Digital twin method for group dispatching and control of road-distributed photovoltaic energy storage systems

Abstract

This invention proposes a digital twin method for group scheduling and control of distributed pavement photovoltaic energy storage systems, including the following steps: real-time collection of environmental data and system status data; construction of a digital twin model based on a multi-level correlation regression network model to dynamically predict system status; prediction of photovoltaic power generation and energy demand in future time periods based on a self-supervised learning framework; dynamic adjustment of the operating modes of power generation and energy storage units based on the prediction results and real-time environmental data; and real-time monitoring of scheduling execution results based on a closed-loop feedback mechanism to ensure stable system operation. This invention is of great significance in achieving efficient energy utilization, reducing energy waste, and improving system flexibility and intelligence. It is particularly suitable for distributed photovoltaic energy storage systems in complex environments such as pavement.

Description

Technical Field

[0001] The present invention belongs to the technical field of digital twins, and in particular relates to a group modulation and group control digital twin method for a road-distributed photovoltaic energy storage system. Background Art

[0002] With the rapid development of renewable energy technologies, photovoltaic power generation, as a clean, efficient, and environmentally friendly energy source, has been widely adopted worldwide. In particular, with the increasing popularity of distributed photovoltaic energy storage systems, an increasing number of photovoltaic systems are being deployed in urban and industrial settings, including on building rooftops, road surfaces, and public facilities. These systems not only provide local power support but also optimize energy efficiency by providing additional power output through energy storage units during peak energy demand. However, in actual operation, distributed photovoltaic energy storage systems face numerous challenges, especially in complex environments and under variable load conditions. Achieving efficient energy scheduling and optimized management remains a major challenge in technological development.

[0003] Currently, most existing PV energy storage system scheduling methods rely primarily on rule-based scheduling mechanisms or traditional optimization algorithms (such as linear programming and dynamic programming) for scheduling decisions. These methods often assume that system operating parameters (such as light intensity, temperature, and traffic flow) are known and unchanging, or use simple empirical formulas for prediction and scheduling, ignoring the system's adaptability and self-optimization capabilities in dynamic environments. This static scheduling approach is often unable to cope with complex environmental changes, leading to problems in actual system operation, such as the following:

[0004] 1) Photovoltaic power generation efficiency is greatly affected by environmental factors (such as weather, temperature, and light), and traditional scheduling methods cannot adjust to environmental changes in real time;

[0005] 2) The charging and discharging scheduling of energy storage units cannot flexibly respond to fluctuations in grid load, resulting in energy waste or instability;

[0006] 3) Most current scheduling algorithms lack sufficient real-time and adaptive capabilities and cannot be dynamically adjusted based on real-time data and future needs, which can easily lead to uneven energy distribution or inefficient system.

[0007] Furthermore, while some more advanced scheduling methods incorporate machine learning and artificial intelligence algorithms, these methods often rely on large amounts of historical data or manual annotations, lacking sufficient self-learning and adaptive adjustment capabilities. For example, scheduling algorithms based on reinforcement learning may not achieve satisfactory results when faced with complex, high-dimensional systems due to insufficient data or inefficient learning processes. Furthermore, traditional machine learning methods require extensive prior knowledge or precisely labeled data, making them inefficient for training without sufficient data.

[0008] Therefore, the main problem with existing technologies is the lack of an intelligent scheduling method that can perceive environmental changes in real time, flexibly dispatch, and optimize energy distribution. Furthermore, most existing energy prediction models and scheduling algorithms rely on artificially designed rules or known external data, lacking adaptability and the ability to cope with complex dynamic environments. This leads to inefficiencies and limitations in energy scheduling and management. In particular, maintaining efficient energy utilization in dynamic environments remains an unaddressed challenge in complex distributed photovoltaic energy storage systems. Summary of the Invention

[0009] The purpose of this invention is to design a digital twin method for group scheduling and control of road-distributed photovoltaic energy storage systems, combined with a distributed energy prediction and scheduling strategy based on self-supervised learning. Through real-time perception of environmental changes and intelligent learning, efficient scheduling and optimized management of distributed photovoltaic energy storage systems in dynamic environments are achieved.

[0010] To achieve the above objectives, the present invention provides a digital twin method for group regulation and control of a road-distributed photovoltaic energy storage system, the method comprising the following steps:

[0011] Collect environmental data and system status data in real time, and generate environmental features through adaptive filtering and standardization;

[0012] A digital twin model based on a multi-level correlation regression network model is constructed to dynamically predict the operating status of the photovoltaic energy storage system by combining environmental characteristics and historical status. The multi-level correlation regression network model is used to capture the spatiotemporal cross-effects of environmental characteristics and historical status as the current system state, and dynamically predict the system state.

[0013] Based on a self-supervised learning framework, the predicted system operating status and environmental characteristics are used to predict photovoltaic power generation and energy demand in the future period;

[0014] Based on the prediction results and real-time environmental data, a multi-objective optimization algorithm is used to generate an environmentally adaptive scheduling strategy to dynamically adjust the operating modes of power generation and energy storage units.

[0015] Based on the closed-loop feedback mechanism, the scheduling execution results are monitored in real time, and the scheduling strategy is dynamically modified according to environmental changes to ensure stable operation of the system.

[0016] Furthermore, the environmental data is collected based on multiple sensors; the sensors include light sensors, temperature sensors, humidity sensors and system status sensors; the environmental data include light intensity, temperature, humidity and traffic flow data; the system status data includes energy storage battery power, photovoltaic power generation power and grid load;

[0017] The environmental characteristics include light intensity, temperature and humidity data.

[0018] Furthermore, the calculation formula of the digital twin model is expressed as:

[0019] S(t)=α1·F e (t)+α2·S(t-1)+γ1·F e (t)⊙S(t-1)+R(t)

[0020] Among them, S(t) is the system state at the current moment; F e (t) is the environmental feature at the current moment; α1, α2 are the weighted coefficients of environmental features and historical states; γ1 is the weighted coefficient of the cross effect, which is used to measure the nonlinear interaction between the environment and historical states; R(t) is the regularization term, which prevents the model from overfitting and enhances its robustness.

[0021] Furthermore, the digital twin model also includes a dynamic adjustment mechanism; the dynamic adjustment mechanism is specifically:

[0022] When the model's prediction error ∈ t When the set threshold is exceeded, the weight coefficients α1, α2 and the cross-effect coefficient γ1 are automatically updated; the regularization term R(t) is determined based on the historical system state mean and the historical environmental feature mean.

[0023] Furthermore, the tasks of the self-supervised learning framework include: power generation prediction and energy demand prediction, which are expressed as:

[0024] P pred (t)=W1·S(t)+W2·F e (t)+C1·S(t-1)+C2·F e (t-1)+b

[0025] Among them, P pred (t) is the predicted power generation or energy demand; S(t) is the system state at the current moment; F e(t) is the environmental characteristic at the current moment; W1, W2 are the weight matrices for the system state and environmental characteristics; C1, C2 are the weight coefficients for the system state and environmental characteristics at the current moment, reflecting the temporal dependence of the system; b is the bias term of the model, representing the constant offset.

[0026] Furthermore, the loss function of the self-supervised learning framework is obtained by weighted summation of the prediction error term, the self-consistency constraint, and the time-varying regularization term;

[0027] The prediction error term is based on the error between the predicted power generation or energy demand and the actual power generation or energy demand, and is determined by the square of the two-norm;

[0028] The self-consistency constraint is based on the error between the predicted power generation or energy demand at the previous time step and the current predicted power generation or energy demand, determined by the square of the two norm;

[0029] The time variation regularization term is used to control the degree of penalty for time series variation;

[0030] The self-supervised learning framework is trained through the back-propagation algorithm to optimize the weight parameters. During the training process, a small batch gradient descent method is used to improve the training efficiency, and a dynamic learning rate adjustment strategy is used to avoid falling into the local optimum.

[0031] Furthermore, the multi-objective optimization algorithm is expressed as:

[0032]

[0033] Among them, P pv (t+k) is the predicted value of photovoltaic power generation at the t+kth moment; P demand (t+k) is the power demand forecast value at time t+k; S(t+k) is the power system state at time t+k; S target is the target system state, i.e. the desired system operating state; ω1 and ω2 are weight coefficients, which control the weight ratio between energy balance and system stability; is the regularization term of the scheduling strategy, which is used to suppress excessive fluctuations; λ(t) is the environmental adaptation coefficient, which indicates the degree to which the system adjusts the scheduling strategy according to changes in the external environment; is the environmental adaptability term, which indicates the impact of environmental changes on scheduling;

[0034] The following constraints are also included in the scheduling optimization process:

[0035] Photovoltaic power generation cannot exceed the maximum value;

[0036] The charging and discharging power of energy storage equipment has upper and lower limits, and the battery storage capacity must be within a certain range;

[0037] Within a certain time window, the error between power generation and demand should not be too large to ensure load matching;

[0038] According to the volatility of environmental data, the adaptability of the system to changes in the external environment is controlled by the environmental adaptability coefficient λ(t).

[0039] Furthermore, a hybrid optimization method based on deep reinforcement learning and evolutionary algorithms is used to solve the multi-objective optimization algorithm, specifically including:

[0040] The state space is composed of the following elements at each time step: the current state of the power system, the predicted PV power generation and power demand, and environmental variables;

[0041] The action space is the photovoltaic strategy;

[0042] The reward function provides feedback based on the difference between the current scheduling strategy and the target, promoting the learning of the optimal scheduling strategy by reducing the energy difference while suppressing excessive fluctuations. It is expressed as:

[0043] R(t+k)=-(P pv (t+k)-P demand (t+k)) 2

[0044] Among them, R(t+k) is the reward function value, P pv (t+k) is the predicted photovoltaic power generation, P demand (t+k) is the predicted electricity demand;

[0045] The evolutionary algorithm is a genetic algorithm, and the operations include:

[0046] Crossover: Mix two different scheduling schemes to form a new candidate solution;

[0047] Mutation: Make small adjustments to the current solution to explore new solutions;

[0048] Selection: Select the best scheduling solution based on fitness evaluation and continue evolution.

[0049] Furthermore, the closed-loop feedback mechanism includes real-time collection of execution feedback data, calculation of scheduling errors, and dynamic adjustment of the learning rate through environmental adaptability factors to correct the scheduling strategy; the environmental adaptability factors are dynamically adjusted based on the real-time environmental change rate to enhance the system's response capability to sudden environmental fluctuations.

[0050] Furthermore, the method also includes continuously updating the digital twin model and the self-supervised learning model based on the newly collected data to adapt to the long-term environmental evolution and changes in the system operating status.

[0051] The beneficial technical effects of the present invention are at least as follows:

[0052] This method acquires real-time environmental data (such as light, temperature and humidity, and traffic flow) as well as system status data (such as energy storage battery charge and photovoltaic power generation), and combines it with digital twin technology to update and simulate the status of each photovoltaic energy storage unit in real time, enabling scheduling decisions to be dynamically adjusted based on varying environmental conditions. Compared to existing rule-based scheduling methods, this innovative approach can more accurately capture the impact of environmental changes on factors such as photovoltaic power generation efficiency and energy storage status, allowing for flexible adjustments to the photovoltaic unit's power output and energy storage strategy, thereby improving the system's overall energy efficiency.

[0053] The present invention also proposes the use of a self-supervised learning algorithm for energy demand forecasting and scheduling optimization. This algorithm, through deep learning of historical data, can automatically identify the intrinsic relationships between parameters such as photovoltaic power generation, energy storage battery charging and discharging efficiency, and load demand, and predict energy demand and power generation in future time periods. Compared with existing supervised learning algorithms, self-supervised learning does not rely on large amounts of labeled data and can still achieve high prediction accuracy even when the amount of data is small or incomplete. This innovation effectively solves the limitations of existing methods that rely on large amounts of historical data and manual labeling, and improves the system's prediction accuracy and adaptability.

[0054] This invention also establishes a closed-loop feedback mechanism, combining the execution of scheduling decisions with real-time data feedback to enable real-time adjustment and optimization of the system. Specifically, when the system detects a deviation between actual operation and prediction, it uses the digital twin model to update and adjust the scheduling strategy in real time, ensuring stable operation in complex environments. This mechanism overcomes the shortcomings of existing scheduling methods, which lack real-time feedback and dynamic optimization, and improves the system's real-time performance and flexibility.

[0055] Through these innovations, the present invention enables real-time sensing and scheduling adjustments based on dynamic environmental changes within photovoltaic energy storage systems. This not only improves the accuracy of energy management but also intelligently adjusts power generation and storage strategies based on system load fluctuations, effectively resolving existing issues such as uneven energy distribution, large prediction errors, and low system efficiency. The technical solution of the present invention is of great significance in achieving efficient energy utilization, reducing energy waste, and improving system flexibility and intelligence. It is particularly suitable for distributed photovoltaic energy storage systems in complex environments such as road surfaces. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative effort.

[0057] Figure 1 This is a flow chart of the digital twin method for group regulation and control of road-distributed photovoltaic energy storage systems according to the present invention. DETAILED DESCRIPTION

[0058] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0059] In one or more embodiments, Figure 1 As shown, a digital twin method for group regulation and control of a road-distributed photovoltaic energy storage system is disclosed, and the method includes the following steps:

[0060] S1. Collect environmental data and system status data in real time, and generate environmental features through adaptive filtering and standardization.

[0061] Specifically, first, the present invention collects relevant environmental variables by deploying multiple sensors in the system operation area. The sensor array should include the following categories:

[0062] Light sensor: collects light intensity data at different locations.

[0063] Temperature sensor: monitors changes in ambient temperature.

[0064] Humidity sensor: Provides humidity data to help predict the degradation of photovoltaic panel performance.

[0065] Traffic flow sensors: In case of urban areas, traffic flow data helps in predicting local load variations.

[0066] System status sensor: including important data such as energy storage battery power and photovoltaic power generation power.

[0067] The present invention expresses the collected data in matrix form:

[0068] D=[D1D2D3…D N ](1)

[0069] Among them, D i Represents the i-th type of data (such as light, temperature, etc.), D is an N×T matrix, N represents the number of sensor types, and T is the time step of the data.

[0070] Furthermore, after data collection, the present invention cleans, denoises, and standardizes the raw data to eliminate measurement errors and external interference, ensuring data quality. An adaptive filter algorithm (AFA) defined in this invention is used to remove short-term random fluctuations and outliers. This method combines weighted mean filtering with a local regression model, reducing the impact of noise on data quality by performing a weighted average and linear regression on data from adjacent time periods. The specific steps are as follows:

[0071] Assume that the collected original data is D, and first smooth the data using the following formula:

[0072]

[0073] Among them, w i is the smoothing weight coefficient, and k is the smoothing window size. The present invention uses a weighted window function w i , weighting the data according to the time interval, and the weight function will decay as the time distance increases. In this way, the present invention can eliminate sudden short-term fluctuations.

[0074] Secondly, local regression analysis is used to identify and correct outliers. For the original data D(t) with a time step of t, if its deviation from the previous and next data exceeds a certain set threshold τ, the data is considered an outlier and corrected through weighted regression:

[0075]

[0076] in, is the predicted value calculated based on historical data, and λ is the correction coefficient, which is usually small (such as 0.1) to ensure the smoothness of data correction.

[0077] Finally, in order to avoid the influence of different dimensions between different variables, the present invention performs normalization on all data. The normalization formula is:

[0078]

[0079] Among them, μ D and σ D are the mean and standard deviation of all collected data, respectively. Through this step, the range of all environmental data is mapped to a standard normal distribution, which improves the stability of model training and optimization in subsequent steps.

[0080] Furthermore, the pre-processed data needs to be further extracted to find the most valuable features for the photovoltaic energy storage system. For example, changes in light intensity and temperature have a significant impact on photovoltaic power generation, so it is necessary to calculate their correlation with the energy storage battery power and generate the environmental feature F. e (t):

[0081] F e (t)=[D light (t),D temp (t),D humidity (t)](5)

[0082] Among them, F e (t) is the processed environmental characteristics, including light intensity, temperature and humidity data.

[0083] Finally, the output data set D processed This data will serve as input for the digital twin model and self-supervised learning algorithm in subsequent steps. After preprocessing, denoising, standardization, and feature extraction, it can accurately reflect the environmental status and system operation, becoming a reliable input for digital twin and energy forecasting.

[0084] S2. Construct a digital twin model based on a multi-level correlation regression network model, and dynamically predict the operating status of the photovoltaic energy storage system by combining environmental characteristics and historical status; the multi-level correlation regression network model is used to capture the spatiotemporal cross-effects of environmental characteristics and historical status as the system status at the current moment, and dynamically predict the system status.

[0085] Specifically, this scheme adopts an innovative multi-level association regression network (MARN) model, which comprehensively considers the environmental characteristics F e The relationship between (t) and the historical state S(t-1) can effectively capture the spatiotemporal cross-effect. The core formula of the model is:

[0086] S(t)=α1·F e (t)+α2·S(t-1)+γ1·F e (t)⊙S(t-1)+R(t)(6)

[0087] Where S(t) is the system state at the current moment (such as photovoltaic power generation, battery charging status, etc.). e (t) is the environmental feature vector at the current moment (such as solar radiation, temperature, etc.). α1 and α2 are the weighting coefficients for environmental characteristics and historical states. γ1 is the weighting coefficient for the cross-effect, which is used to measure the nonlinear interaction between the environmental and historical states. R(t) is the regularization term, which prevents overfitting of the model and enhances its robustness.

[0088] Furthermore, in order to improve the adaptability of the model to sudden environmental changes, a dynamic adjustment mechanism is introduced. Specifically, when the prediction error ∈ 1 of the model exceeds the set threshold, the weight coefficients α1, α2 and the cross-effect coefficient γ1 are automatically updated. t is calculated as follows:

[0089] ∈ t =|S pred (t)-S actual (t)|(7)

[0090] In addition, the regularization term R(t) is introduced to suppress the excessive complexity of the model and maintain the stability of the prediction results. The calculation formula of the regularization term is:

[0091]

[0092] Among them, S avg (t) is the historical system state average, which serves as a comparison benchmark. avg (t) is the mean of historical environmental characteristics. λ is the regularization coefficient, which controls the model complexity.

[0093] Furthermore, the operating status of a photovoltaic energy storage system is significantly affected by temporal and spatial factors. Therefore, an adaptive adjustment mechanism is incorporated into the MARN model. This mechanism dynamically adjusts the parameters α1, α2, and γ1 in the regression model based on environmental characteristics and the rate of change of historical data. This adjustment ensures that the model can adapt to changes in different time periods and spatial regions, improving prediction accuracy. For example, when the system is exposed to extreme weather conditions, the model automatically adjusts to better predict the system status.

[0094] After the model is built, it is validated using historical and real-time data, and the model's prediction error is calculated and optimized based on the feedback. This continuous error assessment ensures the accuracy of the digital twin model and continuously refines the weights and parameters of the regression model, ensuring that it maintains efficient and stable performance over the long term.

[0095] Furthermore, as system operations progress and environmental conditions evolve, the digital twin model requires continuous optimization and updating. Each time new data is collected, the model is retrained using an error feedback mechanism to continuously improve prediction accuracy. Ultimately, through the continuously optimized digital twin model, the system can adapt to diverse environmental changes and operating conditions over the long term, providing accurate decision support for subsequent energy scheduling.

[0096] Through this series of steps, the present invention constructs a digital twin model with high adaptability and high prediction accuracy for the photovoltaic energy storage system. The model can provide real-time status prediction in actual operation and provide reliable data support for subsequent energy scheduling and optimization.

[0097] S3. Based on the self-supervised learning framework, the predicted system operating status and environmental characteristics are used to predict photovoltaic power generation and energy demand in the future period.

[0098] Specifically, to achieve accurate energy demand and power generation forecasts, this step constructs a self-supervised learning framework. Unlike traditional supervised learning methods, this framework does not rely solely on labeled data. Instead, it leverages the system's historical behavior and environmental characteristics to automatically construct state transition relationships through self-supervised learning.

[0099] Among them, the framework includes two core tasks: power generation prediction P pv (t) and energy demand prediction P demand (t). The goal is to learn the historical system states S(t-1), S(t-2), ... and environmental features F through the model e (t-1),F e (t-2),..., and combined with a certain timing delay, the power generation and energy demand at future moments are predicted through self-supervised learning.

[0100] Furthermore, the core prediction function of the model is as follows:

[0101] P pred (t)=W1·S(t)+W2·F e (t)+C1·S(t-1)+C2·F e (t-1)+b(9)

[0102] Among them, P pred (t) is the predicted power generation or energy demand. S(t) is the system state at the current moment (from the digital twin model in the previous step). F e (t) represents the environmental characteristics at the current moment (such as light intensity and temperature). W1 and W2 are weight matrices for the system state and environmental characteristics. C1 and C2 are weight coefficients for the system state and environmental characteristics at the previous moment, reflecting the temporal dependencies of the system. b is the model bias term, representing a constant offset.

[0103] Furthermore, this model framework combines the current state S(t) and the historical data S(t-1) at the previous moment, F e (t-1), introducing the characteristics of time series. This design not only enhances the system's ability to remember historical states, but also effectively captures the system's response to environmental changes at different time points.

[0104] Furthermore, in order to ensure that the model can be effectively trained without labels, it is necessary to design a reasonable loss function that can simultaneously optimize the accuracy of power generation prediction and energy demand prediction and ensure the temporal consistency of the model.

[0105] This paper designs a dual loss function that combines prediction error and self-consistency constraints. The prediction error term measures the difference between the model's predicted value and the actual value, while the self-consistency constraint ensures that the model can maintain consistency over consecutive time steps, reducing prediction fluctuations. The prediction error term formula is:

[0106]

[0107] Among them, P true (t) is the actual generated power or energy demand. is the square of the two norm, which is used to calculate the prediction error.

[0108] The self-consistency constraint formula is:

[0109]

[0110] P pred (t-1) is the predicted value at the previous time step.

[0111] Furthermore, in order to improve the model's adaptability to time series data, the present invention also designs an innovative term based on regularization to penalize the model's instability in time series, especially when the system state undergoes a sudden change, so that the model's prediction output should remain stable. The innovative regularization term formula is:

[0112]

[0113] Among them, α is the regularization coefficient, which is used to control the degree of penalty for time series changes. It is the derivative of the predicted value over time, reflecting the rate of change of the system state.

[0114] The final loss function is:

[0115]

[0116] Among them, λ1, λ2, and λ3 are weighted coefficients of the loss function, which adjust the relative importance of different loss terms.

[0117] The design of this loss function ensures that the model not only focuses on prediction accuracy but also prevents overfitting and time series fluctuations through self-consistency constraints and regularization terms. This is of great significance for energy systems, as electricity demand and generation are often affected by multiple complex factors, requiring the model to maintain stability and reliability.

[0118] Furthermore, in this step, the present invention trains the above model through the back-propagation algorithm to optimize all parameters (i.e., W1, W2, C1, C2, b). During the training process, mini-batch gradient descent (Mini-batch SGD) is used to improve training efficiency, while a dynamic learning rate adjustment strategy is used to avoid falling into local optimality.

[0119] During the training process, the present invention also uses "online learning" technology to ensure that the model can continuously adapt to changes in new data. Through periodic verification, the prediction accuracy and stability of the model are monitored.

[0120] Once the model is trained and verified, it can be used in practical applications to predict future photovoltaic power generation and energy demand. e (t), and the real-time system state S(t), the model will generate P pred (t) serves as the basis for further scheduling optimization and energy allocation. This predicted value provides decision support for the energy management system, helping the dispatch system predict future power demand and match it with photovoltaic power generation, thereby optimizing system operating efficiency. The model's prediction accuracy directly affects the system's energy allocation and resource utilization, ensuring rational scheduling and energy conservation.

[0121] S4. Based on the prediction results and real-time environmental data, an environmental adaptive scheduling strategy is generated through a multi-objective optimization algorithm to dynamically adjust the operating modes of the power generation and energy storage units.

[0122] The input of this step comes from the results of the previous step (self-supervised learning of energy demand and power generation forecast), which mainly includes:

[0123] Energy demand forecast: P demand (t), the future electricity demand is predicted by historical data and self-supervised learning model.

[0124] Power generation forecast: P pv (t), PV power generation obtained through weather forecast and system model prediction.

[0125] Environmental characteristics: F e (t), such as weather, temperature, humidity and other environmental data.

[0126] System status: S(t), including battery energy storage status, grid load, etc.

[0127] At this time, the input data are power demand, power generation, environmental characteristics and system status, which will serve as the core input of the scheduling optimization model.

[0128] Furthermore, in this step, the present invention designs an environment-adaptive scheduling optimization framework. The core of this framework is to achieve the best match between power demand and power generation capacity through scheduling strategies, while ensuring the stability and reliability of the power system.

[0129] Furthermore, in order to achieve this goal, the objective function of the scheduling optimization model is These include the following:

[0130] Energy balance term: balances power generation and demand, minimizing the difference between power generation and demand.

[0131] System stability item: Ensures that the system can remain within the target range by controlling the changes in system state.

[0132] Environmental adaptability item: By adjusting the scheduling strategy, the system can adapt to changes in the external environment and avoid unstable system performance due to environmental factors.

[0133] The mathematical expression of the objective function is:

[0134]

[0135] Among them, P pv (t+k) is the predicted value of photovoltaic power generation at the t+kth moment. demand (t+k) is the power demand forecast at time t+k. S(t+k) is the power system status (such as battery storage, load, etc.) at time t+k. target is the target system state, i.e. the desired system operating state. ω1 and ω2 are weight coefficients that control the weight ratio between energy balance and system stability. is the regularization term of the scheduling strategy, which is used to suppress excessive fluctuations. λ(t) is the environmental adaptation coefficient, which indicates the degree to which the system adjusts the scheduling strategy according to changes in the external environment (such as temperature, humidity, etc.). It is an environmental adaptability item, which indicates the impact of environmental changes on scheduling. The impact of environmental changes can be quantified through model calculation.

[0136] Among them, the innovation of this objective function lies in:

[0137] Environmental adaptability item: By introducing the environmental adaptability item This allows the system to dynamically adjust its scheduling strategy to cope with environmental fluctuations. λ(t) adjusts the sensitivity of environmental adaptability, ensuring that the scheduling system can respond appropriately to environmental changes at different times.

[0138] Regularization term: introduced This regularization term can effectively suppress high-frequency fluctuations in the scheduling process, ensure the stability of the scheduling plan, and prevent excessively frequent scheduling changes.

[0139] Design of regularization term: In order to ensure the stability of scheduling, the regularization term A new design is adopted that not only considers the error between power generation and demand, but also incorporates a frequency penalty term:

[0140]

[0141] Among them, α and β are penalty coefficients that control the penalties for changes in power generation and demand.

[0142] Frequency penalty term: By calculating the difference between power generation at each moment and the previous moment, and the difference between demand and the previous moment, high-frequency fluctuations can be suppressed to avoid overly frequent scheduling adjustments.

[0143] Furthermore, in the scheduling optimization process, in order to ensure the stability of the system and the feasibility of the constraints, a series of constraints must be introduced:

[0144] Power generation limit: Photovoltaic power generation cannot exceed the maximum value P max ,Right now:

[0145] P pv (t+k)≤P max (t+k)(16)

[0146] Battery energy storage limitations: Energy storage equipment has upper and lower limits on charge and discharge power, and battery energy storage must be within a certain range:

[0147] E(t+k)∈[E min ,E max ](17)

[0148] Demand matching error: Within a certain time window, the error between power generation and demand should not be too large to ensure load matching:

[0149] |P pv (t+k)-P demand (t+k)|≤∈(18)

[0150] Environmental adaptability adjustment: According to the volatility of environmental data, λ(t) is used to control the adaptability of the system to changes in the external environment:

[0151] P adjust (t+k)=P pred (t+k)·λ(t)(19)

[0152] Furthermore, the core task of this step is to solve the scheduling optimization problem. Because scheduling optimization problems often involve complex constraints and nonlinear characteristics, traditional optimization methods (such as linear programming and integer programming) cannot effectively address them. To overcome this challenge, this paper adopts a hybrid optimization method based on deep reinforcement learning (DQN) and evolutionary algorithms.

[0153] State space: In scheduling optimization, the state space contains all possible environmental variables, system states, and forecast data within a time period. Specifically, the state s(t+k) at each time step consists of the following elements:

[0154] The current state of the power system: such as battery energy storage state E(t+k), load state, etc.

[0155] Predicted photovoltaic power generation P pv (t+k) and power demand P demand (t+k).

[0156] Environmental variables: such as temperature F e (t+k), humidity, etc.

[0157] Action space: At each time step, the system can choose different scheduling strategies P adjust (t+k). This includes strategies such as adjusting the use of photovoltaic power generation and controlling battery charging and discharging.

[0158] Reward function: The core of DQN is the reward function, which provides feedback based on the difference between the current scheduling strategy and the target. In this scenario, the present invention designs a reward function based on the target function. The reward function is designed to maximize the energy balance and minimize the system fluctuation and environmental adaptability differences.

[0159] Reward function:

[0160] R(t+k)=-(P pv (t+k)-P demand (t+k)) 2 (20)

[0161] The reward function promotes the learning of the optimal scheduling policy by reducing the energy difference while suppressing excessive fluctuations.

[0162] Furthermore, although deep Q networks perform well in local optimization, they can easily get stuck in local optimal solutions. To avoid this problem and accelerate the search for the global optimal solution, this paper introduces an evolutionary algorithm that effectively explores the scheduling strategy space by combining global search and local optimization.

[0163] Genetic Algorithm: This invention uses an evolutionary strategy based on genetic algorithm, where individuals represent different scheduling solutions. Through operations such as crossover, mutation and selection, the quality of candidate solutions is continuously optimized. The fitness of each individual is expressed by the objective function Evaluation,operations of the genetic algorithm include:

[0164] Crossover: Mix two different scheduling schemes to form a new candidate solution.

[0165] Mutation: Make small adjustments to the current solution to explore new solutions.

[0166] Selection: Select the best scheduling solution based on fitness evaluation and continue evolution.

[0167] By combining DQN and evolutionary algorithms, the present invention implements an efficient hybrid optimization strategy that can find the global optimal solution in a complex, high-dimensional scheduling space and dynamically adjust the system based on real-time data.

[0168] Furthermore, after the scheduling optimization algorithm determines the initial scheduling strategy, the system must be able to dynamically and adaptively update according to changes in the external environment. This process mainly relies on incremental learning and environmental adaptability modules.

[0169] According to environmental characteristics F e The system should dynamically adjust its scheduling strategy based on the changes in (t) (such as temperature, humidity, sunshine intensity, etc.). The environmental adaptability update mechanism is based on the environmental adaptability term in the objective function. to make adjustments.

[0170] Incremental learning: The system adjusts scheduling strategies in real time based on moment-by-moment environmental data and historical scheduling results. Through continuous learning, scheduling decisions are gradually optimized. For example, if temperature fluctuations cause PV power generation to fluctuate, the system automatically adjusts the scheduling strategies for battery energy storage and grid power using the environmental adaptability adjustment coefficient λ(t).

[0171] Environmental adaptability adjustment: This invention defines the environmental adaptability item To quantify the impact of environmental changes on the scheduling strategy. By dynamically adjusting λ(t), the system can adapt to fluctuations in the external environment.

[0172] The process of environmental adaptability update is as follows:

[0173] P adjust (t+k)=P pred (t+k)·λ(t)(21)

[0174] Scheduling History Update: Whenever the external environment changes, the system will gradually update the scheduling strategy based on historical scheduling decisions and feedback. This mechanism improves the system's adaptability and robustness, prevents over-reliance on single decisions, and ensures that the system can respond to external changes in real time.

[0175] Once environmental changes trigger a dispatch update, the system immediately adopts the new dispatch strategy within the next time step. By continuously adapting to environmental changes, the system ensures optimal power dispatch over the long term and optimizes the stability and economy of the power system.

[0176] Furthermore, after the scheduling strategy is optimized and adapted to the environment, the final scheduling strategy P adjust (t+k) is passed as input to the Energy Management System (EMS). The EMS is responsible for actually executing the scheduling task and collecting feedback data during the execution process to further optimize the scheduling decision.

[0177] Dispatch Execution: The EMS executes energy dispatch tasks according to the current dispatch plan, including controlling photovoltaic power generation, adjusting battery storage status, and dispatching grid load. During real-time execution, the EMS uses sensors and monitoring equipment to collect new environmental data (such as changes in temperature and humidity, light intensity, etc.) as well as system status data (such as battery charge and grid load).

[0178] Feedback mechanism: The EMS provides real-time feedback during execution (such as the discrepancy between scheduling results and actual demand) to the scheduling optimization model. Based on this feedback, the optimization model fine-tunes the scheduling strategy to ensure more accurate future scheduling decisions.

[0179] Fine-tuning and optimization: When actual dispatch results deviate from predicted demand and generation, the EMS adjusts its dispatch strategy based on feedback signals and updates dispatch decisions using optimization algorithms. This feedback mechanism enables the system to self-optimize and adapt over the long term, continuously improving the efficiency and stability of the power system.

[0180] In this step, the present invention innovatively designs an environmentally adaptive scheduling optimization framework, combining reinforcement learning with evolutionary algorithms to achieve optimal scheduling under varying environmental conditions. The introduction of regularization terms, environmental adaptability adjustments, and the combination of reinforcement learning and evolutionary algorithms address the inability of traditional scheduling methods to flexibly respond to environmental fluctuations and the system instability caused by frequent fluctuations. This solution not only ensures a balance between power generation and demand but also gradually improves the stability and economic efficiency of the power system over the long term.

[0181] S5. Based on the closed-loop feedback mechanism, the scheduling execution results are monitored in real time, and the scheduling strategy is dynamically modified according to environmental changes to ensure stable operation of the system.

[0182] Specifically, after power dispatch is executed, the system will collect new environmental data and system status information through real-time monitoring to form a closed-loop feedback loop. This feedback includes:

[0183] Environmental changes: such as temperature, humidity, light intensity and other environmental characteristics F e Real-time changes of (t+k).

[0184] System status: Real-time monitoring of system status S(t+k), including battery energy storage status and grid load changes.

[0185] Scheduling execution feedback: Scheduling strategy P adjust The feedback after the execution of (t+k) is mainly the deviation between the actual power generation and the demand.

[0186] This feedback information will be passed to the next stage through the following process to achieve dynamic adjustment.

[0187] Furthermore, the present invention first calculates the scheduling strategy P that is actually executed adjust (t+k) and the system expected target P target The error ΔP(t+k) between (t+k) is used as the basis for adjustment. The calculation formula of the error term is as follows:

[0188] ΔP(t+k)=P adjust (t+k)-P target (t+k)(22)

[0189] Among them, P adjust (t+k) is the scheduling strategy actually executed by the system at time t+k. target (t+k) is the target scheduling strategy predicted by the model at time t+k (ideally, it should be consistent with P demand (t+k) is similar).

[0190] Based on the error ΔP(t+k), the system determines whether the dispatch strategy needs to be adjusted. If the error is large, it indicates that the current dispatch strategy is too far away from the actual demand / generation, and parameter adjustments are needed to bridge the gap.

[0191] Furthermore, the system will calculate the error ΔP(t+k) and the current environmental characteristics F e (t+k) and the system state S(t+k), and uses an adaptive adjustment algorithm to fine-tune the scheduling strategy. Here, the present invention designs an "Error-Weighted Adjustment" (EWA) algorithm to adjust the strategy. This algorithm takes into account the combined influence of scheduling errors and environmental factors:

[0192]

[0193] Among them, P adjust (t+k) new is the new scheduling policy after adjustment. γ(t+k) is the learning rate (or adjustment speed), which controls the magnitude of the adjustment and can be adjusted dynamically based on real-time feedback. ΔP(t+k) is the error term, representing the difference between the actual and target scheduling policies. is the environmental adaptability factor, which indicates the impact of environmental changes on the scheduling strategy. Specifically, It is a regulation function based on environmental characteristics. Its calculation can refer to:

[0194]

[0195] Where α is the environmental change sensitivity parameter, which reflects the sensitivity of environmental changes to scheduling adjustments.

[0196] Through the above correction strategy, the system will dynamically adjust P adjust (t+k), and optimize the dispatching strategy at each moment to gradually approach the expected power demand and power generation targets.

[0197] Furthermore, the adjusted dispatch strategy is fed back to the Energy Management System (EMS), which executes the new dispatch command and continues to collect new feedback data during the execution process. In this way, the system will continuously optimize itself through closed-loop feedback during long-term operation.

[0198] Among them, the iterative process of the feedback correction mechanism:

[0199] Initial scheduling policy P adjust (t+k) is obtained and executed through the optimization algorithm.

[0200] Collect system feedback data in real time, including environmental characteristics F e (t+k) and system state S(t+k).

[0201] Calculate the error ΔP(t+k) between the actual execution scheduling strategy and the target.

[0202] Based on the error and environmental adaptability, the error weighted correction algorithm is used to adjust the scheduling strategy P adjust (t+k).

[0203] The adjusted scheduling strategy is passed to the EMS again for execution, and feedback is continued to be collected.

[0204] This closed-loop adjustment process will be carried out at each time step to ensure that the system gradually optimizes scheduling decisions under changing environmental conditions, and ultimately achieves stable and economical operation of the system.

[0205] Furthermore, in order to ensure that the power system can adapt to changes in the external environment more effectively, the adjusted dispatching strategy P adjust (t+k) new It will be readjusted based on real-time environmental data. For example, when the intensity of sunlight changes dramatically, the photovoltaic power generation may fluctuate greatly. To strengthen the response to environmental fluctuations and ensure that the scheduling strategy is more in line with actual power generation and demand conditions.

[0206] Through this dynamic feedback mechanism, the system can quickly respond to environmental changes and make corresponding adjustments, further optimizing the stability and economic efficiency of power dispatch. Through closed-loop feedback and dynamic adjustment strategies, the system can adjust the power dispatch plan in real time to adapt to environmental changes and fluctuations in system status. Specifically, the Error Weighted Correction Algorithm (EWA) adjusts the dispatch strategy at each moment based on real-time feedback and environmental characteristics, making the dispatch process more stable and achieving system optimization over the long term. This solution can effectively address the uncertainties and environmental changes in actual power systems, ensuring the reliability and efficiency of power dispatch.

[0207] In summary, the present invention enables real-time sensing and scheduling adjustments for dynamic environmental changes within photovoltaic energy storage systems. This not only improves the accuracy of energy management but also intelligently adjusts power generation and storage strategies based on system load fluctuations, effectively resolving existing issues such as uneven energy distribution, large prediction errors, and low system efficiency. The technical solution of the present invention is of great significance in achieving efficient energy utilization, reducing energy waste, and improving system flexibility and intelligence. It is particularly suitable for distributed photovoltaic energy storage systems in complex environments such as road surfaces.

[0208] The above disclosures are merely some preferred embodiments of the present invention, and certainly cannot be used to limit the scope of the present invention. A person skilled in the art will understand that all or part of the processes of the above embodiments and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A digital twin method for group dispatching and group control of road-distributed photovoltaic energy storage systems, characterized by: The method comprises the following steps: Collect environmental data and system status data in real time, and generate environmental features through adaptive filtering and standardization; A digital twin model based on a multi-level correlation regression network model is constructed to dynamically predict the operating status of the photovoltaic energy storage system by combining environmental characteristics and historical status. The multi-level correlation regression network model is used to capture the spatiotemporal cross-effects of environmental characteristics and historical status as the current system state, and dynamically predict the system state. Based on a self-supervised learning framework, the predicted system operating status and environmental characteristics are used to predict photovoltaic power generation and energy demand in the future period; Based on the prediction results and real-time environmental data, a multi-objective optimization algorithm is used to generate an environmentally adaptive scheduling strategy to dynamically adjust the operating modes of power generation and energy storage units. Based on the closed-loop feedback mechanism, the scheduling execution results are monitored in real time, and the scheduling strategy is dynamically modified according to environmental changes to ensure stable system operation. The environmental data is collected based on multiple sensors; the sensors include light sensors, temperature sensors, humidity sensors and system status sensors; the environmental data include light intensity, temperature, humidity and traffic flow data; the system status data includes energy storage battery power, photovoltaic power generation power and grid load; The environmental characteristics include light intensity, temperature and humidity data; The calculation formula of the digital twin model is expressed as: ； in, is the system status at the current moment; is the environmental characteristics at the current moment; is the weighting coefficient of environmental characteristics and historical status; is the weighted coefficient of the cross effect, which is used to measure the nonlinear interactive influence of the environment and historical status; is a regularization term that prevents the model from overfitting and enhances its robustness.

2. The group adjustment and control digital twin method for a road-distributed photovoltaic energy storage system according to claim 1 is characterized in that: The digital twin model also includes a dynamic adjustment mechanism; the dynamic adjustment mechanism is specifically: When the model's prediction error When the set threshold is exceeded, the weight coefficient is automatically updated is the cross-effect coefficient ; The regularization term Determined based on the historical system state mean and historical environmental characteristic mean.

3. The group adjustment and control digital twin method for a road-distributed photovoltaic energy storage system according to claim 1 is characterized in that: The tasks of the self-supervised learning framework include power generation prediction and energy demand prediction, which are expressed as: ； in, Forecasted power generation or energy demand; is the system status at the current moment; is the environmental characteristics of the current moment; is the weight matrix for system state and environment characteristics; is the weight coefficient for the current system state and environmental characteristics, reflecting the temporal dependency of the system; is the bias term of the model, which represents a constant offset.

4. The digital twin method for group regulation and control of a road-distributed photovoltaic energy storage system according to claim 3 is characterized in that: The loss function of the self-supervised learning framework is obtained by weighted summation of the prediction error term, the self-consistency constraint, and the time-varying regularization term; The prediction error term is based on the error between the predicted power generation or energy demand and the actual power generation or energy demand, and is determined by the square of the two-norm; The self-consistency constraint is based on the error between the predicted power generation or energy demand at the previous time step and the current predicted power generation or energy demand, determined by the square of the two norm; The time variation regularization term is used to control the degree of penalty for time series variation; The self-supervised learning framework is trained through the back-propagation algorithm to optimize the weight parameters. During the training process, a small batch gradient descent method is used to improve the training efficiency, and a dynamic learning rate adjustment strategy is used to avoid falling into the local optimum.

5. The digital twin method for group regulation and control of a road-distributed photovoltaic energy storage system according to claim 3 is characterized in that: The multi-objective optimization algorithm is expressed as: ； in, For the The predicted value of photovoltaic power generation at the moment; For the The predicted power demand value at the moment; For the The power system status at the moment; is the target system state, i.e. the expected system operation state; 、 is the weight coefficient, which controls the weight ratio between energy balance and system stability; It is the regularization term of the scheduling strategy, used to suppress excessive fluctuations; is the environmental adaptability coefficient, which indicates the degree to which the system adjusts its scheduling strategy according to changes in the external environment; is the environmental adaptability term, which indicates the impact of environmental changes on scheduling; The following constraints are also included in the scheduling optimization process: Photovoltaic power generation cannot exceed the maximum value; The charging and discharging power of energy storage equipment has upper and lower limits, and the battery storage capacity must be within a certain range; Within a certain time window, the error between power generation and demand should not be too large to ensure load matching; According to the volatility of environmental data, the environmental adaptation coefficient The adaptability of the control system to changes in the external environment.

6. The digital twin method for group regulation and control of a road-distributed photovoltaic energy storage system according to claim 5 is characterized in that: A hybrid optimization method based on deep reinforcement learning and evolutionary algorithms is used to solve multi-objective optimization algorithms, including: The state space is composed of the following elements at each time step: the current state of the power system, the predicted PV power generation and power demand, and environmental variables; The action space is the photovoltaic strategy; The reward function provides feedback based on the difference between the current scheduling strategy and the target, promoting the learning of the optimal scheduling strategy by reducing the energy difference while suppressing excessive fluctuations. It is expressed as: ； in, is the reward function value, is the predicted photovoltaic power generation, For forecasted electricity demand; The evolutionary algorithm is a genetic algorithm, and the operations include: Crossover: Mix two different scheduling schemes to form a new candidate solution; Mutation: Make small adjustments to the current solution to explore new solutions; Selection: Select the best scheduling solution based on fitness evaluation and continue evolution.

7. The digital twin method for group regulation and control of a road-distributed photovoltaic energy storage system according to claim 1, characterized in that: The closed-loop feedback mechanism includes real-time collection of execution feedback data, calculation of scheduling errors, and dynamic adjustment of the learning rate through the environmental adaptability factor to correct the scheduling strategy; the environmental adaptability factor is dynamically adjusted based on the real-time environmental change rate to enhance the system's response ability to sudden environmental fluctuations.

8. The digital twin method for group regulation and control of a road-distributed photovoltaic energy storage system according to claim 1, characterized in that: The method also includes continuously updating the digital twin model and the self-supervised learning model based on newly collected data to adapt to long-term environmental evolution and changes in system operating status.

Images