Energy storage system used in V2G application scene and optimization management and control strategy thereof

Abstract

The invention discloses an energy storage system used in a V2G application scene and an optimization management and control strategy thereof, the energy storage system is a combined prediction model based on time synchronization pulse per second, fused with ARIMA and LSTM time sequence prediction algorithms, and matched with a double-layer optimization architecture of an MPC controller and a genetic algorithm, so that a multi-layer composite optimization control strategy is formed. The method is used for stabilizing supply and demand fluctuation and space-time mismatch by using energy buffering according to real-time data of a power grid through an EV, an energy storage system and the power grid under the condition that a vehicle owner performs random charging behaviors. Through accurate time synchronization, an advanced prediction model and an optimization algorithm, benefit balance among a vehicle owner, an energy storage operator and a power grid can be realized under the condition that charging random behaviors of the vehicle owner are considered.

Description

Technical Field

[0001] The present invention relates to an energy storage system in an electric power system, and in particular to an energy storage system and an optimization control strategy thereof in an application scenario where an electric vehicle provides energy storage to a power grid (V2G). Technical Background

[0002] V2G (Vehicle-to-Grid) technology refers to the interaction between electric vehicles (EVs) and the power grid, which achieves dynamic balance of energy and resource optimization through the two-way energy flow between the battery of the electric vehicle and the power grid. Vehicles can not only charge from the power grid, but also feed back electricity to the grid when the electricity price is high, realizing the two-way flow of energy, which is gradually becoming an effective way of energy management. However, with the large-scale application of electric vehicles (EVs) and renewable energy, the load management and power quality problems faced by the power grid are becoming increasingly serious. At the same time, with the increase in the number of EVs, the randomness of the charging behavior of car owners has brought many challenges and complexity to the two-way orderly interaction of the power grid, and has affected the interests of car owners, energy storage operators, the power grid and the stability of the power grid.

[0003] As a buffer between EV and renewable energy to the power grid, the energy storage system needs to smooth the power output curve by energy storage at the power supply end, and on the other hand, it needs to adjust the power consumption end by flexible and adjustable load. At the same time, considering factors such as time-varying electricity prices and charging demand forecasts (charging timing characteristics), how to ensure the operational efficiency and economic benefits of the overall power system, so that the energy storage system can provide such a capability, that is, to use energy buffering to smooth supply and demand fluctuations and time and space mismatches according to the real-time data of the power grid, is one of the technical contents that the applicant has been committed to developing.

[0004] Therefore, in the V2G application scenario, the energy storage system acts as a buffer between EV and renewable energy to the power grid. How to ensure a balance of interests between car owners, energy storage operators and the power grid, and take into account the randomness of car owners' charging behavior, is a practical problem currently facing the energy storage system. Summary of the invention

[0005] The present invention provides an energy storage system and an optimization management and control strategy for a V2G application scenario. The system is based on a time-synchronized second pulse technology, a prediction model integrating ARIMA (Autoregressive Integrated Moving Average) and LSTM (Long Short-Term Memory), and a multi-layer composite optimization control strategy of MPC (Model Predictive Control) and GA (Genetic Algorithm). The control strategy can achieve a balance of interests among car owners, energy storage operators and power grids by taking into account the random charging behavior of car owners through precise time synchronization, advanced prediction models and optimization algorithms.

[0006] The technical solution to achieve the above purpose is: an energy storage system for V2G application scenarios, characterized in that: the energy storage system is based on time-synchronized second pulses, integrates a combined prediction model of two time series prediction algorithms, ARIMA and LSTM, and is equipped with an MPC controller and a genetic algorithm double-layer optimization architecture, thereby forming a multi-layer composite optimization control strategy, which is used for car owners to use EV, energy storage system and power grid to use energy buffering according to real-time data of the power grid to smooth supply and demand fluctuations and time and space mismatches in the case of random charging behavior.

[0007] In the above energy storage system for V2G application scenario, the time synchronization second pulse uses a high-precision timing reference to ensure that the data collection and control operations of each node of the energy storage system are performed under the same time reference, ensuring the timing consistency of the data collection, processing and control decision of the entire system.

[0008] The system hardware of the time synchronization second pulse includes: BD / GPS dual-mode receiver, high-precision clock synchronization module and taming clock submodule,

[0009] The BD / GPS dual-mode receiver receives the standard time signal transmitted by the satellite, and generates a 1PPS signal through the high-precision clock synchronization module. The BD / GPS dual-mode receiver and the high-precision clock synchronization module are connected and deployed in the energy storage system. The time synchronization interface of the high-precision clock synchronization module outputs the 1PPS signal to the corresponding interface of the energy control mainboard of the energy storage system to ensure that all control modules of the energy storage system maintain clock synchronization under a unified and high-precision time reference.

[0010] The clock taming submodule is deployed in the V2G bidirectional charging pile and connected to its control module, and is connected to the time synchronization interface of the high-precision clock synchronization module of the energy storage system. It is calibrated and tamed by receiving the 1PPS signal provided by it to achieve clock synchronization consistent with the energy storage system.

[0011] In the above energy storage system for V2G application scenario, the ARIMA forecasting model is used to analyze and forecast the linear time series data in the time series, and to identify and capture the trend and seasonal changes of the data through the historical charging data of electric vehicles;

[0012] The LSTM prediction model is used to learn and predict time series data with nonlinearities and long-term dependencies in time series. It can effectively capture the randomness of car owners' charging behavior, including changes with time, temperature, and weekdays / holidays.

[0013] The combined prediction model of the two prediction algorithms can provide more accurate charging load prediction for the energy storage system, making the energy storage system more intelligent when facing random and volatile demand changes at different time scales, enabling the system operation to be synchronized with grid load fluctuations, accurately optimizing charging and discharging strategies, using the ARIMA model to handle shorter-term linear dependencies, and using the LSTM model to capture nonlinear patterns such as long-term dependencies and complex patterns. The role of the time-synchronized second pulse in this combined prediction model is to ensure that the time scale of all time series data is synchronized with the location when the actual event occurs, thereby improving the accuracy and reliability of the prediction.

[0014] In the above energy storage system for V2G application scenario, in the MPC controller and genetic algorithm two-layer optimization architecture, wherein:

[0015] The MPC controller generates the optimal charging / discharging strategy in real time based on the results provided by the combined prediction model of ARIMA and LSTM to minimize operating costs and meet grid demand;

[0016] The genetic algorithm is used to optimize the parameters of the MPC controller to ensure that the best control strategy is found in the V2G environment. That is, by using the control strategy generated by the MPC controller as the first generation population of the genetic algorithm, randomly generating an initial parameter set, and calculating the appropriate fitness function for each individual, the economic benefits of the car owner, energy storage operator, and power grid are evaluated, and individuals with high fitness are selected, crossed over, and mutated to generate a new generation of populations, that is, to meet the optimal control strategy for EV, energy storage system, and power grid.

[0017] In the above energy storage system for V2G application scenario, the energy storage system receives a time synchronization signal at the beginning of each full second, compares and calibrates the signal with the clocks of each control module in the system, and ensures that all control modules in the system operate on the same time standard. In each synchronization cycle, the high-precision clock synchronization module 1PPS updates the clock of each tamed clock submodule to perform time calibration and deviation correction. The specific algorithm is:

[0018] ΔTi =T BD / GPS -T local,i

[0019]

[0020] Where: T BD / GPS Represents the BD / GPS standard time received by the high-precision clock synchronization module, i.e. 1PPS, T local,i is the local clock time of the ith subsystem, ΔT i is the time deviation of the measurement, is the corrected local time.

[0021] The present invention also provides an optimization management and control strategy for an energy storage system, comprising the following steps:

[0022] Step 1, receiving BD / GPS time synchronization signal and performing second pulse time synchronization;

[0023] Step 2, collecting charging and discharging time series data;

[0024] Step 3: Combined prediction model training, using the ARIMA model to process the short-term linear relationship of the charging behavior data, and then using these prediction results as input combined with other features, including temperature and date type, using the LSTM model to capture and predict long-term, nonlinear patterns;

[0025] Step 4: Real-time load forecasting, which gradually reduces the forecast error by iteratively updating weights and biases to achieve more accurate forecasts of future demand;

[0026] Step 5: MPC control strategy generation. The MPC controller generates the optimal charge / discharge control strategy for the energy storage system based on the prediction results of the time series relationship generated by the ARIMA and LSTM models. It can be adjusted according to real-time data to ensure stable operation and optimized control of the system in a dynamic environment.

[0027] Step 6: Genetic algorithm optimization: The genetic algorithm optimizes the parameters of the MPC controller to avoid falling into the local optimum. Through the iterative evolution of the genetic algorithm, the control strategy is adaptively adjusted to improve the control effect.

[0028] Step 7: Generate a charge / discharge strategy and execute the charge / discharge strategy.

[0029] In the above-mentioned optimization control strategy for an energy storage system, the ARIMA model in step 3 is specifically as follows:

[0030] Use the ARIMA model for short-term linear forecasting: The ARIMA model consists of three parameters p, d, and q: p is the autoregressive order; d is the difference order; q is the moving average order. The ARIMA model parameters d, p, and q are determined by stationarity testing, calculation, and observation of autocorrelation functions and partial autocorrelation function graphs.

[0031] d is determined by the stationarity test and ADF test, that is, the stationarity of the time series is determined using the ADF test;

[0032] ADF test formula: Δy t =α+βt+γy t-1 +δ1Δy t-1 +δ2Δy t-2 +…+δ k Δy t-k +ε t

[0033] Among them, Δy t is the difference term, α is a constant, βt is the time trend term, γy t-1 is the lag term, δ i Δy t-i is the lagged difference term, ε t is the random error term,

[0034] The autoregressive order p and the moving average order q are determined by observing the autocorrelation function ACF and the partial autocorrelation function PACF.

[0035] ACF formula:

[0036] Where: k is the autocorrelation coefficient at lag k; y t is the value of the time series at time t; y t-k is the value of the time series at time tk, which means the value at time t pushed forward k time points. is the mean of the time series; T is the total length of the time series;

[0037] PACF formula:

[0038] Where: φ kk is the partial autocorrelation coefficient at lag k, which indicates the autocorrelation of the sequence excluding the intermediate lag term at lag k; ρ k is the autocorrelation coefficient at lag k, which indicates the linear correlation between the value of the sequence at the current moment and the value at lag k moments; φ k-1,i is the partial autocorrelation number at lag k-1, which is the intermediate term to be considered when calculating the partial autocorrelation coefficient at the current lag k; ρ k-iis the autocorrelation coefficient at lag ki, which is used to adjust the value of the partial autocorrelation coefficient at the current lag k.

[0039] ARIMA model construction and prediction: The estimated parameter values ​​are input into the ARIMA model to establish the ARIMA (p, d, q) model. Through the combined effects of difference, regression and moving average, short-term linear prediction of time series data can be performed, which can be used to predict charging demand.

[0040] In the above-mentioned optimization control strategy of an energy storage system, the LSTM model in step 3 is specifically as follows:

[0041] Use LSTM for long-term nonlinear prediction: The LSTM model controls the flow of information through three "gate" structures, namely, forget gate, input gate, and output gate, thereby effectively preserving long-term dependencies;

[0042] First, the forget gate f t , which is used to decide what information to discard from the cell state,

[0043] f t =σ(W f ·[h t-1 ,X t ]+b f )

[0044] Where: W f is the weight matrix of the forget gate; [h t-1 ,X t ] is to connect the hidden state of the previous time point and the current input; b f is the bias of the forget gate; σ is the sigmoid activation function, which outputs a scalar from 0 to 1, indicating the retained ratio;

[0045] Second, input gate i t and candidate cell states

[0046] i t =σ(W i ·[h t-1 ,X t ]+b i ), Function: decide what new information to store in the cell state;

[0047]

[0048] Where W i ,W c is the weight matrix of the input gate; b i ,b cis the bias term of the input gate; tanh is the hyperbolic tangent activation function, which is used to generate candidate cell states, and the output range is between -1 and 1;

[0049] Third, the cell state updates Y t , used to update the cell state, combining forgotten and newly input information,

[0050]

[0051] where Y t is the cell state at the current time point; Y t-1 is the cell state at the previous time point;

[0052] Fourth, output gate o t and the hidden state h t :

[0053] o t =σ(W o ·[h t-1 ,X t ]+b o ), which is used to decide what information to output as the output at this time point, h t =o t *tanh(Y t ), that is, the actual demand forecast is finally output through the conversion function

[0054] Where: W o is the weight matrix of the output gate; b o is the bias term of the output gate.

[0055] In the above-mentioned optimization control strategy for an energy storage system, the LSTM model calculation steps in step 3 are as follows:

[0056] Step A, initial steps:

[0057] Collect features related to charging demand including temperature, time, and holidays, and create feature set X t , determine the hidden layer, which contains several LSTM units, each of which captures the temporal dependencies in the data.

[0058] Step B, data processing:

[0059] The forecast result of the ARIMA model C t+1 Combine with other features X t , including the data from time tn to time t, to construct the training set of LSTM model [X t-n ,…,X t ,C t+1 ];

[0060] Step C, LSTM model training:

[0061] The training of the LSTM model is an iterative process that uses historical time series data to train the LSTM network through the steps of forward propagation, loss function, back propagation, and parameter update to determine the weight matrix and bias value;

[0062] Among them, the goal of the loss function L is to minimize the error between the predicted output and the actual demand, using the mean square error MSE:

[0063]

[0064] Where: Z t+1 is the actual charging demand value, is a charging demand prediction value output by LSTM based on the training set using the current weights and biases at each step of the iteration. N is the number of samples.

[0065] Backpropagation is used to calculate the gradient of the loss function with respect to each model parameter (weight and bias). The gradient represents the rate of increase of the loss function near the current value of the parameter and is used to update the weights to reduce the error. The parameter update is calculated using the gradient descent method:

[0066]

[0067] Among them, α is the learning rate, L is the loss function, and are the gradients of the weight matrix and bias, W old and b old They are the unified expressions of the weight matrices and bias values ​​of the three "gates", and They are the updated weight matrix and bias values, which are updated and adjusted according to the gradient of the loss function in each training iteration, so that the model can better learn and capture the underlying patterns of the data;

[0068] Step D, prediction:

[0069] Input the training set [X t-n ,…,X t ,C t+1 ] to the LSTM formula after updating the weight matrix and bias value and calculating the final prediction at time t+1 through the conversion function

[0070] In the above-mentioned optimization control strategy of an energy storage system, the MPC controller in step 5 performs control through the following process:

[0071] First, the dynamic behavior of the energy storage system can be described by a linear state space model:

[0072]

[0073] The output equation is: k =Dx k +v k

[0074] Among them, x k is the system state vector at time k; u k is the control input vector at time k; is the predicted electricity demand at time k+1 predicted by the ARIMA and LSTM models; A is the state transition matrix; B is the control matrix; C is the input matrix; y k is the system output vector at time k; w k is the process noise; D is the output matrix; v k is the measurement noise;

[0075] Second, the goal of MPC is to optimize the control input u in the time domain predicted by the ARIMA and LSTM models. k , so that the system state x k Tracking the desired state x ref ; The prediction time domain length is N, then the control model of the predicted state and control input in the next N moments can be expressed as:

[0076]

[0077] in is the sequence of predicted state vectors from time k+1 to k+N; i = 1, 2, 3...N, is the predicted electricity demand at time k+i, predicted by the ARIMA and LSTM models; the model recursively calculates the future state, where each predicted state x k+i All based on the current state x k+i-1 , control input u k+i-1 and forecast demand Calculated,

[0078] Third, the goal of the MPC controller is to minimize the operating cost and meet the grid demand. The objective function is expressed as:

[0079]

[0080] Among them, the vector set of all control inputs in the prediction domain is expressed as: x i -x ref represents the state deviation at time i, that is, the gap between the current state and the expected state; (x i -x ref )T Q(x i -x ref ) represents the weighted quadratic norm of the state deviation, that is, the gap between the current state and the expected state; Q represents the state deviation weighted matrix; u i represents the control input vector at time i; represents the weighted quadratic norm of the control input, that is, the size of the control input; R represents the control input weighting matrix; the control input u i and state x i Certain constraints need to be met:

[0081]

[0082] The beneficial effects of the present invention are as follows: the technical solution of the present invention applies the time-synchronized second pulse technology to the energy storage system to ensure high synchronization inside and outside the system, improve the overall operating efficiency, and integrate the ARIMA and LSTM prediction models with a unified time base, which can not only capture the linear trend of the data, but also process complex nonlinear characteristics, and improve the load forecasting accuracy. Then, by combining MPC and genetic algorithm based on the prediction results, the timely optimization control of the system is realized, the effectiveness and stability of the control strategy are improved, and the control signal can be delivered and executed at the right time, thereby achieving higher prediction accuracy and adaptive control of the energy storage system. At the same time, the second pulse technology ensures that all devices operate in the same time frame in complex and large-scale application scenarios, so that the coordination and synchronization of the overall power system can be achieved. Compared with the prior art, the scheme significantly improves the operating efficiency and stability of the energy storage system, reduces the overcharging and discharging behavior of the battery of the energy storage system by optimizing the control strategy, prolongs the battery life, and optimizes the energy management strategy, reduces energy loss and operating costs, and improves economic benefits, thereby achieving better performance in the dynamically changing market environment and participant behavior, and providing more flexible support for the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 is a system architecture diagram of the energy storage system of the present invention;

[0084] Figure 2 It is a flow chart of the optimization control strategy of the energy storage system of the present invention. DETAILED DESCRIPTION

[0085] The present invention will be further described below in conjunction with the accompanying drawings.

[0086] 1. The technical solution of the energy storage system of the present invention is as follows (see Figure 1 ):

[0087] 1. Time synchronization second pulse technology

[0088] The time-synchronized second pulse technology plays a cascade role in the solution of the present invention, using a high-precision timing reference to ensure that the data acquisition and control operations of each node in the system are performed under the same time reference, ensuring the timing consistency of the data acquisition, processing and control decisions of the entire system, and laying the foundation for the implementation of efficient prediction and optimization control.

[0089] 2. ARIMA and LSTM combined prediction model

[0090] The energy storage system in this invention aims to balance the load demand of the power grid. The system needs to predict fluctuations in the power grid to optimize storage and discharge operations. The scheme proposes a combined prediction model combining ARIMA and LSTM time series prediction algorithms for charging behavior prediction:

[0091] The ARIMA model is used to analyze and predict linear time series data in time series, and the historical charging data of electric vehicles is used to identify and better capture the trends and seasonal changes of the data.

[0092] The LSTM model is used to learn and predict time series data with nonlinearities and long-term dependencies in time series, and can effectively capture the randomness of car owners' charging behavior, such as changes with time, temperature, weekdays / holidays, and other factors.

[0093] The combination of the two models can provide more accurate charging load forecasts for energy storage systems, making energy storage systems more intelligent when facing random and fluctuating demand changes at different time scales, enabling system operation to be synchronized with grid load fluctuations, and accurately optimizing charging and discharging strategies, bringing economic and environmental benefits to the power system. The ARIMA model is used to handle shorter-term linear dependencies (seasonality and trend), and the LSTM model is used to capture nonlinear patterns such as long-term dependencies and complex patterns. For example, the ARIMA model is used to obtain preliminary charging demand forecasts for trends and seasonality, and then this value and other related features (such as temperature, date type, etc.) are input into the LSTM model to learn and further iterate and refine these data to obtain a final more accurate forecast. The role of time-synchronized second pulse technology in this model is to ensure that the time scale of all time series data is synchronized with the location when the actual event occurs, thereby improving the accuracy and reliability of the forecast.

[0094] 3. MPC controller and genetic algorithm two-layer optimization architecture

[0095] The MPC controller plays a core role in the solution of the present invention. It generates the optimal charge / discharge strategy in real time according to the results provided by the ARIMA and LSTM combined prediction model to minimize the operating cost while meeting the grid demand (objective function), achieving the dual goals of economy and safety.

[0096] GA genetic algorithm is used to optimize MPC controller parameters to ensure the best control strategy is found in a complex and uncertain V2G environment. By using the control strategy generated by MPC as the first generation population of the genetic algorithm, the initial parameter set is randomly generated, and the appropriate fitness function of each individual is calculated (in this case, the fitness function is defined as an inverse relationship between the objective function, that is, when the objective function is minimum, the fitness is maximum), the economic benefits of the car owner, energy storage operator, and power grid are evaluated, and individuals with high fitness are selected, crossed, and mutated to generate a new generation of populations, that is, the optimal control strategy that meets the needs of multiple stakeholders.

[0097] The combination of the two algorithms can provide the optimal control strategy for the energy storage system in the V2G scenario, which can achieve the best balance between economic benefits and power supply and demand in the operation of the energy storage system. MPC uses dynamic models to predict future outputs and calculate the optimal control decisions. GA is used to search for the optimal MPC configuration, thereby achieving efficient operation of the system in the face of uncertainty and dynamic market conditions. The precise time mark provided by the time-synchronized second pulse technology ensures that the control signals of the MPC controller and the genetic algorithm optimization output are delivered and executed at the right time, so that the coordination and synchronization of the overall system can be achieved.

[0098] 4. Implementation steps

[0099] -Synchronize data flow in V2G system using time synchronization pulse-per-second technology.

[0100] -Training a combined ARIMA and LSTM model using historical data.

[0101] -Use the trained model for real-time load forecasting.

[0102] -Generate optimal control strategies using MPC and genetic algorithms.

[0103] -Apply control strategies to manage the charging / discharging process of energy storage systems.

[0104] The above steps can be found in Figure 2 flow chart.

[0105] II. Specific implementation of the energy storage system of the present invention

[0106] 1. Implementation of time-synchronized pulse-per-second technology

[0107] Time synchronization pulse-per-second technology is the key to ensuring accurate data synchronization in V2G energy storage systems. This technology uses high-precision clock synchronization signals to coordinate the time of all devices in the system. This is essential for accurately implementing data collection, processing and control decisions.

[0108] The hardware part of the time synchronization second pulse system in the solution of the present invention includes: a BD / GPS dual-mode receiver, a high-precision clock synchronization module, and a taming clock submodule. The BD / GPS dual-mode receiver receives the standard time signal transmitted by the satellite, and generates a 1PPS signal through processing by the high-precision clock synchronization module. The BD / GPS dual-mode receiver and the high-precision clock synchronization module are connected and deployed in an energy storage system with an energy storage cabinet as the main body. The time synchronization interface of the high-precision clock synchronization module outputs a 1PPS signal to the corresponding interface of the energy control mainboard of the energy storage system to ensure that all control modules of the energy storage system maintain clock synchronization under a unified and high-precision time reference. The taming clock submodule is deployed in the V2G bidirectional charging pile and connected to its control module, and is connected to the time synchronization interface of the high-precision clock synchronization module of the energy storage system. It is calibrated and tamed by receiving the 1PPS signal provided by it to achieve clock synchronization consistent with the energy storage system.

[0109] The entire energy storage system receives a time synchronization signal at the beginning of each full second, and compares and calibrates this signal with the clocks of each control module in the system to ensure that all control modules in the system operate on the same time standard. In each synchronization cycle, the high-precision clock synchronization module 1PPS updates the clock of each tamed clock subsystem, performs time calibration and deviation correction algorithms to represent:

[0110] ΔT i =T BD / GPS -T local,i

[0111]

[0112] Among them, T BD / GPS Represents the BD / GPS standard time received by the high-precision clock synchronization module, i.e. 1PPS. local,i is the local clock time of the ith subsystem, ΔT i is the time deviation of the measurement, is the corrected local time.

[0113] 2. Implementation of ARIMA and LSTM combined forecasting model

[0114] When predicting future electricity consumption behavior, the energy storage system needs to use a prediction function with time as the independent variable, that is, the time parameters of each control part of the energy storage system are all used As a benchmark. The ARIMA and LSTM combined prediction model in the solution of the present invention uses their respective advantages to improve the accuracy of charging load prediction. First, the ARIMA model is used to process the short-term linear relationship of the charging behavior data, and then these prediction results are used as input combined with other features such as temperature, date type, etc., and the LSTM model is used to capture and predict long-term, potentially more complex nonlinear patterns.

[0115] Step 1: Data preparation and pre-processing

[0116] 1. Collect data: Obtain historical charging data, including the charging time and charging amount of electric vehicles, as well as other external factors that may affect charging demand, such as temperature, day of the week, holidays, and electricity prices during different time periods.

[0117] 2. Data cleaning: handle missing values, remove outliers, and standardize data to eliminate the impact of dimensions.

[0118] 3. Feature engineering: construct new features, extract weekdays and weekends from dates, extract peak and trough periods from time-period electricity prices, etc.

[0119] Step 2: Short-term linear forecasting using ARIMA

[0120] After completing step 1, use ARIMA for short-term linear forecasting. The ARIMA model captures the characteristics of the time series by describing the autoregressive (AR), differential (I) and moving average (MA) parts of the time series. The ARIMA (p, d, q) model consists of three main parameters: p (autoregressive order, representing how many past values ​​it depends on), d (difference order, the number of differences required to make the time series stable), and q (moving average order, representing how many past error terms it depends on). The calculation process is as follows:

[0121] 1. Model selection and parameter estimation

[0122] The form of establishing the ARIMA model is ARIMA(p,d,q), where

[0123] d is determined by the stationarity test, ADF test (Augmented Dickey-Fuller Test), that is, the stationarity of the time series is determined by the ADF test. If the time series is not stationary, it is stabilized by the difference operation. If the statistic of the ADF test is less than the significance level (usually 0.05), the time series can be considered to be stationary, otherwise the difference operation is performed.

[0124] ADF test formula: Δy t =α+βt+γy t-1 +δ1Δy t-1 +δ2Δy t-2 +…+δ k Δy t-k +ε t

[0125] Among them, Δy t is the difference term, α is a constant, βt is the time trend term, γy t-1 is the lag term, δi Δy t-i is the lagged difference term, ε t is a random error term.

[0126] Difference operation: If the original sequence is non-stationary (i.e., the null hypothesis cannot be rejected), perform a difference operation and then re-perform the ADF test. Perform the difference operation in sequence until the sequence is stationary. The number of differences is the parameter d.

[0127] p and q are determined by the autocorrelation (ACF) and partial autocorrelation (PACF):

[0128] The autoregressive order p and the moving average order q can be determined by observing the autocorrelation function (ACF) and the partial autocorrelation function (PACF).

[0129] ACF is used to describe the autocorrelation of time series data at different lags k, that is, a sequence y t With its own lagged value y t-k ACF helps identify the seasonality and periodicity of the series by plotting the autocorrelation coefficient at different lag values.

[0130] ACF formula:

[0131] Where: k is the autocorrelation coefficient at lag k; y t is the value of the time series at time t; y t-k It is the value of the time series at time tk, which means the value at k time points forward from time t. is the mean of the time series; T is the total length of the time series.

[0132] PACF is used to describe the partial autocorrelation of the measured time series at different lags k, aiming to exclude the influence of the intermediate lag values ​​to determine the direct correlation. PACF is used to identify the lag order of direct correlation in the time series, that is, the order of the AR model.

[0133] PACF formula:

[0134] Where: φ kk is the partial autocorrelation coefficient at lag k, which indicates the autocorrelation of the sequence excluding the intermediate lag term at lag k; ρ k is the autocorrelation coefficient at lag k, which indicates the linear correlation between the value of the sequence at the current moment and the value at lag k moments; φ k-1,i is the partial autocorrelation number at lag k-1, which is the intermediate term to be considered when calculating the partial autocorrelation coefficient at the current lag k; ρ k-i is the autocorrelation coefficient at lag ki, which is used to adjust the value of the partial autocorrelation coefficient at the current lag k.

[0135] According to the above formula, the ACF and PACF at each lag k are calculated

[0136] Determine the autoregressive order p:

[0137] Characteristics of the autoregressive model (AR): the ACF graph gradually decays; the PACF graph is truncated after order p (that is, the partial autocorrelation coefficient is significantly zero after a significant lag order, and if it becomes zero after being significantly non-zero at lag k, then p = k).

[0138] By observing the PACF diagram, when PACF is truncated at a certain lag order p, this lag order is the order p of the AR model.

[0139] Determine the moving average order q:

[0140] Characteristics of the moving average model (MA): the PACF graph gradually decays; the ACF graph is truncated after the order q (that is, the autocorrelation coefficient is significantly zero after a significant lag order, and if it becomes zero after being significantly non-zero at lag k, then q = k).

[0141] By observing the ACF graph, when the ACF is cut off at a certain lag order q, this lag order is the order q of the MA model.

[0142] In summary, the ARIMA model parameters d, p and q are determined by stationary test (ADF test) and calculation and observation of autocorrelation function (ACF) and partial autocorrelation function (PACF) graphs. Through these steps, the model parameters suitable for time series data can be reasonably determined and applied to prediction.

[0143] 2. ARIMA model construction and forecasting

[0144] The estimated parameter values ​​are input into the ARIMA model to establish the ARIMA (p, d, q) model for short-term linear forecasting.

[0145] Based on current and historical data, the charging demand C at time point t+1 is predicted t+1 , establish the model formula:

[0146]

[0147] Among them, C t is the value of the time series at time point t, and L is the lag operator, which moves the time series data backward by k time points, that is, L k C t =C t-k ,φ i is the autoregressive coefficient (expressed as the linear relationship between the current value and its past values ​​in the time series), i represents the number of lags, and θi is the moving average coefficient (expressed as the effect of past forecast errors on current forecast errors), ε t is a white noise error term, representing the random error at time point t.

[0148] The mathematical expression of the AR part is:

[0149]

[0150] The mathematical expression for part I is:

[0151] I:(1-L) d C t

[0152] The mathematical expression of the MA part is:

[0153]

[0154] According to the above formulas, the ARIMA (1,1,1) model can be used for calculation and interpretation. First, a difference operation is required, that is, the time series data {C t} Perform d = 1 times of difference to make it stable, and obtain the new time series data {ΔC t}, ΔC t =(1-L)C t =C t -C t-1 . Then, the ARIMA model is fitted based on the new time series data after differentiation. The solution of the present invention uses the Yule-Walker equation to solve the autoregressive coefficient φ1 and the maximum likelihood estimation method to solve the moving average coefficient θ1. Finally, the ARIMA (p, d, q) model formula is iteratively updated based on the obtained φ1 and θ1 to construct a practical prediction model for prediction of future time points. According to the ARIMA (1,1,1) model, the prediction difference value ΔC at the next moment t+1 can be obtained t+1 =φ1ΔC t +ε t+1 +θ1ε t , where we assume that ε t and ε t+1 is a value close to zero. Then, the predicted difference value is converted back to the predicted value of the original sequence, that is, the next moment value C t+1 =C t +ΔC t+1 Make predictions.

[0155] In summary, the ARIMA model can model and predict time series data through the combined effects of difference, regression and moving average, and thus can be used to predict charging demand.

[0156] Step 3: Long-term nonlinear prediction using LSTM

[0157] In the solution of the present invention, the charging demand prediction is divided into two steps, namely, the charging demand prediction value C output by the above ARIMA model t+ 1 Combine other relevant features (such as temperature, date type, and user behavior trends) and input them into the LSTM model for training to obtain the final charging demand prediction value

[0158] The LSTM model controls the flow of information by introducing three "gate" structures (forget gate, input gate, and output gate), thereby effectively preserving long-term dependencies. The model formula is as follows:

[0159] 1. Forget gate f t , Function: Decide what information to discard from the cell state.

[0160] f t =σ(W f ·[h t-1 ,X t ]+b f ).

[0161] Where W f is the weight matrix of the forget gate; [h t-1 ,X t ] is to connect the hidden state of the previous time point and the current input; b f is the bias of the forget gate; σ is the sigmoid activation function, which outputs a scalar from 0 to 1, indicating the retained ratio.

[0162] 2. Input gate i t and candidate cell states

[0163] i t =σ(W i ·[h t-1 ,X t ]+b i ), Function: decide what new information to store in the cell state;

[0164]

[0165] Where W i ,W c is the weight matrix of the input gate; b i ,b c is the bias term of the input gate; tanh is the hyperbolic tangent activation function, which is used to generate candidate cell states, and the output range is between -1 and 1.

[0166] 3. Cell state update Y t, Function: Update the cell state, combining forgotten and newly input information.

[0167]

[0168] where Y t is the cell state at the current time point; Y t-1 is the cell state at the previous time point.

[0169] 4. Output gate o t and the hidden state h t :

[0170] o t =σ(W o ·[h t-1 ,X t ]+b o ), function: determine what information to output as the output at this time point. t =o t *tanh(Y t ), that is, finally through the conversion function (such as linear function, a and b are linear transformation parameters, and the actual demand forecast is obtained by training the system using linear regression to minimize the mean square error between the actual output and the predicted output.

[0171] Where W o is the weight matrix of the output gate; b o is the bias term of the output gate.

[0172] The specific calculation steps of the LSTM model are as follows:

[0173] 1. Initial steps:

[0174] Collect features related to charging demand: temperature, time (hours, days of the week), holidays, etc., and create feature set X t The hidden layer is determined to contain several LSTM units, each of which captures the temporal dependencies in the data.

[0175] 2. Data processing:

[0176] The forecast result of the ARIMA model C t+1 Combine with other features X t (including data from time tn to time t), construct the training set of LSTM model [X t-n ,…,X t ,C t+1 ].

[0177] 3. LSTM model training:

[0178] The training of the LSTM model is an iterative process that uses historical time series data to train the LSTM network through the steps of forward propagation, loss function, backpropagation, and parameter update to determine the weight matrix and bias value.

[0179] Among them, the goal of the loss function L is to minimize the error between the predicted output and the actual demand, using the mean square error (MSE):

[0180] Where Z t+1 is the actual charging demand value, It is a charging demand prediction value output by LSTM based on the training set using the current weights and biases at each step of the iteration, and N is the number of samples.

[0181] Backpropagation is used to calculate the gradient of the loss function with respect to each model parameter (weight and bias). The gradient represents the rate at which the loss function rises near the current value of the parameter and is used to update the weights to reduce the error. Parameter updates are calculated using the gradient descent method:

[0182]

[0183] Among them, α is the learning rate, L is the loss function, and are the gradients of the weight matrix and bias, W old and b old They are the unified expressions of the weight matrices and bias values ​​of the three "gates" respectively. and They are the updated weight matrix and bias values, which are updated and adjusted according to the gradient of the loss function in each training iteration, so that the model can better learn and capture the underlying patterns of the data.

[0184] 4. Prediction:

[0185] Input the training set [X t-n ,…,X t ,C t+1 ] to the LSTM formula after updating the weight matrix and bias value and calculating the final prediction at time t+1 through the conversion function

[0186] In summary, the solution of the present invention combines the ARIMA model and the LSTM model, and uses the ability of the ARIMA model to capture linear and short-term dependency features and the LSTM model to capture complex long-term dependency features, thereby solving the problem that a single model is difficult to capture complex patterns. By iteratively updating weights and biases to gradually reduce prediction errors, a more accurate prediction of future demand can be achieved. In practical applications, the combined model can more accurately predict the charging demand load of electric vehicles in the future, and can optimize the scheduling and operation strategies of the energy storage system by scheduling energy storage resources in advance, thereby improving its operating efficiency and reliability, and reducing operating costs.

[0187] 3. Implementation of the MPC controller and genetic algorithm two-layer optimization architecture

[0188] In the solution of the present invention, the MPC controller generates the optimal charge / discharge control strategy for the energy storage system based on the prediction results of the time series relationship generated by the ARIMA and LSTM models, and can be adjusted according to real-time data to ensure the stable operation and optimized control of the system in a dynamic environment. The genetic algorithm optimizes the parameters of the MPC controller to avoid falling into the local optimum. Through the iterative evolution of the genetic algorithm, the control strategy is adaptively adjusted to improve the control effect. In the entire energy storage system, prediction and control are based on time series data, and time synchronization ensures the accuracy and real-time synchronization of data collection, analysis, control signal transmission and execution.

[0189] The implementation process is as follows:

[0190] The MPC controller adjusts the control strategy in real time through online rolling optimization to minimize operating costs and meet grid demand, while ensuring that the charge level soc of the energy storage system is within a safe operating range. In the scheme of the present invention, the dynamic behavior of the energy storage system can be described by a linear state space model:

[0191] The output equation is: k =Dx k +v k

[0192] Among them, x k is the system state vector at time k, representing the current state of the energy storage system, such as the charge level and charge / discharge rate of the energy storage battery; u k is the control input vector at time k, representing the current control action on the energy storage system, such as charging and discharging power; (Equivalent to above ) is the predicted electricity demand at time k+1 predicted by the ARIMA and LSTM models; A is the state transfer matrix, which describes the autoregressive relationship of the system state (such as the relationship between the battery power and the charge and discharge rate); B is the control matrix, which describes the impact of the control input on the system state (such as the impact of the control input on the power and rate); C is the input matrix, which describes the impact of the predicted charge and discharge demand on the system state (A, B, C can be determined by regression analysis of experimental data or historical data); y k is the system output vector at time k, including the monitored battery status, etc., which is used for feedback control and state observation in the MPC control process to help adjust the control input of the system; w k is the process noise, which represents the uncertainty of the system and can be assumed to be zero-mean Gaussian white noise; D is the output matrix, which describes the influence of the system state on the output; v k is the measurement noise, which represents the uncertainty in the measurement process and can be assumed to be zero-mean Gaussian white noise.

[0193] Next, the goal of MPC is to optimize the control input u in the time domain predicted by the ARIMA and LSTM models. k , so that the system state x k Tracking the desired state x ref (represents the desired system state, such as target power and charge / discharge rate). The prediction time domain length is N, and the control model of the predicted state and control input in the next N moments can be expressed as:

[0194]

[0195] in is the sequence of predicted state vectors from time k+1 to k+N;

[0196] (i=1, 2, 3…N) is the predicted electricity demand at time k+i, which is predicted by the ARIMA and LSTM models. The model recursively calculates the future state, where each predicted state x k+i All based on the current state x k+i-1 , control input u k+i-1 and forecast demand Calculated.

[0197] Next, the goal of the MPC controller is to minimize operating costs and meet grid demand. The objective function (in quadratic form) can be expressed as:

[0198]

[0199] The vector set of all control inputs in the prediction time domain is expressed as: x i -x refrepresents the state deviation at time i, that is, the gap between the current state and the expected state; (x i -x ref ) T Q(x i -x ref ) represents the weighted quadratic norm of the state deviation, that is, the gap between the current state and the expected state; Q represents the state deviation weighting matrix, usually a diagonal matrix, used to weight the deviation of each state; u i represents the control input vector at time i; u i T Ru i represents the weighted quadratic norm of the control input, that is, the size of the control input; R represents the control input weighting matrix, usually a diagonal matrix, used to weight the size of each control input (Q, R are determined through empirical data or tuning process). At the same time, in order to ensure the safety and stability of the system, the control input u i and state x i Certain constraints need to be met (in linear form):

[0200]

[0201] Next, in the current step, this optimization framework uses genetic algorithms to optimize the weighted matrices Q and R of MPC in order to achieve optimal control in the prediction domain. The genetic algorithm needs to go through the following specific steps: initializing the population, fitness evaluation, selection operation, crossover operation, mutation operation, generating a new population, until convergence.

[0202] 1. Initialize the population: Generate several individuals based on empirical data, each of which represents a set of MPC parameters {Q, R}.

[0203] 2. Fitness function: Calculate the fitness value of each individual, that is, the control performance under given MPC parameters. The fitness function is defined as f(X * )=-J, where the optimal control input sequence

[0204] 3. Selection: Selecting better individuals for reproduction according to fitness values. The scheme of the present invention adopts a roulette selection method, which makes individuals with high fitness have a higher probability of being selected.

[0205] First, according to the fitness value f(X * ) Calculate the total fitness Then calculate the probability of each individual being selected Finally, the selection probability of each individual is mapped to the roulette wheel, and a new breeding population is formed by generating random numbers between [0,1) multiple times and selecting individuals according to the position of the random numbers on the roulette wheel.

[0206] 4. Crossover: Perform crossover operations on some genes of the selected individuals to generate new individuals in order to increase the diversity of the population and combine the excellent characteristics of the parent individuals.

[0207] For example: Q new =αQ1+(1-α)Q2,R new =βR1+(1-β)R2, where α and β are cross coefficients.

[0208] 5. Mutation: Perform small-scale random mutations on the new individuals generated by crossover to increase diversity and prevent the algorithm from falling into local optimality. For example: Q mut =Q new +μQ,R mut =R new +μR, where μQ,μR are random disturbance momentum.

[0209] 6. Iteration: Repeat the steps of item selection, crossover and mutation several times to gradually optimize the population until the fitness value no longer changes significantly. In the actual application scenario of the energy storage system, the calculation time and optimization effect can be balanced according to the specific requirements of the energy storage system. If necessary, a smaller maximum number of iterations can be set to achieve real-time optimization of the architecture.

[0210] After the iteration, select the individual with the highest fitness {Q * ,R *} as the optimal solution, and then combine the objective function and the constraints, and solve the MPC optimization problem using the quadratic programming (QP) solver to obtain the optimal control input sequence in the prediction time domain Then the optimal control input sequence is applied to the energy storage system according to the PPS time base to update the system state The system state deviation and control input cost are minimized in the prediction time domain. Finally, the MPC controller continues to iterate and update, collecting new state data x k+1 , and based on new forecast demand Reformulate the QP problem.

[0211] 4. Implementation of control strategy:

[0212] Through the above process, based on the unified time benchmark of PPS, the optimized control strategy operation is efficiently and accurately applied to the energy storage system, realizing real-time adjustment of the system operation status and monitoring of the system operation effect. In addition, the prediction model and control strategy are periodically adjusted according to the feedback control strategy and market changes, thus satisfying the balance of interests between car owners, energy storage operators and power grids.

[0213] In summary, in V2G application scenarios, the energy storage system acts as a buffer between EV and renewable energy to the power grid. How to ensure a balance of interests between car owners, energy storage operators and the power grid, and take into account the randomness of car owners' charging behavior, is a practical problem currently facing the energy storage system.

[0214] The technical solution of the present invention applies time-synchronized second pulse technology to the energy storage system to ensure high synchronization inside and outside the system, improve the overall operating efficiency, and integrate the ARIMA and LSTM prediction models with a unified time base, which can not only capture the linear trend of the data, but also process complex nonlinear characteristics, thereby improving the accuracy of load prediction.

[0215] Then, by combining MPC and genetic algorithms based on the prediction results, the system's timely optimization control was achieved, the effectiveness and stability of the control strategy were improved, and it was ensured that the control signal could be delivered and executed at the right time, thereby achieving higher prediction accuracy and adaptive control of the energy storage system.

[0216] At the same time, the pulse-per-second technology ensures that all devices operate within the same time frame in complex application scenarios with a large number of devices, enabling coordination and synchronization of the overall power system.

[0217] Compared with existing technologies, this solution significantly improves the operating efficiency and stability of the energy storage system. By optimizing the control strategy, it reduces the over-charging and discharging behavior of the energy storage system battery, extends the battery life, and optimizes the energy management strategy, reduces energy loss and operating costs, and improves economic benefits, thereby achieving better performance in a dynamically changing market environment and participant behavior, providing more flexible support for the power grid.

[0218] The present invention is described in detail above in conjunction with the embodiments of the accompanying drawings. A person skilled in the art can make various variations of the present invention according to the above description. Therefore, certain details in the embodiments should not constitute a limitation of the present invention, and the scope of protection of the present invention shall be defined by the scope of the attached claims.

Claims

1. An energy storage system for V2G application scenarios, characterized by: The energy storage system is based on time-synchronized second pulses, a combined prediction model that integrates two time series prediction algorithms, ARIMA and LSTM, and is equipped with a two-layer optimization architecture of MPC controller and genetic algorithm, thus forming a multi-layer composite optimization control strategy. It is used to enable EV, energy storage system and power grid to use energy buffering according to the real-time data of the power grid to smooth supply and demand fluctuations and time and space mismatches when car owners have random charging behavior.

2. The energy storage system for V2G application scenario according to claim 1, characterized in that: The time-synchronized second pulse uses a high-precision timing reference to ensure that the data collection and control operations of each node of the energy storage system are performed under the same time reference, ensuring the timing consistency of data collection, processing and control decisions of the entire system. The system hardware of the time synchronization second pulse includes: BD / GPS dual-mode receiver, high-precision clock synchronization module and taming clock submodule, The BD / GPS dual-mode receiver receives the standard time signal transmitted by the satellite, and generates a 1PPS signal through the high-precision clock synchronization module. The BD / GPS dual-mode receiver and the high-precision clock synchronization module are connected and deployed in the energy storage system. The time synchronization interface of the high-precision clock synchronization module outputs the 1PPS signal to the corresponding interface of the energy control mainboard of the energy storage system to ensure that all control modules of the energy storage system maintain clock synchronization under a unified and high-precision time reference. The clock taming submodule is deployed in the V2G bidirectional charging pile and connected to its control module, and is connected to the time synchronization interface of the high-precision clock synchronization module of the energy storage system. It is calibrated and tamed by receiving the 1PPS signal provided by it to achieve clock synchronization consistent with the energy storage system.

3. An energy storage system for V2G application scenarios according to claim 1 or 2, characterized in that: ARIMA forecasting model is used to analyze and forecast linear time series data in time series, and to identify and capture trends and seasonal changes in data through historical charging data of electric vehicles; The LSTM prediction model is used to learn and predict time series data with nonlinearities and long-term dependencies in time series. It can effectively capture the randomness of car owners' charging behavior, including changes with time, temperature, and weekdays / holidays. The combined prediction model of the two prediction algorithms can provide more accurate charging load prediction for the energy storage system, making the energy storage system more intelligent when facing random and volatile demand changes at different time scales, enabling the system operation to be synchronized with grid load fluctuations, accurately optimizing charging and discharging strategies, using the ARIMA model to handle shorter-term linear dependencies, and using the LSTM model to capture nonlinear patterns such as long-term dependencies and complex patterns. The role of the time-synchronized second pulse in this combined prediction model is to ensure that the time scale of all time series data is synchronized with the location when the actual event occurs, thereby improving the accuracy and reliability of the prediction.

4. The energy storage system for V2G application scenario according to claim 3, characterized in that: In the MPC controller and genetic algorithm two-level optimization architecture, wherein: The MPC controller generates the optimal charging / discharging strategy in real time based on the results provided by the combined prediction model of ARIMA and LSTM to minimize operating costs and meet grid demand; The genetic algorithm is used to optimize the parameters of the MPC controller to ensure that the best control strategy is found in the V2G environment. That is, by using the control strategy generated by the MPC controller as the first generation population of the genetic algorithm, randomly generating an initial parameter set, and calculating the appropriate fitness function for each individual, the economic benefits of the car owner, energy storage operator, and power grid are evaluated, and individuals with high fitness are selected, crossed over, and mutated to generate a new generation of populations, that is, to meet the optimal control strategy for EV, energy storage system, and power grid.

5. The energy storage system for V2G application scenario according to claim 2, characterized in that: The energy storage system receives a time synchronization signal at the beginning of each full second, and compares and calibrates the signal with the clocks of each control module in the system to ensure that all control modules in the system operate on the same time standard. In each synchronization cycle, the high-precision clock synchronization module 1PPS updates the clock of each tamed clock submodule to perform time calibration and deviation correction. The specific algorithm is: ΔT i =T BD / GPS -T local,i Where: T BD / GPS Represents the BD / GPS standard time received by the high-precision clock synchronization module, i.e. 1PPS, T local,i is the local clock time of the ith subsystem, ΔT i is the time deviation of the measurement, is the corrected local time.

6. An optimization management and control strategy for the energy storage system according to claim 1, comprising the following steps: Step 1, receiving BD / GPS time synchronization signal and performing second pulse time synchronization; Step 2, collecting charging and discharging time series data; Step 3: Combined prediction model training, using the ARIMA model to process the short-term linear relationship of the charging behavior data, and then using these prediction results as input combined with other features, including temperature and date type, using the LSTM model to capture and predict long-term, nonlinear patterns; Step 4: Real-time load forecasting, which gradually reduces the forecast error by iteratively updating weights and biases to achieve more accurate forecasts of future demand; Step 5: MPC control strategy generation. The MPC controller generates the optimal charge / discharge control strategy for the energy storage system based on the prediction results of the time series relationship generated by the ARIMA and LSTM models. It can be adjusted according to real-time data to ensure stable operation and optimized control of the system in a dynamic environment. Step 6: Genetic algorithm optimization: The genetic algorithm optimizes the parameters of the MPC controller to avoid falling into the local optimum. Through the iterative evolution of the genetic algorithm, the control strategy is adaptively adjusted to improve the control effect. Step 7: Generate a charge / discharge strategy and execute the charge / discharge strategy.

7. According to the optimization management and control strategy of the energy storage system of claim 6, the ARIMA model in step 3 is specifically as follows: Use the ARIMA model for short-term linear forecasting: The ARIMA model consists of three parameters p, d, and q: p is the autoregressive order; d is the difference order; q is the moving average order. The ARIMA model parameters d, p and q are determined by stationarity test, calculation and observation of autocorrelation function and partial autocorrelation function graphs. d is determined by the stationarity test and ADF test, that is, the stationarity of the time series is determined using the ADF test; ADF test formula: Δy t =α+βt+γy t-1 +δ1Δy t-1 +δ2Δy t-2 +…+d k Δy t-k +e t Among them, Δy t is the difference term, α is a constant, βt is the time trend term, γy t-1 is the lag term, δ i Δy t-i is the lagged difference term, ε t is the random error term, The autoregressive order p and the moving average order q are determined by observing the autocorrelation function ACF and the partial autocorrelation function PACF. ACF Official: Where: k is the autocorrelation coefficient at lag k; y t is the value of the time series at time t; y t-k is the value of the time series at time tk, which means the value at time t pushed forward k time points. is the mean of the time series; T is the total length of the time series; PACF Official: Where: φ kk is the partial autocorrelation coefficient at lag k, which indicates the autocorrelation of the sequence excluding the intermediate lag term at lag k; ρ k is the autocorrelation coefficient at lag k, which indicates the linear correlation between the value of the sequence at the current moment and the value at lag k moments; φ k-1,i is the partial autocorrelation number at lag k-1, which is the intermediate term to be considered when calculating the partial autocorrelation coefficient at the current lag k; ρ k-i is the autocorrelation coefficient at lag ki, which is used to adjust the value of the partial autocorrelation coefficient at the current lag k. ARIMA model construction and prediction: The estimated parameter values ​​are input into the ARIMA model to establish the ARIMA (p, d, q) model. Through the combined effects of difference, regression and moving average, short-term linear prediction of time series data can be performed, which can be used to predict charging demand.

8. According to the optimization management and control strategy of an energy storage system according to claim 6 or 7, the LSTM model in step 3 is specifically as follows: Use LSTM for long-term nonlinear prediction: The LSTM model controls the flow of information through three "gate" structures, namely, forget gate, input gate, and output gate, thereby effectively preserving long-term dependencies; First, the forget gate f t , which is used to decide what information to discard from the cell state, f t =σ(W f ·[h t-1 ,X t ]+b f ) in: W f is the weight matrix of the forget gate; [h t-1 ,X t ] is to connect the hidden state of the previous time point and the current input; b f is the bias of the forget gate; σ is the sigmoid activation function, which outputs a scalar from 0 to 1, indicating the retained ratio; Second, input gate i t and candidate cell states i t =σ(W i ·[h t-1 ,X t ]+b i ), Function: decide what new information to store in the cell state; Where W i ,W c is the weight matrix of the input gate; b i ,b c is the bias term of the input gate; tanh is the hyperbolic tangent activation function, which is used to generate candidate cell states, and the output range is between -1 and 1; Third, the cell state updates Y t , used to update the cell state, combining forgotten and newly input information, where Y t is the cell state at the current time point; Y t-1 is the cell state at the previous time point; Fourth, output gate o t and the hidden state h t : o t =σ(W o ·[h t-1 ,X t ]+b o ), which is used to decide what information to output as the output at this time point, h t =o t *tanh(Y t ), that is, the actual demand forecast is finally output through the conversion function Where: W o is the weight matrix of the output gate; b o is the bias term of the output gate.

9. According to the optimization management and control strategy of an energy storage system according to claim 8, the LSTM model calculation steps in step 3 are as follows: Step A, initial steps: The characteristics collected related to charging needs include: Temperature, time and holidays, create feature set X t , determine the hidden layer, which contains several LSTM units, each of which captures the temporal dependencies in the data. Step B, data processing: The forecast result of ARIMA model C t+1 Combine with other features X t , including the data from time tn to time t, to construct the training set of LSTM model [X t-n ,…,X t ,C t+1 ]; Step C, LSTM model training: The training of the LSTM model is an iterative process that uses historical time series data to train the LSTM network through the steps of forward propagation, loss function, back propagation, and parameter update to determine the weight matrix and bias value; Among them, the goal of the loss function L is to minimize the error between the predicted output and the actual demand, using the mean square error MSE: Where: Z t+1 is the actual charging demand value, It is a charging demand prediction value output by LSTM based on the training set using the current weights and biases at each step of the iteration. N is the number of samples. Backpropagation is used to calculate the gradient of the loss function with respect to each model parameter (weight and bias). The gradient represents the rate of increase of the loss function near the current value of the parameter and is used to update the weights to reduce the error. The parameter update is calculated using the gradient descent method: Among them, α is the learning rate, L is the loss function, and are the gradients of the weight matrix and bias, W old and b old They are the unified expressions of the weight matrices and bias values ​​of the three "gates", and They are the updated weight matrix and bias values, which are updated and adjusted according to the gradient of the loss function in each training iteration, so that the model can better learn and capture the underlying patterns of the data; Step D, prediction: Input the training set [X t-n ,…,X t ,C t+1 ] to the LSTM formula after updating the weight matrix and bias value and calculating the final prediction at time t+1 through the conversion function 10. According to the optimization management and control strategy of an energy storage system according to claim 8, the MPC controller in step 5 performs control through the following process: First, the dynamic behavior of the energy storage system can be described by a linear state space model: The output equation is: k =Dx k +v k in, x k is the system state vector at time k; u k is the control input vector at time k; is the predicted electricity demand at time k+1 predicted by the ARIMA and LSTM models; A is the state transition matrix; B is the control matrix; C is the input matrix; y k is the system output vector at time k; w k is the process noise; D is the output matrix; v k is the measurement noise; Second, the goal of MPC is to optimize the control input u in the time domain predicted by the ARIMA and LSTM models. k , so that the system state x k Tracking the desired state x ref ; The prediction time domain length is N, then the control model of the predicted state and control input in the next N moments can be expressed as: in is the sequence of predicted state vectors from time k+1 to k+N; i = 1, 2, 3...N, is the predicted electricity demand at time k+i, predicted by the ARIMA and LSTM models; the model recursively calculates the future state, where each predicted state x k+i All based on the current state x k+i-1 , control input u k+i-1 and forecast demand Calculated, Third, the goal of the MPC controller is to minimize the operating cost and meet the grid demand. The objective function is expressed as: Among them, the vector set of all control inputs in the prediction domain is expressed as: x i -x ref represents the state deviation at time i, that is, the gap between the current state and the expected state; (x i -x ref ) T Q(x i -x ref ) represents the weighted quadratic norm of the state deviation, that is, the gap between the current state and the expected state; Q represents the state deviation weighted matrix; u i represents the control input vector at time i; represents the weighted quadratic norm of the control input, that is, the size of the control input; R represents the control input weighting matrix; the control input u i and state x i Certain constraints need to be met:

Images