Operation optimization method and system of source-grid-load-storage system based on digital twin

Abstract

The present application belongs to the field of smart grid construction technology, and is a method and system for optimizing the operation of a source-grid-load-storage system based on digital twins, including: a test entity, which collects information about a physical entity through various sensors and inputs the information about the physical entity into a virtual entity; a virtual entity, which includes a prediction model for predicting the power generation of renewable energy based on the information about the physical entity; a multi-objective optimization model for real-time optimization and determination of the charging and discharging power of energy storage, and the interconnection line power between the source-grid-load-storage system and the distribution network; and a health index calculation model for calculating the health index of the physical entity and dividing the health status level according to the health index; a digital twin model for mapping the physical entity based on the output results of the virtual entity. The present application can compensate the prediction value based on the similar day meteorological search algorithm, thereby determining the final power prediction value of renewable energy power generation, thus avoiding the situation where the prediction error gradually increases after long-term operation.

Description

Technical Field

[0001] The present application belongs to the technical field of smart grid construction, and specifically relates to a method and system for optimizing the operation of a source-grid-load-storage system based on digital twins. Background Art

[0002] The integration of power source, grid, load and storage is an important part of smart grid construction. It refers to the realization of efficient energy utilization and intelligent scheduling through the comprehensive use of power source, grid, load and energy storage technology to cope with the increasingly complex electricity market demand and environmental protection requirements. It is of great significance to improve energy utilization, promote the consumption of new energy and promote energy transformation.

[0003] As the scale of source-grid-load-storage systems continues to grow, the amount of panoramic real-time information such as load, weather, and equipment continues to increase, greatly increasing the dimension of the information matrix, slowing down the time for data processing, and introducing a time lag in the transmission of information. Therefore, traditional modeling and simulation technologies are difficult to meet the operational needs of source-grid-load-storage systems. Traditional modeling methods are limited by the scale of the problem, the dimension of the variables, the deterministic boundaries, and the reliance on precise physical models. When dealing with high-dimensional, time-varying, and nonlinear problems, it is difficult to ensure the reliability of calculations based on computational timeliness. Summary of the Invention

[0004] An embodiment of the present application provides a source-grid-load-storage system operation optimization system based on digital twins.

[0005] On the other hand, an embodiment of the present application provides a source-grid-load-storage system operation optimization method based on digital twins to solve the lag problem of traditional methods in high-dimensional time-varying scenarios.

[0006] This application is implemented in this way.

[0007] An embodiment of the present application provides a digital twin-based source-grid-load-storage system operation optimization system, which includes:

[0008] The test entity collects information about the physical entity through various sensors and inputs the information about the physical entity into the virtual entity. The information about the physical entity includes meteorological data, historical renewable energy power generation, and equipment parameters in the physical entity.

[0009] The virtual entity includes a prediction model for predicting renewable energy power generation based on information from the physical entity; a multi-objective optimization model for real-time optimization and determination of the charging and discharging power of energy storage, and the power of the interconnection line between the source-grid-load-storage system and the distribution network; and a health index calculation model for calculating the health index of the physical entity and classifying the health status according to the health index;

[0010] The digital twin model is used to map physical entities based on the output results of virtual entities, including wind power generation models, photovoltaic power generation models, and energy storage models.

[0011] Furthermore, the physical entity is a source-grid-load-storage system, which adopts an AC / DC hybrid architecture and includes a photovoltaic array, a wind turbine, an external power grid, an AC load, a DC load, a battery energy storage and a converter. The photovoltaic array and the battery energy storage are connected to the DC bus of the external power grid, the wind turbine and the AC load are connected to the AC bus of the external power grid, and the DC bus and the AC bus are coupled through a converter.

[0012] Furthermore, the prediction model is used to predict the power generation of renewable energy based on the information of the physical entity, including:

[0013] Decompose the renewable energy power generation into several intrinsic mode function components and residual components;

[0014] Reconstruct the intrinsic mode function components and residual components with similar sample entropy values ​​into multiple power components;

[0015] Taking the meteorological variable data related to power components and renewable energy power generation as input, a long short-term memory neural network is used to make a preliminary forecast of renewable energy power generation to obtain a preliminary forecast value.

[0016] Search the historical meteorological database for meteorological day data that is most similar to meteorological variable data related to renewable energy power generation, obtain the actual power output power value of the most similar meteorological day data, and use the long short-term memory neural network to predict the most similar meteorological day data to obtain the predicted power output power value of the most similar meteorological day data;

[0017] When the deviation between the initial forecast value of renewable energy power generation at a certain moment and the actual output power value of the most similar meteorological day data exceeds a threshold, the compensation mechanism is automatically triggered. The forecast error between the actual output power value of the most similar meteorological day data and the forecast value of the output power of the most similar meteorological day data is used to compensate the initial forecast value to obtain the final power forecast value.

[0018] Furthermore, the multi-objective optimization model performs multi-objective optimization on the objective function using constraint conditions to obtain an optimal Pareto solution set, which determines the charging and discharging power of the energy storage and the interconnection line power between the source-grid-load-storage system and the distribution network.

[0019] Furthermore, the objective function using constraint conditions is based on power balance constraints, equipment processing upper and lower limit constraints, grid transformer output constraints and energy storage equipment constraints, and establishes an objective function with the minimum operating cost of the source-grid-load-storage system and the minimum environmental protection cost, wherein the operating cost of the source-grid-load-storage system includes the cost of new energy power generation calculated by the final power forecast value.

[0020] The embodiment of the present application also provides a method for optimizing the operation of a source-grid-load-storage system based on digital twins, the method comprising:

[0021] Collect information about physical entities;

[0022] Predict renewable energy generation power based on physical entity information; perform multi-objective optimization to determine the charging and discharging power of energy storage, and the power of the interconnection line between the source-grid-load-storage system and the distribution network; and calculate the health index of the physical entity and classify the health status according to the health index.

[0023] Map physical entities based on the output of virtual entities.

[0024] Furthermore, the renewable energy power generation is predicted based on the information of the physical entity, including: decomposing the renewable energy power generation into a number of intrinsic mode function components and residual components;

[0025] Reconstruct the intrinsic mode function components and residual components with similar sample entropy values ​​into multiple power components;

[0026] Taking the meteorological variable data related to power components and renewable energy power generation as input, a long short-term memory neural network is used to make a preliminary forecast of renewable energy power generation to obtain a preliminary forecast value.

[0027] Search the historical meteorological database for meteorological day data that is most similar to meteorological variable data related to renewable energy power generation, obtain the actual power output power value of the most similar meteorological day data, and use the long short-term memory neural network to predict the most similar meteorological day data to obtain the predicted power output power value of the most similar meteorological day data;

[0028] When it is monitored that the deviation between the preliminary predicted value of renewable energy power generation at a certain moment and the actual value of power generation output power of the most similar meteorological day data exceeds a threshold, the compensation mechanism is automatically triggered, and the prediction error between the actual value of power generation output power of the most similar meteorological day data and the predicted value of power generation output power of the most similar meteorological day data is used to compensate the preliminary predicted value to obtain the final power prediction value.

[0029] Furthermore, multi-objective optimization is performed to determine the charging and discharging power of the energy storage, and the power of the interconnection line between the source-grid-load-storage system and the distribution network, including:

[0030] Using power balance constraints, equipment processing upper and lower limits, grid transformer output constraints, and energy storage equipment constraints as constraints, an objective function is established to minimize the operating cost and environmental protection cost of the source-grid-load-storage system. The operating cost of the source-grid-load-storage system includes the cost of renewable energy power generation calculated from the final power forecast value.

[0031] Multi-objective optimization is performed on the objective function with constraints to obtain the optimal Pareto solution set, which determines the charging and discharging power of the energy storage and the power of the interconnection line between the source-grid-load-storage system and the distribution network.

[0032] Furthermore, the meteorological variable data related to the renewable energy power generation power is judged based on the size of the Pearson correlation coefficient of the meteorological variable data, and the strength of the linear relationship between the meteorological variable data is judged, and the meteorological variable data that is more correlated with the renewable energy power generation power is selected. The more correlated refers to completely positive linear correlation and completely negative linear correlation.

[0033] Furthermore, the meteorological daily data that is most similar to the meteorological variable data related to renewable energy power generation is searched from the historical meteorological database, including:

[0034] A similar day meteorological objective function is established, wherein the similar day meteorological objective function is to minimize the sum of the Euclidean distances between each data point in the historical meteorological database and each meteorological variable data.

[0035] Furthermore, the prediction error between the actual power output value of the most similar meteorological day data and the predicted power output value of the most similar meteorological day data is used to compensate the preliminary prediction value to obtain the final power prediction value, including:

[0036] The final power forecast value is obtained by weighting the forecast error and superimposing it with the preliminary forecast value. The weighted value is the ratio of the similar day meteorological objective function to the maximum allowable correction amount.

[0037] Furthermore, multi-objective optimization is performed on the objective function using constraints, including:

[0038] An initial population of size N is randomly generated, and a progeny population is generated through non-dominated sorting, selection, crossover, and Gaussian mutation, which uses random values ​​generated by a Gaussian distribution to change individual genes. The two populations are then combined to form a population of size 2N.

[0039] Perform fast non-dominated sorting and calculate the crowding degree of individuals in each non-dominated layer. Select individuals to form a new parent population based on the non-dominated relationship and the crowding degree of the individuals.

[0040] Generate a new offspring population through genetic algorithm, compare the offspring population with the parent population, use the dithering local search strategy for the non-dominated solution of the current population, and repeat the above operation until the end condition is met;

[0041] A dithered local search strategy is used for the non-dominated solutions of the current population, including:

[0042] Set the maximum number of jitter attempts;

[0043] Find the non-dominated solutions in the current population;

[0044] For each nondominated solution, calculate and record the excess capacity;

[0045] The non-dominated solution is perturbed by Gaussian mutation and the supercapacity index is calculated. If the current supercapacity index is larger than the previous one, the old solution is eliminated and the new solution after Gaussian mutation perturbation is inserted into the population.

[0046] Compared with the prior art, the embodiments of the present application have at least the following beneficial effects:

[0047] Through digital twin technology, this application successfully constructs a closed-loop optimization mechanism of "perception-prediction-optimization-feedback" for the source-grid-load-storage system, realizes the dynamic mapping of the physical system throughout its life cycle and the dynamic coupling between the virtual system and the physical system. The optimization method proposed in this application can fully exploit the time series characteristics of time series data for preliminary prediction, and when the prediction result error is too large due to meteorological fluctuations and other conditions, it can compensate the prediction value according to the similar day meteorological search algorithm, thereby determining the final power prediction value of new energy power generation, avoiding the situation where the prediction error gradually increases after long-term operation; based on the new energy power generation power prediction result, through multi-objective optimization of the objective function, the system operating cost and environmental protection cost can be effectively reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of the architecture structure of the digital twin control system of the source-grid-load-storage system provided in an embodiment of the present application;

[0049] Figure 2 A flow chart of a method for optimizing the operation of a source-grid-load-storage system based on digital twins provided in an embodiment of the present application;

[0050] Figure 3 A flowchart of a method for predicting renewable energy power generation based on physical entity information provided in an embodiment of the present application. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0052] This application is used for source-grid-load-storage systems to optimize their operation and address the problem that traditional modeling methods are limited by the scale of the problem, the dimension of the variables, the boundaries of determinism, and the reliance on precise physical models, making it difficult to ensure computational reliability based on computational timeliness when dealing with high-dimensional, time-varying, and nonlinear problems.

[0053] See also Figure 1, which is a schematic diagram of the architecture structure of the digital twin control system of the source-grid-load-storage system provided in an embodiment of the present application, and a schematic diagram of a source-grid-load-storage system operation optimization system based on digital twin, including:

[0054] The test entity collects information about the physical entity through various sensors and inputs the information about the physical entity into the virtual entity. The information about the physical entity includes meteorological data, historical renewable energy power generation, and equipment parameters in the physical entity.

[0055] Among them, the physical entity is the physical entity of the source-grid-load-storage system. The system adopts an AC / DC hybrid architecture, including photovoltaic arrays, wind turbines, external power grids, AC loads, DC loads, battery energy storage and converters. Among them, the photovoltaic arrays and battery energy storage are connected to the DC bus of the external power grid, and the wind turbines and AC loads are connected to the AC bus of the external power grid. The DC bus and the AC bus are coupled through the converter. This physical entity significantly improves the utilization rate of renewable energy by coordinating the combined output of photovoltaic and wind power. At the same time, relying on the stable power support of the external power grid and configuring battery energy storage devices of appropriate capacity, the charging and discharging strategies can be flexibly adjusted according to load demand and wind and solar power generation output.

[0056] The virtual entity includes a prediction model for predicting renewable energy power generation based on information from the physical entity; a multi-objective optimization model for real-time optimization and determination of the charging and discharging power of energy storage, and the power of the interconnection line between the source-grid-load-storage system and the distribution network; and a health index calculation model for calculating the health index of the physical entity, classifying the health status according to the health index, and conducting health status assessments.

[0057] The digital twin model, or digital twin modeling of the power generation, grid, load, and storage system, is a high-fidelity mapping of the physical entities of the power generation, grid, load, and storage system. It includes models for wind power generation, photovoltaic power generation, battery storage, external grid, AC load, and DC load. This model mapping corresponds to the physical entities and can be visualized based on virtual entity predictions of renewable energy generation power, energy storage charge and discharge power, the power of the tie lines between the power generation, grid, load, and storage system and the distribution network, and the health status.

[0058] The working principle of a digital twin-based source-grid-load-storage system operation optimization system provided in the embodiment of the present application is as follows:

[0059] On the one hand, the test entity collects meteorological data of the physical entity, including data such as light intensity, through various sensors, and inputs it into the virtual entity along with data such as historical renewable energy power generation and equipment parameters. The virtual entity performs real-time calculation and analysis based on its own algorithm, and obtains the predicted value of renewable energy power generation through decomposition-reconstruction-prediction-correction power prediction. On the other hand, based on the predicted power results and equipment parameters, the operating status of the source-grid-load-storage system is simulated and calculated in combination with a multi-objective optimization algorithm to obtain the system operation optimization results. The optimization results and health status assessment results are input into the digital twin model together. Since it is a high-fidelity mapping of the physical entity, it can provide a reference for the regulation and operation, comprehensive evaluation, intelligent operation and maintenance, and fault recovery of the physical entity of the source-grid-load-storage system. At the same time, the operating status of the actual equipment can also be fed back to improve and optimize the digital twin model.

[0060] The prediction model is used to predict the power generation of renewable energy based on the information of physical entities, including decomposing the power generation of renewable energy into several intrinsic mode function components and residual components;

[0061] Used to reconstruct the intrinsic mode function components and residual components with similar sample entropy values ​​into multiple power components;

[0062] It is used to take the meteorological variable data related to the power component and the renewable energy power generation as input, and use the long short-term memory neural network to make a preliminary prediction of the renewable energy power generation to obtain a preliminary prediction value;

[0063] It is used to search the historical meteorological database for the meteorological day data that is most similar to the meteorological variable data related to the power generation of renewable energy, obtain the actual value of the power generation output power of the most similar meteorological day data, and use the long short-term memory neural network to predict the most similar meteorological day data to obtain the predicted value of the power generation output power of the most similar meteorological day data;

[0064] When the deviation between the preliminary forecast value of renewable energy power generation at a certain moment and the actual value of power generation output power of the most similar meteorological day data exceeds a threshold, a compensation mechanism is automatically triggered, that is, the forecast error between the actual value of power generation output power of the most similar meteorological day data and the forecast value of power generation output power of the most similar meteorological day data is used to compensate the preliminary forecast value to obtain the final power forecast value;

[0065] The multi-objective optimization algorithm uses the objective function of the constraint conditions to perform multi-objective optimization to obtain the optimal Pareto solution set, which determines the charging and discharging power of the energy storage and the power of the interconnection line between the source-grid-load-storage system and the distribution network;

[0066] The objective function using the constraint conditions is to establish an objective function with the minimum operating cost of the source-grid-load-storage system and the minimum environmental protection cost, with power balance constraint, equipment processing upper and lower limit constraint, grid transformer output constraint and energy storage equipment constraint as constraint conditions, wherein the operating cost of the source-grid-load-storage system includes the cost of renewable energy power generation calculated from the final power forecast value;

[0067] The health index calculation model quantifies the real-time health status of the equipment and predicts potential failure risks through multi-dimensional data analysis and model calculation.

[0068] See Figure 2 , a flow chart of a method for optimizing the operation of a source-grid-load-storage system based on digital twins according to an embodiment of the present application, a method for optimizing the operation of a source-grid-load-storage system based on digital twins includes:

[0069] S201 collects information about physical entities;

[0070] S202 predicts the power generation of renewable energy based on the information of the physical entity; performs multi-objective optimization to determine the charging and discharging power of the energy storage, and the power of the interconnection line between the source-grid-load-storage system and the distribution network; and calculates the health index of the physical entity and classifies the health status according to the health index.

[0071] S203 maps the physical entity according to the output result of the virtual entity.

[0072] Among them, see Figure 3 The flowchart of the method for predicting the power generation of renewable energy based on the information of physical entities shown includes:

[0073] S301 decomposes the renewable energy power generation power to obtain a number of intrinsic mode function components and residual components;

[0074] Various sensors are used to collect meteorological variable data related to renewable energy power generation. These meteorological variable data include solar irradiance, temperature, humidity, and air pressure. The collected meteorological variable data related to renewable energy power generation includes a large amount of data, which is computationally intensive. To reduce the subsequent computational effort and improve computational speed, in one embodiment, the Pearson analysis method is used to select meteorological variable data that is more relevant to renewable energy power generation from the collected meteorological variable data, and decompose it to obtain several intrinsic mode function components and residual components.

[0075] In the Pearson analysis method, the Pearson correlation coefficient is used to measure the statistic of the linear relationship between two random variables, including perfect positive linear correlation, that is, when one variable increases, the other variable also increases proportionally; perfect negative linear correlation, that is, when one variable increases, the other variable decreases proportionally; no linear correlation, that is, there is no linear correlation between the two variables.

[0076] Using the Pearson analysis method, we can determine the strength of the linear relationship between variables based on the Pearson correlation coefficient and select meteorological variable data that is most correlated with renewable energy power generation. This correlation refers to completely positive linear correlation and completely negative linear correlation.

[0077] The decomposition method can be adaptive noise complete ensemble empirical mode decomposition (EMD). This method introduces adaptive noise and multiple iterations, adding white noise to the intrinsic mode functions (IMFs) after EMD decomposition. This gradual addition of noise effectively mitigates the impact of white noise on the decomposition results, improving the accuracy and stability of the decomposition.

[0078] Adaptive noise complete ensemble empirical mode decomposition (ANEDM) introduces adaptive auxiliary decomposition white noise at each decomposition stage and integrates and averages the decomposition results of each stage. This method overcomes the problems of low decomposition efficiency and residual white noise in integrated empirical mode decomposition (EMD), while further suppressing the occurrence of modal aliasing. The uncertainty and randomness of meteorological variables lead to the uncertainty and randomness of renewable energy power generation. By decomposing renewable energy power generation into several intrinsic mode function components and residual components through ANEDM, the complexity of the original renewable energy power generation data can be reduced.

[0079] S302 reconstructs the intrinsic mode function components and residual components with similar sample entropy values ​​into multiple power components;

[0080] By comparing the complexity of the intrinsic mode function components and residual components of the power time series obtained by the adaptive noise complete set empirical mode decomposition using sample entropy values, and reconstructing components with similar sample entropy values ​​into new components, the number of components can be reduced, thereby improving the overall efficiency of renewable energy power generation prediction. The calculation steps for reconstructing the intrinsic mode function components and residual components with similar sample entropy values ​​into multiple power components are as follows:

[0081] Step 1: Set the time length to Time series Reconstruct according to the following formula dimensional vector, using Indicates that:

[0082] ,

[0083] in, , is the selected spatial dimension vector, i.e. the window length, indicating the Continuous starting from The time series is the intrinsic mode function component and the residual component of the power time series.

[0084] Step 2 Calculation and Euclidean distance between , and define the maximum distance between each component as the maximum contribution component distance , as shown below:

[0085] ,

[0086] in, and are two arbitrary components, , .

[0087] Step 3: Using the given metric , statistics show Each corresponding Number And use the following formula to calculate With all Probability of matching :

[0088] ,

[0089] Step 4: Use the following formula to solve 、 、…、 、 Corresponding average value:

[0090] ,

[0091] Step 5: Convert the time series Refactored to Dimension, and repeat steps 2 to 4 to solve , use the following formula to calculate the sample entropy:

[0092] ,

[0093] In practical applications, Generally, 1 or 2 are selected; Take 0.10-0.25 times the standard deviation of the time series.

[0094] S303 uses the power component and meteorological variable data related to the renewable energy power generation as input, and uses a long short-term memory neural network to perform a preliminary prediction of the renewable energy power generation to obtain a preliminary prediction value;

[0095] A preliminary forecast of renewable energy power generation is generated using a long-short-term memory (LSTM) neural network. The input data is meteorological variables related to power components and renewable energy power generation, and the output is the power predicted for each power component. Finally, the output power predictions for each power component are summed to obtain the preliminary forecast for renewable energy power generation.

[0096] The long short-term memory neural network used includes a forget gate, an input gate, and an output gate.

[0097] S304 searches the historical meteorological database for meteorological day data that is most similar to meteorological variable data related to renewable energy power generation, obtains the actual value of power generation output power of the most similar meteorological day data, and uses a long short-term memory neural network to predict the most similar meteorological day data to obtain the predicted value of power generation output power of the most similar meteorological day data.

[0098] The historical meteorological database contains daily meteorological data related to renewable energy power generation. Based on the meteorological variable data, the historical meteorological database is searched for the meteorological daily data most similar to the meteorological variable data related to renewable energy power generation. The meteorological daily data includes solar irradiance, temperature, humidity, and air pressure, and the renewable energy power generation corresponding to the meteorological daily data is stored in the historical meteorological database. These are obtained by statistically analyzing historical data. Generally speaking, due to the regional nature of meteorology, the historical meteorological database used is the historical meteorological database for the region where the source-grid-load-storage system is optimized. The actual power generation output power value obtained from the most similar meteorological daily data is also obtained from the historical meteorological database. The actual power generation output power value here refers to the power generation power of the renewable energy source.

[0099] Among them, the long short-term memory neural network is used to predict the most similar meteorological day data to obtain the most similar meteorological day data power output prediction value;

[0100] The long short-term memory neural network here is the same as that in S303. There may be an error between the predicted power output value of the most similar meteorological day data and the actual power output power value of the most similar meteorological day data. The most similar meteorological day data is processed to correct or compensate for the preliminary predicted value obtained from the meteorological variable data related to the new energy power generation power in the next step.

[0101] S305: When the deviation between the preliminary forecast value of renewable energy power generation at a certain moment and the actual output power value of the most similar meteorological day data exceeds a threshold, a compensation mechanism is automatically triggered, and the forecast error between the actual output power value of the most similar meteorological day data and the forecast value of the output power generation of the most similar meteorological day data is used to compensate the preliminary forecast value to obtain the final power forecast value;

[0102] In one embodiment, a multi-objective optimization is performed to determine the charging and discharging power of the energy storage and the power of the tie line between the source-grid-load-storage system and the distribution network, including: using power balance constraints, upper and lower limits of equipment processing, grid transformer output constraints, and energy storage equipment constraints as constraints, to establish an objective function that minimizes the operating cost of the source-grid-load-storage system and minimizes the environmental protection cost, wherein the operating cost of the source-grid-load-storage system includes the cost of renewable energy power generation calculated from the final power forecast value;

[0103] In one embodiment, a multi-objective optimization is performed on an objective function using constraint conditions to obtain an optimal Pareto solution set, which determines the charging and discharging power of the energy storage and the power of the tie line between the source-grid-load-storage system and the distribution network;

[0104] In one embodiment, key parameters that affect lifespan are selected based on the device type in the source-grid-load-storage system, and the data is normalized to obtain standardized values ​​for each parameter. Based on a defined health index calculation model, the device's health index is calculated, and health status levels are classified based on the health index. This is visually displayed in a digital twin virtual model. When the health index falls below a preset threshold, a maintenance alarm or adaptive control instruction is triggered. The health status assessment of each device in the source-grid-load-storage system generally includes: selecting key parameters for each device; normalizing them; calculating the health index; determining the health status level of each device and taking corresponding countermeasures.

[0105] In one embodiment, searching a historical meteorological database for meteorological daily data that is most similar to meteorological variable data related to renewable energy power generation includes:

[0106] A similar day meteorological objective function is established, wherein the similar day meteorological objective function is to minimize the sum of the Euclidean distances between each data point in the historical meteorological database and each meteorological variable data.

[0107] Similar day meteorological objective function:

[0108] ,

[0109] The objective function must meet the following conditions:

[0110] ,

[0111] Where: is the similar day meteorological objective function, is a meteorological dataset; The data include solar irradiance, temperature, humidity and air pressure; is the Pearson correlation coefficient, ; Each data point to each The Euclidean distance of is the number of meteorological data types; is the number of data points; This is the maximum allowable correction amount. If it exceeds this range, the correction will be invalid.

[0112] By solving the similar day meteorological objective function, the most similar meteorological day data in the historical meteorological database and the currently collected meteorological variable data related to the renewable energy power generation power are searched.

[0113] In one embodiment, when the deviation between the predicted value of renewable energy power generation at a certain moment and the actual value of power generation output power of the most similar meteorological data exceeds a threshold, a compensation mechanism is automatically triggered, i.e., a similar day meteorological search algorithm is used to compensate the predicted value at that moment, wherein the threshold function is defined as:

[0114] ,

[0115] is the actual value of the power output of the most similar meteorological day data at this moment, is the initial forecast value at this moment. Trigger correction.

[0116] In one embodiment, the initial prediction value of the prediction error is weighted and corrected, and the weight value is determined according to the degree of meteorological similarity, so as to obtain the final power prediction value of the new energy output. , and its calculation formula is:

[0117] ,

[0118] in, is the initial forecast value, is the actual value of the power output of the most similar meteorological day data, is the predicted value of power output from the most similar meteorological day data,

[0119] The predicted value of the final power of renewable energy power generation is part of the operating cost of the source-grid-load-storage system. The uncertain renewable energy power generation power is used as the predicted value of the final power of renewable energy power generation to calculate the cost of renewable energy power generation.

[0120] With power balance constraints, equipment processing upper and lower limit constraints, grid transformer output constraints and energy storage equipment constraints as constraints, an objective function is established to minimize the operating cost and environmental protection cost of the source-grid-load-storage system.

[0121] The lowest operating cost of the source-grid-load-storage system refers to the cost of using grid electricity, the operating costs of other distributed power equipment, the operating and maintenance costs of power generation equipment, and the operating and maintenance costs of energy storage equipment during the operation of the source-grid-load-storage system. The operating cost of the source-grid-load-storage system is:

[0122] ,

[0123] in, yes Distributed power generation at all times Operation and maintenance cost coefficient; yes Distributed power generation at all times The output power, is the number of distributed generation sources; yes Purchase electricity at any time; yes The electricity price at the time; is the amount of electricity purchased by the source-grid-load-storage system, The cost of renewable energy generation is the product of the predicted final power output of renewable energy generation and the operation and maintenance cost coefficient.

[0124] The environmental protection cost is the main pollutants in the operation of the source-grid-load-storage system, including 、 and The processing cost is as follows:

[0125] ,

[0126] Where, is the number of pollutant types ( 、 or ), is a distributed power source The pollutant type is Unit processing cost per hour; is the emission factor.

[0127] The power balance constraint refers to maintaining a dynamic balance between the real-time load power of all energy-consuming devices in the source-grid-load-storage system, the real-time power generation power of the power supply equipment, the real-time power of the grid input system, and the real-time charge and discharge power of the energy storage equipment. The sum of the power of the four types of equipment is always zero. Only under this premise can the entire source-grid-load-storage system be kept in a constant voltage state, allowing the source-grid-load-storage system to operate stably. The formula is as follows:

[0128] ,

[0129] in, Indicates the Load equipment in Operating power at all times, Indicates that the power grid is Real-time power at all times, Indicates the A power supply device Real-time power at all times, For energy storage equipment Real-time power at all times.

[0130] The upper and lower limits of device output constraints refer to the fact that all adjustable devices have their own output ranges during the control process, such as the charging power of charging pile equipment and the output power of air conditioning equipment. When designing the constraints, the power ranges of these devices must be fully considered to ensure that after the optimization calculation, the control parameters of the corresponding devices are within the upper and lower limits of the device output to ensure the normal operation of the device. The formula is as follows:

[0131] ,

[0132] in, Indicates the Equipment (including power supply equipment, grid equipment and energy storage equipment) The output power at the moment, Indicates the The minimum output power of a device, Indicates the The maximum output power of a device.

[0133] The grid transformer output constraint refers to the fact that the energy that a park can request from the grid and the energy that it can send to the grid are both limited. The device that limits the energy transmission and reception is the substation transformer in the area. Therefore, during the optimization calculation process, the output capacity of the grid transformer must be considered as a constraint to ensure stable system operation. The formula is as follows:

[0134] ,

[0135] in, For transformer Output power at any moment, is the minimum output power of the transformer, is the maximum output power of the transformer.

[0136] The energy storage device constraint refers to the constraints related to the SOC value of the energy storage device. Through research on products of various energy storage device manufacturers, it is found that when the battery SOC value is less than 20%, the energy storage device will no longer continue to discharge, and when the battery SOC value is greater than 95%, the energy storage device will no longer charge. The formula is as follows:

[0137] ,

[0138] in, For energy storage equipment Real-time battery remaining percentage at the moment, is the minimum remaining battery power of the energy storage device, The maximum remaining battery power of the energy storage device. For energy storage equipment Real-time operating power at all times, is the minimum output power of the energy storage device, is the maximum output power of the energy storage device.

[0139] Based on the objective function and related constraint characteristics, an optimization algorithm is selected, an improved multi-objective non-dominated sorting genetic algorithm is constructed, and multi-objective optimization decision-making is performed to obtain the optimal Pareto solution set, determine the charging and discharging power of the energy storage, the interconnection line power between the system and the distribution network, etc.

[0140] The characteristics of the objective function and related constraints are that since the designed objective function includes the two objectives of the source-grid-load-storage system operating cost and the environmental protection cost, multi-objective optimization is selected. Since the source-grid-load-storage system requires a large number of types of equipment and parameters to be regulated, the optimization algorithm has high requirements for its optimization ability.

[0141] In one embodiment, multi-objective optimization is performed on an objective function using constraints, including:

[0142] An initial population of size A is randomly generated, and after non-dominated sorting, selection, crossover, and Gaussian mutation, a subpopulation is generated, and the two populations are combined to form a population of size 2A.

[0143] Perform fast non-dominated sorting and calculate the crowding degree of individuals in each non-dominated layer. Select appropriate individuals to form a new parent population based on the non-dominated relationship and the crowding degree of the individuals.

[0144] The basic operations of the genetic algorithm generate a new offspring population, which is then compared to the parent population. A dithered local search strategy is used for the non-dominated solutions of the current population, and the above operations are repeated until the termination condition is met. The basic operations here refer to initialization of the population, fitness evaluation, selection, crossover, and mutation, and finally the generation of a new offspring population.

[0145] Based on the characteristics of the proposed optimization objective function and related constraints, a suitable algorithm is selected, which is required to be a multi-objective optimization algorithm, with high optimization ability and as stable as possible, and as little computational complexity as possible; in the embodiment of the present application, Gaussian mutation is used instead of polynomial mutation to reproduce the next generation of individuals; a jittered local search strategy is adopted, and Gaussian mutation is used to perturb the non-dominated solution to calculate the supercapacity index. If the supercapacity index is larger than before, it proves that the current population still has room for improvement. Then the old solution is eliminated and the perturbed new solution is inserted into the population, which helps the algorithm to escape from the local optimum.

[0146] Gaussian mutation uses random values ​​generated by Gaussian distribution to change individual genes. The details of Gaussian mutation are as follows: It is A single vector encoded in real numbers in the algebra An element of .

[0147] ,

[0148] In the formula, Obey Gaussian distribution , is called the scale parameter.

[0149] Adopting a dithered local search strategy and using Gaussian mutation to perturb the non-dominated solution, it can be decomposed into the following steps:

[0150] Set the maximum number of jitter attempts;

[0151] Find the non-dominated solutions in the current population;

[0152] For each nondominated solution, calculate and record the excess capacity;

[0153] The non-dominated solution is perturbed by Gaussian mutation and the supercapacity index is calculated. If the current supercapacity index is larger than the previous one, the old solution is eliminated and the new solution after Gaussian mutation perturbation is inserted into the population.

[0154] Multi-objective optimization is performed on the objective function with constraints to obtain the optimal Pareto solution set, which determines the charging and discharging power of the energy storage and the power of the interconnection line between the source-grid-load-storage system and the distribution network.

[0155] In one embodiment, key parameters that affect lifespan are selected based on the type of equipment in the source-grid-load-storage system, where the key parameters include:

[0156] Battery energy storage system: state of charge, health status, internal resistance, temperature;

[0157] Photovoltaic inverter: output efficiency, harmonic distortion rate, radiator temperature;

[0158] Fan bearings: vibration amplitude, spectrum characteristics, and lubricant viscosity.

[0159] In one embodiment, based on a defined health index calculation model, the health index of the device is calculated, the health status level is classified according to the health index, and the result is returned to the digital twin model. When the health index falls below a preset threshold, a maintenance alarm instruction is triggered. The health index calculation model is:

[0160] ,

[0161] Where, is the health index, For the The normalized values ​​of the key parameters; taking the health index calculation of the battery energy storage system as an example, there are four core parameters, then is 4; is the weight, ,The weight determination method can adopt the statistical regression method based on fault history.

[0162] The following grading standards can be used to divide the health status into different levels according to the health index:

[0163] when The range is between 0.6 and 1.0, the status level is healthy, and the response measure is normal operation;

[0164] when The range is between 0.3 and 0.6, the status level is warning, and the response measure is planned maintenance;

[0165] when The range is between 0-0.3, the status level is dangerous, and the response measure is to immediately shut down and repair.

[0166] By assessing the health status of each device and returning the assessment results, we can achieve a shift from "post-fault processing" to "pre-fault prevention" and reduce unplanned downtime.

[0167] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A digital twin-based source-grid-load-storage system operation optimization system, characterized by: The system includes: The test entity collects information about the physical entity through various sensors and inputs the information about the physical entity into the virtual entity. The information about the physical entity includes meteorological data, historical renewable energy power generation, and equipment parameters in the physical entity. The virtual entity includes a prediction model for predicting renewable energy power generation based on information from the physical entity; a multi-objective optimization model for real-time optimization and determination of the charging and discharging power of energy storage, and the power of the interconnection line between the source-grid-load-storage system and the distribution network; and a health index calculation model for calculating the health index of the physical entity and classifying the health status according to the health index; The prediction model is used to predict the power generation of renewable energy based on information of physical entities, including: Decompose the renewable energy power generation into several intrinsic mode function components and residual components; Reconstruct the intrinsic mode function components and residual components with similar sample entropy values ​​into multiple power components; Taking the meteorological variable data related to power components and renewable energy power generation as input, a long short-term memory neural network is used to make a preliminary forecast of renewable energy power generation to obtain a preliminary forecast value. Search the historical meteorological database for meteorological day data that is most similar to meteorological variable data related to renewable energy power generation, obtain the actual power output power value of the most similar meteorological day data, and use the long short-term memory neural network to predict the most similar meteorological day data to obtain the predicted power output power value of the most similar meteorological day data; When the deviation between the initial forecast value of renewable energy power generation at a certain moment and the actual output power value of the most similar meteorological day data exceeds the threshold, the compensation mechanism is automatically triggered. The forecast error between the actual output power value of the most similar meteorological day data and the forecast value of the output power of the most similar meteorological day data is used to compensate the initial forecast value to obtain the final power forecast value. The calculation formula of the final power forecast value is: ,in, is the final power prediction value, is the initial forecast value, is the actual value of the power output of the most similar meteorological day data, is the predicted value of power output from the most similar meteorological day data, is the similar day meteorological objective function; The similar day meteorological objective function is expressed as: , the following conditions are met: ， Where: is a meteorological dataset; The data include solar irradiance, temperature, humidity and air pressure; is the Pearson correlation coefficient, ; Each data point to each The Euclidean distance of is the number of meteorological data types; is the number of data points; The maximum allowable correction amount. If the maximum allowable correction amount is exceeded, the correction will be invalid. The digital twin model is used to map physical entities based on the output results of virtual entities, including wind power generation model, photovoltaic power generation model, battery energy storage model, external power grid model, AC load model and DC load model.

2. The digital twin-based source-grid-load-storage system operation optimization system according to claim 1 is characterized in that: The physical entity is a source-grid-load-storage system, which adopts an AC / DC hybrid architecture and includes a photovoltaic array, a wind turbine, an external power grid, an AC load, a DC load, a battery energy storage and a converter. Among them, the photovoltaic array and the battery energy storage are connected to the DC bus of the external power grid, and the wind turbine and the AC load are connected to the AC bus of the external power grid. The DC bus and the AC bus are coupled through the converter.

3. The digital twin-based source-grid-load-storage system operation optimization system according to claim 1 is characterized in that: The multi-objective optimization model performs multi-objective optimization on the objective function using constraint conditions to obtain the optimal Pareto solution set, which determines the charging and discharging power of the energy storage and the interconnection line power between the source-grid-load-storage system and the distribution network.

4. The digital twin-based source-grid-load-storage system operation optimization system according to claim 3 is characterized in that: The objective function using constraint conditions is based on power balance constraints, equipment processing upper and lower limit constraints, grid transformer output constraints and energy storage equipment constraints, and establishes an objective function with the minimum operating cost of the source-grid-load-storage system and the minimum environmental protection cost, where the operating cost of the source-grid-load-storage system includes the cost of new energy power generation calculated by the final power forecast value.

5. A method for optimizing the operation of a source-grid-load-storage system based on digital twins, characterized in that: The method includes: Collect information about physical entities; Predict renewable energy generation power based on physical entity information; perform multi-objective optimization to determine the charging and discharging power of energy storage, and the power of the interconnection line between the source-grid-load-storage system and the distribution network; and calculate the health index of the physical entity and classify the health status according to the health index. The power generation of renewable energy is predicted based on the information of physical entities, including: decomposing the power generation of renewable energy to obtain a number of intrinsic mode function components and residual components; Reconstruct the intrinsic mode function components and residual components with similar sample entropy values ​​into multiple power components; Taking the meteorological variable data related to power components and renewable energy power generation as input, a long short-term memory neural network is used to make a preliminary forecast of renewable energy power generation to obtain a preliminary forecast value. Search the historical meteorological database for meteorological day data that is most similar to meteorological variable data related to renewable energy power generation, obtain the actual power output power value of the most similar meteorological day data, and use the long short-term memory neural network to predict the most similar meteorological day data to obtain the predicted power output power value of the most similar meteorological day data; When the deviation between the initial forecast value of renewable energy power generation at a certain moment and the actual output power value of the most similar meteorological day data exceeds the threshold, the compensation mechanism is automatically triggered. The forecast error between the actual output power value of the most similar meteorological day data and the forecast value of the output power of the most similar meteorological day data is used to compensate the initial forecast value to obtain the final power forecast value. The calculation formula of the final power forecast value is: ,in, is the final power prediction value, is the initial forecast value, is the actual value of the power output of the most similar meteorological day data, is the predicted value of power output from the most similar meteorological day data, is the similar day meteorological objective function; The similar day meteorological objective function is expressed as: , the following conditions are met: ， Where: is a meteorological dataset; The data include solar irradiance, temperature, humidity and air pressure; is the Pearson correlation coefficient, ; Each data point to each The Euclidean distance of is the number of meteorological data types; is the number of data points; The maximum allowable correction amount. If the maximum allowable correction amount is exceeded, the correction will be invalid. Map physical entities based on the output of virtual entities.

6. The method for optimizing the operation of a source-grid-load-storage system based on digital twins according to claim 5, characterized in that: Multi-objective optimization is performed to determine the charging and discharging power of the energy storage, and the power of the interconnection line between the source-grid-load-storage system and the distribution network, including: Using power balance constraints, equipment processing upper and lower limits, grid transformer output constraints, and energy storage equipment constraints as constraints, an objective function is established to minimize the operating cost and environmental protection cost of the source-grid-load-storage system. The operating cost of the source-grid-load-storage system includes the cost of renewable energy power generation calculated from the final power forecast value. Multi-objective optimization is performed on the objective function with constraints to obtain the optimal Pareto solution set, which determines the charging and discharging power of the energy storage and the power of the interconnection line between the source-grid-load-storage system and the distribution network.

7. The method for optimizing the operation of a source-grid-load-storage system based on digital twins according to claim 5, characterized in that: The meteorological variable data related to the new energy power generation power is judged based on the size of the Pearson correlation coefficient of the meteorological variable data, the strength of the linear relationship between the meteorological variable data is judged, and the meteorological variable data that is more correlated with the new energy power generation power is selected. The more correlated refers to completely positive linear correlation and completely negative linear correlation.

8. The method for optimizing the operation of a source-grid-load-storage system based on digital twins according to claim 5, characterized in that: Search the historical meteorological database for meteorological daily data that is most similar to meteorological variable data related to renewable energy power generation, including: A similar day meteorological objective function is established, wherein the similar day meteorological objective function is to minimize the sum of the Euclidean distances between each data point in the historical meteorological database and each meteorological variable data.

9. The method for optimizing the operation of a source-grid-load-storage system based on digital twins according to claim 5, characterized in that: The prediction error between the actual power output value of the most similar meteorological day data and the predicted power output value of the most similar meteorological day data is used to compensate the preliminary prediction value to obtain the final power prediction value, including: The final power forecast value is obtained by weighting the forecast error and superimposing it with the preliminary forecast value. The weighted value is the ratio of the similar day meteorological objective function to the maximum allowable correction amount.

10. The method for optimizing the operation of a source-grid-load-storage system based on digital twins according to claim 5, characterized in that: Perform multi-objective optimization of objective functions with constraints, including: An initial population of size N is randomly generated, and a progeny population is generated through non-dominated sorting, selection, crossover, and Gaussian mutation, which uses random values ​​generated by a Gaussian distribution to change individual genes. The two populations are then combined to form a population of size 2N. Perform fast non-dominated sorting and calculate the crowding degree of individuals in each non-dominated layer. Select individuals to form a new parent population based on the non-dominated relationship and the crowding degree of the individuals. Generate a new offspring population through genetic algorithm, compare the offspring population with the parent population, use the dithering local search strategy for the non-dominated solution of the current population, and repeat the above operation until the end condition is met; A dithered local search strategy is used for the non-dominated solutions of the current population, including: Set the maximum number of jitter attempts; Find the non-dominated solutions in the current population; For each nondominated solution, calculate and record the excess capacity; The non-dominated solution is perturbed by Gaussian mutation and the supercapacity index is calculated. If the current supercapacity index is larger than the previous one, the old solution is eliminated and the new solution after Gaussian mutation perturbation is inserted into the population.

Images