Weight and summarize all hidden states to generate a context vector.

[0145] ;

[0146] First dimension: ;

[0147] Second dimension: ;

[0148] The final result is: ;

[0149] Step 4: Final Prediction Output

[0150] ;

[0151] Assumption:

[0152] ,

[0153]

[0154] Traditional LSTM computation process:

[0155] .

[0156] With the addition of an attention mechanism, the core advantage of the LSTM model lies in its ability to dynamically focus on key historical information, significantly improving the accuracy and robustness of predictions. Traditional LSTM treats all historical data equally, while the attention mechanism automatically calculates weights to prioritize extracting features most relevant to the current prediction, while ignoring irrelevant noise.

[0157] By leveraging IoT technology and an LSTM model, incorporating attention mechanisms and graph neural network algorithms, real-time data collection and accurate prediction of distributed energy resources are achieved. A global optimization model is then constructed to realize rational energy allocation. This effectively solves the problems of fragmented and uncoordinated distributed energy resources, avoids energy waste, ensures power supply stability and reliability, and improves energy utilization efficiency.

[0158] Specifically, step S3 includes:

[0159] S31: Establish dynamic equations for different SOCs based on different energy storage technologies;

[0160] Specifically, taking lithium batteries as an example, the dynamic equation for their state of charge (SOC) is:

[0161]

[0162] in, Let V be the derivative of the state of charge (SOC) with respect to time, reflecting the rate of change of the state of charge over time. I is the battery charging / discharging current, and Q is the battery's rated capacity. For a supercapacitor, the relationship between its voltage V and the battery's rated capacity Q is: Where C is the capacitance value.

[0163] S32: To obtain the MPC-RL algorithm by performing MPC-RL modeling on the power grid system, the following steps are included:

[0164] S321: Upper-level global optimization: An optimization model is established with the objective function of minimizing the overall operating cost of the virtual power plant and maximizing energy utilization. This considers constraints such as charging / discharging power limitations and state-of-charge limitations of the energy storage system. Objective function It can be represented as:

[0165]

[0166] in, for The cost of distributed energy generation at all times for Operating costs of real-time energy storage systems This indicates taking the minimum value. Total duration.

[0167] The constraints include:

[0168] (Power balance constraints);

[0169] (State of charge constraints);

[0170] (Energy storage discharge power constraint);

[0171] (Energy storage charging power constraints);

[0172] In the formula, for Total power generation at all times for The discharge power of the energy storage system at all times. for Real-time energy storage system charging power, for Constant load power, This is the minimum state of charge value. for State of charge value at time, This is the maximum state of charge value. This represents the minimum discharge power value of the energy storage system. This represents the maximum discharge power of the energy storage system. This represents the minimum charging power value for the energy storage system. The maximum charging power value for the energy storage system.

[0173] The model predictive control (MPC) method is used to solve this optimization problem, and a reference value for the discharge power of the energy storage system at a future time is obtained. Reference value for charging power of energy storage system .

[0174] S323: Each energy storage device performs local control based on the reference values ​​sent down from the upper level and its own status. For example, a lithium battery energy storage system adjusts the current of the charging circuit according to the reference charging power to achieve charging control; a supercapacitor rapidly releases energy according to the reference discharging power.

[0175] S33: The dynamic equation and prediction results of the SOC are input into the MPC-RL algorithm to obtain the reference value of the charging and discharging power, including the following steps:

[0176] S331: Define variables:

[0177] Let control variables :

[0178] Let the state variable be... :

[0179] RL_state(T) represents the auxiliary state provided by RL:

[0180] Prediction error :

[0181] In the formula, Let t be the actual power generation value. Let t be the predicted power generation value. This represents the modulo operation of a vector.

[0182] Power grid fluctuations: ;

[0183] Historical control instructions ;

[0184] RL Dynamic Weights : ;

[0185] The dynamic weight (0~1) of the cost of distributed energy generation is adjusted by RL based on electricity price and forecast error;

[0186] The dynamic weight (0~1) of energy storage operating costs is adjusted by RL based on SOC and grid stability.

[0187] S332: Modeling based on the optimization problem:

[0188] Corrected objective function The expression is:

[0189]

[0190] in, This represents the control variable corresponding to the minimum value. Penalty terms generated for RL.

[0191]

[0192] in, For extreme SOC coefficients, This is the frequency fluctuation coefficient.

[0193] S333: Cooperative solution process:

[0194] Real-time calculation by the RL module:

[0195] Input: Prediction result ;

[0196] Output: Dynamic weight λ;

[0197] MPC solution:

[0198] Input: New optimization function, constraints, and dynamic weights λ;

[0199] Output: Optimal control sequence;

[0200] Finally, a closed-loop solution is obtained, and the process is repeated to find a charging and discharging strategy that balances economy and stability under the guidance of RL.

[0201] Precise mathematical models of different energy storage technologies are established, and a hierarchical control strategy is adopted. The upper-level MPC-RL algorithm performs global optimization, while the lower-level energy storage devices implement precise local control. This approach takes into account grid power fluctuations, the stability of energy storage charging and discharging, and battery health, thereby improving the adaptability of the energy storage system under different time scales and power demands, and fully leveraging the characteristics of different energy storage technologies.

[0202] The optimization control aims to minimize the overall operating cost and maximize the energy utilization rate of the virtual power plant. By rationally arranging the generation and discharging of distributed energy generation and energy storage systems, energy loss and operating costs are reduced, thereby improving the economic efficiency of the virtual power plant.

[0203] Table 1. Comparison of the present invention and traditional methods

[0204]

[0205] As shown in Table 1, which is a comparison table between the present invention and the traditional method, it can be seen that the present invention has great advantages over the traditional method.

[0206] like Figure 2 As shown, the present invention also provides a virtual power plant control system based on energy cluster and energy storage collaboration, comprising:

[0207] Data preprocessing module 101: Collects real-time data from energy equipment and preprocesses the real-time data to obtain preprocessed real-time data;

[0208] Data prediction module 102: Based on the preprocessed real-time data, it uses an LSTM attention graph neural network to perform prediction and obtain the prediction result;

[0209] Data collaborative optimization module 103: Based on the prediction results, it performs multi-element energy storage system configuration and collaborative optimization to obtain charging and discharging power reference values;

[0210] Generator set adjustment module 104: Adjusts the generator set in the virtual power plant according to the charging and discharging power reference value.

[0211] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 can call logical instructions in the memory 830 to execute a control method for a virtual power plant based on energy cluster and energy storage collaboration, the method including:

[0212] S1: Collect real-time data from energy equipment and preprocess the real-time data to obtain preprocessed real-time data;

[0213] S2: Based on the preprocessed real-time data, an LSTM attention graph neural network is used to make predictions to obtain prediction results;

[0214] S3: Based on the prediction results, perform multi-element energy storage system configuration and collaborative optimization to obtain a reference value for charging and discharging power;

[0215] S4: Adjust the generator sets in the virtual power plant according to the charging and discharging power reference value.

[0216] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0217] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, the computer is able to execute the virtual power plant control method based on energy cluster and energy storage collaboration provided by the above methods.

[0218] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the aforementioned control methods for virtual power plants based on energy clusters and energy storage collaboration.

[0219] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0220] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0221] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0222] It should be noted that the embodiments of this disclosure can be implemented using hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a programmable memory or a data carrier such as an optical or electronic signal carrier.

[0223] Furthermore, although the operation of the methods of this disclosure is described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Rather, the steps depicted in the flowcharts may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps. It should also be noted that the features and functions of two or more devices according to this disclosure may be embodied in one device. Conversely, the features and functions of one device described above may be further divided and embodied by multiple devices.

[0224] While this disclosure has been described with reference to several specific embodiments, it should be understood that this disclosure is not limited to the specific embodiments disclosed. This disclosure is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims

1. A control method for a virtual power plant based on the coordinated operation of energy clusters and energy storage, characterized in that, include: S1: Collect real-time data from energy equipment and preprocess the real-time data to obtain preprocessed real-time data. The preprocessing methods include: S11: Sort the collected real-time data in chronological order to obtain the raw real-time data; S12: Filter out the raw real-time data whose length is less than the length threshold to obtain the filtered raw real-time data; S13: Fill in missing values ​​in the filtered raw real-time data to obtain the preprocessed real-time data; S2: Based on the preprocessed real-time data, an LSTM attention map neural network is used to perform prediction to obtain the prediction result. Step S2 includes: S21: Optimize the preprocessed real-time data input by dynamic gated convolution to obtain optimized real-time data; S22: Use the LSTM algorithm to predict and preprocess real-time data to obtain the hidden state; S23: Use graph neural networks and attention mechanisms to dynamically allocate and optimize the hidden states to obtain prediction results; S3: Based on the prediction results, perform multi-element energy storage system configuration and collaborative optimization to obtain a reference value for charging and discharging power. Step S3 includes: S31: Establish dynamic equations for different SOCs based on different energy storage technologies; S32: MPC-RL modeling of the power grid system yields the MPC-RL algorithm; S321: Upper-level global optimization objective function It can be represented as: in, for The cost of distributed energy generation at all times for Operating costs of real-time energy storage systems This indicates taking the minimum value, where T is the total duration; The constraints include: ； ； ； ； In the formula, for Total power generation at all times for The discharge power of the energy storage system at all times. for Real-time energy storage system charging power, for Constant load power, This is the minimum state of charge value. for State of charge value at time, This is the maximum state of charge value. This represents the minimum discharge power value of the energy storage system. This represents the maximum discharge power of the energy storage system. This represents the minimum charging power value for the energy storage system. The maximum charging power value for the energy storage system; The optimization problem is solved using model predictive control methods to obtain a reference value for the discharge power of the energy storage system at a future time. Reference value for charging power of energy storage system ; S323: Each energy storage device performs local control based on the reference values ​​issued by the upper level and its own status; S33: The dynamic equation and prediction results of the SOC are input into the MPC-RL algorithm to obtain the reference value of charging and discharging power; S331: Define variables: Let control variables : Let the state variable be... : RL_state() represents the auxiliary state provided by RL: Prediction error : In the formula, Let t be the actual power generation value. Let t be the predicted power generation value. Indicates the modulus of a vector; Power grid fluctuations: ; Historical control instructions ; RL Dynamic Weights : ; Dynamic weighting of distributed energy generation costs; : Dynamic weighting of energy storage operating costs; S332: Modeling based on the optimization problem: Corrected objective function The expression is: in, This represents the control variable corresponding to the minimum value. Penalty terms generated for RL; in, For extreme SOC coefficients, This is the frequency fluctuation coefficient; S333: Cooperative solution process: Real-time calculation by the RL module: Input: Prediction result ; Output: Dynamic weight λ; MPC solution: Input: New optimization function, constraints, and dynamic weights λ; Output: Optimal control sequence; Finally, the closed-loop solution is obtained, and the process is repeated to find a charging and discharging strategy that balances economy and stability under the guidance of RL. S4: Adjust the generator sets in the virtual power plant according to the charging and discharging power reference value.

2. The virtual power plant control method based on energy cluster and energy storage collaboration according to claim 1, characterized in that, The real-time data includes power generation, equipment status, and environmental parameters.

3. The virtual power plant control method based on energy cluster and energy storage collaboration according to claim 1, characterized in that, Step S21 includes: S211: Optimizes prediction input using dynamic gated convolution: in, For the final convolution kernel, As the first activation function, For multilayer perceptron functions, For the past Preprocessed real-time data at each moment, At this moment, Rewind time step, Window start time, Use the initial convolution kernel; S212: Calculate the learnable expansion rate : in, For the floor function, For the learning parameters of the expansion rate, To predict the scaling factor function, This represents the preprocessed real-time data at the current moment. The maximum expansion rate; S213: Calculate channel gating: in, For the first Access control for individual energy devices For the ordinal number of the energy equipment, , The total number of energy devices, For normalized weight parameters, This is the global average pooling function; S214: Fusion LSTM input: in, Adding value to LSTM It is a one-dimensional convolution function. For the first Energy equipment , For the first Energy equipment , This is the optimized real-time data.

4. The virtual power plant control method based on energy cluster and energy storage collaboration according to claim 3, characterized in that, Step S23 includes: S231: Optimize the hidden state using a graph neural network: in, For the updated power station The hidden state, For the second activation function, For power station The neighboring power station, For power station The collection of neighboring power stations, For power station The degree, For power station The degree; It is a learnable matrix; For neighboring power stations The original hidden state at time t; S232: Calculate the importance score for each time step: in, Score the importance of the current moment. The initial weight matrix for attention. Indicates matrix transpose. The hyperbolic tangent activation function is used. For trainable weights, This is the bias value. This is the updated hidden state; S233: Calculate the context vector : in, For context weights, Score the importance of neighboring power stations; S234: Combine the context vector with the updated hidden state to obtain the prediction result: in, For the predicted results, To output trainable weights, This is the output bias.

5. A virtual power plant control system based on energy clustering and energy storage collaboration, used to execute the virtual power plant control method based on energy clustering and energy storage collaboration as described in any one of claims 1 to 4, characterized in that, include: Data preprocessing module: Collects real-time data from energy equipment and preprocesses the real-time data to obtain preprocessed real-time data; Data prediction module: Based on the preprocessed real-time data, an LSTM attention graph neural network is used to perform predictions to obtain prediction results; Data collaborative optimization module: Based on the prediction results, it performs multi-element energy storage system configuration and collaborative optimization to obtain charging and discharging power reference values; Generator set adjustment module: Adjusts the generator sets in the virtual power plant according to the charging and discharging power reference value.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the virtual power plant control method based on energy cluster and energy storage collaboration as described in any one of claims 1 to 4.

Images