A Single-Scenario Optimization Method and System for Energy Storage Based on Multi-Objective Rule Fusion
By using a multi-objective rule fusion-based single-scenario optimization method for energy storage, combined with real-time demand response signals and historical data, the discharge strategy of the energy storage system is predicted and optimized. This solves the problems of substandard response, battery aging, and assessment risks in existing technologies, and achieves efficient operation and battery protection of the energy storage system.
Patent Information
- Application Number
- CN202611525273.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-29
- Publication Date
- 2026-11-10
AI Technical Summary
Existing energy storage real-time demand response control methods fail to effectively balance discharge capacity, battery life, and market performance during the response period, resulting in substandard response, accelerated battery aging, and decreased operational efficiency.
A single-scenario optimization method for energy storage based on multi-objective rule fusion is adopted. By acquiring real-time demand response signals, combining historical response events and current battery status, the response demand sequence is predicted and the target SOC lower limit is determined. Pre-adjustment and multi-objective function optimization are then performed to control the operation of energy storage to balance response efficiency, battery dynamic loss, and default penalty.
It achieves reliable compliance during the response period, protects battery life, reduces assessment risks, and improves the operational efficiency and adaptability of the energy storage system.
Smart Images

Figure CN123073897A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of smart grid technology, and in particular to a single-scenario optimization method and system for energy storage based on multi-objective rule fusion. Background Technology
[0002] With the increasing proportion of renewable energy generation, the pressure on power grid peak shaving and supply guarantee is growing, and demand response has become the main means for the power grid to utilize user-side flexible resources. Energy storage, due to its adjustable power and fast response, is widely used to undertake real-time demand response tasks issued by the power grid. The power grid dispatch channel issues instructions at a certain moment, specifying the start and end times of the response and the minimum discharge power to be achieved during the response period. Meanwhile, energy storage is an expensive asset, and its available capacity and lifespan directly depend on the battery state during charging and discharging. Therefore, when undertaking each response task, the energy storage aggregation and management terminal must ensure that the discharge capacity meets the grid requirements, avoid over-discharge which accelerates aging, and reserve sufficient power for subsequent response tasks.
[0003] Existing real-time demand response control for energy storage largely relies on preset, fixed rules. Upon receiving a real-time dispatch signal from the grid, this approach only checks if the energy storage's current State of Charge (SOC) has reached a preset pre-response threshold. If it does, it discharges at a minimum discharge power and prohibits charging. After the response ends, normal operation resumes. This approach fails to consider SOC margin, battery degradation, market performance evaluations, and subsequent demand. Consequently, insufficient SOC margin leads to the inability to maintain minimum discharge power, resulting in performance evaluations; maintaining a high SOC long-term to meet targets accelerates battery aging; and the power allocation after the response is unpredictable, often leading to a lack of power for subsequent tasks, thus reducing overall operational efficiency. Summary of the Invention
[0004] This invention provides a single-scenario optimization method and system for energy storage based on multi-objective rule fusion. It improves the existing response control in real-time demand response scenarios of energy storage, which only judges whether the SOC has reached the threshold before the response. If it does, it discharges at a power no less than the minimum discharge power and prohibits charging. After the event ends, it resumes normal operation. However, it does not take into account the SOC margin, battery loss, market assessment and subsequent response requirements, which leads to the difficulty in balancing response compliance and battery life and frequent default penalties.
[0005] The present invention discloses the following technical solution: In a first aspect, the present invention provides a single-scenario optimization method for energy storage based on multi-objective rule fusion, the method comprising: Acquire real-time demand response signals, analyze response time periods, minimum discharge power, and pre-response SOC threshold, obtain response demand sequences through a response demand prediction model based on historical response events and current battery status, and determine the target SOC lower limit after the response ends. Before the response period begins, the SOC is pre-adjusted based on the current SOC, the pre-response SOC threshold, and the target SOC lower limit, and the pre-adjusted SOC is used as the initial SOC for the response period. A multi-objective function is constructed based on the initial SOC of the response period, which integrates response performance, battery dynamic loss and default penalty. The minimum discharge power of the response period, charging prohibition, SOC threshold before response and target SOC lower limit are used as constraints. The nonlinear terms in the multi-objective function and constraints are linearized and solved to obtain the charging and discharging power at each time during the response period. The energy storage operation is controlled based on the solution results. After the response ends, the response demand prediction model is updated based on the actual response deviation and battery degradation, and the target SOC lower limit is corrected.
[0006] Secondly, this invention provides a single-scenario optimization system for energy storage based on multi-objective rule fusion, comprising: The response demand analysis module is used to acquire real-time demand response signals, analyze the response period, minimum discharge power and SOC threshold before response, obtain the response demand sequence through the response demand prediction model based on historical response events and current battery status, and determine the target SOC lower limit after the response ends. The SOC pre-adjustment module is used to pre-adjust the SOC based on the current SOC, the pre-response SOC threshold, and the target SOC lower limit before the start of the response period, and to use the pre-adjusted SOC as the initial SOC for the response period. The multi-objective solution module is used to construct a multi-objective function that integrates response performance, battery dynamic loss and default penalty based on the initial SOC of the response period. The module uses the minimum discharge power of the response period, charging prohibition, SOC threshold before response and target SOC lower limit as constraints to linearize the nonlinear terms in the multi-objective function and constraints and solve them to obtain the charging and discharging power at each time during the response period. The control correction module is used to control the operation of energy storage based on the solution results. After the response ends, it updates the response demand prediction model based on the actual response deviation and battery degradation, and corrects the target SOC lower limit.
[0007] One or more technical solutions provided in this invention have at least the following technical effects or advantages: The energy storage single-scenario optimization method based on multi-objective rule fusion provided by the present invention first obtains real-time demand response signals and analyzes the response period, minimum discharge power, and pre-response SOC threshold. Then, it combines historical response events and the current battery state to obtain the response demand sequence through a response demand prediction model and determine the target SOC lower limit. This enables the determination of the required discharge capacity for this response and the energy boundary after the response before the response arrives. It solves the problem in existing methods that only the pre-response SOC threshold is known, but not how much power needs to be discharged during the response or how much power needs to be retained after the response. This provides clear boundary conditions for subsequent pre-adjustment and multi-objective solution.
[0008] Furthermore, by pre-adjusting based on the current SOC and the target SOC lower limit before the start of the response period, the energy storage capacity is pre-adjusted to a level that balances the current target achievement and subsequent needs before the response arrives. This solves the problem that existing practices only determine the threshold once before the response and cannot take into account the power allocation after the response ends, thus providing a reasonable starting power capacity for power allocation during the response period.
[0009] Furthermore, by constructing a multi-objective function that integrates response performance, battery dynamic loss, and default penalty based on the initial SOC of the response period, and linearizing and solving it under constraints such as minimum discharge power and prohibition of charging, the system integrates the three requirements of response compliance, battery protection, and default risk into a single optimization problem. This solves the problem that existing practices only pursue whether discharge can occur within the response period and cannot achieve a balance among multiple operational objectives, providing directly executable values for the charging and discharging power at each moment within the response period.
[0010] Furthermore, by controlling the operation of energy storage based on the solution results, and updating the response demand prediction model and correcting the target SOC lower limit based on the actual response deviation and battery degradation after the response ends, the prediction model and the energy boundary are continuously evolved with the actual operating data. This solves the problem that existing practices are set once and then not adjusted, making it difficult to adapt to subsequent changes in response demand. It provides adaptive capability for energy storage to participate in real-time demand response for a long time.
[0011] Therefore, the single-scenario optimization method and system for energy storage based on multi-objective rule fusion provided by the embodiments of the present invention can achieve reliable compliance during the response period, protect battery life and reduce assessment risks by completing boundary determination and power pre-adjustment before the response arrives, coordinating response efficiency, battery loss and default risk with multi-objective functions during the response period, and making closed-loop correction based on actual deviation and attenuation after the response ends. Attached Figure Description
[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0013] Figure 1 This is a schematic diagram of the overall process of the single-scenario optimization method for energy storage based on multi-objective rule fusion provided in an embodiment of the present invention.
[0014] Figure 2 This is a schematic diagram of the pre-adjustment process of the single-scenario optimization method for energy storage based on multi-objective rule fusion provided in an embodiment of the present invention.
[0015] Figure 3 This is a schematic diagram of the structure of a single-scenario optimization system for energy storage based on multi-objective rule fusion, provided in an embodiment of the present invention.
[0016] The components represented by each number in the attached diagram are described as follows: Response Demand Analysis Module 11, SOC Pre-adjustment Module 12, Multi-Objective Solving Module 13, and Control Correction Module 14. Detailed Implementation
[0017] This invention provides a single-scenario optimization method and system for energy storage based on multi-objective rule fusion, which is used to solve the technical problem that in the real-time demand response scenario of energy storage, the interval between the issuance time of the grid dispatch signal and the response period is uncertain, the actual discharge capacity during the response period depends on the current SOC, health status and cycle number of the battery, and whether there will be subsequent response demand after the response ends is also uncertain in advance. This makes it difficult to simultaneously meet the minimum discharge power requirement, protect battery life and avoid market assessment during the response period.
[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. It should be noted that the numerical values in the embodiments are for illustrative purposes only and do not constitute a limitation on the present invention.
[0019] Example 1, as Figure 1 As shown, this invention provides a single-scenario optimization method for energy storage based on multi-objective rule fusion, the method comprising the following steps: S100: Acquires real-time demand response signals, analyzes the response period, minimum discharge power, and pre-response SOC threshold, obtains the response demand sequence based on historical response events and current battery state through a response demand prediction model, and determines the target SOC lower limit after the response ends. This invention solves the technical problem that when energy storage participates in the real-time demand response of the power grid, the response control is based only on preset fixed rules, which leads to substandard or defaulted responses, accelerated battery aging, and low operating efficiency.
[0020] Before the start of the response period, the energy storage aggregation and management terminal receives a real-time demand response signal from the power grid. It extracts the response period, minimum discharge power, and pre-response SOC threshold from the signal. Then, it combines historical response events with the battery's current SOC, health status, and cycle count to predict the power that needs to be released at each moment during the response period and after the response ends. Based on this, it calculates the lower limit of the amount of electricity that should be retained after the response ends, providing a basis for subsequent pre-adjustment and power calculation.
[0021] Step S100 in the method provided in this embodiment of the invention includes: The real-time demand response signal is obtained through the real-time dispatch channel of the power grid. The issuance time, response start time and response end time in the real-time demand response signal are analyzed. The time period between the response start time and the response end time is taken as the response period. The advance notification time is determined according to the interval between the issuance time and the response start time. The minimum discharge power required by the power grid dispatch instruction is extracted from the real-time demand response signal. The SOC threshold before response is extracted first from the real-time demand response signal. If the signal does not carry the SOC threshold before response, the lowest SOC before response for successful responses in the same response period in historical response events is corrected based on the SOC and health status of the current battery state to obtain the SOC threshold before response. Historical response events include response time period, response power, SOC before response and SOC after response. Current battery status includes SOC, health status and cycle count. Health status is determined based on the cycle count; the higher the cycle count, the lower the health status.
[0022] Specifically, the response period refers to the time interval between the start and end times of discharge required by the power grid dispatching command, which is determined by the response start time and response end time in the real-time demand response signal.
[0023] Minimum discharge power represents the lowest value that the energy storage discharge power must reach within the response period, and is given by the grid dispatch command.
[0024] The pre-response SOC threshold represents the minimum energy storage capacity required before the start of the response period. It is primarily determined by the real-time demand response signal, and if the signal does not carry this threshold, it is corrected based on historical response events.
[0025] The SOC before a successful response indicates the lowest battery level among all successful responses occurring within the same timeframe as the current response, prior to the start of the response.
[0026] Historical response events represent records of energy storage's past participation in demand response. Each record includes the response period, response power, pre-response SOC, and post-response SOC.
[0027] The current battery status indicates the operating state of the energy storage battery before the start of this response, including SOC, health status, and cycle count.
[0028] The health status represents the ratio of the battery's current usable capacity to its brand-new capacity. It is determined by the number of cycles; the higher the number of cycles, the lower the health status.
[0029] The cycle count represents the cumulative number of charge-discharge cycles the battery has completed since it was put into operation, and is counted by the battery management system.
[0030] In this step, the energy storage aggregation and management terminal first obtains real-time demand response signals through the real-time grid dispatch channel.
[0031] Furthermore, the issuance time, response start time, and response end time are extracted from the real-time demand response signal. The time period between the response start time and the response end time is taken as the response period, and the interval between the issuance time and the response start time is taken as the advance notification time. This time determines the available duration for subsequent pre-adjustment.
[0032] Furthermore, the minimum discharge power required by the power grid dispatching command is extracted from the real-time demand response signal, and it is determined whether the signal carries the pre-response SOC threshold. If the signal carries it, it is used directly; if the signal does not carry it, it is corrected according to the following formula: .
[0033] The first factor is to adjust the historical minimum threshold upwards based on the degree of battery degradation; the lower the battery health status, the larger the coefficient. The second factor is to add a margin based on the difference between the current battery level and the full charge; the lower the current SOC, the more margin is added. The threshold supplement coefficient is determined based on the historical response records of the site, and is 0.15 in this example.
[0034] For example, a certain energy storage power station is equipped with a lithium iron phosphate energy storage with a rated power of 1MW and a rated capacity of 2MWh. The current SOC is 0.55, the number of cycles is 600, and the rated cycle life is 6000. Based on this, its health status is determined to be 1-600÷6000=1-0.10=0.90.
[0035] The real-time demand response signal was issued at 17:00, the response started at 18:00, and the response ended at 20:00. Therefore, the response period is from 18:00 to 20:00, the advance notification time is 18:00-17:00=1h, and the minimum discharge power is 0.30MW.
[0036] The signal did not carry the pre-response SOC threshold, so it was corrected according to historical response events. There are a total of 12 historical response events for this site, of which 4 had a response period from 18:00 to 20:00. The lowest pre-response SOC for successful response in these 4 events was 0.45, while the lowest pre-response SOC for all 12 historical response events was 0.41.
[0037] The SOC threshold before response = 0.45 × (2 - 0.90) + (1 - 0.55) × 0.15 = 0.45 × 1.10 + 0.45 × 0.15 = 0.495 + 0.0675 = 0.5625.
[0038] Step S100 in the method provided in this embodiment of the invention further includes: Filter historical response events that are in the same period as the response time from the historical response events, and extract the corresponding response power, pre-response SOC and post-response SOC; The response power, SOC before response, and SOC after response are combined with the current SOC, health status, and cycle count in the battery state and input into the response demand prediction model. The output response demand sequence includes the response power at each time point during the response period and the response power at each time point after the response ends. The discharge capacity during the response period is calculated based on the response power at each moment and the duration of the response period. The discharge capacity after the response end time is calculated based on the response power at each moment and the corresponding duration. Based on the discharge capacity during the response period and the discharge capacity after the response ends, the SOC in the current battery state is reduced, and then adjusted according to the current battery health status and cycle count to obtain the target lower limit of SOC after the response ends.
[0039] More specifically, the response demand prediction model is a machine learning model that takes historical response power, pre-response SOC and post-response SOC, and current battery state as inputs, and outputs response power during the response period and after the response ends. It is trained from historical response events.
[0040] The response demand sequence represents the output of the response demand prediction model, including the response power at each time point within the response period and the response power at each time point after the response ends.
[0041] The discharge capacity represents the electrical energy accumulated according to the response power and corresponding duration in the response demand sequence.
[0042] The target SOC lower limit represents the minimum battery charge that must be retained after all responses have been executed. It is obtained by subtracting the predicted discharge charge from the current SOC and then correcting for it.
[0043] In this step, we first select historical response events that are in the same period as the current response from the historical response events, and extract the response power, SOC before response and SOC after response for these events.
[0044] Furthermore, the extracted response power, pre-response SOC and post-response SOC, along with the current battery SOC, health status and cycle count, are input into the response demand prediction model, and the model outputs a response demand sequence.
[0045] Furthermore, based on the response power and corresponding duration in the response demand sequence, the discharge capacity during the response period and after the response ends are accumulated separately, as calculated by the following formula: , .
[0046] Then, the discharge capacity of the two parts mentioned above is converted into a deduction from the State of Charge (SOC), and then corrected according to the health status to obtain the target lower limit of SOC after the response ends. The calculation formula is as follows: .
[0047] The number of cycles is taken into account along with the health status; the lower the health status, the larger the correction coefficient in the calculation formula, and the higher the target SOC lower limit, so as to retain more power for subsequent responses after the response ends.
[0048] For example, in the response demand sequence output by the response demand prediction model, the response power at eight times during the response period is 0.32, 0.32, 0.30, 0.30, 0.30, 0.30, 0.30 and 0.30MW respectively, and the response power at the two times after the response ends is 0.18MW.
[0049] The discharge capacity during the response period = (0.32 + 0.32 + 0.30 + 0.30 + 0.30 + 0.30 + 0.30) × 0.25 = 2.44 × 0.25 = 0.6100 MWh. The discharge capacity after the response ends = (0.18 + 0.18) × 0.25 = 0.36 × 0.25 = 0.0900 MWh. The sum of the two discharge capacities = 0.6100 + 0.0900 = 0.7000 MWh. The target SOC lower limit = 0.55 - (0.7000 ÷ (2 × 0.90)) × (2 - 0.90) = 0.55 - 0.3889 × 1.10 = 0.55 - 0.4278 = 0.1222.
[0050] Step S100 in the method provided in this embodiment of the invention further includes: Based on historical response events, collect multiple sets of sample response power, sample pre-response SOC, and sample post-response SOC; Collect different sets of sample response power, sets of SOC before sample response, and sets of SOC after sample response, corresponding sets of SOC, health status, and number of cycles in the sample battery state; The annotation yields a set of sample response demand sequences, which includes the response power at each time point within the response period and the response power at each time point after the response end time. Based on machine learning, a response demand prediction model is constructed and trained by taking multiple sets of sample response power, sample pre-response SOC, sample post-response SOC, and corresponding sets of SOC, health status and cycle count in the sample battery state as inputs and a set of sample response demand sequences as outputs. Configure the demand forecasting model in the energy storage aggregation and management terminal.
[0051] More specifically, the sample response power set represents the set of response power at each moment of the same response period in multiple historical response events, and the sample response pre-response SOC set and sample response post-response SOC set represent the sets of electricity generated before and after the response of these historical response events, respectively.
[0052] The sample response demand sequence set represents the set consisting of the response power at each moment within the labeled response period corresponding to each historical response event, as well as the response power at each moment after the response end time.
[0053] Energy storage aggregation and control terminal refers to the terminal equipment deployed on the energy storage site side, which is responsible for receiving grid dispatch instructions and controlling the charging and discharging of energy storage.
[0054] In this step, multiple sets of sample response power, sample pre-response SOC, and sample post-response SOC are first collected based on historical response events.
[0055] Furthermore, the SOC set, health state set, and cycle count set of the corresponding sample battery states under different sample response power sets, sample pre-response SOC set, and sample post-response SOC set are collected.
[0056] Furthermore, the annotation obtains a set of sample response demand sequences, which includes the response power at each time point within the response period and the response power at each time point after the response ends.
[0057] Then, based on machine learning, a response demand prediction model is trained by taking the response power, pre-response SOC, post-response SOC and corresponding SOC, health status and number of cycles of each sample as input and the set of sample response demand sequences as output.
[0058] Finally, the demand forecasting model is configured on the energy storage aggregation and management terminal so that it can be directly invoked when a subsequent response arrives.
[0059] For example, the training data is taken from 8,000 historical response events from multiple energy storage sites under the aggregated management and control platform. Each event includes response power, SOC before response, SOC after response, SOC in the sample battery state, health status and cycle count, as well as the labeled sample response demand sequence. The data is divided into two sets according to time: the 7,200 earlier events are used as the training set, and the 800 later events are used as the test set. The division by time is to avoid data from the same response falling into both sets at the same time.
[0060] The response demand prediction model used in this step is a multilayer perceptron. The input layer has a dimension of 13, corresponding to the response power, pre-response SOC, post-response SOC, current SOC, health status, and number of iterations at eight time points within the response period. The hidden layers consist of two layers: a 64-dimensional first layer and a 32-dimensional second layer, both using linear rectified functions. The output layer has a dimension of 10, corresponding to the response power at eight time points within the response period and at two time points after the response ends. The loss function is calculated using the following formula: .
[0061] Training was performed using an adaptive moment estimation optimizer with an initial learning rate of 0.001 and a batch size of 32. Training stopped when the loss value no longer decreased for 20 consecutive epochs or when the total number of training epochs reached 500. In this example, the loss value converged to 0.0021 in the 168th epoch, and the model's mean absolute error on the test set was 0.012MW.
[0062] Therefore, this step, by analyzing the real-time demand response signal before the start of the response period and combining historical response events with the current battery state to predict the response demand, achieves the determination of the response period, minimum discharge power, pre-response SOC threshold, response demand sequence, and target SOC lower limit. This solves the problem in existing methods where only a fixed pre-response threshold is known, and it is impossible to determine how much power needs to be discharged during the response period and how much power needs to be retained after the response. It provides complete boundary conditions for subsequent pre-adjustment and power calculation.
[0063] S200: Before the start of the response period, a pre-adjustment is made based on the current SOC, the pre-response SOC threshold, and the target SOC lower limit, and the pre-adjusted SOC is used as the initial SOC for the response period. When an energy storage device receives a response signal, its energy level may not be suitable for handling the response.
[0064] If the power level is below the threshold, the grid will face insufficient discharge capacity at the beginning of the response period; if the power level has exceeded the threshold, but the grid still has subsequent demand after the response ends, releasing the power level too low during the response period will also affect the next response.
[0065] Once the response period begins, the restriction prohibiting charging will take effect immediately, and it will be impossible to replenish the power at that time. Therefore, the power adjustment can only be completed before the response period begins, and it is necessary to take into account both the entry threshold for this response and the power arrangement after the response ends.
[0066] The pre-adjustment process for this step is as follows: Figure 2 As shown.
[0067] Within the advance notification period, the energy storage aggregation and management terminal compares the current power level with the SOC threshold before the response. If the power level is insufficient, it will charge until the threshold is reached.
[0068] When the battery level has exceeded the threshold, it is necessary to check whether there is still a need for a follow-up response after the response ends: if there is a need for a follow-up response and the current battery level is below the target SOC lower limit, continue charging until the target SOC lower limit is reached; if there is no need for a follow-up response or the battery level is already sufficient, keep the current battery level unchanged.
[0069] The adjusted power level is used as the power level at the beginning of the response period, and is used to establish the power level recursion relationship and solve the charging and discharging power.
[0070] Step S200 in the method provided in this embodiment of the invention includes: Compare the current state of charge (SOC) of the battery with the SOC threshold before the response; When the SOC in the current battery state is lower than the SOC threshold before the response, the charging power is determined based on the difference between the SOC in the current battery state and the SOC threshold before the response, the health status of the current battery state, and the number of cycles, and the energy storage is controlled to charge until the SOC threshold before the response is reached. When the SOC in the current battery state is not lower than the SOC threshold before the response, determine whether the response power at each time after the response end time in the response demand sequence is greater than zero. If the response power is greater than zero at each time point after the response ends, compare the current SOC of the battery state with the target SOC lower limit. When the current SOC of the battery is lower than the target SOC lower limit, the charging power is determined based on the difference between the current SOC and the target SOC lower limit, the current battery health status, and the number of cycles, and the energy storage is controlled to charge until the target SOC lower limit is reached. When the current state of charge (SOC) is not lower than the target lower limit of SOC, the current state of charge (SOC) remains unchanged. If the response power is zero at each time point after the response ends, the current state of charge (SOC) of the battery remains unchanged. The pre-adjusted SOC is used as the initial SOC for the response period.
[0071] Specifically, pre-adjustment refers to the process of adjusting the energy storage capacity to an appropriate level before the start of the response period, based on the current state of the battery and the boundary conditions required for this response.
[0072] Charging power represents the power used to control energy storage charging during the pre-adjustment process. It is determined by the difference between the current SOC and the target capacity, the health status, and the number of cycles. The larger the difference, the lower the health status, and the larger the number of cycles, the greater the charging power.
[0073] The initial SOC during the response period represents the battery charge at the start of the response period after the pre-adjustment is completed. It is the starting point for establishing the SOC recursive relationship and solving for the charge and discharge power.
[0074] In this step, the SOC in the current battery state is first compared with the SOC threshold before the response.
[0075] Furthermore, when the current SOC of the battery is lower than the pre-response SOC threshold, the charging power is determined based on the difference between the current SOC and the pre-response SOC threshold, the battery health status, and the number of cycles. The energy storage is then controlled to charge until the pre-response SOC threshold is reached, ensuring that the battery level at the start of the response period is not lower than the entry threshold for this response. The charging power is calculated using the following formula: .
[0076] The difference represents the proportion of the difference between the current SOC and the target charge. The calculation formula converts the difference into the required charge and spreads it over the unit time according to the advance notice time. Then it is amplified according to the health status and the number of cycles, so that batteries with more severe degradation and more cycles can use a larger charging power under the same difference, so as to leave a margin.
[0077] Furthermore, when the SOC in the current battery state is not lower than the SOC threshold before the response, it is determined whether the response power at each time after the response end time in the response demand sequence is greater than zero. If it is greater than zero, it means that there is still a subsequent response demand after the end of this response. At this time, the current SOC is further compared with the target SOC lower limit. When the current SOC is lower than the target SOC lower limit, the charging power is determined according to the difference between the current SOC and the target SOC lower limit, the health status and the number of cycles, and the energy storage is controlled to charge until the target SOC lower limit is reached. When the current SOC is not lower than the target SOC lower limit, the current SOC remains unchanged.
[0078] Then, when the response power is zero at each time point after the response ends, it indicates that there is no subsequent response requirement after the response ends, and the current SOC remains unchanged.
[0079] Finally, the pre-adjusted SOC is used as the initial SOC for the response period.
[0080] For example, a certain energy storage power station has a current SOC of 0.55, a pre-response SOC threshold of 0.5625, a target lower limit SOC of 0.1222, a 1-hour advance notification time, a rated capacity of 2MWh, a rated cycle life of 6000 cycles, a cycle count of 600, and a health status of 0.90. Since the current SOC of 0.55 is lower than the pre-response SOC threshold of 0.5625, it enters the charging branch. The difference is 0.5625 - 0.55 = 0.0125.
[0081] Charging power = (0.0125 × 2 × 0.90 ÷ 1) × (2 - 0.90) × (1 + 600 ÷ 6000) = 0.0225 × 1.10 × 1.10 = 0.027225 MW. Since the charging power of 0.027225 MW is less than the rated power of 1 MW, the energy storage is controlled to charge at 0.027225 MW until the pre-response SOC threshold of 0.5625 is reached.
[0082] If the current SOC is 0.60, it will not fall below the pre-response SOC threshold of 0.5625, so charging will stop. Instead, it will determine whether the response power at each time point after the response end in the response demand sequence is greater than zero. In this example, the response power at the two times after the response end is 0.18MW, which is greater than zero. Therefore, the current SOC is further compared with the target SOC lower limit of 0.1222. Since the current SOC is 0.60 and not lower than the target SOC lower limit of 0.1222, the current SOC is kept unchanged, and the initial SOC of the response period is 0.60.
[0083] Therefore, this step, by comparing the current SOC with the pre-response SOC threshold and the target SOC lower limit before the start of the response period, and determining whether to charge and to what extent based on the comparison results, achieves the goal of adjusting the energy storage capacity in advance to a level that can meet the entry threshold of this response and also take into account the subsequent response needs after the response ends. This solves the problem of existing practices that only judge the threshold once before the response and do not distinguish whether there are still response needs after the response ends, thus making it impossible to arrange the energy capacity. It provides a reasonable starting energy capacity for power calculation during the response period.
[0084] S300: A multi-objective function is constructed based on the initial SOC during the response period, integrating response performance, battery dynamic loss, and default penalty. Constraints include the minimum discharge power during the response period, charging prohibition, the pre-response SOC threshold, and the target SOC lower limit. The nonlinear terms in the multi-objective function and constraints are linearized and solved to obtain the charging and discharging power at each moment within the response period. Once the response period begins, the restriction prohibiting charging takes effect, and energy storage can only be adjusted on the discharge side. The amount of discharge affects three things simultaneously: whether the discharge is sufficient, which relates to whether the response can meet the standard; how deep the discharge is, which relates to the battery's lifespan; and how much the discharge power deviates from the required level of charge at the end of the response, which relates to whether the assessment will be triggered.
[0085] These three things are not consistent with each other. The more you release, the easier it is to meet the target, but the loss and deviation will also increase. If you only focus on one of them to arrange the discharge, the other two will inevitably be sacrificed. Therefore, it is necessary to weigh the three together under the same set of constraints and ensure that the calculated discharge power can be directly executed at each moment.
[0086] The energy storage aggregation and management terminal takes the initial SOC of the response period obtained through pre-adjustment as the starting point, treats the response efficiency as a positive term, and treats the battery dynamic loss and default penalty as negative terms, assembling them into a multi-objective function. Then, with the minimum discharge power, prohibition of charging, the SOC threshold before response and the lower limit of the target SOC as constraints, the nonlinear terms in it are transformed into a linear form using auxiliary variables, and then handed over to a mixed integer linear programming solver to solve for the power that should be released at each moment within the response period.
[0087] Step S300 in the method provided in this embodiment of the invention includes: The response efficiency is calculated by taking the average of the response efficiency at each moment within the response period and the ratio of the response power to the minimum discharge power. Based on the difference between the initial SOC and the target lower limit of SOC during the response period, combined with the current battery health status and cycle count, the battery dynamic loss is calculated. The battery dynamic loss increases with the increase of the difference, increases with the decrease of the health status, and increases with the increase of the cycle count. The penalty for breach of contract is calculated based on the deviation between the charging and discharging power and the minimum discharge power at each moment during the response period, and the deviation between the SOC at the end of the response period and the target lower limit of SOC. The larger the deviation between the charging and discharging power and the minimum discharge power, and the larger the deviation between the SOC at the end of the response period and the target lower limit of SOC, the greater the penalty for breach of contract. A multi-objective function is constructed using response performance, battery dynamic loss, and default penalty. The multi-objective function has response performance as a positive term and battery dynamic loss and default penalty as negative terms.
[0088] Specifically, response performance represents the degree to which the discharge power at each moment within the response period is achieved relative to the minimum discharge power. It is obtained by taking the average of the ratios at each moment, and the larger the value, the more complete the response.
[0089] Battery dynamic loss represents the lifespan loss caused by the depth of discharge and the degree of battery aging in this response. It increases with the difference between the initial SOC and the lower limit of the target SOC during the response period and the number of cycles, and increases with the decrease of the battery health status.
[0090] The penalty for breach of contract refers to the assessment cost corresponding to the discharge power not reaching the minimum discharge power and the charge deviating from the target SOC lower limit at the end of the response period. The greater the deviation, the greater the cost.
[0091] The multi-objective function represents the objective to be optimized with response performance as the positive term and battery dynamic loss and default penalty as the negative terms. The larger the value, the better the overall effect of this response in terms of meeting the standard, protecting the battery and avoiding the assessment.
[0092] In this step, the response efficiency at each moment is first calculated based on the ratio of the response power to the minimum discharge power within the response time period. Then, the average value is taken to obtain the response efficiency. The calculation formula is as follows: .
[0093] Furthermore, based on the difference between the initial SOC and the target lower limit of SOC during the response period, combined with the health status and cycle count, the battery dynamic loss is calculated as follows: .
[0094] The difference between the initial SOC and the lower limit of the target SOC during the response period is the proportion of the dischargeable capacity during the response period. The larger the difference, the more capacity is to be discharged in this operation. The lower the health status and the greater the number of cycles, the higher the cost for the battery to withstand the same discharge capacity. Therefore, these values are amplified by two coefficients.
[0095] Furthermore, based on the deviation between the charging / discharging power and the minimum discharge power at each moment during the response period, and the deviation between the SOC and the target SOC lower limit after the response period ends, the penalty for breach of contract is calculated as follows: .
[0096] The first item measures the degree to which the discharge power deviates from the minimum discharge power at each moment, and the second item measures the degree to which the charge deviates from the target SOC lower limit at the end of the response period. The power deviation coefficient and the charge deviation coefficient are calibrated based on the historical assessment records of the station, and are set to 1 and 2 respectively in this example.
[0097] Finally, a multi-objective function is constructed with response performance as the positive term and battery dynamic loss and default penalty as the negative terms, and the calculation formula is as follows: .
[0098] For example, the response period consists of eight moments, with response power at each moment being 0.32, 0.32, 0.30, 0.30, 0.30, 0.30, 0.30 and 0.30MW respectively, and the minimum discharge power being 0.30MW.
[0099] Response efficiency = (0.32 ÷ 0.30 + 0.32 ÷ 0.30 + 0.30 ÷ 0.30 × 6) ÷ 8 = (1.0667 + 1.0667 + 6.0000) ÷ 8 = 8.1333 ÷ 8 = 1.0167. Initial SOC during response period = 0.5625, target SOC lower limit = 0.1222, health status = 0.90, number of cycles = 600, rated cycle life = 6000.
[0100] Battery dynamic loss = (0.5625 - 0.1222) × (2 - 0.90) × (1 + 600 ÷ 6000) = 0.4403 × 1.10 × 1.10 = 0.4403 × 1.21 = 0.5328. The sum of the absolute values of the deviations between the charging / discharging power and the minimum discharge power at each moment = |0.32 - 0.30| + |0.32 - 0.30| + |0.30 - 0.30| × 6 = 0.0200 + 0.0200 + 0.0000 = 0.0400MW.
[0101] The SOC at the end of the response period is 0.2236, and the absolute value of its deviation from the target SOC lower limit is |0.2236 - 0.1222| = 0.1014. The penalty for breach of contract is 0.0400 × 1 + 0.1014 × 2 = 0.0400 + 0.2028 = 0.2428. The multi-objective function is 1.0167 - (0.5328 + 0.2428) = 1.0167 - 0.7756 = 0.2411.
[0102] Step S300 in the method provided in this embodiment of the invention further includes: Based on the initial SOC during the response period, the charging and discharging power at each moment during the response period, the duration of the response period, and the current health status of the battery, a recursive relationship is established between the SOC at each moment during the response period and the charging and discharging power at each moment during the response period. Based on the minimum discharge power and the prohibition of charging, it is determined that the discharge power at each moment during the response period is not lower than the minimum discharge power, and the charging power is zero. The pre-response SOC threshold is used as a constraint that the initial SOC during the response period is not lower than the threshold. The lower bound of SOC at each time point during the response period is determined based on the target SOC lower limit, the current battery health status, and the number of cycles. The lower bound of SOC is determined based on the target SOC lower limit, and the pre-response SOC threshold is not used as the lower bound of SOC at other times during the response period except for the initial time point. The upper bound of SOC at each moment within the response period is determined based on the upper limit of battery SOC, the current health status of the battery, and the number of cycles. The lower the health status and the higher the number of cycles, the narrower the upper and lower bounds of SOC. For the absolute value of the deviation between the charging and discharging power and the minimum discharge power at each moment during the response period, and the mutual exclusion of charging and discharging power at each moment during the response period, auxiliary variables are introduced to transform them into linear constraints. A mixed-integer linear programming solver is used with a multi-objective function as the objective. The solution is obtained under the following conditions: recursion relationship, discharge power is not lower than the minimum discharge power, charging power is zero, the initial SOC during the response period is not lower than the SOC threshold before the response, and the SOC at each time point during the response period satisfies the upper and lower bounds and linear constraints. The charging and discharging power at each time point during the response period is obtained.
[0103] More specifically, the recursive relationship represents the conversion relationship between SOC and charge / discharge power between two adjacent moments, based on the current available capacity of the battery.
[0104] The lower bound of SOC represents the minimum battery charge level that must not fall below a certain threshold during the response period, and is derived from the target lower SOC limit adjusted for health status. The upper bound of SOC represents the maximum battery charge level that must not exceed a certain threshold during the response period, and is derived from the battery SOC upper bound adjusted for health status.
[0105] A mixed-integer linear programming solver is a computational program capable of solving optimization problems that contain both continuous and integer variables, and whose objectives and constraints are linear.
[0106] Auxiliary variables are intermediate variables introduced to transform the deviation term containing absolute values and the charging / discharging mutual exclusion condition into linear constraints.
[0107] In this step, the recursive relationship between SOC and charge / discharge power at each moment within the response period is first established, and the calculation formula is as follows: .
[0108] The product of rated capacity and state of health represents the battery's current usable capacity; the lower the state of health, the smaller the usable capacity, and the greater the change in SOC caused by the same amount of charging and discharging.
[0109] Furthermore, based on the minimum discharge power and the prohibition of charging, it is determined that the discharge power at each moment during the response period is not lower than the minimum discharge power, and the charging power is zero.
[0110] Furthermore, the pre-response SOC threshold is used as a constraint that the initial SOC during the response period must not be lower than this threshold, ensuring that the energy level at the start of the response period meets the entry threshold for this response. The lower bound of the SOC at each moment during the response period is determined based on the target SOC lower limit and the health status, calculated as follows: .
[0111] The SOC threshold before response only constrains the initial moment of the response period and is not used as the lower bound of SOC at other moments during the response period. The reason is that continuous discharge is required during the response period. If the entry threshold is used as the lower bound throughout the entire period, the SOC will be exceeded as soon as the charge begins to decrease, and there will be no intersection between the constraints.
[0112] Then, the upper bound of SOC at each moment within the response period is determined based on the battery SOC upper limit, health status, and cycle count. The calculation formula is as follows: .
[0113] Among them, the lower the health status and the greater the number of cycles, the lower the upper bound and the higher the lower bound, and the narrower the usable range of SOC.
[0114] Then, auxiliary variables are introduced to transform the absolute values of the deviations between the charging and discharging power and the minimum discharging power at each moment during the response period, as well as the mutual exclusion of charging and discharging, into linear constraints.
[0115] Finally, a mixed-integer linear programming solver is used with a multi-objective function as the objective. Under the conditions of the above recursive relationship, the discharge power is not lower than the minimum discharge power, the charging power is zero, the initial SOC of the response period is not lower than the SOC threshold before the response, and the SOC at each time point in the response period satisfies the upper and lower bounds, the charging and discharging power at each time point in the response period is obtained.
[0116] For example, during the response period, the initial SOC is 0.5625, the rated capacity is 2MWh, and the health status is 0.90, so the available capacity is 2 × 0.90 = 1.80MWh.
[0117] The SOC at time 1 is 0.5625 - 0.32 × 0.25 ÷ 1.80 = 0.5625 - 0.0444 = 0.5181. The SOC at time 2 is 0.518056 - 0.32 × 0.25 ÷ 1.80 = 0.518056 - 0.044444 = 0.473611.
[0118] The charging and discharging power at times 3 to 8 is 0.30 MW, causing the State of Charge (SOC) to decrease by 0.30 × 0.25 ÷ 1.80 = 0.041667 at each time step. The SOC at times 3 to 8 are 0.431944, 0.390278, 0.348611, 0.306944, 0.265278, and 0.223611, respectively, with the minimum SOC during the response period being 0.223611.
[0119] The lower bound of SOC is 0.1222 × (2 - 0.90) = 0.1222 × 1.10 = 0.1344. The upper bound of SOC is 0.95 × 0.90 = 0.8550. During the response period, the SOC at each moment lies between 0.1344 and 0.8550, and both the recursive relationship and the upper and lower bound constraints of SOC hold. The initial SOC of the response period (0.5625) is not lower than the pre-response SOC threshold of 0.5625, thus this threshold constraint holds. During the response period, the discharge power at each moment is not lower than the minimum discharge power of 0.30 MW, and the charging power is zero.
[0120] Step S300 in the method provided in this embodiment of the invention further includes: For the absolute value of the deviation between the charging / discharging power and the minimum discharge power at each moment during the response period, a first non-negative auxiliary variable is introduced, wherein the first non-negative auxiliary variable is not less than the difference between the charging / discharging power and the minimum discharge power, and is not less than the difference between the minimum discharge power and the charging / discharging power. To address the deviation between the SOC at the end of the response period and the target lower limit of SOC, a second non-negative auxiliary variable is introduced. The second non-negative auxiliary variable is not less than the difference between the SOC at the end of the response period and the target lower limit of SOC, and is not less than the difference between the target lower limit of SOC and the SOC at the end of the response period. Substitute the first and second non-negative auxiliary variables into the default penalty in the multi-objective function; Based on the prohibition of charging, the charging power at each moment during the response period is determined to be zero, and the zero charging power is used as a linear constraint for mutual exclusion of charging and discharging. The inequalities corresponding to the first non-negative auxiliary variable, the inequalities corresponding to the second non-negative auxiliary variable, and the zero charging power are used as linear constraints and added to the mixed integer linear programming solver.
[0121] More specifically, the first non-negative auxiliary variable represents the non-negative variable used to replace the absolute value of the deviation between the charging / discharging power and the minimum discharge power. Its value is not less than the two directions of the deviation, and the minimum value is the absolute value of the deviation.
[0122] The second non-negative auxiliary variable represents the absolute value of the deviation between the SOC at the end of the response period and the target lower limit of SOC. Its value is not less than the two directions of the deviation, and the minimum value is the absolute value of the deviation.
[0123] In this step, a first non-negative auxiliary variable is introduced, which takes the absolute value of the deviation between the charging / discharging power and the minimum discharge power at each moment within the response period. Its value satisfies the following two inequalities: , .
[0124] Furthermore, to address the deviation between the SOC at the end of the response period and the target lower limit of SOC, a second non-negative auxiliary variable is introduced, whose value satisfies the following two inequalities: , .
[0125] Furthermore, by substituting the first and second non-negative auxiliary variables into the penalty for breach of contract in the multi-objective function, and replacing the two absolute value terms, the multi-objective function is transformed into a linear form.
[0126] Then, based on the prohibition of charging, the charging power at each moment within the response period is determined to be zero; since the charging power is always zero, charging and discharging will not occur simultaneously, so the zero charging power is directly used as a linear constraint for mutual exclusion of charging and discharging.
[0127] Finally, the four inequalities corresponding to the two non-negative auxiliary variables, along with the zero charging power, are treated as linear constraints and added to the mixed-integer linear programming solver.
[0128] For example, the charge / discharge power at the first moment is 0.32MW, and the minimum discharge power is 0.30MW.
[0129] The first non-negative auxiliary variable satisfies the following conditions: the first non-negative auxiliary variable ≥ 0.32 - 0.30 = 0.02, and the first non-negative auxiliary variable ≥ 0.30 - 0.32 = -0.02. The first non-negative auxiliary variable takes the smallest non-negative value that satisfies the above two conditions, which is 0.02.
[0130] Similarly, at time 2, the first non-negative auxiliary variable = 0.02; the charging and discharging power at times 3 to 8 equals the minimum discharge power, and the first non-negative auxiliary variable = 0. The SOC at the end of the response period is 0.223611, and the target SOC lower limit is 0.1222.
[0131] The second non-negative auxiliary variable satisfies the following: the second non-negative auxiliary variable ≥ 0.223611 - 0.1222 = 0.1014, and the second non-negative auxiliary variable ≥ 0.1222 - 0.223611 = -0.1014. The second non-negative auxiliary variable takes the smallest non-negative value that satisfies the above two equations, which is 0.1014.
[0132] Substituting the above values into the default penalty calculation = (0.02 + 0.02 + 0 × 6) × 1 + 0.1014 × 2 = 0.0400 + 0.2028 = 0.2428, the result is consistent with the previous result calculated directly based on the absolute value of the deviation, indicating that linearization did not change the value of the default penalty. The charging power is zero at all times during the response period, and the linear constraint of mutual exclusion between charging and discharging holds. .
[0133] Therefore, this step, starting from the initial SOC of the response period, assembles a multi-objective function with response performance as the positive term and battery dynamic loss and default penalty as the negative terms. Under the constraints of minimum discharge power, prohibition of charging, pre-response SOC threshold and target SOC lower limit, the nonlinear terms are linearized and then solved by a mixed integer linear programming solver. This achieves a comprehensive balance between the three requirements of response compliance, battery protection and avoidance of assessment in the same problem. It solves the problem of existing practices that only focus on whether discharge can be achieved within the response period and cannot achieve a balance among multiple operational objectives, and provides directly executable charging and discharging power for each moment within the response period.
[0134] S400: Controls energy storage operation based on the solution results. After the response ends, updates the response demand prediction model based on the actual response deviation and battery degradation, and corrects the target SOC lower limit. After a response is completed, there will always be a difference between the actual power released and the scheduling requirements, and the battery degradation caused by the discharge in this response has become a given fact.
[0135] If this post-event information is only used for this settlement, and the next response still uses the original prediction model and the lower limit of battery capacity, then the model deviation will accumulate successively, and the lower limit of battery capacity will become increasingly distant from the actual state of the battery.
[0136] The window of opportunity between the end of the response and the arrival of the next response is precisely the period during which these actual results can be used to correct the model. The problem that needs to be solved is to convert both the "size of the deviation" and the "size of the battery degradation" into corrections to the model and the lower limit of the battery capacity.
[0137] The energy storage aggregation and control terminal controls the charging and discharging of energy storage according to the solution results at each moment. After the response ends, it obtains the actual discharge power at each moment, compares it with the minimum discharge power to obtain the actual response deviation, and calculates the battery degradation based on the discharge capacity.
[0138] The response demand prediction model is then updated based on these two quantities, and the lower limit of the target SOC is adjusted upwards. The adjustment result must not exceed the upper limit of the battery SOC for use in the next round of response.
[0139] Step S400 in the method provided in this embodiment of the invention includes: The energy storage charging and discharging is controlled according to the charging and discharging power at each moment within the response period. After the response ends, the actual discharge power at each moment during the response period is obtained, and the absolute value of the difference between the actual discharge power and the minimum discharge power is calculated to obtain the actual response deviation. The discharge capacity is calculated based on the charging and discharging power at each moment within the response period and the duration of the response period. Combined with the current battery health status and cycle count, the battery degradation is determined. The larger the discharge capacity, the larger the cycle count, and the lower the health status, the greater the battery degradation. The response demand prediction model is updated based on the actual response deviation and battery degradation. The larger the actual response deviation and the larger the battery degradation, the larger the update magnitude. The target SOC lower limit is corrected based on the ratio of the actual response deviation to the sum of the response power at each moment in the response demand sequence, and the battery degradation. The larger the ratio and the greater the battery degradation, the higher the corrected target SOC lower limit, and the corrected target SOC lower limit does not exceed the battery SOC upper limit.
[0140] Specifically, the actual response deviation represents the cumulative absolute value of the deviation between the actual power released at each moment during the response period and the minimum discharge power, reflecting the degree to which the response deviates from the scheduling requirements.
[0141] Battery degradation represents the proportion of capacity loss caused by discharge in this response. It increases with the increase of discharge capacity and cycle number, and increases with the decrease of battery health.
[0142] The update magnitude indicates the degree to which the response demand forecasting model has been adjusted in this update, and is determined by both the actual response deviation and the amount of battery degradation.
[0143] In this step, the energy storage charging and discharging is first controlled according to the charging and discharging power at each moment within the response period.
[0144] Furthermore, after the response ends, the actual discharge power at each moment during the response period is obtained, and the actual response deviation is calculated using the following formula: .
[0145] Furthermore, the discharge capacity is calculated by accumulating the charging and discharging power at each moment within the response period and the duration of the response period. This, combined with the battery health status and cycle count, determines the battery degradation. The calculation formula is as follows: .
[0146] The proportion of capacity in this discharge is obtained by dividing the discharge capacity by the available capacity. The lower the health status, the greater the corresponding degradation. The more cycles the battery has, the more sensitive it is to this discharge. Therefore, it is amplified by two coefficients.
[0147] Then, the response demand prediction model is updated based on the actual response deviation and battery degradation. The larger the values of the two, the greater the update magnitude, making the model approach the actual results more quickly.
[0148] Finally, the target SOC lower limit is adjusted upward based on the ratio of the actual response deviation to the sum of the response power at each moment within the response period in the response demand sequence and the battery degradation. The calculation formula is as follows: .
[0149] The ratio represents the proportion of the actual response deviation to the power scale of the current response. The larger the ratio, the more the current response deviates from the expectation, and the more power needs to be reserved in the next response. If the lower limit of the corrected target SOC exceeds the upper limit of the battery SOC, the upper limit of the battery SOC is taken.
[0150] For example, the energy storage is controlled to charge and discharge according to the charging and discharging power at times 1 to 8, and the discharge power at each time is 0.32, 0.32, 0.30, 0.30, 0.30, 0.30, 0.30 and 0.30MW respectively, and the charging power at each time is zero.
[0151] After the response ends, the actual discharge power at each moment during the response period is obtained, which is consistent with the scheduling result and remains the same as the above values. Actual response deviation = |0.32-0.30|+|0.32-0.30|+|0.30-0.30|×6=0.0200+0.0200+0.0000=0.0400MW. Discharge capacity = (0.32+0.32+0.30×6)×0.25=2.44×0.25=0.6100MWh.
[0152] Battery degradation = (0.6100 ÷ (2 × 0.90)) × (1 - 0.90) × (1 + 600 ÷ 6000) = (0.6100 ÷ 1.80) × 0.10 × 1.10 = 0.3389 × 0.10 × 1.10 = 0.0373. Model update magnitude = Actual response deviation + Battery degradation = 0.0400 + 0.0373 = 0.0773. Ratio = Actual response deviation ÷ Sum of response power at each moment within the response period in the response demand sequence = 0.0400 ÷ 2.44 = 0.016393.
[0153] The revised target SOC lower limit = 0.1222 × (1 + 0.016393) × (1 + 0.0373) = 0.1222 × 1.016393 × 1.0373 = 0.1288. Since the revised target SOC lower limit of 0.1288 does not exceed the battery SOC upper limit of 0.95, 0.1288 is chosen as the target SOC lower limit for the next round of response.
[0154] Therefore, this step controls the charging and discharging of energy storage according to the solution results, and updates the response demand prediction model and corrects the target SOC lower limit based on the actual response deviation and battery degradation after the response ends. This enables the prediction model and the lower limit of energy capacity to continuously evolve with each actual response result, solving the problem that existing practices are not adjusted after being set once and are difficult to adapt to subsequent changes in response demand. This provides adaptive capability for energy storage to participate in real-time demand response for a long time.
[0155] Through the above specific implementation methods, the embodiments of the present invention achieve the following technical effects: In summary, this embodiment obtains the response demand sequence and the target SOC lower limit by parsing the real-time demand response signal before the start of the response period and combining historical response events with the current battery state to predict the response demand; by pre-adjusting according to the current SOC and the target SOC lower limit, the energy storage capacity can meet the entry threshold while also taking into account subsequent demand.
[0156] Furthermore, by linearizing and solving a multi-objective function with response performance as the positive term and battery dynamic loss and default penalty as the negative terms under the constraints of minimum discharge power, prohibition of charging and upper and lower limits of energy, the charging and discharging power is obtained time-by-time; after the response ends, the prediction model is updated and the target SOC lower limit is corrected based on the actual response deviation and battery degradation.
[0157] Example 2, as Figure 3 As shown, based on the inventive concept of the energy storage single-scenario optimization method based on multi-objective rule fusion provided in Embodiment 1, this invention also provides an energy storage single-scenario optimization system based on multi-objective rule fusion, specifically including: The response demand analysis module 11 is used to acquire real-time demand response signals, analyze the response period, minimum discharge power and SOC threshold before response, obtain the response demand sequence through the response demand prediction model based on historical response events and current battery status, and determine the target SOC lower limit after the response ends. SOC pre-adjustment module 12 is used to pre-adjust the SOC based on the current SOC, the pre-response SOC threshold and the target SOC lower limit before the start of the response period, and use the pre-adjusted SOC as the initial SOC of the response period. The multi-objective solution module 13 is used to construct a multi-objective function that integrates response performance, battery dynamic loss and default penalty based on the initial SOC of the response period. The module uses the minimum discharge power of the response period, charging prohibition, SOC threshold before response and target SOC lower limit as constraints to linearize and solve the nonlinear terms in the multi-objective function and constraints, and obtain the charging and discharging power at each time during the response period. The control correction module 14 is used to control the operation of energy storage based on the solution results. After the response ends, it updates the response demand prediction model based on the actual response deviation and battery degradation, and corrects the target SOC lower limit.
[0158] In one embodiment, the response demand parsing module 11 is further configured to: obtain the real-time demand response signal through the real-time dispatch channel of the power grid, parse the issuance time, response start time and response end time in the real-time demand response signal, take the time period between the response start time and the response end time as the response time period, determine the advance notification time according to the interval between the issuance time and the response start time, and extract the minimum discharge power required by the power grid dispatch instruction from the real-time demand response signal. The SOC threshold before response is extracted first from the real-time demand response signal. If the signal does not carry the SOC threshold before response, the lowest SOC before response for successful responses in the same response period in historical response events is corrected based on the SOC and health status of the current battery state to obtain the SOC threshold before response. Historical response events include response time period, response power, SOC before response and SOC after response. Current battery status includes SOC, health status and cycle count. Health status is determined based on the cycle count; the higher the cycle count, the lower the health status.
[0159] Furthermore, the response demand analysis module 11 is also used to: filter historical response events that are the same as the response period from historical response events, and extract the corresponding response power, pre-response SOC and post-response SOC; The response power, SOC before response, and SOC after response are combined with the current SOC, health status, and cycle count in the battery state and input into the response demand prediction model. The output response demand sequence includes the response power at each time point during the response period and the response power at each time point after the response ends. The discharge capacity during the response period is calculated based on the response power at each moment and the duration of the response period. The discharge capacity after the response end time is calculated based on the response power at each moment and the corresponding duration. Based on the discharge capacity during the response period and the discharge capacity after the response ends, the SOC in the current battery state is reduced, and then adjusted according to the current battery health status and cycle count to obtain the target lower limit of SOC after the response ends.
[0160] Furthermore, the response demand analysis module 11 is also used to: collect multiple sample response power sets, sample pre-response SOC sets, and sample post-response SOC sets based on historical response events; Collect different sets of sample response power, sets of SOC before sample response, and sets of SOC after sample response, corresponding sets of SOC, health status, and number of cycles in the sample battery state; The annotation yields a set of sample response demand sequences, which includes the response power at each time point within the response period and the response power at each time point after the response end time. Based on machine learning, a response demand prediction model is constructed and trained by taking multiple sets of sample response power, sample pre-response SOC, sample post-response SOC, and corresponding sets of SOC, health status and cycle count in the sample battery state as inputs and a set of sample response demand sequences as outputs. Configure the demand forecasting model in the energy storage aggregation and management terminal.
[0161] In one embodiment, the SOC pre-adjustment module 12 is further configured to: compare the SOC in the current battery state with the SOC threshold before the response; When the SOC in the current battery state is lower than the SOC threshold before the response, the charging power is determined based on the difference between the SOC in the current battery state and the SOC threshold before the response, the health status of the current battery state, and the number of cycles, and the energy storage is controlled to charge until the SOC threshold before the response is reached. When the SOC in the current battery state is not lower than the SOC threshold before the response, determine whether the response power at each time after the response end time in the response demand sequence is greater than zero. If the response power is greater than zero at each time point after the response ends, compare the current SOC of the battery state with the target SOC lower limit. When the current SOC of the battery is lower than the target SOC lower limit, the charging power is determined based on the difference between the current SOC and the target SOC lower limit, the current battery health status, and the number of cycles, and the energy storage is controlled to charge until the target SOC lower limit is reached. When the current state of charge (SOC) is not lower than the target lower limit of SOC, the current state of charge (SOC) remains unchanged. If the response power is zero at each time point after the response ends, the current state of charge (SOC) of the battery remains unchanged. The pre-adjusted SOC is used as the initial SOC for the response period.
[0162] In one embodiment, the multi-objective solution module 13 is further configured to: calculate the response performance at each moment based on the ratio of the response power to the minimum discharge power at each moment within the response period, and take the average of the response performance at each moment to obtain the response performance. Based on the difference between the initial SOC and the target lower limit of SOC during the response period, combined with the current battery health status and cycle count, the battery dynamic loss is calculated. The battery dynamic loss increases with the increase of the difference, increases with the decrease of the health status, and increases with the increase of the cycle count. The penalty for breach of contract is calculated based on the deviation between the charging and discharging power and the minimum discharge power at each moment during the response period, and the deviation between the SOC at the end of the response period and the target lower limit of SOC. The larger the deviation between the charging and discharging power and the minimum discharge power, and the larger the deviation between the SOC at the end of the response period and the target lower limit of SOC, the greater the penalty for breach of contract. A multi-objective function is constructed using response performance, battery dynamic loss, and default penalty. The multi-objective function has response performance as a positive term and battery dynamic loss and default penalty as negative terms.
[0163] Furthermore, the multi-objective solution module 13 is also used to: establish a recursive relationship between the SOC at each moment during the response period and the charge / discharge power at each moment during the response period, based on the initial SOC during the response period, the charge / discharge power at each moment during the response period, the duration of the response period, and the current health status of the battery. Based on the minimum discharge power and the prohibition of charging, it is determined that the discharge power at each moment during the response period is not lower than the minimum discharge power, and the charging power is zero. The pre-response SOC threshold is used as a constraint that the initial SOC during the response period is not lower than the threshold. The lower bound of SOC at each time point during the response period is determined based on the target SOC lower limit, the current battery health status, and the number of cycles. The lower bound of SOC is determined based on the target SOC lower limit, and the pre-response SOC threshold is not used as the lower bound of SOC at other times during the response period except for the initial time point. The upper bound of SOC at each moment within the response period is determined based on the upper limit of battery SOC, the current health status of the battery, and the number of cycles. The lower the health status and the higher the number of cycles, the narrower the upper and lower bounds of SOC. For the absolute value of the deviation between the charging and discharging power and the minimum discharge power at each moment during the response period, and the mutual exclusion of charging and discharging power at each moment during the response period, auxiliary variables are introduced to transform them into linear constraints. A mixed-integer linear programming solver is used with a multi-objective function as the objective. The solution is obtained under the following conditions: recursion relationship, discharge power is not lower than the minimum discharge power, charging power is zero, the initial SOC during the response period is not lower than the SOC threshold before the response, and the SOC at each time point during the response period satisfies the upper and lower bounds and linear constraints. The charging and discharging power at each time point during the response period is obtained.
[0164] Furthermore, the multi-objective solution module 13 is also used to: introduce a first non-negative auxiliary variable for the absolute value of the deviation between the charging and discharging power and the minimum discharging power at each moment within the response period, wherein the first non-negative auxiliary variable is not less than the difference between the charging and discharging power and the minimum discharging power, and is not less than the difference between the minimum discharging power and the charging and discharging power. To address the deviation between the SOC at the end of the response period and the target lower limit of SOC, a second non-negative auxiliary variable is introduced. The second non-negative auxiliary variable is not less than the difference between the SOC at the end of the response period and the target lower limit of SOC, and is not less than the difference between the target lower limit of SOC and the SOC at the end of the response period. Substitute the first and second non-negative auxiliary variables into the default penalty in the multi-objective function; Based on the prohibition of charging, the charging power at each moment during the response period is determined to be zero, and the zero charging power is used as a linear constraint for mutual exclusion of charging and discharging. The inequalities corresponding to the first non-negative auxiliary variable, the inequalities corresponding to the second non-negative auxiliary variable, and the zero charging power are used as linear constraints and added to the mixed integer linear programming solver.
[0165] In one embodiment, the control correction module 14 is further configured to: control the energy storage charging and discharging according to the charging and discharging power at each moment within the response period; After the response ends, the actual discharge power at each moment during the response period is obtained, and the absolute value of the difference between the actual discharge power and the minimum discharge power is calculated to obtain the actual response deviation. The discharge capacity is calculated based on the charging and discharging power at each moment within the response period and the duration of the response period. Combined with the current battery health status and cycle count, the battery degradation is determined. The larger the discharge capacity, the larger the cycle count, and the lower the health status, the greater the battery degradation. The response demand prediction model is updated based on the actual response deviation and battery degradation. The larger the actual response deviation and the larger the battery degradation, the larger the update magnitude. The target SOC lower limit is corrected based on the ratio of the actual response deviation to the sum of the response power at each moment in the response demand sequence, and the battery degradation. The larger the ratio and the greater the battery degradation, the higher the corrected target SOC lower limit, and the corrected target SOC lower limit does not exceed the battery SOC upper limit.
[0166] The energy storage single-scenario optimization method and system based on multi-objective rule fusion provided by this invention can be applied to the energy storage operation control in the real-time demand response scenario of the power grid. It can be integrated into the energy storage aggregation and management terminal, complete the boundary condition determination and power pre-adjustment before the start of the response period, and provide directly executable charging and discharging power at each moment during the response period.
[0167] The energy storage single-scenario optimization method and system based on multi-objective rule fusion provided by this invention analyzes real-time demand response signals and predicts response demand by combining historical response events and current battery status. It constructs a multi-objective function with response efficiency as the positive term and battery dynamic loss and default penalty as the negative terms and solves it linearly. This achieves the balance between response compliance, battery protection and avoidance of market assessment during the response period, which is conducive to improving the reliability of energy storage in real-time demand response and the full life cycle operation benefits. For specific implementation details of this invention, please refer to Embodiment 1.
[0168] It should be noted that the order of the embodiments in this invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Specific embodiments of the present invention have been described above. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
Claims
1. A single-scenario optimization method for energy storage based on multi-objective rule fusion, characterized in that, The method includes: Acquire real-time demand response signals, analyze response time periods, minimum discharge power, and pre-response SOC threshold, obtain response demand sequences through a response demand prediction model based on historical response events and current battery status, and determine the target SOC lower limit after the response ends. Before the response period begins, the SOC is pre-adjusted based on the current SOC, the pre-response SOC threshold, and the target SOC lower limit, and the pre-adjusted SOC is used as the initial SOC for the response period. A multi-objective function is constructed based on the initial SOC of the response period, which integrates response performance, battery dynamic loss, and default penalty. The minimum discharge power of the response period, charging prohibition, the SOC threshold before the response, and the target SOC lower limit are used as constraints. The nonlinear terms in the multi-objective function and constraints are linearized and solved to obtain the charging and discharging power at each time point within the response period. The energy storage operation is controlled based on the solution results. After the response ends, the response demand prediction model is updated based on the actual response deviation and battery degradation, and the target SOC lower limit is corrected.
2. The energy storage single-scenario optimization method based on multi-objective rule fusion according to claim 1, characterized in that, The real-time demand response signal is acquired through the real-time grid dispatch channel. The issuance time, response start time, and response end time of the real-time demand response signal are analyzed. The time period between the response start time and the response end time is taken as the response period. The advance notification time is determined based on the interval between the issuance time and the response start time. The minimum discharge power required by the grid dispatch command is extracted from the real-time demand response signal. The pre-response SOC threshold is preferentially extracted from the real-time demand response signal. If the signal does not carry the pre-response SOC threshold, the lowest pre-response SOC of successful responses during the same response period in historical response events is corrected based on the current battery state's SOC and health status to obtain the pre-response SOC threshold. The historical response events include the response period, response power, pre-response SOC, and post-response SOC. The current battery state includes SOC, health status, and cycle count. The health status is determined based on the cycle count; the higher the cycle count, the lower the health status.
3. The energy storage single-scenario optimization method based on multi-objective rule fusion according to claim 1, characterized in that, Based on historical response events and the current battery state, a response demand sequence is obtained through a response demand prediction model, and the target SOC lower limit after the response ends is determined, including: Filter historical response events that are the same as the response time period from the historical response events, and extract the corresponding response power, pre-response SOC and post-response SOC; The response power, pre-response SOC, and post-response SOC, combined with the current battery state's SOC, health status, and cycle count, are input into the response demand prediction model to output a response demand sequence. The response demand sequence includes the response power at each moment within the response period and the response power at each moment after the response end. The discharge capacity during the response period is calculated based on the response power at each moment within the response period and the duration of the response period. The discharge capacity after the response end time is calculated based on the response power at each moment after the response end time and the corresponding duration. Based on the discharge capacity during the response period and the discharge capacity after the response ends, the SOC in the current battery state is reduced, and then corrected by combining the health status and cycle count in the current battery state to obtain the target lower limit of SOC after the response ends.
4. The energy storage single-scenario optimization method based on multi-objective rule fusion according to claim 3, characterized in that, The construction of the response demand prediction model includes: Based on the historical response events, collect multiple sets of sample response power, pre-response SOC set, and post-response SOC set; Collect different sets of sample response power, sets of SOC before sample response, and sets of SOC after sample response, corresponding sets of SOC, health status, and number of cycles in the sample battery state; The sample response demand sequence set is obtained by labeling, wherein the sample response demand sequence set includes the response power at each time point within the response period and the response power at each time point after the response end time; Based on machine learning, the multiple sample response power sets, sample pre-response SOC sets, sample post-response SOC sets, and corresponding sample battery state sets, health state sets, and cycle count sets are used as inputs, and the sample response demand sequence sets are used as outputs to construct and train a response demand prediction model. Configure the response demand prediction model on the energy storage aggregation and management terminal.
5. The energy storage single-scenario optimization method based on multi-objective rule fusion according to claim 1, characterized in that, The pre-adjustment includes: Compare the SOC in the current battery state with the SOC threshold before the response; When the SOC in the current battery state is lower than the SOC threshold before the response, the charging power is determined based on the difference between the SOC in the current battery state and the SOC threshold before the response, the health status of the current battery state, and the number of cycles, and the energy storage is controlled to charge until the SOC threshold before the response is reached. When the SOC in the current battery state is not lower than the SOC threshold before the response, determine whether the response power at each time after the response end time in the response demand sequence is greater than zero. If the response power is greater than zero at each time after the response end time, compare the SOC in the current battery state with the target SOC lower limit; When the SOC in the current battery state is lower than the target SOC lower limit, the charging power is determined based on the difference between the SOC in the current battery state and the target SOC lower limit, the health status of the current battery state, and the number of cycles, and the energy storage is controlled to charge until the target SOC lower limit is reached. When the SOC in the current battery state is not lower than the target SOC lower limit, the SOC in the current battery state remains unchanged; If the response power is zero at each time point after the response end time, the SOC in the current battery state remains unchanged; The pre-adjusted SOC is used as the initial SOC for the response period.
6. The energy storage single-scenario optimization method based on multi-objective rule fusion according to claim 1, characterized in that, A multi-objective function is constructed based on the initial SOC during the response period, integrating response performance, battery dynamic degradation, and default penalty, including: The response performance at each moment is calculated based on the ratio of the response power to the minimum discharge power at each moment within the response period, and the average of the response performance at each moment is taken to obtain the response performance. Based on the difference between the initial SOC and the target lower limit of SOC during the response period, combined with the current battery health status and cycle count, the battery dynamic loss is calculated, wherein the battery dynamic loss increases with the increase of the difference, increases with the decrease of the health status, and increases with the increase of the cycle count. The penalty for breach of contract is calculated based on the deviation between the charging / discharging power and the minimum discharge power at each moment during the response period, and the deviation between the SOC at the end of the response period and the target lower limit of SOC. The larger the deviation between the charging / discharging power and the minimum discharge power, and the larger the deviation between the SOC at the end of the response period and the target lower limit of SOC, the greater the penalty for breach of contract. A multi-objective function is constructed using the response performance, the battery dynamic loss, and the default penalty, wherein the multi-objective function has the response performance as a positive term and the battery dynamic loss and the default penalty as negative terms.
7. The energy storage single-scenario optimization method based on multi-objective rule fusion according to claim 1, characterized in that, Using the minimum discharge power during the response period, charging prohibition, the pre-response SOC threshold, and the target SOC lower limit as constraints, the nonlinear terms in the multi-objective function and constraints are linearized and solved to obtain the charging and discharging power at each moment during the response period, including: Based on the initial SOC of the response period, the charging and discharging power at each moment within the response period, the duration of the response period, and the health status of the current battery state, a recursive relationship is established between the SOC at each moment within the response period and the charging and discharging power at each moment within the response period. Based on the minimum discharge power and the prohibition of charging, it is determined that the discharge power at each moment during the response period is not lower than the minimum discharge power, and the charging power is zero. Based on the pre-response SOC threshold, the target SOC lower limit, the current battery health status, and the number of cycles, the upper and lower bounds of SOC at each moment within the response period are determined. The lower bound of SOC is the larger of the pre-response SOC threshold and the target SOC lower limit. The lower the health status and the larger the number of cycles, the narrower the upper and lower bounds of SOC. The absolute value of the deviation between the charging and discharging power and the minimum discharge power at each moment within the response period, and the mutual exclusion of charging and discharging power at each moment within the response period, are transformed into linear constraints by introducing auxiliary variables. A mixed-integer linear programming solver is used to solve the problem under the following conditions: the multi-objective function is used as the objective; the recursive relationship is met; the discharge power is not lower than the minimum discharge power; the charging power is zero; the upper and lower bounds of the SOC are met; and the linear constraints are met. This allows the charging and discharging power at each moment within the response period to be obtained.
8. The energy storage single-scenario optimization method based on multi-objective rule fusion according to claim 7, characterized in that, The absolute values of the deviations between the charging / discharging power and the minimum discharge power at each moment within the response period, and the mutual exclusion of charging / discharging power at each moment within the response period, are transformed into linear constraints by introducing auxiliary variables, including: For the absolute value of the deviation between the charging / discharging power and the minimum discharge power at each moment during the response period, a first non-negative auxiliary variable is introduced, wherein the first non-negative auxiliary variable is not less than the difference between the charging / discharging power and the minimum discharge power, and is not less than the difference between the minimum discharge power and the charging / discharging power. To address the deviation between the SOC at the end of the response period and the target lower limit of SOC, a second non-negative auxiliary variable is introduced, wherein the second non-negative auxiliary variable is not less than the difference between the SOC at the end of the response period and the target lower limit of SOC, and is not less than the difference between the target lower limit of SOC and the SOC at the end of the response period. Substitute the first non-negative auxiliary variable and the second non-negative auxiliary variable into the default penalty in the multi-objective function; Based on the prohibition of charging, the charging power at each moment during the response period is determined to be zero, and the zero charging power is used as a linear constraint for mutual exclusion of charging and discharging. The inequalities corresponding to the first non-negative auxiliary variable, the inequalities corresponding to the second non-negative auxiliary variable, and the zero charging power are used as the linear constraints and added to the mixed integer linear programming solver.
9. The energy storage single-scenario optimization method based on multi-objective rule fusion according to claim 1, characterized in that, The energy storage operation is controlled based on the solution results. After the response ends, the response demand prediction model is updated based on the actual response deviation and battery degradation, and the target SOC lower limit is corrected, including: The energy storage charging and discharging is controlled according to the charging and discharging power at each moment within the response period. After the response ends, the actual discharge power at each moment during the response period is obtained, and the absolute value of the difference between the actual discharge power and the minimum discharge power is calculated to obtain the actual response deviation. The discharge capacity is calculated based on the charging and discharging power at each moment during the response period and the duration of the response period. Combined with the current battery health status and cycle count, the battery degradation is determined. The larger the discharge capacity, the larger the cycle count, and the lower the health status, the greater the battery degradation. The response demand prediction model is updated based on the actual response deviation and the battery degradation, wherein the larger the actual response deviation and the larger the battery degradation, the larger the update magnitude. The target SOC lower limit is corrected based on the ratio of the actual response deviation to the sum of the response power at each moment in the response demand sequence, and the battery degradation. The larger the ratio and the greater the battery degradation, the higher the corrected target SOC lower limit, and the corrected target SOC lower limit does not exceed the battery SOC upper limit.
10. A single-scenario optimization system for energy storage based on multi-objective rule fusion, characterized in that, The system is used to execute the energy storage single-scenario optimization method based on multi-objective rule fusion as described in any one of claims 1 to 9, and the system comprises: The response demand analysis module is used to acquire real-time demand response signals, analyze the response period, minimum discharge power and SOC threshold before response, obtain the response demand sequence through the response demand prediction model based on historical response events and current battery status, and determine the target SOC lower limit after the response ends. The SOC pre-adjustment module is used to pre-adjust the SOC based on the current SOC, the pre-response SOC threshold, and the target SOC lower limit before the start of the response period, and use the pre-adjusted SOC as the initial SOC for the response period. The multi-objective solution module is used to construct a multi-objective function that integrates response performance, battery dynamic loss and default penalty based on the initial SOC of the response period. The module uses the minimum discharge power of the response period, the prohibition of charging, the SOC threshold before the response and the target SOC lower limit as constraints to linearize the nonlinear terms in the multi-objective function and constraints and solve them to obtain the charging and discharging power at each time during the response period. The control correction module is used to control the operation of energy storage based on the solution results. After the response ends, it updates the response demand prediction model based on the actual response deviation and battery degradation, and corrects the target SOC lower limit.
