Power distribution network optimization operation method considering electrochemical energy storage operation performance degradation

Abstract

The invention relates to a power distribution network optimization operation method considering electrochemical energy storage operation performance degradation. The method comprises the following steps: forming a degradation sample data set; preprocessing the degraded sample data set; constructing a battery degradation model based on a neural network; the battery degradation amount of the whole scheduling period is obtained; and constructing a power distribution network dispatching model and solving to obtain a dispatching scheme. According to the method, the model training process is more stable, short-term fluctuation is effectively restrained, the accumulated degradation amount is not changed on the long-term statistical scale such as 24 hours, and unification of high precision and robustness is achieved; a dynamic and segmented-multiplying-power moving track is converted into a fixed-cycle-form input, and deviation caused by inconsistent distribution of training data and actual operating data is effectively eliminated; a degradation result is embedded into a target function f of a power distribution network dispatching model in a cost form, organic coupling of degradation evaluation and economical efficiency optimization is achieved, and collaborative optimization of the operation cost and the service life utilization rate is achieved.

Description

Technical Field

[0001] This invention relates to the field of power system energy storage life technology, and in particular to a distribution network optimization operation method that takes into account the performance degradation of electrochemical energy storage. Background Technology

[0002] As the penetration rate of renewable energy in microgrids continues to rise, the randomness and intermittency of power output are significantly enhanced, and the system's dependence on battery energy storage systems (BESS) is rapidly increasing. However, lithium-ion batteries inevitably degrade under the coupled effects of multiple factors such as different temperatures, rates, SOC (state of charge), DOD (depth of discharge), and SOH (state of health), and the degradation mechanism is complex and difficult to predict directly. In this context, ignoring or roughly handling battery degradation in energy management and day-ahead dispatch will lead to distorted cost assessments and an imbalance in lifetime utilization, making it difficult to balance economic efficiency and sustainability.

[0003] Existing degradation handling methods commonly used in the dispatching field have three key shortcomings: First, linear or heuristic models based solely on SOC / DOD ignore key factors such as charge / discharge rate and ambient temperature, making degradation estimation prone to systematic biases; second, many lifetime prediction methods rely on experimental settings of "fixed charge / discharge cycles," making it difficult to reflect the real-world usage scenarios where "power-SOC-rate" changes dynamically over time in microgrids; and third, many models are difficult to efficiently embed into the day-ahead dispatching (MDS) optimization framework, resulting in high computational overhead and poor engineering usability. Summary of the Invention

[0004] To address the issues of insufficient consideration of battery degradation in existing day-ahead dispatching of distribution networks and the nonlinear difficulties arising from the embedding of neural network degradation models, the present invention aims to provide a distribution network optimization operation method that considers the performance degradation of electrochemical energy storage, achieving an organic coupling between degradation assessment and economic optimization, and synergistically optimizing operating costs and lifetime utilization.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a distribution network optimization operation method considering the performance degradation of electrochemical energy storage, the method comprising the following sequential steps:

[0006] (1) Obtain cyclic aging data of the battery energy storage system under different operating conditions and form a degradation sample dataset, wherein the cyclic aging data includes ambient temperature, charge / discharge rate, state of charge, depth of discharge and health status data;

[0007] (2) Preprocess the degraded sample dataset to obtain the preprocessed dataset;

[0008] (3) Construct a battery degradation model based on neural networks;

[0009] (4) Input the preprocessed dataset into the battery degradation model based on neural networks to obtain the battery degradation amount for the entire scheduling cycle;

[0010] (5) By using a cycle-based aggregation method, the continuous time periods of the battery energy storage system under the same working state are merged into a single equivalent cycle, and the average operating characteristics of the single equivalent cycle are used as the input of the battery degradation model based on the neural network.

[0011] (6) Based on the amount of battery degradation throughout the entire scheduling cycle, construct and solve the distribution network scheduling model to obtain the scheduling scheme.

[0012] Step (1) specifically refers to: conducting cyclic charging and discharging tests on the battery energy storage system under various combinations of ambient temperature, charge / discharge rate, state of charge, and depth of discharge until the health status drops to a preset lifespan threshold, thereby obtaining a degradation sample dataset covering multiple operating conditions. The preset lifespan threshold is 80% of the battery's maximum rated capacity. In each cycle, ambient temperature, charge / discharge rate, state of charge, depth of discharge, and health status data are collected as input features. The difference in health status between two adjacent cycles is defined as the degradation amount in a single cycle, which serves as the output label for supervised learning.

[0013] Step (2) specifically refers to: First, the degraded sample dataset is processed in three ways: The first way is to leave it unchanged to obtain the first sample; the second way is to use a smoothing method, which removes outliers by filtering or moving average, reduces short-term random fluctuations, and improves data continuity to obtain the second sample; the third way is to use linear regression on the second sample to extract the long-term trend of the degradation curve and eliminate residual noise, and then perform zero mean variance normalization to obtain the third sample.

[0014] ;

[0015] In the formula: express Expectations; express The variance; Represents the original data; This represents the normalized input value;

[0016] The first, second, and third samples constitute the preprocessed dataset, which is then divided into a training dataset and a validation dataset in a 4:1 ratio.

[0017] Step (3) specifically refers to: the neural network adopts a fully connected feedforward neural network, which includes an input layer, a first hidden layer, a second hidden layer and an output layer; wherein, the input layer contains five neurons, which correspond to five key degradation features, namely temperature, charge / discharge rate, state of charge, depth of discharge and health status; the first hidden layer has twenty neurons, which are used to extract complex feature combinations; the second hidden layer has ten neurons, which are used to further compress and fuse degradation features; the output layer has one neuron, which is used to output the degradation rate of a single cycle; both the first hidden layer and the second hidden layer use the ReLU activation function to enhance the nonlinear mapping ability, and the output layer uses a linear activation function to maintain the continuity of degradation prediction;

[0018] The neural network-based battery degradation model employs mini-batch gradient descent for parameter optimization during training, determining the optimal training configuration through multiple batch size trials. The mean squared error (MSE) is used as the loss function to measure the deviation between predicted and actual degradation values.

[0019] ;

[0020] In the formula: This represents the true degradation value of the sample; is the neural network prediction value; n is the total number of samples.

[0021] Step (4) specifically refers to: the data in the preprocessed dataset forming the input vector, inputting the input vector into the neural network-based battery degradation model, and outputting the time step. The degradation results are then summed for all time steps to obtain the total degradation amount for the entire scheduling cycle.

[0022] ;

[0023] ;

[0024] ;

[0025] ;

[0026] In the formula: Indicates the battery energy storage system at time step The depth of discharge refers to the magnitude of the change in the state of charge between the current time step and the previous time step; Indicates the battery energy storage system at time step The state of charge of the battery; Indicates the battery energy storage system at time step The charge / discharge rate; Indicates the length of the scheduling interval; This represents the input feature vector of a neural network-based battery degradation model; This indicates the temperature of the battery energy storage system; This indicates the cumulative degradation of the battery energy storage system throughout the entire dispatch cycle; This represents a battery degradation model based on a neural network. Indicates the health status of the battery energy storage system; This indicates the setting of the time interval.

[0027] Step (5) specifically refers to: using a cycle-based aggregation method, merging consecutive time periods of the battery energy storage system under the same operating state, i.e., charging or discharging, into a single equivalent cycle, and using the average operating characteristics of the single equivalent cycle as the input to the neural network-based battery degradation model; specifically, for any consecutive time interval, if the operating state, i.e., charging or discharging, does not change, then the consecutive time intervals are aggregated into a single charging or discharging cycle, i.e., the aggregation cycle; for the aggregation cycle, the charging power or discharging power will be the average value of the aggregation time period;

[0028] The total degradation of the battery energy storage system is:

[0029] ;

[0030] In the formula: This represents the input feature vector of the aggregation cycle in the neural network-based battery degradation model; AC is the set of all aggregation cycles. This indicates the cumulative degradation of the battery energy storage system throughout the entire dispatch cycle; This represents a battery degradation model based on a neural network. This indicates the health status of the battery energy storage system.

[0031] In step (6), the objective function f of the distribution network scheduling model is:

[0032] ;

[0033] ;

[0034] ;

[0035] ;

[0036] In the formula: This represents the operating cost of a conventional power distribution network; This represents the degradation cost of the battery energy storage system. This indicates the initial investment cost of the battery energy storage system; This represents the residual value of a battery energy storage system at the end of its lifespan. This represents the health status threshold of a battery energy storage system at the end of its lifespan; It is a collection of all time periods of a day; The set of all nodes; It refers to the collection of all transmission lines in the distribution network; A collection of typical scenarios; The total number of days in a year; Let be the probability distribution of scenario s; Let be the resistance of line ij; This indicates the cost per unit of load failure caused by exceeding limits; Indicates network loss cost; This represents the voltage over-limit penalty coefficient of node i at time t in scenario s; The electricity purchase price of the distribution network at time t; The unit cost of electricity generated by photovoltaic power; The penalty cost for forfeiting optical power; The cost of charging and discharging the battery energy storage system; This is a Boolean variable representing the deployment status of energy storage devices; Let be a continuous variable, representing the curtailed power of photovoltaic power at node i at time t in scenario s; Let be the square of the current flowing through line ij at time t in scenario s; This represents the square of the voltage at node i at time t in scenario s; This represents the active power of the electrical load at node i at time t in scenario s; Let t be the power purchased by the distribution network from the next higher level grid in scenario s; Let be the active power emitted by photovoltaic node i at time t in scenario s; Let be the charging power of the energy stored at node i in scenario s at time t; Let be the discharge power of the energy stored at node i in scenario s at time t; The square of the maximum node voltage; The square of the minimum node voltage; This represents the square of the voltage at node i at time t in scenario s.

[0037] In step (6), the constraints of the distribution network dispatch model include node power balance constraints, voltage constraints at both ends of the line, line transmission capacity constraints, AC distribution network operation safety constraints, voltage and current safety constraints, and energy storage constraints.

[0038] The node power balance constraints include node active power balance constraints and node reactive power balance constraints:

[0039] (1);

[0040] (2);

[0041] In the formula: Let be the reactance of line ij; , These represent the active power and reactive power flowing into the distribution network from the substation at node i at time t in scenario s; Let be the active power flowing through line ji at time t in scenario s; Let be the reactive power flowing through line ji at time t in scenario s; Let be the active power flowing through line ij at time t in scenario s; Let be the reactive power flowing through line ij at time t in scenario s; This represents the reactive power of the electrical load at node i at time t in scenario s; Let be the reactive power emitted by photovoltaic node i at time st in scenario; Let be the active power emitted by photovoltaic node i at time t in scenario s; Let be a continuous variable, representing the curtailed power of photovoltaic power at node i at time t in scenario s; It is a collection of all time periods of a day; The set of all nodes; This refers to the collection of all transmission lines in a battery energy storage system. A collection of typical scenarios; Let be the resistance of line ij; Let be the square of the current flowing through line ji at time t in scenario s; This represents the active power of the electrical load at node i at time t in scenario s; Let be the charging power of the energy stored at node i in scenario s at time t;

[0042] Equation (1) is the active power balance constraint of the node, and Equation (2) is the reactive power balance constraint of the node.

[0043] The voltage constraints at both ends of the line are:

[0044] (3);

[0045] After incorporating the power flow direction of the line, equation (3) becomes:

[0046] ;

[0047] ;

[0048] ;

[0049] ;

[0050] ;

[0051] In the formula: M is a very large positive number; This is a Boolean variable; a value of 1 indicates that the power flow is from node i to node j. This is a Boolean variable; a value of 1 indicates that the power flow is from node j to node i. A Boolean variable representing the state of the line; its value is 1 when the line is closed and 0 when the line is open. This represents the square of the voltage at node i at time t in scenario s; Represents the square of the voltage at node j at time t in scenario s;

[0052] The line transmission capacity constraint is:

[0053] ;

[0054] The line transmission capacity constraint is transformed into a convex constraint using second-order cone relaxation:

[0055] ;

[0056] The operational safety constraints of the AC distribution network are as follows:

[0057] ;

[0058] In the formula: This represents the maximum transmission capacity of line ij;

[0059] The voltage and current safety constraints are as follows:

[0060] ;

[0061] ;

[0062] In the formula: The maximum current flowing through line ij; The square of the maximum node voltage; The square of the minimum node voltage;

[0063] The energy storage constraint is:

[0064] ;

[0065] ;

[0066] ;

[0067] ;

[0068] ;

[0069] ;

[0070] ;

[0071] ;

[0072] In the formula: This is a Boolean variable, representing a value of 1 when the battery energy storage system of node i at time t in scenario s is in the construction state, and 0 otherwise. This is a Boolean variable, representing a value of 1 when the battery energy storage system of node i at time t in scenario s is in a charging state, and 0 otherwise. This is a Boolean variable, representing a value of 1 when the battery energy storage system of node i at time t in scenario s is in a discharging state, and 0 otherwise. and These are the charging and discharging efficiencies of node i, respectively; and These are the upper and lower limits of the energy storage capacity of node i, respectively; and These are the maximum charging and discharging power of the battery energy storage system at node i, respectively; Let be the battery level of node i in scenario s at time t; and These represent the energy storage capacity of node i at the beginning and end of the scheduling cycle, respectively. Let be the discharge power of the energy stored at node i in scenario s at time t.

[0073] As can be seen from the above technical solution, the beneficial effects of the present invention are as follows: First, the present invention makes the model training process more stable by using data preprocessing methods such as smoothing and regression, effectively suppressing short-term fluctuations, and not changing the cumulative degradation amount on long-term statistical scales such as 24 hours, thus achieving a balance between high precision and robustness; Second, the present invention adopts a cycle-based battery usage data processing method to convert the dynamic, segmented rate operation trajectory into a fixed cycle input, effectively eliminating the deviation caused by the inconsistency between the distribution of training data and actual operation data; Third, the present invention achieves the organic coupling of degradation assessment and economic optimization by embedding the degradation result in the form of cost into the objective function f of the distribution network scheduling model, thereby achieving the synergistic optimization of operating cost and life utilization rate. Attached Figure Description

[0074] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0075] like Figure 1As shown, a distribution network optimization operation method considering the performance degradation of electrochemical energy storage is presented. The method includes the following sequential steps:

[0076] (1) Obtain cyclic aging data of the battery energy storage system under different operating conditions and form a degradation sample dataset, wherein the cyclic aging data includes ambient temperature, charge / discharge rate, state of charge, depth of discharge and health status data;

[0077] (2) Preprocess the degraded sample dataset to obtain the preprocessed dataset;

[0078] (3) Construct a battery degradation model based on neural networks;

[0079] (4) Input the preprocessed dataset into the battery degradation model based on neural networks to obtain the battery degradation amount for the entire scheduling cycle;

[0080] (5) By using a cycle-based aggregation method, the continuous time periods of the battery energy storage system under the same working state are merged into a single equivalent cycle, and the average operating characteristics of the single equivalent cycle are used as the input of the battery degradation model based on the neural network.

[0081] (6) Based on the amount of battery degradation throughout the entire scheduling cycle, construct and solve the distribution network scheduling model to obtain the scheduling scheme.

[0082] Step (1) specifically refers to: conducting cyclic charging and discharging tests on the battery energy storage system under various combinations of ambient temperature, charge / discharge rate, state of charge, and depth of discharge until the health status drops to a preset lifespan threshold, thereby obtaining a degradation sample dataset covering multiple operating conditions. The preset lifespan threshold is 80% of the battery's maximum rated capacity. In each cycle, ambient temperature, charge / discharge rate, state of charge, depth of discharge, and health status data are collected as input features. The difference in health status between two adjacent cycles is defined as the degradation amount in a single cycle, which serves as the output label for supervised learning.

[0083] Step (2) specifically refers to: First, the degraded sample dataset is processed in three ways: The first way is to leave it unchanged to obtain the first sample; the second way is to use a smoothing method, which removes outliers by filtering or moving average, reduces short-term random fluctuations, and improves data continuity to obtain the second sample; the third way is to use linear regression on the second sample to extract the long-term trend of the degradation curve and eliminate residual noise, and then perform zero mean variance normalization to obtain the third sample.

[0084] ;

[0085] In the formula: express Expectations; express The variance; Represents the original data; This represents the normalized input value;

[0086] The first, second, and third samples constitute the preprocessed dataset, which is then divided into a training dataset and a validation dataset in a 4:1 ratio.

[0087] Step (2) is used to perform quality optimization processing on degraded data before model training in order to improve the training accuracy and stability of neural networks.

[0088] Step (3) specifically refers to: the neural network adopts a fully connected feedforward neural network, which includes an input layer, a first hidden layer, a second hidden layer and an output layer; wherein, the input layer contains five neurons, which correspond to five key degradation features, namely temperature, charge / discharge rate, state of charge, depth of discharge and health status; the first hidden layer has twenty neurons, which are used to extract complex feature combinations; the second hidden layer has ten neurons, which are used to further compress and fuse degradation features; the output layer has one neuron, which is used to output the degradation rate of a single cycle; both the first hidden layer and the second hidden layer use the ReLU activation function to enhance the nonlinear mapping ability, and the output layer uses a linear activation function to maintain the continuity of degradation prediction;

[0089] The neural network-based battery degradation model employs mini-batch gradient descent for parameter optimization during training, determining the optimal training configuration through multiple batch size trials. The mean squared error (MSE) is used as the loss function to measure the deviation between predicted and actual degradation values.

[0090] ;

[0091] In the formula: This represents the true degradation value of the sample; is the neural network prediction value; n is the total number of samples.

[0092] Step (4) specifically refers to: the data in the preprocessed dataset forming the input vector, inputting the input vector into the neural network-based battery degradation model, and outputting the time step. The degradation results are then summed for all time steps to obtain the total degradation amount for the entire scheduling cycle.

[0093] ;

[0094] ;

[0095] ;

[0096] ;

[0097] In the formula: Indicates the battery energy storage system at time step The depth of discharge refers to the magnitude of the change in the state of charge between the current time step and the previous time step; Indicates the battery energy storage system at time step The state of charge of the battery; Indicates the battery energy storage system at time step The charge / discharge rate; Indicates the length of the scheduling interval; This represents the input feature vector of a neural network-based battery degradation model; This indicates the temperature of the battery energy storage system; This indicates the cumulative degradation of the battery energy storage system throughout the entire dispatch cycle; This represents a battery degradation model based on a neural network. Indicates the health status of the battery energy storage system; This indicates the setting of the time interval.

[0098] Step (5) specifically refers to the following: In actual operation, the power change of the battery energy storage system often does not follow a fixed charge-discharge cycle. If the model is directly input according to the time step, it will lead to inconsistency with the training data of the fixed cycle. Therefore, by using a cycle-based aggregation method, the continuous time periods of the battery energy storage system under the same working state, i.e., charging or discharging, are merged into a single equivalent cycle, and the average operating characteristics of the single equivalent cycle are used as the input of the battery degradation model based on the neural network. Specifically, for any continuous time interval, if the operating state, i.e., charging or discharging, does not change, the continuous time intervals are aggregated into a single charging or discharging cycle, i.e., the aggregation cycle. For the aggregation cycle, the charging power or discharging power will be the average value of the aggregation time period.

[0099] The total degradation of the battery energy storage system is:

[0100] ;

[0101] In the formula: This represents the input feature vector of the aggregation cycle in the neural network-based battery degradation model; AC is the set of all aggregation cycles. This indicates the cumulative degradation of the battery energy storage system throughout the entire dispatch cycle; This represents a battery degradation model based on a neural network. This indicates the health status of the battery energy storage system.

[0102] In step (6), the objective function f of the distribution network scheduling model is:

[0103] ;

[0104] ;

[0105] ;

[0106] ;

[0107] In the formula: This represents the operating cost of a conventional power distribution network; This represents the degradation cost of the battery energy storage system. This indicates the initial investment cost of the battery energy storage system; This represents the residual value of a battery energy storage system at the end of its lifespan. This represents the health status threshold of a battery energy storage system at the end of its lifespan; It is a collection of all time periods of a day; The set of all nodes; It refers to the collection of all transmission lines in the distribution network; A collection of typical scenarios; The total number of days in a year; Let be the probability distribution of scenario s; Let be the resistance of line ij; This indicates the cost per unit of load failure caused by exceeding limits; Indicates network loss cost; This represents the voltage over-limit penalty coefficient of node i at time t in scenario s; The electricity purchase price of the distribution network at time t; The unit cost of electricity generated by photovoltaic power; The penalty cost for forfeiting optical power; The cost of charging and discharging the battery energy storage system; This is a Boolean variable representing the deployment status of energy storage devices; Let be a continuous variable, representing the curtailed power of photovoltaic power at node i at time t in scenario s; Let be the square of the current flowing through line ij at time t in scenario s; This represents the square of the voltage at node i at time t in scenario s; This represents the active power of the electrical load at node i at time t in scenario s; Let t be the power purchased by the distribution network from the next higher level grid in scenario s; Let be the active power emitted by photovoltaic node i at time t in scenario s; Let be the charging power of the energy stored at node i in scenario s at time t; Let be the discharge power of the energy stored at node i in scenario s at time t; The square of the maximum node voltage; The square of the minimum node voltage; This represents the square of the voltage at node i at time t in scenario s.

[0108] In step (6), the constraints of the distribution network dispatch model include node power balance constraints, voltage constraints at both ends of the line, line transmission capacity constraints, AC distribution network operation safety constraints, voltage and current safety constraints, and energy storage constraints.

[0109] The node power balance constraints include node active power balance constraints and node reactive power balance constraints:

[0110] (1);

[0111] (2);

[0112] In the formula: Let be the reactance of line ij; , These represent the active power and reactive power flowing into the distribution network from the substation at node i at time t in scenario s; Let be the active power flowing through line ji at time t in scenario s; Let be the reactive power flowing through line ji at time t in scenario s; Let be the active power flowing through line ij at time t in scenario s; Let be the reactive power flowing through line ij at time t in scenario s; This represents the reactive power of the electrical load at node i at time t in scenario s; Let be the reactive power emitted by photovoltaic node i at time st in scenario; Let be the active power emitted by photovoltaic node i at time t in scenario s; Let be a continuous variable, representing the curtailed power of photovoltaic power at node i at time t in scenario s; It is a collection of all time periods of a day; The set of all nodes; This refers to the collection of all transmission lines in a battery energy storage system. A collection of typical scenarios; Let be the resistance of line ij; Let be the square of the current flowing through line ji at time t in scenario s; This represents the active power of the electrical load at node i at time t in scenario s; Let be the charging power of the energy stored at node i in scenario s at time t;

[0113] Equation (1) is the active power balance constraint of the node, and Equation (2) is the reactive power balance constraint of the node.

[0114] The voltage constraints at both ends of the line are:

[0115] (3);

[0116] After incorporating the power flow direction of the line, equation (3) becomes:

[0117] ;

[0118] ;

[0119] ;

[0120] ;

[0121] ;

[0122] In the formula: M is a very large positive number; This is a Boolean variable; a value of 1 indicates that the power flow is from node i to node j. This is a Boolean variable; a value of 1 indicates that the power flow is from node j to node i. A Boolean variable representing the state of the line; its value is 1 when the line is closed and 0 when the line is open. This represents the square of the voltage at node i at time t in scenario s; Represents the square of the voltage at node j at time t in scenario s;

[0123] The line transmission capacity constraint is:

[0124] ;

[0125] The line transmission capacity constraint is transformed into a convex constraint using second-order cone relaxation:

[0126] ;

[0127] The operational safety constraints of the AC distribution network are as follows:

[0128] ;

[0129] In the formula: This represents the maximum transmission capacity of line ij;

[0130] The voltage and current safety constraints are as follows:

[0131] ;

[0132] ;

[0133] In the formula: The maximum current flowing through line ij; The square of the maximum node voltage; The square of the minimum node voltage;

[0134] The energy storage constraint is:

[0135] ;

[0136] ;

[0137] ;

[0138] ;

[0139] ;

[0140] ;

[0141] ;

[0142] ;

[0143] In the formula: This is a Boolean variable, representing a value of 1 when the battery energy storage system of node i at time t in scenario s is in the construction state, and 0 otherwise. This is a Boolean variable, representing a value of 1 when the battery energy storage system of node i at time t in scenario s is in a charging state, and 0 otherwise. This is a Boolean variable, representing a value of 1 when the battery energy storage system of node i at time t in scenario s is in a discharging state, and 0 otherwise. and These are the charging and discharging efficiencies of node i, respectively; and These are the upper and lower limits of the energy storage capacity of node i, respectively; and These are the maximum charging and discharging power of the battery energy storage system at node i, respectively; Let be the battery level of node i in scenario s at time t; and These represent the energy storage capacity of node i at the beginning and end of the scheduling cycle, respectively. Let be the discharge power of the energy stored at node i in scenario s at time t.

[0144] In step (6), solving the distribution network dispatch model specifically includes the following steps:

[0145] (6a) Initialization parameters:

[0146] Before optimization calculation, the battery energy storage system parameters and grid operation data are imported. The grid operation data includes node voltage, current, active load, reactive load, line topology, time step, electricity purchase price, and curtailment penalty coefficient. The battery energy storage system parameters include rated capacity, upper limit of charging and discharging power, charging and discharging efficiency, lifetime threshold, initial health, investment cost, and residual value.

[0147] (6b) Microgrid dispatch and energy storage power calculation:

[0148] Without considering the effects of degradation, the day-ahead scheduling problem of the microgrid is first solved to obtain the charging and discharging power trajectory and energy state of the battery energy storage system in each time period. The charging and discharging power trajectory reflects the operating behavior of the battery energy storage system under optimal economic scheduling, providing input for degradation analysis.

[0149] (6c) Cyclic-based runtime data aggregation and degradation assessment:

[0150] The obtained energy storage operation data is input into a cycle-based aggregation method, which aggregates continuous time intervals with consistent operating states into equivalent cycles, and calculates the ambient temperature, charge / discharge rate, state of charge, depth of discharge, and health status of the battery energy storage system. Subsequently, the aggregated cycle data is input into a neural network-based battery degradation model to predict the degradation amount of each equivalent cycle, and calculates the corresponding equivalent degradation cost according to the formula.

[0151] (6d) Update the constraints of the distribution network dispatch model and iteratively optimize it;

[0152] (6e) Output the optimal scheduling result:

[0153] Once the iteration converges, the optimal scheduling scheme for the battery energy storage system is output, including the charging and discharging power, SOC change, cumulative degradation amount, and degradation cost for each time period.

[0154] In summary, this invention utilizes data preprocessing methods such as smoothing and regression to make the model training process more stable, effectively suppressing short-term fluctuations and maintaining the cumulative degradation amount on long-term statistical scales such as 24 hours, thus achieving a balance between high accuracy and robustness. This invention employs a cyclic-based battery usage data processing method, converting the dynamic, segmented operating trajectory into a fixed cyclic input, effectively eliminating the bias caused by the inconsistency between the distribution of training data and actual operating data. Furthermore, by embedding the degradation results as costs into the objective function f of the distribution network scheduling model, this invention achieves the organic coupling of degradation assessment and economic optimization, reaching a synergistic optimal balance between operating costs and lifetime utilization.

[0155] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for optimizing the operation of a distribution network considering the performance degradation of electrochemical energy storage, characterized in that: The method includes the following steps in sequence: (1) Obtain cyclic aging data of the battery energy storage system under different operating conditions and form a degradation sample dataset, wherein the cyclic aging data includes ambient temperature, charge / discharge rate, state of charge, depth of discharge and health status data; (2) Preprocess the degraded sample dataset to obtain the preprocessed dataset; (3) Construct a battery degradation model based on neural networks; (4) Input the preprocessed dataset into the battery degradation model based on neural networks to obtain the battery degradation amount for the entire scheduling cycle; (5) By using a cycle-based aggregation method, the continuous time periods of the battery energy storage system under the same working state are merged into a single equivalent cycle, and the average operating characteristics of the single equivalent cycle are used as the input of the battery degradation model based on the neural network. (6) Based on the amount of battery degradation throughout the entire scheduling cycle, construct and solve the power distribution network scheduling model to obtain the scheduling scheme.

2. The distribution network optimization operation method considering the performance degradation of electrochemical energy storage according to claim 1, characterized in that: Step (1) specifically refers to: conducting charge-discharge cycle tests on the battery energy storage system under various combinations of ambient temperature, charge-discharge rate, state of charge and depth of discharge until it drops to a preset lifespan threshold in a healthy state, and obtaining a degradation sample dataset covering multiple operating conditions, wherein the preset lifespan threshold is 80% of the battery's maximum rated capacity; In each cycle, ambient temperature, charge / discharge rate, state of charge, depth of discharge, and health status data are collected as input features. The difference in health status between two adjacent cycles is defined as the degradation amount in a single cycle, and is used as the output label for supervised learning.

3. The distribution network optimization operation method considering the performance degradation of electrochemical energy storage according to claim 1, characterized in that: Step (2) specifically refers to: First, three processing methods are used on the degraded sample dataset: The first method is to not modify it and obtain the first sample; the second method is to use the data smoothing method, remove outliers by filtering or moving average, reduce short-term random fluctuations, improve data continuity, and obtain the second sample. The third approach involves applying linear regression to the second sample to extract the long-term trend of the degradation curve and eliminate residual noise, followed by zero-mean-variance normalization to obtain the third sample. ； In the formula: express Expectations; express The variance; Represents the original data; This represents the normalized input value; The first, second, and third samples constitute the preprocessed dataset, which is then divided into a training dataset and a validation dataset in a 4:1 ratio.

4. The distribution network optimization operation method considering the performance degradation of electrochemical energy storage according to claim 1, characterized in that: Step (3) specifically refers to: the neural network adopts a fully connected feedforward neural network, which includes an input layer, a first hidden layer, a second hidden layer and an output layer; wherein, the input layer contains five neurons, which correspond to five key degradation features, namely temperature, charge / discharge rate, state of charge, depth of discharge and health status; the first hidden layer has twenty neurons, which are used to extract complex feature combinations; the second hidden layer has ten neurons, which are used to further compress and fuse degradation features; the output layer has one neuron, which is used to output the degradation rate of a single cycle; both the first hidden layer and the second hidden layer use the ReLU activation function to enhance the nonlinear mapping ability, and the output layer uses a linear activation function to maintain the continuity of degradation prediction; The neural network-based battery degradation model employs mini-batch gradient descent for parameter optimization during training, determining the optimal training configuration through multiple batch size trials. The mean squared error (MSE) is used as the loss function to measure the deviation between predicted and actual degradation values. ； In the formula: This represents the true degradation value of the sample. is the neural network prediction value; n is the total number of samples.

5. The distribution network optimization operation method considering the performance degradation of electrochemical energy storage according to claim 1, characterized in that: Step (4) specifically refers to: the data in the preprocessed dataset forming the input vector, inputting the input vector into the neural network-based battery degradation model, and outputting the time step. The degradation results are then summed for all time steps to obtain the total degradation amount for the entire scheduling cycle. ； ； ； ； In the formula: Indicates the battery energy storage system at time step The depth of discharge refers to the magnitude of the change in the state of charge between the current time step and the previous time step; Indicates the battery energy storage system at time step The state of charge of the battery; Indicates the battery energy storage system at time step The charge / discharge rate; Indicates the length of the scheduling interval; This represents the input feature vector of a neural network-based battery degradation model; This indicates the temperature of the battery energy storage system; This indicates the cumulative degradation of the battery energy storage system throughout the entire dispatch cycle; This represents a battery degradation model based on a neural network. Indicates the health status of the battery energy storage system; This indicates the setting of the time interval.

6. The distribution network optimization operation method considering the performance degradation of electrochemical energy storage according to claim 1, characterized in that: Step (5) specifically refers to: using a cycle-based aggregation method, merging consecutive time periods of the battery energy storage system under the same operating state, i.e., charging or discharging, into a single equivalent cycle, and using the average operating characteristics of the single equivalent cycle as the input to the neural network-based battery degradation model; specifically, for any consecutive time interval, if the operating state, i.e., charging or discharging, does not change, then the consecutive time intervals are aggregated into a single charging or discharging cycle, i.e., the aggregation cycle; for the aggregation cycle, the charging power or discharging power will be the average value of the aggregation time period; The total degradation of the battery energy storage system is: ； In the formula: This represents the input feature vector of the aggregation cycle in the neural network-based battery degradation model; AC is the set of all aggregation cycles. This indicates the cumulative degradation of the battery energy storage system throughout the entire dispatch cycle; This represents a battery degradation model based on a neural network. This indicates the health status of the battery energy storage system.

7. The distribution network optimization operation method considering the performance degradation of electrochemical energy storage according to claim 1, characterized in that: In step (6), the objective function f of the distribution network scheduling model is: ； ； ； ； In the formula: This represents the operating cost of a conventional power distribution network; This represents the degradation cost of the battery energy storage system. This indicates the initial investment cost of the battery energy storage system; This represents the residual value of a battery energy storage system at the end of its lifespan. This represents the health status threshold of a battery energy storage system at the end of its lifespan; It is a collection of all time periods of a day; The set of all nodes; It refers to the collection of all transmission lines in the distribution network; A collection of typical scenarios; The total number of days in a year; Let be the probability distribution of scenario s; Let be the resistance of line ij; This indicates the cost per unit of load failure caused by exceeding limits; Indicates network loss cost; This represents the voltage over-limit penalty coefficient of node i at time t in scenario s; The electricity purchase price of the distribution network at time t; The unit cost of electricity generated by photovoltaic power; The penalty cost for forfeiting optical power; The cost of charging and discharging the battery energy storage system; This is a Boolean variable representing the deployment status of energy storage devices; Let be a continuous variable, representing the curtailed power of photovoltaic power at node i at time t in scenario s; Let be the square of the current flowing through line ij at time t in scenario s; This represents the square of the voltage at node i at time t in scenario s; This represents the active power of the electrical load at node i at time t in scenario s; Let t be the power purchased by the distribution network from the next higher level grid in scenario s; Let be the active power emitted by photovoltaic node i at time t in scenario s; Let be the charging power of the energy stored at node i in scenario s at time t; Let be the discharge power of the energy stored at node i in scenario s at time t; The square of the maximum node voltage; The square of the minimum node voltage; This represents the square of the voltage at node i at time t in scenario s.

8. The distribution network optimization operation method considering the performance degradation of electrochemical energy storage according to claim 1, characterized in that: In step (6), the constraints of the distribution network dispatch model include node power balance constraints, voltage constraints at both ends of the line, line transmission capacity constraints, AC distribution network operation safety constraints, voltage and current safety constraints, and energy storage constraints. The node power balance constraints include node active power balance constraints and node reactive power balance constraints: （1）； （2）； In the formula: Let be the reactance of line ij; , These represent the active power and reactive power flowing into the distribution network from the substation at node i at time t in scenario s; Let be the active power flowing through line ji at time t in scenario s; Let be the reactive power flowing through line ji at time t in scenario s; Let be the active power flowing through line ij at time t in scenario s; Let be the reactive power flowing through line ij at time t in scenario s; This represents the reactive power of the electrical load at node i at time t in scenario s; Let be the reactive power emitted by photovoltaic node i at time st in scenario; Let be the active power emitted by photovoltaic node i at time t in scenario s; Let be a continuous variable, representing the curtailed power of photovoltaic power at node i at time t in scenario s; It is a collection of all time periods of a day; The set of all nodes; This refers to the collection of all transmission lines in a battery energy storage system. A collection of typical scenarios; Let be the resistance of line ij; Let be the square of the current flowing through line ji at time t in scenario s; This represents the active power of the electrical load at node i at time t in scenario s; Let be the charging power of the energy stored at node i in scenario s at time t; Equation (1) is the active power balance constraint of the node, and Equation (2) is the reactive power balance constraint of the node. The voltage constraints at both ends of the line are: （3）； After incorporating the power flow direction of the line, equation (3) becomes: ； ； ； ； ； In the formula: M is a very large positive number; This is a Boolean variable; a value of 1 indicates that the power flow is from node i to node j. This is a Boolean variable; a value of 1 indicates that the power flow is from node j to node i. A Boolean variable representing the state of the line; its value is 1 when the line is closed and 0 when the line is open. This represents the square of the voltage at node i at time t in scenario s; Let represent the square of the voltage at node j at time t in scenario s; The line transmission capacity constraint is: ； The line transmission capacity constraint is transformed into a convex constraint using second-order cone relaxation: ； The operational safety constraints of the AC distribution network are as follows: ； In the formula: This represents the maximum transmission capacity of line ij; The voltage and current safety constraints are as follows: ； ； In the formula: The maximum current flowing through line ij; The square of the maximum node voltage; The square of the minimum node voltage; The energy storage constraint is: ； ； ； ； ； ； ； ； In the formula: This is a Boolean variable, representing a value of 1 when the battery energy storage system of node i at time t in scenario s is in the construction state, and 0 otherwise. This is a Boolean variable, representing a value of 1 when the battery energy storage system of node i at time t in scenario s is in a charging state, and 0 otherwise. This is a Boolean variable, representing a value of 1 when the battery energy storage system of node i at time t in scenario s is in a discharging state, and 0 otherwise. and These are the charging and discharging efficiencies of node i, respectively; and These are the upper and lower limits of the energy storage capacity of node i, respectively; and These are the maximum charging and discharging power of the battery energy storage system at node i, respectively; Let be the battery level of node i in scenario s at time t; and These represent the energy storage capacity of node i at the beginning and end of the scheduling cycle, respectively. Let be the discharge power of the energy stored at node i in scenario s at time t.

Images