Smart energy system based on electric power communication and scheduling method

Abstract

The invention relates to the field of energy scheduling, and particularly provides a smart energy system and scheduling method based on electric power communication, the system collects operation data of field equipment such as a charging pile, a V2G charging and discharging pile, a photovoltaic inverter and the like through a smart energy unit, and a master station sends the operation data to a master station on the basis of the data and external meteorological information. And obtaining energy demand and photovoltaic power generation power prediction values through the energy demand prediction model and the photovoltaic power generation power prediction model. And finally, establishing a target optimization model for solving at least by taking the total operation cost of the system as a target, generating an optimal scheduling instruction, and issuing the optimal scheduling instruction to the field equipment for execution through the smart energy unit. According to the invention, power grid load fluctuation is effectively stabilized, and consumption of new energy is promoted.

Description

Technical Field

[0001] This invention relates to the field of energy dispatching, specifically to a smart energy system and dispatching method based on power communication. Background Technology

[0002] With the transformation of the energy structure and the increasing popularity of electric vehicles and other electrical devices, traditional energy management systems face challenges. Firstly, at the system communication and data acquisition level, when various types of electrical devices, such as charging piles, V2G charging and discharging piles, and photovoltaic inverters, are connected to the grid, the system lacks a unified and efficient communication coordination mechanism. This results in inaccurate and untimely collection of energy data on the operating status of various devices, preventing the upper-level management system from comprehensively and accurately grasping the real-time operating status of the system, thus seriously affecting the scientific nature of subsequent energy management decisions. Secondly, at the energy dispatch and resource utilization level, the control methods of related schemes are limited. Faced with peak and valley changes in grid load, the system struggles to effectively coordinate and dispatch the abundant distributed resources within the region. In particular, it cannot fully utilize flexible energy storage resources such as electric vehicles, which combine electricity consumption and generation characteristics, to help the grid smooth load pressure, exacerbating the grid's operational risks and limiting the absorption capacity of intermittent new energy sources such as photovoltaics, leading to energy waste. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides a smart energy system and scheduling method based on power communication. It incorporates V2G charging and discharging piles into an optimized scheduling system to control their charging and discharging behavior. This allows for discharging support during peak grid load periods and charging and energy storage during off-peak periods, effectively mitigating grid load fluctuations and promoting the absorption of new energy sources.

[0004] In a first aspect, the technical solution of the present invention provides a smart energy system based on power communication, comprising: Multiple field devices, including charging piles, electrical loads, V2G charging and discharging piles, and photovoltaic inverters, all have built-in communication modules for data interaction with the smart energy unit; Smart energy unit: used to collect operating data of field equipment and upload it to the main station of the energy management system, as well as to receive and execute energy dispatch and control instructions issued by the main station of the energy management system; The energy management system master station is configured to perform the following operations: a) Based on operational data, as well as meteorological forecast data and / or date type for future periods, predict future energy demand and photovoltaic power generation using energy demand forecasting models and photovoltaic power generation forecasting models respectively, and use the forecast results as input parameters for solving the optimization model; b) Establish and solve an optimization model with the objective function of minimizing the total operating cost of the system; the constraints of the model shall include at least system power balance constraints and equipment operation constraints; c) Based on the solution results of the optimization model, generate energy dispatch control commands for one or more devices in the field.

[0005] Secondly, the technical solution of the present invention provides a smart energy dispatching method based on power communication. This method is based on the above-mentioned system implementation and includes the following steps: The smart energy unit collects the operating data of the field equipment and sends the operating data to the main station of the energy management system; The energy management system master station executes scheduling decisions, generates energy scheduling control commands, and sends the energy scheduling control commands to the smart energy units, which then control the corresponding field equipment. The execution of scheduling decisions includes: a) Based on operational data, as well as meteorological forecast data and / or date type for future periods, predict future energy demand and photovoltaic power generation using energy demand forecasting models and photovoltaic power generation forecasting models respectively, and use the forecast results as input parameters for solving the optimization model; b) Establish and solve an optimization model with the objective function of minimizing the total operating cost of the system; the constraints of the model shall include at least system power balance constraints and equipment operation constraints; c) Based on the solution results of the optimization model, generate energy dispatch control commands for one or more devices in the field.

[0006] As can be seen from the above technical solutions, this application has the following advantages: (1) By constructing a collaborative communication architecture consisting of field devices (built-in communication modules), smart energy units, and the main station of the energy management system, a unified and efficient data interaction channel is provided for various heterogeneous devices, ensuring that the system can collect all field device operation data comprehensively, accurately, and in a timely manner, providing a data foundation for implementing scientific energy management decisions, and overcoming the defects of inaccurate and untimely data collection in the existing system; (2) Through the main station of the energy management system, the future energy demand and photovoltaic power generation are accurately predicted based on the collected data, and an optimization model is established based on this to make scheduling decisions. It can dynamically generate refined control instructions for charging piles, V2G charging and discharging piles, photovoltaic inverters and other equipment, thereby actively managing and allocating the system energy flow, significantly enhancing the ability to regulate the peak and valley loads of the power grid. In particular, it can make full use of the flexible resource of V2G charging and discharging piles to discharge and support the power grid during peak loads and charge and consume surplus power during low loads, effectively easing load pressure and promoting the consumption of new energy sources, thereby improving the stability of power grid operation and energy utilization efficiency. Attached Figure Description

[0007] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a schematic diagram of a smart energy system structure based on power communication, provided as an embodiment of the present invention.

[0009] Figure 2 This is a flowchart illustrating a smart energy dispatching method based on power communication, provided as an embodiment of the present invention.

[0010] Figure 3 A flowchart illustrating the process of the energy management system master station executing scheduling decisions. Detailed Implementation

[0011] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0013] Figure 1 A schematic diagram of a smart energy system structure based on power communication is provided as an embodiment of the present invention, such as... Figure 1As shown, the system includes an energy management system master station, a smart energy unit, and multiple field devices.

[0014] Multiple field devices, including charging piles, electrical loads, V2G charging and discharging piles, and photovoltaic inverters, all have built-in communication modules for data interaction with the smart energy unit.

[0015] Smart Energy Unit: Used to collect operating data from field equipment and upload it to the main station of the energy management system, as well as to receive and execute energy dispatch and control commands issued by the main station of the energy management system.

[0016] The energy management system master station is configured to execute scheduling decisions.

[0017] In one specific embodiment, the charging pile and V2G charging / discharging pile of the field equipment are respectively a 7kW AC charging pile and a 7kW V2G charging / discharging pile.

[0018] 7kW AC charging station: mainly used for AC charging services of electric vehicles in home or small scenarios to meet users' daily charging needs.

[0019] Electrical loads: covering Class I, Class II and Class III loads. Different types of loads have different power supply reliability requirements, and the system needs to carry out differentiated energy allocation and management according to their characteristics.

[0020] 7kW V2G charging and discharging pile: Equipped with bidirectional charging and discharging function, it can guide electric vehicles to charge and store electrical energy when the grid load is low; when the grid load is high or there is other energy demand, the electrical energy in the vehicle can be discharged back to the grid, thereby helping the grid to smooth the load pressure, while promoting the effective storage and consumption of new energy.

[0021] Photovoltaic inverters: They can efficiently transmit the electricity generated by photovoltaic panels to the power grid, and are an important link in realizing the grid connection of new energy sources.

[0022] The energy management system master station issues background energy scheduling tasks to the entire system through remote communication, and at the same time receives the collected data uploaded by the smart energy units. It is the control core and data aggregation center of the entire system and can make reasonable energy scheduling decisions based on big data analysis. Specifically, the energy management system master station is configured to perform the following operations to make scheduling decisions.

[0023] a) Based on operational data, as well as meteorological forecast data and / or date type for future periods, predict future energy demand and photovoltaic power generation using energy demand forecasting models and photovoltaic power generation forecasting models respectively, and use the forecast results as input parameters for solving the optimization model; b) Establish and solve an optimization model with the objective function of minimizing the total operating cost of the system; the constraints of the model shall include at least system power balance constraints and equipment operation constraints; c) Based on the solution results of the optimization model, generate energy dispatch control commands for one or more devices in the field.

[0024] The specific process of scheduling decisions will be explained in detail in subsequent method embodiments, and will not be repeated here.

[0025] In some optional implementations, the communication module built into the field equipment is a dual-mode carrier module. That is, all equipment such as 7kW AC charging piles, electrical loads, 7kW V2G charging and discharging piles, and photovoltaic inverters are equipped with dual-mode carrier modules. Through the dual-mode communication function of this module, these devices can achieve efficient data interaction with the smart energy unit without the need to lay dedicated data transmission lines, which greatly simplifies the system construction process and reduces cost investment.

[0026] In some optional implementations, the smart energy unit uses built-in data acquisition modules and sensors to collect comprehensive data from field equipment. For a 7kW AC charging pile, it collects data such as charging voltage Vc, charging current Ic, and cumulative charging capacity Ec; for electrical loads, it collects real-time power P1, power consumption E1, and load category identifier C1; for a 7kW V2G charging / discharging pile, it collects data including charging / discharging power Pv2g and remaining battery capacity Sv2g; and for photovoltaic inverters, it collects key data such as output power Ppv and conversion efficiency ηpv. The data acquisition frequency is set to a comprehensive acquisition every t=5 minutes to ensure timely capture of changes in equipment operating status.

[0027] Meanwhile, the energy management system master station and the smart energy unit are connected via 5G communication. The smart energy unit collects and integrates various operating data and status information of the field equipment in the entire smart energy system, and then sends the data to the energy management system master station via 5G communication. This ensures that the energy management system master station can grasp the detailed status of each device in the system in real time, providing data support for the efficient operation of the smart energy system.

[0028] Figure 2 This is a flowchart illustrating a smart energy dispatching method based on power communication, provided by an embodiment of the present invention. The method is based on the system implementation of the above embodiment, as follows: Figure 2 As shown, the method includes the following steps.

[0029] S1 collects operating data from field equipment through the smart energy unit and sends the operating data to the main station of the energy management system.

[0030] S2, the energy management system master station executes scheduling decisions to generate energy scheduling control commands, and sends the energy scheduling control commands to the smart energy unit, which then controls the corresponding field equipment.

[0031] Figure 3 A flowchart illustrating the process of the energy management system master station executing scheduling decisions, as shown below. Figure 3 As shown, the execution of scheduling decisions includes the following steps.

[0032] S201, based on operational data, as well as meteorological forecast data and / or date type for future periods, predicts future energy demand and photovoltaic power generation using energy demand prediction models and photovoltaic power generation prediction models, respectively, and uses the prediction results as input parameters for solving the optimization model.

[0033] This embodiment predicts two targets: future energy demand and photovoltaic power generation. Both target sequences are non-stationary and nonlinear time series, and both contain fluctuations at multiple time scales. Therefore, the same prediction model architecture and prediction method can be used. The prediction model architecture includes a long short-term memory network for high-frequency components and an ARIMA model for low-frequency components. It is understood that although both use long short-term memory networks and ARIMA models, the parameters of their prediction models are different due to different prediction purposes. The model parameters are determined according to their respective training processes.

[0034] Predicting future energy demand using energy demand forecasting models involves the following steps.

[0035] S2011, extract relevant power data from the operating data, calculate the total load power at that moment based on the relevant power data, and construct a system total load power time series from the total load power at multiple sampling moments.

[0036] The relevant power data extracted from the operational data includes the real-time power of each electrical load. Charging power of each charging station Charging power of each V2G charging and discharging station With discharge power .

[0037] All ,all ,all Add them together, then subtract all of them. Obtain the total load power at that moment.

[0038] The total load power at multiple sampling times is arranged in chronological order to obtain the system total load power time series.

[0039] S2012 uses the variational mode decomposition algorithm to decompose the system's total load power time series into multiple intrinsic mode function components.

[0040] In practice, the system total load power time series is first preprocessed, including handling missing and outlier values ​​and normalizing. Then, the Variational Mode Decomposition (VMD) algorithm is used to adaptively decompose the preprocessed system total load power time series into K quasi-orthogonal eigenmode function components. This decomposes complex sequences into a series of relatively stationary subsequences. The VMD solution process involves constructing and solving the following constrained variational model:

[0041] in, The total system load power time series, and These are the K IMF components and their corresponding center frequencies, δ(t) is the Dirac function, and * denotes convolution operation.

[0042] S2013 adapts different prediction models for intrinsic mode function components of different frequencies. Specifically, for high-frequency components, the first long short-term memory network model is used for prediction by integrating meteorological forecast data and date type for future periods. For low-frequency components, the first ARIMA model is used for prediction by integrating meteorological forecast data and date type for future periods.

[0043] For high-frequency IMF components (which typically contain random fluctuations and noise), a Long Short-Term Memory (LSTM) network is used for prediction. LSTM effectively learns long-term dependencies in time series through its gating mechanism. That is, during prediction, the high-frequency IMF components, future weather forecast data, and date type are used to construct input parameters, which are then fed into the first LSTM network model to output a prediction of energy demand.

[0044] For the low-frequency IMF component (representing the long-term trend and stable pattern of the series), an autoregressive moving average model with external input is used for prediction. This model introduces external variables as regression terms based on the standard ARIMA model structure. The structure of the autoregressive moving average model will be detailed later and will not be repeated here.

[0045] In other words, when making predictions, low-frequency IMF components and future weather forecast data are used to construct input parameters, which are then input into the ARIMA model to output a prediction of energy demand.

[0046] It should be noted that the total system load power time series is a non-stationary composite signal, containing variations at different time scales. Variational mode decomposition (VMD) algorithms can separate the long-term trends and periodic patterns (corresponding to low-frequency intrinsic mode function components) from the short-term, sudden random fluctuations (corresponding to high-frequency intrinsic mode function components). Adapting different prediction models to the characteristics of different components essentially involves modeling the different intrinsic driving forces of load changes, thereby achieving a more refined and accurate prediction of the overall load.

[0047] The first Long Short-Term Memory (LSTM) network model employs a stacked structure to predict high-frequency intrinsic mode function (EMF) components obtained after variational mode decomposition. The model structure specifically includes: Input layer: Receives fixed-length historical sequence data. The input shape is (n, f), where n is the input time step, such as the past 24 time points, and f is the feature dimension, which must at least contain the historical values ​​of the high-frequency IMF component and can be extended to external features such as meteorological data. Stacked LSTM layers: This method contains at least two LSTM layers. The first LSTM layer receives the input sequence and returns its entire output sequence; the second LSTM layer further processes the output of the previous layer to learn more complex temporal dynamic features. Dropout layer: A Dropout layer is introduced after each LSTM layer to randomly "drop" a portion of the output of neurons during training, as a regularization method to prevent the model from overfitting. Fully connected output layer: Maps the LSTM output of the last time step to the desired prediction dimension.

[0048] The training process is as follows: a training dataset is constructed from historical operational data, including time series of high-frequency IMF components and corresponding meteorological data and external features such as date type; mean squared error is used as the loss function in the training process; the Adam optimizer is used to minimize the loss function. The Adam optimizer can adaptively adjust the learning rate of each parameter and can usually achieve fast and stable convergence when processing time series data; in multiple training cycles, data is input into the model, gradients are calculated through the backpropagation algorithm, and the Adam optimizer updates the model weight parameters until the model's performance on the validation set no longer improves significantly.

[0049] The first ARIMA model is used to predict the low-frequency intrinsic mode function components obtained after variational mode decomposition. Its structure is determined by three core parameters (p, d, q): AR(p) — Autoregressive term: Represents a linear relationship between the current value and the values ​​at the past p time points. The model structure is reflected in... ; I(d) — Difference order: The d-th order difference operation performed to make the time series stationary. This is a preprocessing step and is not the parameter to be estimated. MA(q) – Moving Average: Represents a linear relationship between the current value and the random errors over the past q time points. The model structure is reflected in… .

[0050] The complete ARIMA(p,d,q) model formula is as follows:

[0051] In the formula, For lag operators, and For model parameters, It is a white noise sequence. Represents time The For energy demand forecasting, these external variables could be temperature and humidity; for photovoltaic forecasting, they could be irradiance and cloud cover. These are the coefficients that the model needs to estimate, used to measure the impact of this external variable on the prediction target. The extent of influence of (i.e., the IMF component).

[0052] When the model does not contain differences (d=0), it is essentially an ARX or ARMAX model, where c represents a constant intercept term. This indicates the sequence value when all autoregressive terms, external variables, and historical error terms are zero. The long-term average level or baseline. Specifically, for a stationary IMF component, c can be understood as the long-term average level of the fluctuations around which the component revolves after excluding its own historical inertia (AR term), external influences (X term), and random shocks (MA term).

[0053] When the model includes differencing (d≥1), it is a true ARIMAX model, and the formula applies to the stationary sequence after differencing. In this case, c is called the "drift term". It represents the original non-stationary sequence. The average trend change per unit time is called the average slope. Specifically, if a low-frequency IMF component exhibits a stable linear growth or decline trend, then the constant term 'c' quantifies the strength and direction of this trend. A positive 'c' indicates an upward trend in the sequence, while a negative 'c' indicates a downward trend.

[0054] During training, the parameter estimation process of the ARIMA model relies not only on historical sequences of low-frequency components but also on external historical data aligned with the sequence time (temperature and date types for energy demand forecasting; irradiance and cloud cover for photovoltaic power generation forecasting). The autoregressive parameters, moving average parameters, and regression coefficients of external variables are determined using maximum likelihood estimation.

[0055] The training process, i.e., the parameter estimation process, is as follows: The training data includes not only historical IMF component sequences. It must also include historical data of external variables that are time-aligned with it. The maximum likelihood estimation method is used to simultaneously estimate all parameters, including the autoregressive coefficients of the ARIMA component. Moving average coefficient and the regression coefficients of external variables After the model is fitted, it is necessary to check whether the residuals are white noise to ensure that the model has fully captured the information. S2014 integrates the prediction results of each component through adaptive weighting to obtain the final energy demand forecast.

[0056] S2014a, for the first For each intrinsic mode function component of the prediction model, its in-sample goodness of fit and out-of-sample prediction accuracy are weighted and summed to obtain a comprehensive score. The comprehensive score is then normalized to obtain the initial dynamic weight of the component. .

[0057] In-sample goodness of fit The ability of a model to interpret historical data is measured as follows:

[0058] In the formula, The true value It is the first k The predicted value of each component, It is the mean of the true values.

[0059] Out-of-sample prediction accuracy The reciprocal of the symmetric mean absolute percentage error (sMAPE) is used to measure the model's ability to predict future data.

[0060] The two indicators are weighted and summed to obtain the comprehensive score. Then, the overall score is calculated. Normalization is performed to obtain the initial dynamic weight of this component. , is represented as:

[0061] S2014b, regarding the first For a prediction model with individual intrinsic mode function components, obtain the uncertainty measure of its predicted values. For the components predicted using a Long Short-Term Memory (LSTM) network model, the Monte Carlo Dropout method is used to perform several forward propagations to obtain the distribution of the predicted values, and the coefficient of variation of this distribution is used as a measure of its uncertainty. For components predicted using the ARIMA model, the uncertainty is measured by calculating the confidence interval of the model at the prediction time point and using the ratio of the width of the confidence interval to the predicted value. .

[0062] For the LSTM model, Monte Carlo Dropout is used to estimate the prediction uncertainty. During prediction, Dropout is kept on, and T forward propagations are performed to obtain T predicted values. The uncertainty of this prediction component is then calculated. The uncertainty can be measured using the coefficient of variation of these T predicted values, i.e., a measure of uncertainty. This is the ratio of the standard deviation to the mean of the T predicted values.

[0063] S2014c, using uncertainty measurement Calculate the weight correction factor In the formula This is the preset attenuation coefficient.

[0064] The higher the uncertainty, the smaller the correction factor; therefore, an exponential decay function is used to construct this factor. It is a decay coefficient greater than 0, used to control the penalty force of uncertainty on the weight.

[0065] S2014e, initial dynamic weights With weighting correction factor Multiply and normalize to obtain the first... The final fusion weight of each component .

[0066] S2014f, using the final fusion weights The prediction results of all components are weighted and summed to output the final energy demand forecast.

[0067] This embodiment combines goodness of fit and prediction accuracy to more comprehensively evaluate the performance of each component model. At the same time, it uses Monte Carlo Dropout technology to quantify the prediction uncertainty of the deep learning model and adjust the weights accordingly to improve the robustness and reliability of the fusion prediction.

[0068] The photovoltaic power generation prediction model is used to predict photovoltaic power generation, which includes the following steps.

[0069] S2015: Extract the output power of each photovoltaic inverter from the operating data. The sum of the output power of each photovoltaic inverter constitutes the system photovoltaic power generation at that moment. Construct the system photovoltaic power generation time series from multiple sampling moments.

[0070] S2016 uses the variational mode decomposition algorithm to decompose the system's photovoltaic power generation time series into multiple intrinsic mode function components.

[0071] S2017 uses different prediction models to predict intrinsic mode function components of different frequencies. Specifically, for high-frequency components, the second long short-term memory network model is used for prediction by fusing meteorological forecast data for future periods, while for low-frequency components, the second ARIMA model is used for prediction by fusing meteorological forecast data for future periods.

[0072] S2018 integrates the prediction results of each component through adaptive weighting to obtain the final photovoltaic power generation prediction value.

[0073] In step S2015, the output power of all photovoltaic inverters is extracted from the operating data, and these powers are summed to obtain the total photovoltaic power generation of the system at that moment. The total photovoltaic power generation of the system calculated at each sampling moment is arranged in chronological order to obtain the system photovoltaic power generation time series.

[0074] The other steps are similar to those for forecasting future energy demand, and will not be repeated here.

[0075] S202, with the objective function of minimizing the total operating cost of the system, establish and solve an optimization model; the constraints of the model shall include at least system power balance constraints and equipment operation constraints.

[0076] In this embodiment, the optimization model established by the energy management system master station is a multi-objective optimization model. Its optimization objectives include minimizing the total operating cost of the system and minimizing the total lifespan loss cost of electric vehicle batteries participating in V2G. This allows the system to weigh the impact of charging and discharging behavior on battery lifespan when making decisions. Consequently, each generated scheduling scheme is no longer simply driven by economics, but seeks an optimal balance between economic benefits and battery health to protect user assets.

[0077] Let the scheduling period be T time intervals. The decision variables defined in time interval t (t=1,2,…,T) are as follows: : Power (kW) purchased from the grid during time period t, positive indicates power purchase, negative indicates power sale; : Total charging power of V2G charging and discharging piles during time period t (kW); : Total discharge power of V2G charging and discharging piles during time period t (kW); : Power generated by the photovoltaic inverter to the system bus during time period t (kW); Interruptible load power (kW) that is reduced during period t.

[0078] Total system operating cost The calculation formula is:

[0079] In the formula, for t Electricity price during the specified time period The unit maintenance cost for V2G charging and discharging, including equipment wear and tear, The number of hours for each scheduling period.

[0080] Minimizing the total lifespan loss cost of electric vehicle batteries refers to calculating the effective lifespan loss of batteries caused by V2G charging and discharging behavior based on the battery lifespan model and quantifying it into economic costs. The battery lifespan model comprehensively considers the impact of battery cycle depth, average state of charge, and charge / discharge rate on lifespan.

[0081] Specifically, minimizing the total battery life loss cost of V2G Represented as:

[0082] In the formula, This represents the total number of charging and discharging stations. The replacement cost of a single electric vehicle battery. This refers to the total throughput that the battery can withstand over its lifespan. In order to be in Time period, the The effective lifespan loss of a V2G charging / discharging station due to charging and discharging is expressed as:

[0083] in, The aging factor is a factor greater than 1. When the charging and discharging behavior is under harsh conditions, this factor increases, indicating that the same amount of power throughput causes greater lifespan loss. for The remaining battery charge percentage during the specified time period. for Charge / discharge rate over time period for Time period The charging power of each V2G charging and discharging station for Time period The discharge power of a V2G charging and discharging station This is the scheduling time interval.

[0084] The optimization model must meet the following physical and safety constraints, including system power balance constraints and equipment operation constraints.

[0085] The system power balance constraint is expressed as:

[0086] In the formula, for Forecasted load demand for the period for Total network loss power during a given time period represents the total power loss generated in transmission and distribution lines, transformers, converters, and other equipment during the entire process of transmitting electrical energy from the power supply side to the power consumption side.

[0087] Equipment operation constraints include the following constraints.

[0088] Power constraints for grid interconnection: ; V2G charging and discharging power constraints: , ; V2G simultaneous charging and discharging constraint: To prevent a single charging station from charging and discharging simultaneously, a binary variable is introduced, or simplified to... ; Photovoltaic inverter power constraints: The actual power generation did not exceed the predicted value; Load reduction constraints: , This represents the maximum permissible reduction percentage.

[0089] Since the model may involve multiple objectives and nonlinearity, linearization or intelligent optimization algorithms (such as NSGA-II) can be used to solve it. Specifically, a non-dominated sorting genetic algorithm with an elitist strategy is used to solve the optimization model to obtain the Pareto optimal solution set. Then, based on the fuzzy membership function method, the solution with the highest overall satisfaction is selected from the Pareto optimal solution set as the final energy scheduling scheme. Specifically, for each objective function... Calculate their satisfaction (Values ​​are between 0 and 1), and the solution with the highest overall satisfaction is selected as the final scheduling scheme, expressed as:

[0090]

[0091] Through the above process, the energy management system master station can generate scheduling instructions that achieve the best balance between economic benefits and battery health, thus realizing intelligent energy optimization scheduling.

[0092] S203 generates energy dispatch control commands for one or more devices in the field based on the solution results of the optimization model.

[0093] S2031 transmits the control command data frame to the smart energy unit via the network; after receiving the data frame, the smart energy unit performs CRC cyclic redundancy check.

[0094] The energy management system master station will optimize the solution results, namely the power setpoints of each field device in the next scheduling cycle, and convert them into specific control commands. This process follows the standard Modbus TCP / IP communication protocol to ensure compatibility and universality.

[0095] The master station encodes control commands according to the Modbus protocol's application data units. Each data unit contains the following key information: Device address: Used to uniquely identify the target device; Function code: Specifies the operation to be performed, such as "write a single register"; Register address: Specifies the register location corresponding to the parameter for setting the internal control power of the device.

[0096] Data field: Contains specific control parameter values. For example, for V2G charging and discharging piles, the data field is a signed integer representing the set charging and discharging power value (unit: W), with positive numbers representing charging and negative numbers representing discharging.

[0097] The encoded application data units are encapsulated into TCP / IP data packets to form complete control command data frames that can be transmitted over the network.

[0098] The encapsulated control command data frame is sent from the energy management system master station to the corresponding smart energy unit via the network. After receiving the data frame, the smart energy unit first performs a CRC cyclic redundancy check to verify whether the data is complete and error-free during transmission.

[0099] S2032, if the verification passes, the smart energy unit parses the data frame and converts the parsed power setting value into a control signal that the device can recognize, and sends it to the corresponding field device to drive it to operate at the set power.

[0100] If the verification passes, the smart energy unit parses the data frame and extracts key parameters such as the target device address, function code, register address, and power setting. Based on the parsed parameters, the smart energy unit generates control signals that the device can execute.

[0101] Specifically, the smart energy unit acts as a protocol conversion gateway, converting standard Modbus commands into signals that can be recognized by the field devices' own communication protocols. The smart energy unit then sends control signals to the target devices through its communication interface.

[0102] For example, a 7kW V2G charging and discharging pile: after receiving an instruction containing the parameter "-5000" (indicating a discharge of 5kW), its internal power electronic converter and control module work together to adjust the duty cycle and phase of the switching devices to precisely control the AC power output to the grid to 5kW.

[0103] Photovoltaic inverter: After receiving the instruction, its maximum power point tracking controller will adjust the operating point according to the instruction to stabilize its output power at the set value, rather than the maximum power under natural conditions.

[0104] S2033, the smart energy unit monitors the actual operating power of the field equipment in real time and compares it with the issued power setting value; if the deviation between the actual operating power and the setting value continues to exceed the preset threshold, it will report the abnormal information to the energy management system master station, triggering the master station to re-make energy scheduling decisions and generate new energy scheduling control instructions.

[0105] After the control command is executed, the system enters a closed-loop monitoring state to ensure that the scheduling objectives are achieved. The smart energy unit collects the actual operating data of the controlled equipment in real time and compares the actual operating data with the issued control commands. If the deviation continues to exceed the preset allowable threshold or the equipment reports a fault status, it is determined to be a control anomaly. The smart energy unit immediately sends the anomaly status information to the energy management system master station. After receiving the feedback, the master station takes this as the new system state, restarts the energy demand forecasting and optimization scheduling process, generates new control commands and issues them, thus forming a complete adaptive closed-loop control system to ensure the dynamic optimization of the energy scheduling strategy and the stable operation of the system.

[0106] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A smart energy system based on power communication, characterized in that, include: Multiple field devices, including charging piles, electrical loads, V2G charging and discharging piles, and photovoltaic inverters, all have built-in communication modules for data interaction with the smart energy unit; Smart energy unit: used to collect operating data of field equipment and upload it to the main station of the energy management system, as well as to receive and execute energy dispatch and control instructions issued by the main station of the energy management system; The energy management system master station is configured to perform the following operations: a) Based on operational data, as well as meteorological forecast data and / or date type for future periods, predict future energy demand and photovoltaic power generation using energy demand forecasting models and photovoltaic power generation forecasting models respectively, and use the forecast results as input parameters for solving the optimization model; b) Establish and solve an optimization model with the objective function of minimizing the total operating cost of the system; the constraints of the model shall include at least system power balance constraints and equipment operation constraints; c) Based on the solution results of the optimization model, generate energy dispatch control commands for one or more devices in the field.

2. The smart energy system based on power communication according to claim 1, characterized in that, The communication module built into the field equipment is a dual-mode carrier module, and the energy management system master station and the smart energy unit are connected via 5G communication.

3. A smart energy dispatching method based on power communication, characterized in that, This method is implemented based on the system described in claim 1 or 2, and includes the following steps: The smart energy unit collects the operating data of the field equipment and sends the operating data to the main station of the energy management system; The energy management system master station executes scheduling decisions, generates energy scheduling control commands, and sends the energy scheduling control commands to the smart energy units, which then control the corresponding field equipment. The execution of scheduling decisions includes: a) Based on operational data, as well as meteorological forecast data and / or date type for future periods, predict future energy demand and photovoltaic power generation using energy demand forecasting models and photovoltaic power generation forecasting models respectively, and use the forecast results as input parameters for solving the optimization model; b) Establish and solve an optimization model with the objective function of minimizing the total operating cost of the system; the constraints of the model shall include at least system power balance constraints and equipment operation constraints; c) Based on the solution results of the optimization model, generate energy dispatch control commands for one or more devices in the field.

4. The intelligent energy dispatching method based on power communication according to claim 3, characterized in that, Predicting future energy demand through energy demand forecasting models includes: Extract relevant power data from the operating data, calculate the total load power at that moment based on the relevant power data, and construct a time series of the total load power of the system from the total load power at multiple sampling moments; The variational mode decomposition algorithm is used to decompose the total load power time series of the system into multiple intrinsic mode function components; For intrinsic mode function components of different frequencies, different prediction models are adapted for prediction. Specifically, for high-frequency components, the first long short-term memory network model is used for prediction by fusing meteorological forecast data and date type for future periods. For low-frequency components, the first ARIMA model is used for prediction by fusing meteorological forecast data and date type for future periods. The prediction results of each component are fused through adaptive weighting to obtain the final energy demand forecast.

5. The intelligent energy dispatching method based on power communication according to claim 4, characterized in that, The photovoltaic power generation is predicted using a photovoltaic power generation prediction model, specifically including: The output power of each photovoltaic inverter is extracted from the operating data. The sum of the output power of each photovoltaic inverter constitutes the system photovoltaic power generation at that moment. The system photovoltaic power generation at multiple sampling moments is constructed into a system photovoltaic power generation time series. The variational mode decomposition algorithm is used to decompose the system's photovoltaic power generation time series into multiple intrinsic mode function components; For intrinsic mode function components of different frequencies, different prediction models are adapted for prediction. Specifically, for high-frequency components, the second long short-term memory network model is used for prediction by fusing meteorological forecast data for future periods, and for low-frequency components, the second ARIMA model is used for prediction by fusing meteorological forecast data for future periods. The prediction results of each component are fused through adaptive weighting to obtain the final photovoltaic power generation prediction value.

6. The intelligent energy dispatching method based on power communication according to claim 5, characterized in that, The prediction results of each component are fused using adaptive weighting, specifically including: Regarding the first For each intrinsic mode function component of the prediction model, its in-sample goodness of fit and out-of-sample prediction accuracy are weighted and summed to obtain a comprehensive score. The comprehensive score is then normalized to obtain the initial dynamic weight of the component. ; Regarding the first For a prediction model with individual intrinsic mode function components, obtain the uncertainty measure of its predicted values. For the components predicted using a Long Short-Term Memory (LSTM) network model, the Monte Carlo Dropout method is used to perform several forward propagations to obtain the distribution of the predicted values, and the coefficient of variation of this distribution is used as a measure of its uncertainty. For components predicted using the ARIMA model, the uncertainty is measured by calculating the confidence interval of the model at the prediction time point and using the ratio of the width of the confidence interval to the predicted value. ; Using uncertainty measurement Calculate the weight correction factor In the formula The preset attenuation coefficient; Initial dynamic weights With weighting correction factor Multiply and normalize to obtain the first... The final fusion weight of each component ; Use the final fusion weights The prediction results of all components are weighted and summed to output the final energy demand forecast.

7. The intelligent energy dispatching method based on power communication according to claim 6, characterized in that, The optimization model established by the main station of the energy management system is a multi-objective optimization model. Its optimization objectives include minimizing the total operating cost of the system and minimizing the total lifespan loss cost of electric vehicle batteries participating in V2G.

8. The intelligent energy dispatching method based on power communication according to claim 7, characterized in that, The cost of battery life loss is expressed as: In the formula, This represents the total number of charging and discharging stations. The replacement cost of a single electric vehicle battery. This refers to the total throughput that the battery can withstand over its lifespan. In order to be in Time period, the The effective lifespan loss of a V2G charging / discharging station due to charging and discharging is expressed as: in, To accelerate aging factors, for The remaining battery charge percentage during the specified time period. for Charge / discharge rate over time period for Time period The charging power of each V2G charging and discharging station for Time period The discharge power of a V2G charging and discharging station This is the scheduling time interval.

9. The intelligent energy dispatching method based on power communication according to claim 8, characterized in that, When solving the optimization model, a non-dominated sorting genetic algorithm with an elitist strategy is used to solve the optimization model to obtain the Pareto optimal solution set; then, the solution with the highest overall satisfaction is selected from the Pareto optimal solution set based on the fuzzy membership function method as the final energy scheduling scheme.

10. The intelligent energy dispatching method based on power communication according to claim 9, characterized in that, Based on the solution results of the optimization model, energy dispatch control commands are generated for one or more devices in the field, specifically including: The solution results of the optimization model are encoded according to the Modbus TCP / IP protocol and encapsulated into control command data frames containing device addresses, function codes, and power parameters; the solution results include the power setpoints of the corresponding field devices; The control command data frame is transmitted to the smart energy unit via the network; after receiving the data frame, the smart energy unit performs CRC cyclic redundancy check. If the verification passes, the smart energy unit parses the data frame and converts the parsed power setpoint into a control signal that the device can recognize, which is then sent to the corresponding field device to drive it to operate at the set power. The smart energy unit monitors the actual operating power of the field equipment in real time and compares it with the issued power setting value. If the deviation between the actual operating power and the setting value continues to exceed the preset threshold, it will report the abnormal information to the main station of the energy management system, triggering the main station to re-make energy scheduling decisions and generate new energy scheduling control instructions.

Images