Optimization and Control Methods for Energy Storage Configuration in Photovoltaic Power Plants Based on Power Prediction
By using a photovoltaic power plant energy storage configuration optimization method based on power prediction, future power shortage data and operating cost models are generated to optimize energy storage capacity configuration. This solves the problems of overinvestment and insufficient energy storage caused by fixed-ratio capacity setting, and reduces grid penalties and equipment losses.
Patent Information
- Application Number
- CN202611636082.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-22
- Publication Date
- 2026-11-17
AI Technical Summary
The fixed-ratio capacity determination method of existing photovoltaic power plant energy storage systems can easily lead to overinvestment or insufficient energy storage capacity. Furthermore, during long-term cloud cover, the battery may be depleted prematurely or have no power to discharge, resulting in high grid default penalties and equipment lifespan losses.
By acquiring missing power generation data and historical trends of photovoltaic power plants, future missing power data is generated, and extrapolation grid data is constructed to calculate operating costs. Based on the minimum consumption cost, the energy storage capacity configuration is optimized, unnecessary discharge commands are actively cut off, and greedy control is avoided.
It significantly reduces the total amount of penalties for defaults and equipment depreciation expenses throughout the day, avoids overinvestment or insufficient energy storage capacity, and improves the economy and reliability of energy storage systems.
Smart Images

Figure CN123121772A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of photovoltaic power generation and energy storage technology, specifically to a method for optimizing and controlling the configuration of photovoltaic power plant energy storage based on power prediction. Background Technology
[0002] Under the photovoltaic (PV) power grid connection assessment mechanism, PV power plants are generally equipped with energy storage systems to discharge when PV output is insufficient, thereby mitigating power deviations and avoiding penalties. Existing energy storage management systems generally employ an immediate-response "greedy control" strategy, meaning they discharge to compensate immediately upon a power deficit. In situations with prolonged cloud cover or other sustained power drops, this strategy can lead to premature battery depletion in the early stages of shallow shading; when severe deficits occur, the batteries have no charge to discharge, resulting in hefty grid default penalties and unnecessary battery cycle life losses for the power plant. Furthermore, existing static energy storage capacity determination methods based on fixed ratios configure storage according to a fixed percentage (e.g., 10%) of the total installed capacity of the PV power plant, which can easily lead to overinvestment or insufficient energy storage capacity. Summary of the Invention
[0003] To address the technical problem that static energy storage capacity determination methods based on fixed ratios in related technologies are prone to leading to overinvestment or insufficient energy storage capacity, this application provides a method for optimizing and controlling the configuration of photovoltaic power plant energy storage based on power prediction.
[0004] The specific technical solution adopted is as follows: Obtain the current power generation data of the photovoltaic power station, and based on the power generation data and the historical power generation trend, generate the expected power generation data for each future time point. Construct a projection grid data consisting of future time nodes and the power nodes of energy storage batteries in photovoltaic power plants, and calculate the operating cost data of each node in the projection grid data based on preset grid penalty data, discharge loss cost, and expected power loss data. In the simulation grid data, based on the operating cost data, a reverse simulation operation is performed along the time axis from the future to the current moment to calculate the value of the reserved power corresponding to the remaining power of the energy storage battery; When the difference between the preset grid penalty data and the discharge loss cost is less than the reserved power value, the converter's discharge compensation command is cut off and it remains in standby mode. Based on the minimum cost incurred by the discharge compensation command, the optimal energy storage capacity configuration operation is performed on the photovoltaic power station.
[0005] In one possible implementation of this application, obtaining the missing power generation data of the photovoltaic power station at the current moment includes: Obtain the expected target power generation and actual active power of the photovoltaic power station at the current moment; Based on the difference between the expected target power generation and the actual active power, the missing power generation data is determined.
[0006] In one possible implementation of this application, based on missing power generation data and historical power generation trends, expected power loss data for future time points is generated, including: The rate of change of power generation is calculated based on the trend of actual active power change within a preset historical time period. Divide the future preset time period into multiple time nodes and determine the time interval between the time nodes; The missing power generation data is subtracted from the product of the power generation change rate and the time interval to generate the expected power loss data at each future time point.
[0007] In one possible implementation of this application, a projection grid data is constructed, consisting of future time points and the energy level nodes of energy storage batteries in a photovoltaic power station, including: Obtain the upper and lower limits of the operating capacity of energy storage batteries in photovoltaic power plants; Based on the upper and lower limits of the operating power, the power value of the energy storage battery is divided into multiple power nodes; Using electricity nodes as the vertical axis and time nodes as the horizontal axis, a projection grid data is constructed.
[0008] In one possible implementation of this application, the energy storage battery's capacity is divided into multiple capacity nodes based on an upper and lower limit of operating capacity, including: Determine the normal operating capacity range of the energy storage battery; The range between the upper limit of operating power and the normal operating power range is defined as the high power range, and the range between the lower limit of operating power and the normal operating power range is defined as the low power range. By using an equidistant division method, the normal operating power range is divided into multiple power nodes. By using the Chebyshev algorithm, the high power range and the low power range are divided into multiple non-equidistant power nodes.
[0009] In one possible implementation of this application, based on preset grid penalty data and discharge loss costs, as well as expected power loss data, the operating cost data of each node within the grid data is calculated and extrapolated, including: The maximum discharge power of the converter in the photovoltaic power station is obtained, and the power range from zero to the maximum discharge power is divided into multiple discrete discharge powers according to a preset interval. Among them, any node in the extrapolation grid data corresponds to multiple discrete discharge powers. Based on the difference between the expected power missing data and the discrete discharge power, the preset grid penalty data, and the discharge loss cost, the operating cost data of each node in the simulation grid data is calculated.
[0010] In one possible implementation of this application, the operating cost data of each node within the simulation grid data is calculated based on the difference between the expected power loss data and the discrete discharge power, preset grid penalty data, and discharge loss costs, including: The difference between the expected power loss data and the discrete discharge power is multiplied by the preset grid penalty data to obtain the penalty data caused by the power deviation. By superimposing the penalty data with the discharge loss cost, the operating cost data of each node in the extrapolated grid data is obtained.
[0011] In one possible implementation of this application, based on operating cost data, a reverse extrapolation operation is performed along the time axis from the future to the present to calculate the value of the reserved energy corresponding to the remaining energy of the energy storage battery, including: For any discrete discharge power, the target charge data for the next time node is calculated based on the difference between the charge of the energy storage battery at the current time node and the discharge amount corresponding to the discrete discharge power. Identify the two first charge nodes that are adjacent to the target charge data in the simulation grid data; The node cost of the two first energy nodes is weighted and calculated to obtain the target expected cost. The minimum expected value of the current node is obtained by overlaying the target expected value with the operating cost data. Based on the minimum expected generation value of the current node, the value of the reserved energy corresponding to the remaining energy of the energy storage battery is calculated.
[0012] In one possible implementation of this application, the value of reserved energy corresponding to the remaining energy capacity of the energy storage battery is calculated based on the minimum expected replacement value of the current node, including: Identify the two adjacent second power nodes of the current node and extract the remaining power value corresponding to the second power node; The reserved energy value corresponding to the remaining energy of the energy storage battery is calculated based on the difference between the remaining energy values of adjacent second energy nodes and the difference between the minimum expected replacement value of the second energy nodes.
[0013] In one possible implementation of this application, the optimal energy storage capacity configuration operation for the photovoltaic power station is performed based on the minimum consumption cost generated by the discharge compensation command, including: Obtain data on investment costs, cumulative fines, and equipment loss expenditures for photovoltaic power plants within a preset period; The installed capacity of photovoltaic power plants is divided into multiple candidate energy storage capacities; For any candidate energy storage capacity, the investment cost, cumulative penalty data, and equipment loss expenditure data of the current candidate energy storage capacity are superimposed to obtain the comprehensive consumption cost; The candidate energy storage capacity corresponding to the minimum comprehensive consumption cost is selected as the optimal energy storage capacity for configuration.
[0014] This application has, but is not limited to, the following technical effects: In this application, by acquiring the current power generation loss data of the photovoltaic power station, and then generating the expected power loss data for each future time node based on the power generation loss data and historical power generation change trends, the operating cost data corresponding to each time node and different power nodes is determined. This determines the reserved power value corresponding to the remaining power of the energy storage battery. When the difference between the preset grid penalty data and the discharge loss cost is less than the reserved power value, the inverter's discharge compensation command is cut off and it remains in standby mode. Based on the minimum consumption cost generated by the discharge compensation command, the optimal energy storage capacity configuration operation is performed on the photovoltaic power station. This ensures that when the immediate deterministic discharge revenue is lower than the reserved power value, the discharge command is actively cut off, avoiding ineffective power consumption in the early stages of long-term cloud cover, significantly reducing the total daily penalty amount and equipment depreciation expenses. This avoids situations of over-investment or insufficient energy storage capacity. Attached Figure Description
[0015] Figure 1 This is a flowchart illustrating the first embodiment of the photovoltaic power plant energy storage configuration optimization and control method based on power prediction of this application. Figure 2 This is a schematic diagram of the system architecture involved in the photovoltaic power plant energy storage configuration optimization and control method based on power prediction in this application. Detailed Implementation
[0016] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.
[0017] This application provides a method for optimizing and controlling the configuration of energy storage in a photovoltaic power plant based on power prediction. In the first embodiment of this method, referring to... Figure 1 The methods include: Step S10: Obtain the missing power generation data of the photovoltaic power station at the current moment. Based on the missing power generation data and the historical power generation trend, generate the expected missing power data for each future time node.
[0018] As an example, the photovoltaic power plant energy storage configuration optimization and control method based on power prediction can also be applied to the photovoltaic power plant energy storage configuration optimization and control system based on power prediction (hereinafter referred to as the system). The missing power generation data represents the difference between the actual grid-connected active power and the grid dispatch target power at the current moment. The missing expected power data represents the set of power generation shortfalls at each future time point, which is generated by combining the current power generation gap and the historical power generation trend.
[0019] As an example, the system sets the current actual running time as The system defines the discrete time series index for future assessment cycles as follows: ,in, It is an integer, and its value range is 1. , This represents the total number of steps for future projections. The system sets the time interval between adjacent time nodes to be [value missing]. As a parameter configuration example, based on the common 15-minute rolling assessment cycle requirement for power grid dispatching, the system will set the time interval... Set to 1 minute, and then calculate the total number of steps in the simulation. It is set to 15. This configuration establishes a one-dimensional time axis for all subsequent cost extrapolations.
[0020] Step S10, which involves obtaining the missing power generation data of the photovoltaic power station at the current moment, includes: Obtain the expected target power generation and actual active power of the photovoltaic power station at the current moment; Based on the difference between the expected target power generation and the actual active power, the missing power generation data is determined.
[0021] As an example, the system reads the current target power generation from the energy management system via a local communication interface. Simultaneously, the system uses the grid connection point data acquisition terminal to read the current actual active power. Then, the system performs a difference calculation, subtracting the actual active power from the target power generation to obtain the missing current power generation data. .
[0022] It should be noted that when When the value is greater than zero, it indicates that the current actual power generation is lower than the grid dispatch target, that is, there is a shortfall in meeting the target; the larger the value, the greater the actual default deviation faced at the current moment.
[0023] The step S10, which generates expected power shortage data for each future time point based on missing power generation data and historical power generation trends, includes: The rate of change of power generation is calculated based on the trend of actual active power changes within a preset historical time period.
[0024] As an example, the system uses the current time Use the endpoint to extract a preset historical time period. The actual active power time series within the specified period, i.e., the actual active power. This preset historical time period. The length can be adaptively configured manually by the implementers based on the actual local weather fluctuation cycle, for example, set to second-level sampling data from the past 10 or 15 minutes.
[0025] The system first performs a data integrity check on the acquired actual active power. If the system is in its initial power-on state, or if a communication interruption has been detected that resulted in the effective historical data duration not reaching the preset historical time period, the system determines that it is currently in a state of missing historical data. At this time, to prevent subsequent extrapolation calculations from going out of bounds or diverging, the system will use the historical power generation change rate... Forced assignment of zero.
[0026] When the data volume within the preset historical time period passes verification, the system invokes the Savitzky-Golay filtering algorithm to process the actual active power. By configuring the sliding window length and polynomial order (e.g., configuring a second-order fit with a width of 60 sampling points), the system smooths the actual active power. This operation eliminates the jitter and glitches generated by the inverter's maximum power point tracking algorithm, preserving the stable characteristics representing macroscopic meteorological evolution.
[0027] After smoothing, the system extracts the sequence corresponding to the actual active power. The time value corresponds to the first time derivative at the endpoint. The system records this time derivative value as the rate of change of power generation over the historical time period. .
[0028] It is important to understand the rate of change in power generation. The meaning is: if A negative number indicates that cloud cover is worsening, and power generation conditions are gradually deteriorating; if A positive number indicates that the clouds are dissipating and power generation conditions are improving.
[0029] Divide the future preset time period into multiple time nodes and determine the time interval between the time nodes.
[0030] As an example, the future preset time period corresponds to the aforementioned future assessment cycle, which can be 15 minutes, with the time interval set to 1 minute. Based on this, the time period is divided into multiple time nodes.
[0031] The missing power generation data is subtracted from the product of the power generation change rate and the time interval to generate the expected power loss data at each future time point.
[0032] As an example, to eliminate inference distortion caused by data dimension misalignment, the system first adjusts the rate of change of power generation. Convert the time unit to match the time interval. The dimensions must be kept consistent. The specific dimension alignment operation is as follows: If In minutes, and In kilowatts per second, the system will Multiply by 60 to convert it to kilowatts per minute.
[0033] Furthermore, the system introduces an attenuation factor. The range of values for this attenuation factor is: The specific value can be set by technical personnel based on the accuracy statistics of historical weather forecasts (e.g., 0.95). The system iterates through each discrete time node index within the future assessment cycle using a loop statement. The system combines current missing power generation data, rate of change, decay factor, and future time span to calculate the future [number missing]. Missing data on expected power at each time point The mathematical expression for this calculation process is as follows: in, This is missing electrical power data, used to represent the power deviation that already exists at the current moment. This represents the absolute time span from the current time to the target node in the projection. This term is aligned with the dimensions. Together they constitute the linear evolution increment in the case of no decay.
[0034] Indicates index over time The prediction confidence weights decrease exponentially. This term directly affects the evolution increment and constitutes a composite term. The physical meaning of this combined term is that, due to the high uncertainty of long-term weather changes, a single historical slope should not be infinitely amplified. As the number of extrapolation steps increases... The increase, The value of gradually decreases, thereby forcibly limiting the convergence of the long-term forecast increment and preventing the forecast gap from diverging infinitely due to simple linear multiplication.
[0035] This indicates a zero-value lower limit truncation operator. Because in the current grid connection assessment mechanism, exceeding the forecast target in actual power generation does not generate a corresponding financial reward, and a negative shortfall would lead to false negative costs if included in subsequent financial accounting. Therefore, when the result after the above subtraction calculation is lower than zero, this operator forcibly truncates the result to zero; when the result is greater than zero, the original value is retained.
[0036] The system is indexed according to discrete time series. Increasing from 1 to In the order described above, execute the formula completely to generate a set containing the expected power missing data for all future time points: This set is used as the expected power missing data for the entire time period.
[0037] Step S20: Construct a projection grid data consisting of future time nodes and the power nodes of energy storage batteries in photovoltaic power plants. Based on preset grid penalty data, discharge loss costs, and expected power loss data, calculate the operating cost data of each node in the projection grid data.
[0038] As an example, the extrapolation grid data can be a state-space matrix used for dynamic programming extrapolation. The extrapolation grid refers to the specific charge state of the battery at a certain future moment, corresponding to each coordinate point in the grid. As the underlying data structure, it is used to carry out the calculation of action cost and to build the calculation path for reverse state backtracking.
[0039] As an example, operating cost data represents the sum of grid penalty expenses and battery losses incurred after performing a specific discharge action at a specific node of the simulation grid.
[0040] The step S20, which constructs the projection grid data consisting of future time nodes and the power nodes of energy storage batteries in photovoltaic power plants, also includes steps S21 to S23: Step S21: Obtain the upper limit and lower limit of the operating power of the energy storage battery in the photovoltaic power station.
[0041] As an example, the system reads the upper and lower limits of the rated operating capacity of the energy storage battery from the battery management system through a local interface, and records them as the upper limit of operating capacity. and operating power limit .
[0042] Step S22, based on the upper and lower limits of the operating power, divide the power value of the energy storage battery into multiple power nodes, specifically including: Determine the normal operating capacity range of the energy storage battery.
[0043] As an example, to ensure the lifespan of battery hardware, it is common practice to... Set to 90%, The system is set to 10%. The total number of nodes in the discrete charge state is set to [value missing]. (For example, setting) (set to 50), and set the sequence index of the discrete charged state to . ,in, .
[0044] As an example, based on this, the system uses a segmented node generation strategy to construct the power level vertical axis. In actual engineering scenarios where energy storage batteries smooth out fluctuations daily, they operate in the middle power level range most of the time. Therefore, the system first extracts the set normal operating power level range (in this example, this range is defined as 30% to 70%).
[0045] The range between the upper limit of operating power and the normal operating power range is defined as the high power range, and the range between the lower limit of operating power and the normal operating power range is defined as the low power range.
[0046] As an example, the range between the upper limit of the normal operating power range and the upper limit of the operating power range is defined as the high power range, for example, the high power range of 70% to 90%. The range between the lower limit of the normal operating power range and the lower limit of the operating power range is defined as the low power range, for example, the low power range of 10% to 30%. These two ranges are also the boundary ranges extending from the middle normal operating range to both ends.
[0047] By using an equidistant division method, the normal operating power range is divided into multiple power nodes. By using the Chebyshev algorithm, the high power range and the low power range are divided into multiple non-equidistant power nodes.
[0048] As an example, within this typical operating power range, the system uses an arithmetic progression to generate equally spaced nodes. This equidistant partitioning ensures that the battery maintains a constant and dense computational resolution within its most frequently used operating range, avoiding excessive truncation during state transitions.
[0049] As an example, the system uses the Chebyshev polynomial root distribution algorithm to generate non-equidistant distributed charge nodes. Because the battery's terminal voltage exhibits drastic nonlinear changes when approaching depletion or full charge, the characteristics of the Chebyshev algorithm cause the generated nodes to become increasingly distributed as they approach... and When the extreme boundary is reached, the smaller the distribution interval (the denser the distribution), the better. This operation does not increase the total number of nodes. Under the premise of [missing information], the rounding error at the extreme value boundary is effectively reduced.
[0050] Furthermore, the total number of nodes is set to M=50. 26 equally spaced nodes are allocated to the normal operating range [30%, 70%], with a step size of 1.6%. 12 nodes are allocated to each of the low power range [10%, 30%] and the high power range [70%, 90%]. For the high power range […]… , [70%, 90%], which is the j-th Chebyshev node inside. The calculation formula is: Where j = 1, 2, ..., N, N = 12; Simplified, it becomes: Low battery range , [i.e., [10%, 30%]] is mapped to the range of 10% to 30% using the same formula.
[0051] Step S23: Construct the projection grid data with the power node as the vertical axis and the time node as the horizontal axis.
[0052] As an example, the system concatenates and sorts the equidistant nodes in the middle interval with the Chebyshev encrypted nodes in the two end intervals to obtain multiple energy nodes, which are then aggregated to form a complete one-dimensional energy vertical axis. The system extracts the future discrete time series set in the above steps. As a one-dimensional horizontal axis, this horizontal axis includes multiple time nodes. The horizontal axis and the vertical axis are combined perpendicularly to each other, thereby generating inference grid data.
[0053] The step S20, which calculates and extrapolates the operating cost data of each node within the grid data based on preset grid penalty data, discharge loss costs, and expected power loss data, includes: The maximum discharge power of the converter in the photovoltaic power station is obtained. The power range from zero to the maximum discharge power is divided into multiple discrete discharge powers according to a preset interval. Each node in the extrapolation grid data corresponds to multiple discrete discharge powers.
[0054] As an example, the system first obtains the maximum discharge power of the converter. and the minimum power regulation step size allowed by the converter communication control protocol. Among them, the maximum discharge power The preset value for the converter is used, and therefore, the system uses 0 as the lower limit and... With upper limit, The interval is divided into multiple discrete discharge power intervals based on a preset time. This generates a discrete discharge power sequence composed of discrete discharge powers. Each node in the extrapolated grid data represents a defined "future moment + battery charge". In this state, the converter has multiple action options, such as no discharge, low-power discharge, or high-power discharge, thus corresponding to multiple discrete discharge powers.
[0055] Furthermore, the system targets any specific node within the inference grid data. Perform two layers of filtering.
[0056] The first layer is physical over-limit filtering: the system reads the preset total rated energy storage capacity. (Relative calculated value determined by the total installed capacity of the photovoltaic power plant, unit: kilowatt-hours, e.g., 20,000 kilowatt-hours). The system performs virtual deduction calculations on the discharge power in the initial sequence. If the power consumption corresponding to a certain discharge power will lead to the current... Falling below the lower limit of the operating capacity of energy storage batteries The system then removes the discrete discharge power from the sequence.
[0057] The second layer is a computational pruning constraint filter: In cloud-occupied scenarios, the only benefit of discharge is to fill the grid gap. If the discharged power exceeds the gap, the excess not only fails to provide any additional default mitigation but also unnecessarily increases battery physical losses, resulting in a poor solution that inevitably leads to losses. Therefore, to eliminate ineffective over-discharge power combinations and significantly reduce the computational overhead of the derivation dimension, the system uses the available discrete discharge power corresponding to the node. The upper limit is strictly constrained to the current time node. The corresponding missing data of expected power The system will include all values greater than [a certain value]. The discharge values are forcibly discarded.
[0058] The power items retained after the above two layers of filtering are packaged by the system to form a candidate set of available discharge power specific to that node, which is used for the next step of cost calculation.
[0059] Based on the difference between the expected power missing data and the discrete discharge power, the preset grid penalty data, and the discharge loss cost, the operating cost data of each node in the simulation grid data is calculated, specifically including: The difference between the expected power loss data and the discrete discharge power is multiplied by the preset grid penalty data to obtain the penalty data caused by the power deviation.
[0060] By superimposing the penalty data with the discharge loss cost, the operating cost data of each node in the extrapolated grid data is obtained.
[0061] As an example, the system reads preset grid penalty data specified by the regional power grid from a local configuration file. (Unit: Yuan / kWh), which is the grid deviation penalty rate, as well as the discharge loss cost provided by the battery manufacturer. (Unit: Yuan / kWh). A segmented mapping table of power shortfall and penalty unit price is pre-established. Before each calculation, this mapping table is queried based on the expected power shortfall data for the current node to obtain the corresponding preset grid penalty data. When the power shortfall becomes more severe due to long-term declines, the predicted immediate penalty rate for the current node will increase accordingly. The preset grid penalty data includes the immediate assessment penalty rate. Penalty rates for simulation nodes Immediate assessment penalty rate Used to calculate subsequent net discharge revenue and extrapolate node penalty rates. Used to calculate operating cost data.
[0062] Specifically, the local configuration file pre-stores a "Mapping Table of Default Gap Rate and Penalty Unit Price". When calculating the immediate net discharge revenue at the current moment, the table is consulted to obtain a fixed immediate assessment penalty rate based on the ratio of the current power generation missing data to the rated photovoltaic installed capacity (immediate gap rate). .
[0063] When calculating the operating cost data of the k-th node in the simulation grid data, the corresponding simulation node penalty rate is dynamically obtained by looking up a table based on the predicted gap rate corresponding to the expected power loss data of each future node. The larger the gap ratio, the higher the penalty fee rate will be.
[0064] Furthermore, for any node in the inference grid data... And each available discrete discharge power in the dedicated candidate set corresponding to that node. The system calculates its corresponding operating cost data. : in, This indicates that the system determines the discrete discharge power at a specific time and under a specific electrical state. The total financial expenses that must be incurred to carry out the discharge.
[0065] It is a constant transformation term used to convert the extrapolation time step, which is in minutes, into an hourly scale, so that it can be multiplied with the power variable to obtain the standard unit of electricity (kilowatt-hour, unit of electricity).
[0066] The first half of the formula This constitutes the expected penalty term of the system, representing the penalty data obtained due to power deviation, where the subtraction term within parentheses... Indicates that it has been executed After the compensation discharge operation, the photovoltaic power station still cannot fill the residual default power gap; multiplying this residual gap by the time conversion term yields the amount of electricity that did not meet the standard, which is then multiplied by the penalty unit price. Multiplication quantifies the cash penalty expenditure directly paid by the power plant to the grid as a result of this action. This reflects a negative correlation: the discrete discharge power selected by the system... The larger the value, the smaller the remaining deficit, and the less the grid penalty expenditure. Indicates time node The corresponding expected power data is missing.
[0067] The second half of the formula This constitutes the system's equipment loss term. This term will include the selected discrete discharge power. After converting to discharge quantity, simply multiply by the discharge loss cost. Its physical engineering significance lies in quantifying the irreversible damage and depreciation of the battery's chemical materials caused by this extracted power. This reflects a positive correlation: the selected discrete discharge power... The higher the value, the higher the cost of physical wear and tear inside the battery.
[0068] This formula, through simple addition logic, clearly illustrates the core financial game in energy storage scheduling: active discharge can reduce the penalty costs at the front end, but it will simultaneously increase the equipment wear and tear costs at the back end. The total cost of a single operation is the only metric that represents the compromise between the two.
[0069] The system targets each effective node within a two-dimensional grid and its corresponding discrete discharge power. All calculations are performed using the formulas described above.
[0070] Step S30: In the simulation grid data, based on the operating cost data, perform a reverse simulation operation along the time axis from the future to the current moment to calculate the value of the reserved power corresponding to the remaining power of the energy storage battery.
[0071] As an example, the value of reserved electricity represents the expected opportunity cost of using current electricity to mitigate the risk of future power drops. In other words, it's the amount of penalty for default that a power station could avoid by reserving a unit of electricity at the current moment, under the optimal subsequent strategy.
[0072] As an example, the optimal discharge action depends not only on the immediate penalty for default but also on the remaining power capacity after discharge, which is sufficient to address future power shortages. The subsequent cost at the end of the assessment period (i.e., a future preset time period) serves as a definite baseline boundary. Working backward from the end of the timeline, the system iterates through all candidate discharge actions for each node, adding the operating cost data of that action to the known minimum total cost of the target node at the next moment caused by the action. The minimum sum among all actions is extracted as the minimum expected cost of the current node, and this is used to backtrack to the current moment. This approach aims to overcome the short-sightedness of greedy algorithms by discounting all future uncertainties in default risks and equipment depreciation to the current moment, thus forming a globally optimal action evaluation benchmark.
[0073] Step S30 includes: For any discrete discharge power, the target charge data for the next time node is calculated based on the difference between the charge of the energy storage battery at the current time node and the discharge amount corresponding to the discrete discharge power.
[0074] As an example, locate the last time node of the inference timeline in the inference grid data. The system will include all discrete electrical nodes at the very end. Corresponding minimum expected cost All values are forcibly assigned to 0. This operation establishes the initial boundary conditions for the reverse deduction, indicating that subsequent penalties after the end of the assessment period will not be included in this calculation.
[0075] Furthermore, the system indexes the time series according to the time step. from The time increments sequentially to 0, triggering a reverse-order outer loop. At each time node... The system traverses all discrete electrical nodes. For the specific node being computed. The system extracts and iterates through each discrete discharge power in the aforementioned candidate set of available discharge power. .
[0076] For each selected discrete discharge power The system calculates the power consumption caused by the discharge behavior. The system utilizes the current power level. Deducting percentage of discharge The target charge data for the next step (next time node) is calculated and denoted as... .in, This indicates the rated total capacity of energy storage.
[0077] Identify the two first charge nodes that are adjacent to the target charge data in the simulation grid data.
[0078] As an example, due to target charge data Typically, the value is a continuous floating-point number, which cannot fall precisely on discrete vertical axis nodes. The system performs linear interpolation to obtain the accurate subsequent cost. The specific implementation logic is as follows: The system is positioned on a one-dimensional discrete vertical axis. In the interval between two adjacent nodes, the node immediately below it (with lower battery level) is recorded as... Record the node immediately above it (with a larger charge) as... , and All are with The first adjacent electrical node.
[0079] The node value of the two first energy nodes is weighted and calculated to obtain the target expected value.
[0080] As an example, system computation The weighting ratio between the two first energy nodes: the closer to a node, the greater its weight. Based on this calculated weighting ratio, the system determines the weight for the next time step. The known node cost that has already been calculated is also the cost of the first energy node. Minimum expected cost With the first power node of The system performs a weighted summation calculation to obtain the expected cost of the misaligned target state, denoted as . , specifically, The calculation method is as follows: By combining the target expected value with the operating cost data, the minimum expected value of the current node is obtained.
[0081] As an example, the system will display the operational cost data resulting from the action performed at the current moment. The target expected cost calculated by the above interpolation The summation represents the discrete discharge power applied in the current step. And the total lifecycle cost incurred by adopting the optimal strategy in the future. The system traverses all discrete discharge powers. Compare all the summation results, select the discrete discharge power with the smallest value as the optimal choice for the current node, and update the node with this smallest summation value. Minimum expected cost .
[0082] When the outer loop finishes executing in reverse order, it backtracks to the current time point (i.e., completion). When calculating the cost distribution curve at the current time starting point, the system obtains the cost distribution curve.
[0083] Based on the minimum expected generation value of the current node, the value of the reserved capacity corresponding to the remaining capacity of the energy storage battery is calculated, specifically including: Identify the two adjacent second power nodes of the current node and extract the remaining power value corresponding to the second power nodes.
[0084] As an example, the system reads the current remaining capacity (i.e., the actual remaining percentage) of the energy storage battery in real time through the local battery management system (BMS) interface, and records it as follows. The system performs matching on the vertical axis of the simulation grid, selecting nodes that contain the current actual power mapping. The system identifies two adjacent second-level energy nodes. The node with the lower energy value among these two second-level energy nodes is designated as the node. The node with the higher energy value is designated as the node. .
[0085] The reserved energy value corresponding to the remaining energy of the energy storage battery is calculated based on the difference between the remaining energy values of adjacent second energy nodes and the difference between the minimum expected replacement value of the second energy nodes.
[0086] As an example, the system extracts the cost values corresponding to two known adjacent nodes and, in conjunction with the battery's rated parameters, calculates the value of the reserved power capacity. The specific calculation formula is as follows: in, This indicates that at the current moment, if the remaining battery power is at a low level... At that time, the minimum expected cost that must be paid in the future assessment cycle (usually facing huge fines due to the inability to fill subsequent gaps).
[0087] This indicates that at the current moment, if the remaining battery power is at a high level... At that time, the minimum expected cost that must be paid during the future assessment cycle.
[0088] Numerator The calculation logic is the total cost of low power consumption minus the total cost of high power consumption. Its physical engineering significance lies in quantifying the financial difference between the two: since having more power consumption can avoid more default penalties in the future, this numerator difference directly represents the expected amount of default funds saved by having more power consumption in this range.
[0089] denominator In China, due to Always greater than This subtraction operation avoids the calculation risk of producing a negative or zero denominator. This item multiplies the pure percentage difference in electricity by the total rated capacity of the energy storage. This is then converted into specific physical electricity consumption (unit: kilowatt-hour).
[0090] This represents the expected opportunity cost (in yuan / kWh) of retaining a unit of electricity at the current stage. The systemic pattern is that the larger this value, the more severe the future price drop gap, and the greater the financial significance of reserving each unit of electricity to offset hefty fines in the future.
[0091] Step S40: When the difference between the preset grid penalty data and the discharge loss cost is less than the reserved power value, the discharge compensation command of the converter is cut off and the system remains in standby mode.
[0092] Furthermore, the system first determines that current power generation data is missing. The value is [value missing]. When this value is greater than zero, it indicates that there is an assessment gap that needs to be filled. The system retrieves a fixed immediate assessment penalty rate from a table in the local configuration file based on the ratio of the current missing power generation data to the rated photovoltaic installed capacity (immediate gap rate). and obtain discharge loss cost .
[0093] Based on this, the system calculates the net immediate discharge benefit of executing the discharge compensation command at the current moment. The specific calculation formula is as follows: In the above formula for calculating the net benefit of immediate discharge: The penalty rate for immediate assessment represents the amount of objective default penalties that can be waived for photovoltaic power plants for every kilowatt-hour of electricity delivered to the grid by energy storage batteries to fill the gap.
[0094] Discharge loss cost represents the cost of the energy storage battery itself at the rate of lifespan degradation it must withstand for each portion of electricity discharged.
[0095] The physical engineering significance lies in quantifying the book profit of a single action: the difference between the two is the net immediate discharge benefit. The larger this value, the higher the net profit from the current operation. Specifically, this is calculated at this point. Substitute The immediate assessment penalty rate is extracted based on the shallower gap depth at the current moment. Therefore, when the simulation anticipates a severe drop due to dense cloud cover, thus linking it to higher-order expected penalty compensation opportunities, The assessment values are able to cross the current immediate discharge net benefit when making comparisons. .
[0096] Net income from immediate discharge This represents the immediate deterministic marginal benefit that can be obtained by performing a unit of discharge operation at the current moment; while the value of the reserved power input from the preceding module... This represents the long-term expected opportunity cost of reserving a unit of electricity to mitigate future unknown drop penalties. The system compares the magnitudes of these two values to select the action with the best overall economic benefit. Specifically: when This indicates that the immediate, certain net benefit gained from discharging the electricity now is sufficient to cover or even exceed the expected opportunity cost of reserving that portion of the electricity for future use. At this point, the system determines that discharging is economically justifiable and directly issues a discharge compensation command to the energy storage converter.
[0097] when This means that the penalty caused by the current weather fluctuations is relatively light. If the system discharges at this time, the small penalty saved cannot compensate for the huge default losses caused by premature power consumption in the future when dealing with severe drops. At this time, the system triggers the interception mechanism, actively cuts off the discharge command, and sends a standby command to the inverter, forcing it to maintain zero power output to preserve the remaining power.
[0098] Furthermore, the system in this application first determines the power balance state in each sampling period. If the actual active power is greater than or equal to the dispatch target power (generation surplus), the system executes charging control logic to store the surplus power into the energy storage battery. This process does not generate grid default penalty, and therefore does not trigger the reverse solution of the derivation grid. If the actual active power is less than the dispatch target power (generation gap), the system immediately triggers the construction of the simulation grid and the reverse simulation calculation starting from the current real SOC.
[0099] The charging actions performed during daytime periods of surplus power generation are solely for changing the physical remaining capacity of the energy storage batteries. The state transition process; when cloud cover occurs again and a power shortage occurs, the reverse engineering module updates the charging status using the latest power level. As the entry point for the current grid's vertical axis slice, the value of reserved power is recalculated. The two achieve a complete physical closed loop through a "state input - passive triggering" approach.
[0100] Step S50: Based on the minimum consumption cost generated by the discharge compensation command, perform the optimal energy storage capacity configuration operation for the photovoltaic power station.
[0101] As an example, this step reads the historical power generation operation data of the photovoltaic power station over a period of time (such as six months or a whole year), and generates a discrete candidate sequence of battery capacity configurations inside the computer. The system simulates and deduces by repeatedly calling the algorithm of the aforementioned module, and outputs the specific capacity value with the lowest total billing value according to the discharge compensation instruction, thereby executing the optimal energy storage capacity configuration operation.
[0102] Step S50 includes: Obtain data on investment costs, cumulative fines, and equipment loss expenditures for photovoltaic power plants within a preset period.
[0103] The installed capacity of photovoltaic power plants is divided into multiple candidate energy storage capacities.
[0104] As an example, the system reads the total installed capacity of the photovoltaic power plant. As a parameter configuration embodiment, the system configures a capacity lower limit of 5% of the total installed capacity, a capacity upper limit of 25%, and calculates and generates a candidate set of rated energy storage capacity using an arithmetic progression of 1%. The system sets the candidate capacity index of this set to [index missing]. The specific rated energy capacity values of each energy storage unit in the set are recorded as candidate energy storage capacities. .
[0105] As an example, the preset period could be six months or one year. The investment cost represents the hardware investment cost. Since the initial hardware investment covers total expenditures over the next ten years, annualized conversion is necessary to ensure that funds at different time scales can be directly summed. This item is based on candidate energy storage capacity. The corresponding initial hardware investment, the preset bank discount rate, and the designed service life are used to calculate and amortize the fixed cost to a single calendar year using an annuity factor. This value is related to the candidate energy storage capacity. There is a positive correlation: the larger the capacity, the higher the equivalent annual investment cost. The investment cost can be calculated based on the capacity using the annuity factor formula. The formula is: multiply the candidate energy storage capacity by the unit price of energy storage capacity, and then multiply by the annuity factor. This method is a conventional annuity factor discounting method in engineering economics, and will not be elaborated here.
[0106] As an example, cumulative penalty data represents the total amount of grid default penalties triggered during a year-long review period due to unmet power shortfalls caused by depleted energy storage capacity or proactive system shutdowns. This figure is related to candidate energy storage capacity. There is a negative correlation: the larger the configured capacity, the stronger the ability to mitigate penalties, and the lower the total amount of fines collected throughout the year. The cumulative penalty data can be obtained through virtual playback statistics based on a one-year panoramic historical data. The system replays historical scheduling curves in seconds, accumulating the penalties corresponding to all defaults for power gaps that could not be filled due to energy storage depletion or algorithmic truncation within the year.
[0107] As an example, equipment wear and tear expenditure data represents the total cumulative cost of battery hardware cycle wear and tear throughout the year, incurred after the converter performed all uninterrupted valid discharge actions during the playback period. This item reflects the actual equipment wear and tear expenditure in operation. Equipment wear and tear expenditure data can be obtained based on a virtual playback statistical analysis of a one-year panoramic historical data. The system accumulates the battery physical wear and tear depreciation costs corresponding to all uninterrupted discharge commands actually executed by the energy storage converter within one year.
[0108] For any candidate energy storage capacity, the investment cost, cumulative penalty data, and equipment loss expenditure data of the current candidate energy storage capacity are superimposed to obtain the comprehensive consumption cost.
[0109] The candidate energy storage capacity corresponding to the minimum comprehensive consumption cost is selected as the optimal energy storage capacity for configuration.
[0110] As an example, for any candidate energy storage capacity Comprehensive consumption costs The calculation method can be: in, Indicates investment cost, This indicates the cumulative number of fines. This represents data on equipment wear and tear expenses.
[0111] As an example, the system iterates through all capacity indices in the candidate set. Compare all recorded total consumption costs Based on the numerical values, the system extracts the term with the smallest calculated value and assigns its corresponding candidate energy storage capacity. As the final output, this output value represents the globally optimal energy storage configuration capacity that balances construction investment and long-term operational benefits, and can be directly used to guide the actual construction of photovoltaic power plants.
[0112] As an example, in each step cycle, in addition to performing discharge compensation or maintaining standby based on predictions and judgments, when the actual active power of the photovoltaic power station exceeds the expected target power generation, resulting in a power generation surplus, the system issues a charging command to the energy storage battery. Simultaneously, the system combines the inverter's maximum input power limit and the battery's current available remaining capacity, accumulating the charging amount corresponding to the aforementioned surplus power to update the energy storage battery's current remaining capacity value.
[0113] This application provides a method for optimizing and controlling the energy storage configuration of photovoltaic power plants based on power prediction. In this application, by acquiring the current power generation loss data of the photovoltaic power plant, and then generating expected power loss data for future time nodes based on the power generation loss data and historical power generation trends, the operating cost data corresponding to each time node and different power levels is determined. This determines the reserved power value corresponding to the remaining power of the energy storage battery. When the difference between the preset grid penalty data and the discharge loss cost is less than the reserved power value, the inverter's discharge compensation command is cut off and it remains in standby mode. Based on the minimum consumption cost generated by the discharge compensation command, the optimal energy storage capacity configuration operation is performed on the photovoltaic power plant. This ensures that when the immediate deterministic discharge revenue is lower than the reserved power value, the discharge command is actively cut off, avoiding ineffective power consumption in the early stages of long-term cloud cover, significantly reducing the total daily penalty amount and equipment depreciation expenses. This avoids situations of over-investment or insufficient energy storage capacity.
[0114] Reference Figure 2 , Figure 2 This is a schematic diagram of the system architecture of the photovoltaic power plant energy storage configuration optimization and control system based on power prediction involved in the embodiments of this application, specifically including: The acquisition module 101 is used to acquire the missing power generation data of the photovoltaic power station at the current moment, and generate the expected missing power data for each future time node based on the missing power generation data and the historical power generation change trend. The calculation module 102 is used to construct a projection grid data consisting of future time nodes and the power nodes of energy storage batteries in photovoltaic power plants, and to calculate the operating cost data of each node in the projection grid data based on preset grid penalty data, discharge loss cost, and expected power loss data. The calculation module 103 is used to perform a reverse extrapolation operation along the time axis from the future to the present moment based on the operating cost data in the extrapolation grid data, and calculate the value of the reserved power corresponding to the remaining power of the energy storage battery. Control module 104 is used to cut off the converter's discharge compensation command and remain in standby mode when the difference between the preset grid penalty data and the discharge loss cost is less than the reserved power value. Configuration module 105 is used to perform optimal energy storage capacity configuration operation for photovoltaic power plants based on the minimum consumption cost generated by the discharge compensation command.
[0115] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.
[0116] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0117] The above are merely preferred embodiments of this application and do not limit the scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the scope of protection of this application.
Claims
1. A method for optimizing and controlling the energy storage configuration of a photovoltaic power plant based on power prediction, characterized in that, The method includes: Obtain the current power generation missing data of the photovoltaic power station, and based on the power generation missing data and the historical power generation change trend, generate the expected power missing data for each future time node; Construct a projection grid data consisting of future time nodes and the power nodes of energy storage batteries in photovoltaic power plants, and calculate the operating cost data of each node in the projection grid data based on preset grid penalty data, discharge loss cost, and the expected power loss data. In the simulation grid data, based on the operating cost data, a reverse simulation operation is performed along the time axis from the future to the current moment to calculate the value of the reserved power corresponding to the remaining power of the energy storage battery; When the difference between the preset grid penalty data and the discharge loss cost is less than the reserved power value, the discharge compensation command of the converter is cut off and the standby mode is maintained. Based on the minimum cost generated by the discharge compensation command, the photovoltaic power station is configured with the optimal energy storage capacity.
2. The photovoltaic power plant energy storage configuration optimization and control method based on power prediction as described in claim 1, characterized in that, The acquisition of missing power generation data of the photovoltaic power station at the current moment includes: Obtain the expected target power generation and actual active power of the photovoltaic power station at the current moment; Based on the difference between the expected target power generation and the actual active power, the missing power generation data is determined.
3. The photovoltaic power plant energy storage configuration optimization and control method based on power prediction as described in claim 1, characterized in that, The process of generating expected power loss data for future time points based on the missing power generation data and historical power generation trends includes: The rate of change of power generation is calculated based on the trend of actual active power change within a preset historical time period. The future preset time period is divided into multiple time nodes, and the time interval between the time nodes is determined; The missing power generation data is subtracted from the product of the power generation change rate and the time interval to generate the expected missing power data for each future time node.
4. The photovoltaic power plant energy storage configuration optimization and control method based on power prediction as described in claim 1, characterized in that, The constructed projection grid data, consisting of future time points and the energy levels of energy storage batteries in the photovoltaic power plant, includes: Obtain the upper and lower limits of the operating capacity of energy storage batteries in photovoltaic power plants; Based on the upper and lower limits of the operating power, the power value of the energy storage battery is divided into multiple power nodes; Using the electrical energy nodes as the vertical axis and the time nodes as the horizontal axis, a projection grid data is constructed.
5. The photovoltaic power plant energy storage configuration optimization and control method based on power prediction as described in claim 4, characterized in that, The method of dividing the energy storage battery's capacity into multiple capacity nodes based on the upper and lower limits of the operating capacity includes: Determine the normal operating capacity range of the energy storage battery; The power range between the upper limit of the operating power and the normal operating power range is defined as the high power range, and the power range between the lower limit of the operating power and the normal operating power range is defined as the low power range. The normal operating power range is divided into multiple power nodes by using an equidistant division method, and the high power range and the low power range are divided into multiple non-equidistant power nodes by using the Chebyshev algorithm.
6. The photovoltaic power plant energy storage configuration optimization and control method based on power prediction as described in claim 1, characterized in that, The calculation of the operating cost data of each node within the simulation grid data, based on preset grid penalty data, discharge loss costs, and expected power loss data, includes: The maximum discharge power of the converter in the photovoltaic power station is obtained, and the power range from zero to the maximum discharge power is divided into multiple discrete discharge powers according to a preset interval. Each node in the inference grid data corresponds to multiple discrete discharge powers. Based on the difference between the expected power missing data and the discrete discharge power, the preset grid penalty data, and the discharge loss cost, the operating cost data of each node in the simulation grid data is calculated.
7. The photovoltaic power plant energy storage configuration optimization and control method based on power prediction as described in claim 6, characterized in that, The operating cost data of each node in the simulation grid data is calculated based on the difference between the expected power loss data and the discrete discharge power, the preset grid penalty data, and the discharge loss cost, including: The difference between the expected power loss data and the discrete discharge power is multiplied by the preset grid penalty data to obtain the penalty data caused by the power deviation. The penalty data is superimposed with the discharge loss cost to obtain the operating cost data of each node in the simulation grid data.
8. The photovoltaic power plant energy storage configuration optimization and control method based on power prediction as described in claim 6, characterized in that, The step of performing a reverse extrapolation operation along the time axis from the future to the present based on the operating cost data to calculate the value of the reserved energy corresponding to the remaining energy of the energy storage battery includes: For any of the discrete discharge powers, the target charge data for the next time node is calculated based on the difference between the charge of the energy storage battery at the current time node and the discharge amount corresponding to the discrete discharge power. Identify the two first charge nodes that are adjacent to the target charge data in the simulation grid data; The node cost of the two first energy nodes is weighted and calculated to obtain the target expected cost. The minimum expected cost value of the current node is obtained by superimposing the target expected cost value with the operating cost data. Based on the minimum expected generation value of the current node, the value of the reserved energy corresponding to the remaining energy of the energy storage battery is calculated.
9. The method for optimizing and controlling the configuration of photovoltaic power plant energy storage based on power prediction as described in claim 8, characterized in that, The calculation of the reserved energy value corresponding to the remaining energy capacity of the energy storage battery based on the minimum expected replacement value of the current node includes: Determine the two adjacent second power nodes of the current node, and extract the remaining power value corresponding to the second power node; The reserved energy value corresponding to the remaining energy of the energy storage battery is calculated based on the difference between the remaining energy values of adjacent second energy nodes and the difference between the minimum expected replacement value of the second energy nodes.
10. The photovoltaic power plant energy storage configuration optimization and control method based on power prediction as described in claim 1, characterized in that, The step of performing optimal energy storage capacity configuration operation on the photovoltaic power station based on the minimum consumption cost generated by the discharge compensation command includes: Obtain data on investment costs, cumulative fines, and equipment loss expenditures for photovoltaic power plants within a preset period; The installed capacity of photovoltaic power plants is divided into multiple candidate energy storage capacities; For any candidate energy storage capacity, the investment cost, cumulative penalty data, and equipment loss expenditure data of the current candidate energy storage capacity are superimposed to obtain the comprehensive consumption cost; The candidate energy storage capacity corresponding to the minimum of the comprehensive consumption cost is selected as the optimal energy storage capacity for configuration.
