An Optimization Method for Collaborative Peak Shaving Strategy of Molten Salt Thermal Storage Coupled Coal-fired Units
By constructing a two-layer optimization model and using rolling feedback correction, the problem of insufficient coordination between coal-fired power units and molten salt thermal storage systems during peak-shaving operation was solved, improving peak-shaving capacity and thermal storage operation effect, and achieving adaptive optimization for actual operating conditions.
Patent Information
- Application Number
- CN202611250402.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-11-17
AI Technical Summary
The existing coal-fired power units and molten salt thermal storage systems lack coordination during peak-shaving operation, making it difficult to adapt to changes in actual operating conditions. This results in the peak-shaving capacity not being fully utilized, and the thermal storage operation effect is affected.
A two-layer optimization model is constructed. The outer layer aims to maximize the cumulative amount of the reduction in the power output of the generating unit, while the inner layer aims to maximize the net heat storage capacity of the thermal storage system. The thermal storage power is corrected based on the operating status of the thermal storage system, and the optimization model is corrected through rolling feedback to generate optimization strategies for peak power deficit and thermal storage system.
It improved the peak-shaving capacity of coal-fired units and the operating effect of thermal storage systems, enhanced the adaptability to actual operating conditions, and ensured the reliability of optimized control results and the feasibility of engineering applications.
Smart Images

Figure CN123121833A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of collaborative peak shaving optimization technology for coal-fired power units, specifically to a collaborative peak shaving strategy optimization method for molten salt thermal storage coupled coal-fired power units. Background Technology
[0002] With the continuous increase in the installed capacity of new energy power generation, photovoltaic (PV) power generation, due to its characteristics of concentrated output and significant fluctuations, typically generates a large amount of electricity during midday. When PV power generation accounts for a high proportion, the net load of the power grid will decrease significantly, and thermal power units will need to undertake more peak-shaving tasks to maintain the balance between power supply and demand and the stability of system operation. Therefore, improving the peak-shaving capacity of coal-fired units during periods of high PV power generation has become an important research direction for power system operation under the background of high proportion of new energy integration.
[0003] Currently, some coal-fired power units are equipped with molten salt thermal storage systems. When the unit load is low or there is surplus steam, some heat is stored in molten salt and released when needed to improve energy efficiency. Existing thermal storage operation methods are usually controlled according to pre-set operating strategies, such as controlling the operation of the thermal storage system based on fixed time periods, fixed charging power, or preset heat storage capacity, and stopping charging after the set heat storage target is reached.
[0004] However, in actual operation, the grid load and renewable energy output exhibit strong time-varying characteristics, with significant differences in peak-shaving demand under different dates and weather conditions. The thermal storage system, however, operates according to a pre-set charging mode, making it difficult to fully adapt to actual operating conditions. When the grid's peak-shaving demand changes, the timing and rate of the thermal storage system's charging deviate from the actual peak-shaving demand of the grid, affecting the overall operating performance of the coal-fired power unit.
[0005] Furthermore, during continuous operation, the working fluid temperature, storage state, and equipment operating conditions of a thermal storage system will all change dynamically, and the storage capacity and operating status will also change accordingly. If a fixed control strategy is still used for thermal storage operation, it is easy for the thermal storage process to deviate from the actual operating state of the equipment, thereby affecting the stability and economy of the thermal storage system.
[0006] Therefore, how to improve the operational coordination between coal-fired power units and molten salt thermal storage systems during actual operation, so that the thermal storage system can better adapt to complex operating conditions, remains a problem that needs further research in this field. Summary of the Invention
[0007] This invention provides an optimization method for coordinated peak-shaving strategies for molten salt thermal storage coupled coal-fired power units. This addresses the technical problem in the prior art where the coordination between coal-fired power units and molten salt thermal storage systems during peak-shaving operation is insufficient, making it difficult to adapt to changes in actual operating conditions, resulting in the inability to fully utilize peak-shaving capacity and affecting the thermal storage operation effect.
[0008] In view of the above problems, the present invention provides an optimization method for collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units, the method comprising: The system obtains the photovoltaic power forecast sequence, grid load forecast sequence and current power output of the generating units during the noon period, constructs the net load forecast curve, and generates the peak-shaving power deficit based on the maximum power reduction demand during the off-peak period and the adjustable capacity margin between the current power output and the minimum technical output. A two-layer optimization model is constructed, and the peak power deficit is set as the endpoint constraint of the power output reduction depth of the outer layer at the valley time. The outer layer of the two-layer optimization model takes maximizing the cumulative amount of the power output reduction depth of the unit as the optimization objective, and the inner layer takes maximizing the net heat storage of the thermal storage system as the optimization objective. The outer layer has a higher priority than the inner layer. Based on the real-time average working fluid temperature and the percentage of remaining usable volume in the thermal storage tank, the thermal storage efficiency attenuation coefficient when the working fluid temperature deviates from the optimal charging temperature is calculated, and the thermal storage power term in the inner objective function is corrected. Solve the modified bi-level optimization model under joint constraints, output the current cycle's thermal storage working fluid pump speed sequence, and generate the first step speed command; The first speed command is executed, and the two-layer optimization model is corrected based on the deviation between the actual and theoretical remaining available volume ratio. The remaining value of the current sequence is used as the initial solution for the next window, and the solution is rolled until the cumulative deviation converges to the preset tolerance.
[0009] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention generates a peak-shaving power deficit and constructs a two-layer optimization model targeting the peak-shaving capacity of coal-fired units and the thermal storage capacity of thermal storage systems. It corrects the thermal storage power based on the operating status of the thermal storage system and uses actual operating results to perform rolling feedback correction on the optimization model. This enables the peak-shaving process of coal-fired units and the molten salt thermal storage process to operate in synergy, improving the adaptability of the peak-shaving strategy to actual operating conditions, enhancing the peak-shaving capacity of coal-fired units and the operating effect of thermal storage systems, while ensuring the reliability of the optimization control results and the feasibility of engineering applications. Attached Figure Description
[0010] Figure 1 This is a schematic diagram of a collaborative peak-shaving strategy optimization method for molten salt thermal storage coupled coal-fired power units provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of the two-layer optimization model structure provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the rolling feedback optimization closed loop provided in an embodiment of the present invention. Detailed Implementation
[0011] This invention provides an optimization method for coordinated peak-shaving strategies for molten salt thermal storage coupled coal-fired power units. It specifically addresses the technical problem in the prior art where the coordination between coal-fired power units and molten salt thermal storage systems is insufficient during peak-shaving operation, making it difficult to adapt to changes in actual operating conditions, resulting in the inability to fully utilize peak-shaving capacity and affecting the thermal storage operation effect.
[0012] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.
[0013] Examples, such as Figure 1 As shown, this invention provides an optimization method for collaborative peak-shaving strategies in molten salt thermal storage coupled coal-fired power units, comprising: S100: Obtain the photovoltaic power forecast sequence, grid load forecast sequence and current power output of the unit during the noon period, construct the net load forecast curve, and generate the peak-shaving power deficit based on the maximum power reduction demand during the off-peak period and the adjustable capacity margin between the current power output and the minimum technical output. Step S100 in the method provided in this embodiment of the invention includes: Obtain the all-weather irradiance prediction sequence output by the numerical weather prediction system, as well as the longitude and latitude of the location of the photovoltaic power station; Based on longitude and latitude, combined with date information, the solar hour angle and solar declination angle of the day are calculated, and the clock time corresponding to the noon solar hour of the day is determined; Obtain the sunrise and sunset times for the day, calculate the time expansion width, and expand the time expansion width to both sides of the clock time corresponding to the noon solar time as the center to form the initial noon window; Extract the irradiance prediction value corresponding to the initial noon window from the all-weather irradiance prediction sequence, calculate the average value of the irradiance prediction value within the window, and use the preset proportion of the average value as the irradiance threshold. From the irradiance prediction values at each time point within the initial noon window, the continuous period when the irradiance prediction value is consistently higher than the irradiance threshold is determined as the noon period.
[0014] Specifically, the all-weather irradiance prediction sequence output by the numerical weather prediction system is first obtained, preferably generated at a sampling interval of 5 minutes or 15 minutes. Simultaneously, the longitude, latitude, and current date information of the photovoltaic power station's geographical location are obtained.
[0015] The 5-minute or 15-minute sampling interval is chosen because this sampling accuracy can accurately reflect the changing trends of photovoltaic irradiance and output during the midday period, while also taking into account the real-time nature of rolling optimization calculations. When the sampling interval is greater than 15 minutes, it is easy to cause errors in identifying the photovoltaic peak time; when the sampling interval is less than 5 minutes, the amount of data increases significantly, while the improvement in identification accuracy is limited. Therefore, a 5-minute or 15-minute sampling interval is preferred.
[0016] Subsequently, the solar declination angle, solar hour angle, and mean time difference for the day are calculated based on the longitude, latitude, and date information.
[0017] Specifically, first, the corresponding annual day is calculated based on the current date, and the solar declination angle is calculated using an empirical model of the sun's position. The solar declination angle characterizes the degree of deviation of the sun from the Earth's equatorial plane. Then, the sunrise angle is calculated based on the local latitude and solar declination angle, and the solar hour angle is determined by combining this with the local longitude. Further, the corresponding mean time difference is calculated based on the annual day; this mean time difference is used to correct the deviation between solar time and clock time caused by the Earth's orbital eccentricity and axial tilt. Finally, longitude correction is performed based on the difference between the local longitude and the standard meridian of the time zone, and the mean time difference correction is added to obtain the solar time. When solar time reaches 12:00, the corresponding clock time is determined as solar noon for that day.
[0018] For example, a photovoltaic power station in Shaanxi Province is located at 108.9° East longitude. After solar time correction, the solar noon on that day corresponds to approximately 12:36 Beijing time. Therefore, 12:36 is determined as the central noon time of that day.
[0019] Next, the sunrise and sunset times for the day are obtained, and the time extension width is calculated. First, the duration of sunshine is calculated. The duration of sunshine is the time difference between sunrise and sunset, i.e.: Duration of sunshine = Sunset time - Sunrise time.
[0020] Furthermore, the time spread width is calculated based on the duration of sunshine. The time spread width is the product of the duration of sunshine, the time spread coefficient, and half of that value, i.e.: Time extension width = Sunshine duration × Time extension coefficient ÷ 2.
[0021] The time extension factor is determined based on the historical operating data of the target photovoltaic power station.
[0022] Specifically, historical operating data of the target photovoltaic power station for at least one year were collected. The ratio between the duration of the period when the actual photovoltaic output power reached more than 90% of the maximum output power of the day and the duration of sunshine on that day was calculated, and a statistical distribution was established. Further analysis revealed that when the time spread factor is 0.60, it can cover more than 95% of the peak photovoltaic power generation periods, while effectively excluding the low irradiance periods in the morning and evening. Therefore, in this embodiment, 0.60 is preferably used as the time spread factor.
[0023] Subsequently, taking the clock time corresponding to solar noon as the center, the time expansion width is extended forward and backward respectively to form the initial noon window.
[0024] The reason for adopting the above method is that the duration of sunshine can comprehensively reflect the changing patterns of solar radiation time in different seasons and at different latitudes. When the days are longer in summer, the initial window automatically expands; when the days are shorter in winter, the initial window automatically shrinks, thus ensuring that the initial window always covers the entire time range where solar power generation may occur.
[0025] For example, if the sun rises at 06:02 and sets at 19:58 on a certain day, the duration of sunshine is approximately 13 hours and 56 minutes. The time spread factor, determined based on historical statistics, is 0.60. Therefore, the time spread width = (19:58 - 06:02) × 0.6 ÷ 2 ≈ 4 hours and 11 minutes. If solar noon corresponds to 12:36 Beijing time, then an initial noon window is formed from 08:25 to 16:47.
[0026] Subsequently, all predicted irradiance values corresponding to the initial noon window are extracted from the all-weather irradiance prediction sequence, and the window-average irradiance is calculated. The window-average irradiance is the ratio of the sum of all predicted irradiance values within the window to the number of sampling points within the window, i.e.: Window average irradiance = sum of all predicted irradiance values within the window ÷ number of sampling points within the window.
[0027] Furthermore, an irradiance threshold is calculated based on the window's average irradiance. The irradiance threshold is the product of the window's average irradiance and a preset ratio, i.e.: Irradiance threshold = window average irradiance × preset ratio.
[0028] The preset ratio in this embodiment is not a fixed value, but is obtained based on the historical operation data of the target photovoltaic power station.
[0029] Specifically, the statistical analysis involves collecting irradiance data from at least one year of continuous historical operation data of the target photovoltaic power station, specifically the data corresponding to when the actual power generation reached 90% or more of the maximum daily output power. The ratio of the irradiance at each sampling time to the average irradiance within the corresponding window is then calculated. That is: Irradiance relative ratio = Irradiance at the current sampling time ÷ Average irradiance of the corresponding date window.
[0030] By conducting offline statistical analysis of historical operating data of the target photovoltaic power station for more than one year, the coverage rate of the effective output period was gradually improved, and the identification accuracy and window length for the corresponding noon period were calculated. Statistical results show that when the coverage rate reaches 95%, further increasing the coverage rate improves the identification accuracy of the noon period by less than 1%, while the window length increases significantly. Therefore, 95% is the preferred coverage rate for the effective output period to balance identification accuracy and subsequent rolling optimization calculation efficiency.
[0031] For example, statistical analysis of the historical operating data of a 200MW photovoltaic power station over two consecutive years revealed that when the relative irradiance ratio reaches 1.23, it can cover more than 95% of the peak photovoltaic power generation periods. Therefore, the preset ratio was determined to be 1.23.
[0032] Subsequently, the predicted irradiance values at each sampling time within the initial noon window are compared sequentially with the irradiance threshold.
[0033] Specifically, when the predicted irradiance values at three or more consecutive sampling times are not lower than the irradiance threshold, the first sampling time that meets the condition is determined as the start time of the noon period; when the predicted irradiance values at three or more consecutive sampling times are lower than the irradiance threshold, the first sampling time that is lower than the irradiance threshold is determined as the end time of the noon period, and the continuous time interval between the two is determined as the final noon period.
[0034] Preferably, to avoid missegmentation of the midday period due to short-term cloud cover, this embodiment stipulates that the duration for which the above conditions are met continuously is no less than three consecutive sampling periods. The three consecutive sampling periods are determined based on statistical analysis of the target photovoltaic power station's historical irradiance data over several years. Statistical results show that irradiance decreases over one or two sampling periods are mostly caused by short-term cloud cover, while the misjudgment rate drops below 5% when the irradiance decreases over three or more sampling periods. Therefore, three consecutive sampling periods are used as the continuous determination condition.
[0035] For example, the average irradiance for a predicted summer day was calculated to be 780 W / m². Based on a preset ratio of 1.23 determined from historical statistics, the irradiance threshold was calculated to be approximately 959 W / m² (780 × 1.23). A comparison of the predicted irradiance values at various times revealed that all sampling points consistently exceeded 959 W / m² between 11:00 AM and 1:00 PM, while values before 10:45 AM and after 1:15 PM were below 959 W / m². Therefore, the period from 11:00 AM to 1:00 PM was ultimately determined as the noon period for that day.
[0036] It should be noted that the above values are for illustrative purposes only and do not constitute a limitation on the present invention.
[0037] Step S100 in the method provided in this embodiment of the invention further includes: The predicted irradiance values for each moment during the noon period are obtained, and the corresponding predicted photovoltaic power values are obtained after irradiance-power conversion. These values are then arranged by time index to form the photovoltaic power prediction sequence for the noon period. Based on the ultra-short-term load forecast data released by the power grid dispatching system, load forecast values that are consistent with the time index of the photovoltaic power forecast sequence are extracted and arranged in chronological order to form the power grid load forecast sequence. The active power output value of the unit is collected in real time using the power transmitter on the output side of the unit, which is used as the current power output of the unit. The photovoltaic power forecast sequence and the grid load forecast sequence are aligned point by point according to the time index. At each time point, the grid load forecast value is subtracted from the photovoltaic power forecast value to calculate the net load forecast value at each time point. The net load forecast curve is generated by combining the forecast values at each time point.
[0038] Specifically, the irradiance prediction value corresponding to each sampling time within the noon period is first obtained, and the corresponding photovoltaic power prediction value is obtained according to the preset irradiance-power conversion relationship. The photovoltaic power prediction values at each time are arranged in chronological order to form the photovoltaic power prediction sequence for the noon period.
[0039] Subsequently, ultra-short-term load forecast data released by the power grid dispatching system is acquired. Preferably, the ultra-short-term load forecast data covers the entire forecast time range corresponding to noon. When the sampling period of the power grid load forecast data is inconsistent with the sampling period of the photovoltaic power forecast sequence at noon, linear interpolation or cubic spline interpolation methods can be used to resample the data with the larger sampling period, so that the two types of forecast data have a consistent time index.
[0040] Preferably, when the sampling period difference is small and the load change is relatively smooth, a linear interpolation method is used; when the prediction time span is long or the load change has significant fluctuations, a cubic spline interpolation method is used to ensure the continuity and smoothness of the load prediction sequence.
[0041] After unifying the sampling period, load forecast values corresponding to the noon period are extracted from the ultra-short-term load forecast data and arranged in chronological order to form a power grid load forecast sequence.
[0042] Simultaneously, a power transmitter installed on the generator outlet side of the coal-fired unit collects the current active power output value of the unit in real time, and uses this active power output value as the unit's current electrical output. Preferably, the data refresh cycle of the power transmitter is consistent with the data acquisition cycle of the unit's automatic power generation control system to ensure that the acquired current electrical output of the unit can accurately reflect the unit's real-time operating status.
[0043] For example, the data acquisition cycle of the existing thermal power unit automatic power generation control system is usually 1 second. This acquisition cycle can meet the requirements of real-time monitoring and automatic control of the unit. Therefore, in this embodiment, the data refresh cycle of the power transmitter is preferably set to 1 second.
[0044] Since this embodiment adopts a rolling optimization control method, the current power output of the unit is used as the initial operating state of the current rolling optimization window. After the optimization of this window is completed, the real-time power output of the unit is collected again in the next rolling window and the initial optimization state is updated. Therefore, using the current real-time power output of the unit as the optimization starting value can accurately reflect the actual operating conditions of the unit.
[0045] Subsequently, the photovoltaic power forecast sequence and the grid load forecast sequence are aligned using time indices. Specifically, the forecast time is used as a unified time index, and the corresponding grid load forecast value and photovoltaic power forecast value are read at the same time. The net load forecast value is the difference between the corresponding grid load forecast value and the corresponding photovoltaic power forecast value, i.e.: Net load forecast = Grid load forecast at the corresponding time - Photovoltaic power forecast at the corresponding time.
[0046] After calculating the net load forecast values for all sampling times, the net load forecast curves are formed by combining them in chronological order.
[0047] The reason for using the above calculation method is that the grid load forecast reflects the total electricity demand of users, while photovoltaic power generation can directly bear part of the load. Therefore, the actual load that coal-fired units need to bear is the remaining load after deducting photovoltaic power generation. By constructing a net load forecast curve, the peak-shaving pressure faced by coal-fired units during the midday peak of photovoltaic power generation can be reflected more accurately, providing a basis for subsequently determining the maximum power reduction requirement.
[0048] For example, in a certain region, the predicted noon time is from 11:00 to 13:00, with a sampling period of 15 minutes, resulting in 9 predicted time points. At 11:30, the predicted grid load is 8200MW, and the predicted photovoltaic power is 2350MW, so the predicted net load at this time is 8200 - 2350 = 5850MW; at 12:00, the predicted grid load is 8150MW, and the predicted photovoltaic power is 2480MW, so the predicted net load is 8150 - 2480 = 5670MW; at 12:30, the predicted grid load is 8100MW, and the predicted photovoltaic power is 2420MW, so the predicted net load is 8100 - 2420 = 5680MW. After calculating all predicted time points sequentially, the corresponding net load prediction curve is generated. The current active power output value of the generating units is simultaneously acquired. The current real-time power output of the unit is 650MW, and this value is used as the initial power output input for the subsequent adjustable capacity margin of the computer unit and the peak power deficit.
[0049] It should be noted that the above values are for illustrative purposes only and do not constitute a limitation on the present invention.
[0050] Step S100 in the method provided in this embodiment of the invention further includes: The predicted irradiance values at various times during the noon period were obtained, and the standard irradiance under standard test conditions and the installed capacity of the photovoltaic power station were used as the conversion benchmark. Divide the predicted irradiance value at each time point by the standard irradiance value to obtain the irradiance ratio, and combine it with the installed capacity to obtain the initial DC power value; The rated operating temperature and temperature power coefficient of the photovoltaic module are obtained. The predicted irradiance value at each time moment is substituted into the definition of the rated operating temperature to calculate the module operating temperature at each time moment. The difference between the component's operating temperature and the standard temperature under standard test conditions is calculated, and combined with the temperature power coefficient, a temperature correction coefficient is obtained. Multiply the initial DC power value by the temperature correction factor, and then combine it with the overall efficiency coefficient of the photovoltaic power station to obtain the predicted photovoltaic power value at each time. The comprehensive efficiency coefficient is the comprehensive reduction factor of the energy loss of a photovoltaic power plant from the DC side output to the AC grid connection point. The total energy loss includes at least the DC-AC conversion loss, the collector line loss, the inverter loss, and the dust shading loss on the surface of the photovoltaic modules.
[0051] Specifically, the predicted irradiance values for each moment during the noon period are first obtained, and then the installed capacity of the photovoltaic power station is obtained. The installed capacity of the photovoltaic power station is preferably the rated DC-side installed capacity of the photovoltaic array; when the installed capacity published by the power station is the rated AC-side capacity, the corresponding DC-side installed capacity can be calculated based on the rated capacity of the inverter to ensure consistency in power conversion.
[0052] The standard testing conditions are the module power calibration conditions uniformly adopted in the photovoltaic industry, including: standard irradiance of 1000W / m², standard module temperature of 25℃, and solar spectral distribution with air quality AM1.5. The module output power measured under the above conditions is used as the rated power on the module nameplate. Therefore, this embodiment uses a standard irradiance of 1000W / m² as the photovoltaic power conversion benchmark.
[0053] Subsequently, the irradiance ratio is calculated based on the predicted irradiance values at each time point. The irradiance ratio is the ratio of the predicted irradiance value at the current time to the standard irradiance, i.e.: Irradiance ratio = Current irradiance prediction value ÷ 1000.
[0054] Next, the initial DC power value is calculated based on the irradiance ratio. The initial DC power value is the product of the irradiance ratio and the installed capacity of the photovoltaic power station, i.e.: Initial DC power value = irradiance ratio × photovoltaic power station installed capacity.
[0055] The reason for using the above calculation method is that, under standard test conditions, the output power of photovoltaic modules is approximately linearly related to the received irradiance. Therefore, the installed capacity can be linearly converted using the ratio of irradiance to standard irradiance, thereby obtaining the theoretical DC output power without considering the effects of temperature and system losses.
[0056] Subsequently, to improve prediction accuracy, the impact of component temperature on output power was further considered.
[0057] Specifically, the rated operating temperature and temperature power coefficient are obtained from the photovoltaic module nameplate parameters. Both the rated operating temperature and temperature power coefficient are obtained by the photovoltaic module manufacturer according to national standards before shipment and are fixed performance parameters of the module. The rated operating temperature reflects the operating temperature of the module under typical operating conditions; the temperature power coefficient reflects the percentage change in output power for every 1°C increase in module temperature, typically ranging from -0.0035°C to -0.0045°C for monocrystalline silicon modules.
[0058] Simultaneously, the corresponding ambient temperature prediction value output by the numerical weather prediction system is acquired.
[0059] Subsequently, the module operating temperature is calculated according to the definition of rated operating temperature. The module operating temperature is the actual temperature of the module after the combined effects of ambient temperature and irradiance temperature rise, i.e.: Component operating temperature = ambient temperature + (current predicted irradiance value ÷ 800) × (rated operating temperature - 20).
[0060] The ambient temperature is the predicted value of the ambient temperature at the corresponding time output by the numerical weather prediction system; 800 W / m² is the standard reference irradiance corresponding to the rated operating temperature; and 20℃ is the standard reference ambient temperature corresponding to the rated operating temperature. The above calculation method adopts an empirical model of rated operating temperature, which is a commonly used method for estimating the module operating temperature in the field of photovoltaic power prediction.
[0061] The reason for using the above calculation method is that, in addition to being affected by ambient temperature, photovoltaic modules also experience temperature rise due to solar irradiance during operation, and the temperature rise is basically linearly related to the irradiance. Therefore, the actual operating temperature of the module can be estimated more accurately.
[0062] Subsequently, a temperature correction factor is calculated based on the module's operating temperature. This temperature correction factor is a power correction factor calculated jointly from the standard temperature correction value and the module's temperature change, i.e.: Temperature correction factor = 1 + temperature power factor × (module operating temperature - 25).
[0063] Here, 25℃ is the standard module temperature under standard test conditions. Since the temperature power coefficient of most crystalline silicon photovoltaic modules is negative, the temperature correction factor is less than 1 when the module operating temperature is higher than 25℃; and close to 1 when the module operating temperature is close to 25℃, thus accurately reflecting the output power degradation caused by temperature.
[0064] Furthermore, to make the predicted results closer to the actual output power at the AC grid connection point, this embodiment introduces a comprehensive efficiency coefficient. The comprehensive efficiency coefficient is a comprehensive reduction factor for the energy loss along the entire path from the DC side of the photovoltaic power plant to the AC grid connection point. This total energy loss includes at least DC-AC conversion losses, inverter losses, collector line losses, photovoltaic module dust accumulation and shading losses, box-type transformer losses, and module mismatch losses.
[0065] The overall efficiency coefficient is obtained based on historical operating data of the photovoltaic power plant. Specifically, it is calculated by comparing the actual AC output power with the theoretical DC output power under sunny conditions for at least three consecutive months, and the average value is taken as the overall efficiency coefficient. The statistical period of at least three consecutive months can cover different ambient temperatures and operating conditions, while reducing the impact of short-term weather fluctuations on the statistical results, so that the overall efficiency coefficient can more accurately reflect the long-term operating efficiency of the power plant.
[0066] For example, after collecting historical operating data for the past six months, the average ratio between the actual output power on the AC side and the theoretical DC output power of a 100MW photovoltaic power plant is about 0.95. Therefore, the comprehensive efficiency coefficient is determined to be 0.95.
[0067] Subsequently, the final photovoltaic power prediction value is calculated based on the initial DC power value, the temperature correction factor, and the overall efficiency factor. The photovoltaic power prediction value is the product of the initial DC power value, the temperature correction factor, and the overall efficiency factor, i.e.: Photovoltaic power prediction = Initial DC power value × Temperature correction factor × Overall efficiency factor.
[0068] For example, if the predicted irradiance at a certain moment is 920 W / m², and the installed capacity of the photovoltaic power station is 200 MW, then the irradiance ratio = 920 ÷ 1000 = 0.92, corresponding to an initial DC power value of 0.92 × 200 = 184 MW; the predicted ambient temperature is 34℃, and the rated operating temperature of the modules is 45℃, then the calculated module operating temperature = 34 + (920 ÷ 800) × (45 - 20) = 62.75℃; the module temperature power coefficient is -0.0040 / ℃, then the temperature correction coefficient = 1 + (-0.0040) × (62.75 - 25) = 0.849; the comprehensive efficiency coefficient is determined to be 0.95 based on historical statistics. The final calculated predicted photovoltaic power value is approximately 184 × 0.849 × 0.95 = 148.4 MW.
[0069] It should be noted that the above values are for illustrative purposes only and do not constitute a limitation on the present invention.
[0070] Step S100 in the method provided in this embodiment of the invention further includes: The trough period is located from the net load forecast curve, and the time corresponding to the minimum net load forecast value within the trough period is identified as the trough time. The trough period is determined based on the trend of the net load forecast value during the noon period. Extract the predicted net load value at the off-peak time as the off-peak net load value, and calculate the difference between the off-peak net load value and the current power output of the unit as the maximum power demand to be reduced during the off-peak period. The minimum technical output of the unit is obtained, and the difference between the current electrical output of the computer unit and the minimum technical output is used as the adjustable capacity margin. The maximum power demand during off-peak hours is compared with the adjustable capacity margin, and the smaller of the two is taken as the peak-shaving power deficit.
[0071] Specifically, using the noon period as the search scope, the net load forecast values corresponding to each sampling time in the net load forecast curve are traversed in chronological order to identify the minimum value and record the corresponding time as a candidate valley value time. When multiple identical minimum values exist, the center time of the longest continuous valley interval is preferably selected as the valley value time; if the durations are the same, the earliest occurrence time is selected as the valley value time.
[0072] Subsequently, using the trough moment as the center, the net load forecast curve is searched on both sides to determine the boundary of the trough period.
[0073] Specifically, starting from the trough moment, the net load forecast curve is traversed forward moment by moment, and the first-order difference between the net load forecast values of two adjacent sampling moments is calculated. First-order difference value = current net load forecast value - previous net load forecast value.
[0074] When the first-order difference value of two consecutive sampling periods is less than zero, it indicates that the net load is continuously decreasing; when the first-order difference value of two consecutive sampling periods changes from less than zero to greater than or equal to zero for the first time, it is considered that the downward trend of the net load has ended, and the corresponding change point is determined as the starting time of the trough period.
[0075] Similarly, by iterating through the net load prediction curve backward from the trough moment, when the first difference value of two consecutive sampling periods is greater than zero, it indicates that the net load is continuing to rise; when the first difference value of two consecutive sampling periods changes from greater than zero to less than or equal to zero for the first time, it is considered that the net load rise trend has ended, and the corresponding change point is determined as the end time of the trough period.
[0076] The reason for using two consecutive sampling periods as the boundary determination condition is that a single sampling point is easily affected by prediction errors or short-term load fluctuations, while two consecutive sampling periods can effectively filter out random disturbances and improve the stability of valley boundary identification.
[0077] After determining the off-peak hours, the predicted net load value corresponding to the off-peak hours is extracted as the off-peak net load value. The maximum power reduction demand during the off-peak period is the difference between the unit's current power output and the off-peak net load value, i.e.: Maximum power demand reduction during off-peak hours = Current power output of the unit - Off-peak net load value.
[0078] After calculating the maximum power demand to be reduced during off-peak hours, it is further determined whether the calculation result is less than zero. If the maximum power demand to be reduced during off-peak hours is less than zero, it is corrected to zero, indicating that the current unit output is no higher than the off-peak net load, and there is no need to perform peak shaving.
[0079] Subsequently, the minimum technical output of the unit is obtained. The minimum technical output is the lowest output value that a coal-fired unit can sustain continuously while ensuring stable boiler combustion, safe turbine operation, environmental emissions, and normal operation of auxiliary equipment. This parameter is typically determined by unit design parameters, performance test results, or power plant operating procedures and is stored in the unit's control system.
[0080] For example, for a 600MW supercritical coal-fired unit, after deep peak-shaving retrofit, its minimum technical output can typically reach 25% to 35% of the rated capacity; for unretrofitted units, it is generally 40% to 50%. Therefore, this embodiment directly uses the minimum technical output parameters configured in the unit's control system as the basis for calculation.
[0081] Further calculate the adjustable capacity margin. The adjustable capacity margin is the difference between the unit's current electrical output and its minimum technical output, i.e.: Adjustable capacity margin = Current electrical output of the unit - Minimum technical output of the unit.
[0082] Adjustable capacity margin reflects the maximum output capacity that the unit can continue to reduce while meeting safe operating constraints.
[0083] Finally, the peak-shaving power deficit is determined by combining the maximum power demand during off-peak periods with the adjustable capacity margin.
[0084] Among them, the peak-shaving power deficit is the smaller value between the maximum downward power demand during the off-peak period and the adjustable capacity margin, that is: Peak power deficit = Min (maximum power demand to be reduced during off-peak hours, adjustable capacity margin).
[0085] The reason for using the above calculation method is that when the maximum power reduction demand exceeds the unit's adjustable capacity margin, the unit can no longer reduce its output, and the actual peak-shaving capacity is limited by the adjustable capacity margin. When the adjustable capacity margin is greater than the maximum power reduction demand, the unit can meet all peak-shaving demands. Therefore, the actual demand should be used as the peak-shaving target. By taking the smaller value of the two, it can be ensured that the peak-shaving strategy generated by the subsequent two-layer optimization model always meets the unit's safe operation constraints and will not require the unit's output to be lower than the minimum technical output.
[0086] For example, a 600MW coal-fired unit currently has a real-time power output of 520MW. Based on the net load forecast curve, the predicted net load for off-peak hours is 365MW. Therefore, the maximum power demand to be reduced during off-peak hours is 520 - 365 = 155MW. Meanwhile, the minimum technical output recorded by the unit's control system is 180MW. Therefore, the adjustable capacity margin is 520 - 180 = 340MW. Since 155MW is less than 340MW, the peak-shaving power deficit is Min(155, 340) = 155MW.
[0087] It should be noted that the above values are for illustrative purposes only and do not constitute a limitation on the present invention.
[0088] In summary, this step first adaptively identifies the noon time period based on the irradiance prediction data output by the numerical weather prediction system and the geographical location information of the photovoltaic power station. Then, by combining component temperature correction and system comprehensive efficiency correction, a photovoltaic power prediction sequence for the noon time period is generated and time-aligned with the grid load prediction sequence to construct a net load prediction curve. Further, based on the net load prediction curve, the off-peak period and off-peak time are identified, the maximum power reduction demand during the off-peak period is calculated, and the adjustable capacity margin is determined by combining the minimum technical output of the unit. Finally, a peak-shaving power deficit that meets the unit's safe operation constraints is generated, providing accurate boundary conditions and optimization objectives for the subsequent establishment of a two-layer optimization model and rolling peak-shaving control.
[0089] S200: Construct a two-layer optimization model and set the peak power deficit as the endpoint constraint of the power output reduction depth of the outer layer at the valley time. The outer layer of the two-layer optimization model takes maximizing the cumulative amount of the power output reduction depth of the unit as the optimization objective, while the inner layer takes maximizing the net heat storage of the thermal storage system as the optimization objective. The outer layer has a higher priority than the inner layer. Step S200 in the method provided in this embodiment of the invention includes: Using the depth of power output reduction at each discrete sampling time during the noon period as the outer decision variable, and setting the peak power deficit as the lower limit constraint of the reduction depth at the valley time, an outer objective function is established to maximize the cumulative amount of power output reduction during the entire noon period. Using the pump speed of the thermal storage medium at each sampling time as the inner layer decision variable, and based on the inherent characteristic relationship between the pump speed of the thermal storage medium and the thermal storage power, as well as the thermal storage efficiency decay coefficient, an inner layer objective function is established to maximize the cumulative amount of thermal storage power after correction by the thermal storage efficiency decay coefficient during the noon period. Establish a coupled inequality constraint that the power reduction of the unit's electrical output at any given moment is not less than the charging power of the thermal storage system; Under the combined constraint conditions of outer constraints, inner constraints, and coupled inequality constraints, a layered solution mechanism prioritizing outer constraints is adopted for solving the problem, including: Solve the outer optimization problem that satisfies the lower limit constraint of the endpoint to obtain the optimal power output reduction depth sequence. Use the power reduction value at each time point in the optimal power output reduction depth sequence as the upper limit value of the thermal storage power. Solve the inner-layer optimization problem with the pump speed of the thermal storage medium as the variable. Under the premise that the thermal storage power does not exceed the corresponding upper limit at each time, obtain and output the optimal pump speed sequence that maximizes the cumulative amount of the corrected thermal storage power.
[0090] like Figure 2As shown, specifically, the depth of power output reduction at each discrete sampling time within the noon period is first used as the decision variable of the outer optimization model. The power output reduction depth represents the decrease in the current power output of the unit relative to the actual operating power output at the corresponding sampling time, and is used to characterize the peak-shaving capacity released by the unit participating in peak shaving.
[0091] Subsequently, the peak-shaving power deficit obtained in step S100 is used as the lower limit constraint for the depth of power output reduction during off-peak hours. Specifically, the lower limit constraint is that the depth of power output reduction during off-peak hours is not less than the peak-shaving power deficit, i.e.: The depth of power output reduction during off-peak hours is greater than or equal to the peak power deficit.
[0092] The reason for adopting the above constraint method is that the trough time corresponds to the time when the net load is the lowest during the entire noon period, which is also the time when the power grid's peak-shaving demand is the greatest. It is only necessary to ensure that the minimum peak-shaving requirement is met at this time, without limiting the specific change form of the entire peak-shaving trajectory, thereby preserving sufficient optimization freedom for the optimization model.
[0093] Then, the outer objective function is established. The cumulative decrease in electrical output depth is the sum of the decreases in electrical output depth of the units at each sampling time during the noon period, i.e.: Cumulative amount of power output reduction depth = ∑ (power output reduction depth at each sampling time × sampling period).
[0094] When the sampling period is the same at each sampling time, the sampling period can be regarded as a constant, and the objective function can be equivalently expressed as the sum of the power output reduction depth at each sampling time.
[0095] The goal of the outer layer optimization is to maximize the cumulative amount of power output reduction, thereby increasing the peak-shaving capacity released by coal-fired units during the entire solar power generation period and improving the capacity for renewable energy consumption.
[0096] Subsequently, an inner-layer objective function is established. Specifically, the rotational speed of the thermal storage medium pump at each sampling time is used as the inner-layer decision variable. Since the rotational speed of the thermal storage medium pump directly determines the mass of working fluid entering the molten salt thermal storage system per unit time, it can further determine the thermal storage power.
[0097] Preferably, based on historical operating data or performance curves of the thermal storage working fluid pump collected during the commissioning phase of the thermal storage system, a correspondence between the pump speed and the working fluid flow rate is established. This is combined with the specific heat capacity of the working fluid, the temperature difference between the inlet and outlet of the working fluid, and its density to calculate the thermal storage power corresponding to different flow rates, forming a "pump speed - thermal storage power" correspondence table. During operation, for any pump speed, the corresponding thermal storage power is obtained through table lookup or linear interpolation.
[0098] The thermal storage power is further corrected by combining the thermal storage efficiency attenuation coefficient calculated in step S300. Among other things, Corrected thermal storage power = thermal storage power × thermal storage efficiency attenuation coefficient.
[0099] Furthermore, the cumulative thermal storage power is calculated using a discrete integral form, namely: Cumulative thermal storage power = ∑ (corrected thermal storage power at each sampling time × sampling period).
[0100] When the sampling period is fixed, it can be equivalently represented as the sum of the corrected thermal storage power at each sampling time.
[0101] The goal of inner layer optimization is to maximize the cumulative thermal storage power, thereby ensuring the maximum effective heat storage per unit of peak-shaving capacity.
[0102] Subsequently, in order to ensure the energy balance between the coal-fired unit and the thermal storage system, this embodiment establishes a coupling inequality constraint.
[0103] Specifically, the charging power of the thermal storage system is uniformly converted into the corresponding extraction steam power, and the reduced power output of the unit is converted into the corresponding releaseable heat power based on the unit's heat rate. Coupled constraints are established under a unified heat power unit. Among these constraints: The unit's electrical output power reduction is greater than or equal to the thermal storage system's charging power.
[0104] The reason for adopting the above constraints is that the heat required for the thermal storage system to charge comes from the peak-shaving capacity released by the coal-fired power unit. When the charging power of the thermal storage system exceeds the actual power reduction of the unit, it will lead to insufficient heat source. Therefore, it is necessary to ensure that the charging power of the thermal storage system does not exceed the peak-shaving capacity that the unit can actually provide at any sampling time.
[0105] The modified bi-layer optimization model is solved under the conditions of satisfying the outer layer constraints, inner layer constraints, and coupling constraints.
[0106] Specifically, the outer optimization model is solved first. Under the condition of satisfying the lower limit constraint of the power output reduction depth at the trough time, the maximum cumulative amount of power output reduction depth is taken as the optimization objective, and the optimal power output reduction depth sequence during the noon period is obtained.
[0107] Subsequently, the power of power reduction at each sampling time corresponding to the optimal power reduction depth sequence is taken as the upper limit of the allowable charging power of the thermal storage system at the corresponding time. Under the condition that the thermal storage power at each sampling time does not exceed the upper limit of the corresponding allowable charging power, the inner layer optimization model is further solved, and the maximum cumulative amount of the corrected thermal storage power is taken as the optimization objective to obtain the corresponding optimal thermal storage working fluid pump speed sequence.
[0108] In this embodiment, the two-level optimization model can be solved using a sequential quadratic programming algorithm, an interior point method, or a model predictive control rolling optimization algorithm; when the optimization variables are discretized, a mixed integer linear programming solver can also be used to complete the solution, and this invention does not limit this.
[0109] For example, a 600MW coal-fired unit divides the midday period into nine 15-minute sampling times. Step S100 calculates a peak power deficit of 150MW, therefore the power output reduction depth during off-peak hours should be at least 150MW. After outer-layer optimization, the power output reduction depth sequence is 40MW, 65MW, 90MW, 120MW, 150MW, 145MW, 120MW, 85MW, and 50MW, with a cumulative power output reduction depth of 865MW. Subsequently, the reduced power at each time point is used as the upper limit for thermal storage power. Assuming that the thermal storage power calculated from the pump performance curve at a certain time is 118MW and the thermal storage efficiency decay coefficient is 0.93, the corrected thermal storage power = 118 × 0.93 = 109.74MW. Since 109.74MW is less than the allowed upper limit of 120MW for thermal storage power at that time, the coupling constraint is satisfied, and it can be used as an effective solution for inner-layer optimization.
[0110] It should be noted that the above values are for illustrative purposes only and do not constitute a limitation on the present invention.
[0111] In summary, this step constructs a two-layer optimization model with the depth of power output reduction of the unit as the outer decision variable and the speed of the thermal storage working fluid pump as the inner decision variable. On the basis of satisfying the peak-shaving power deficit constraint and the energy coupling constraint of the unit-thermal storage system, it prioritizes ensuring that the coal-fired unit completes the grid peak-shaving target, and further optimizes the thermal storage system charging process, so as to achieve synergistic optimization of the peak-shaving capacity of the coal-fired unit and the thermal storage capacity, improve the new energy consumption capacity and the overall operating efficiency of the thermal storage system.
[0112] S300: Based on the real-time average working fluid temperature and the percentage of remaining usable volume in the thermal storage tank, calculate the thermal storage efficiency attenuation coefficient when the working fluid temperature deviates from the optimal charging temperature, and correct the thermal storage power term in the inner objective function. Step S300 in the method provided in this embodiment of the invention includes: Obtain the optimal charging temperature and the maximum allowable operating temperature of the thermal storage tank, and combine the real-time average working fluid temperature and the percentage of remaining usable volume in the thermal storage tank to calculate the deviation of the current working fluid temperature from the optimal charging temperature. The real-time average temperature of the working fluid in the thermal storage tank is obtained by collecting and weighting the data from temperature measuring points deployed at different heights of the tank. The remaining usable volume percentage is obtained by the ratio of the current molten salt volume measured by the tank's continuous level gauge to the tank's rated usable volume. Based on the ratio of the absolute value of the deviation to the optimal charging temperature, and combined with the nonlinear decay relationship between the deviation and the charging efficiency, a thermal storage efficiency decay coefficient is generated, wherein the thermal storage efficiency decay coefficient is a positive number not greater than 1. Multiply the thermal storage power term in the inner objective function by the thermal storage efficiency decay coefficient to obtain the effective thermal storage power term after temperature correction, and substitute it into the inner objective function to replace the thermal storage power term before correction. The remaining available volume of the thermal storage tank is determined by the proportion of remaining available volume, and the maximum heat storage capacity corresponding to the current remaining available volume is used as the upper limit constraint of the integral value of the thermal storage power term in the inner objective function.
[0113] Specifically, the optimal charging temperature of the thermal storage tank and the maximum allowable operating temperature of the tank body are first obtained.
[0114] The optimal charging temperature is obtained from the thermal storage system design specification or the heat exchanger performance parameter table. It is the design operating temperature corresponding to the minimum heat exchange end difference and the highest overall heat exchange efficiency when the thermal storage medium and steam exchange heat. This parameter is determined during the thermal storage system design phase and is a fixed design parameter.
[0115] The maximum allowable operating temperature of the tank is obtained from the nameplate of the thermal storage tank or the pressure vessel design documents. It represents the highest working fluid temperature that the thermal storage tank can operate continuously for extended periods under the maximum design pressure. This parameter is determined based on the allowable stress of the tank material, wall thickness, and pressure vessel design standards, and is used to limit the safe operating range of the thermal storage tank.
[0116] Subsequently, the average temperature of the working fluid in the thermal storage tank is acquired in real time.
[0117] Specifically, multiple temperature measuring points are arranged along the height direction inside the heat storage tank, such as at the bottom, middle and top of the tank, and the temperature of the molten salt at each measuring point is collected in real time.
[0118] Because temperature stratification exists inside the thermal storage tank during the thermal storage process, and the working fluid temperature varies at different heights, this embodiment uses a weighted average method to calculate the real-time average working fluid temperature of the thermal storage tank. Among other things, The real-time average temperature of the working fluid is the sum of the products of the temperature at each measuring point and its corresponding weight.
[0119] The weight of each measuring point is determined based on the volume distribution of the thermal storage area represented by the corresponding measuring point. When the volume of the area corresponding to each measuring point is basically the same, the same weight is used for each measuring point; when there are differences in the volume of areas at different heights, the weight of each measuring point is determined according to the proportion of the volume of the corresponding area to the effective volume of the thermal storage tank, so as to ensure that the calculated real-time average temperature of the working fluid can more accurately reflect the overall thermal state of the entire thermal storage tank.
[0120] For example, three temperature measuring points are set up at the top, middle and bottom of the thermal storage tank, respectively. The volume of the area corresponding to each measuring point is basically the same, so the weight of each measuring point is 1 / 3. If the real-time temperature collected by the three measuring points is 548℃, 553℃ and 559℃ respectively, then the real-time average temperature of the working fluid = 548×1 / 3 + 553×1 / 3 + 559×1 / 3 = 553.3℃.
[0121] Subsequently, the deviation of the current working fluid temperature from the optimal charging temperature is calculated based on the real-time average working fluid temperature. Working fluid temperature deviation = Real-time average working fluid temperature - Optimal charging temperature.
[0122] Furthermore, the current safe temperature margin is calculated based on the tank's maximum permissible operating temperature. Among these, Safety temperature margin = maximum allowable operating temperature - real-time average working fluid temperature.
[0123] The safety temperature margin characterizes the safe temperature rise range remaining before the maximum allowable operating temperature of the thermal storage tank. When the safety temperature margin gradually decreases, it indicates that the thermal storage tank is gradually approaching its safe operating limit, and subsequent heat charging capacity should be limited to ensure the safe operation of the thermal storage equipment.
[0124] In addition, this embodiment also obtains the percentage of remaining available volume of the thermal storage tank in real time.
[0125] Specifically, the current molten salt level is measured in real time using a continuous level gauge installed on the thermal storage tank, and the current molten salt volume is calculated based on the tank's geometry. Remaining usable volume percentage = (rated usable volume - current molten salt volume) ÷ rated usable volume.
[0126] The rated usable volume of the thermal storage tank is determined by the design documents of the thermal storage tank, and is the maximum allowable thermal storage volume after deducting the safety level and expansion margin.
[0127] Furthermore, the current remaining available volume is calculated based on the percentage of remaining available volume. Among these, Current remaining available volume = percentage of remaining available volume × rated available volume.
[0128] Subsequently, based on the current remaining available volume, the density of the heat storage medium, the specific heat capacity of the medium, and the allowable temperature rise during heat charging, the maximum heat that the remaining space can store is calculated. Maximum heat storage capacity = Current remaining usable volume × Density of heat storage medium × Specific heat capacity of medium × Allowable temperature rise during heat charging.
[0129] The density and specific heat capacity of the thermal storage medium are obtained from the thermal storage medium physical property parameter table; the allowable charging temperature rise is the temperature difference between the maximum allowable operating temperature and the current real-time average temperature of the medium, i.e.: Permissible heat charge temperature rise = maximum permissible operating temperature - real-time average working fluid temperature.
[0130] Subsequently, the maximum storable heat capacity was used as the upper limit constraint for the cumulative heat storage capacity of the inner layer optimization model.
[0131] Specifically, when discrete sampling optimization is used, the cumulative heat storage is expressed as: Cumulative heat storage = Sum of the products of heat storage power at each sampling time and the sampling period. The optimization process satisfies: Cumulative heat storage ≤ maximum heat storage capacity.
[0132] The above method can prevent the cumulative heat storage obtained from optimization from exceeding the current remaining heat storage capacity of the heat storage tank, thereby ensuring that the optimization results meet the actual operating constraints of the heat storage system.
[0133] Furthermore, the thermal storage efficiency attenuation coefficient calculated in step S300 is applied to the thermal storage power term in the inner layer objective function. Wherein, Effective thermal storage power term = thermal storage power term × thermal storage efficiency attenuation coefficient.
[0134] Subsequently, the effective thermal storage power term is used to replace the original thermal storage power term in the inner objective function optimization calculation, so that the optimization objective can comprehensively consider the current thermal state of the thermal storage tank, the charging efficiency and the remaining thermal storage capacity, thereby improving the economic efficiency and safety of the thermal storage system operation.
[0135] For example, the design document of a molten salt thermal storage system specifies an optimal charging temperature of 565℃, a maximum allowable operating temperature of 600℃ for the storage tank, and a rated usable volume of 12000 m³. Currently, the molten salt temperatures measured at the top, middle, and bottom are 548℃, 553℃, and 559℃ respectively. Therefore, the real-time average working fluid temperature is 553.3℃; the corresponding working fluid temperature deviation is 553.3 - 565 = -11.7℃; and the safety temperature margin is 600 - 553.3 = 46.7℃. Simultaneously, the continuous level gauge measures the current molten salt volume as 9300 m³. Therefore, the remaining usable volume percentage is (12000 - 9300) ÷ 12000 = 0.225. Further, the current remaining usable volume is 0.225 × 12000 = 2700 m³. If the density of the molten salt is 1800 kg / m³, the specific heat capacity is 1.5 kJ / (kg·℃), and the allowable temperature rise during charging is 46.7℃, then the calculated maximum heat storage capacity is approximately 3.40 × 10⁻⁶. 8 kJ. During the optimization process, the cumulative heat storage never exceeds the maximum storable heat storage capacity to ensure the safe and stable operation of the thermal storage system.
[0136] It should be noted that the above values are for illustrative purposes only and do not constitute a limitation on the present invention.
[0137] Step S300 in the method provided in this embodiment of the invention further includes: Calculate the ratio of the absolute value of the deviation to the optimal charging temperature to obtain the dimensionless relative temperature deviation. Calculate the squared value of the deviation based on the dimensionless relative temperature deviation. The initial attenuation factor is obtained by performing a nonlinear transformation based on the squared deviation value. The nonlinear transformation causes the initial attenuation factor to decrease monotonically as the squared deviation value increases. By applying upper and lower limit truncation constraints to the initial attenuation factor, the thermal storage efficiency attenuation coefficient is obtained, where the thermal storage efficiency attenuation coefficient is a positive number that is not less than 0 and not greater than 1.
[0138] Specifically, firstly, based on the working fluid temperature deviation obtained in step S300, the ratio between the absolute value of the deviation and the optimal charging temperature is calculated to eliminate the influence of temperature dimensions and obtain the dimensionless relative temperature deviation. The dimensionless relative temperature deviation is the ratio of the absolute value of the working fluid temperature deviation to the optimal charging temperature, i.e.: Dimensionless relative temperature deviation = |working fluid temperature deviation| ÷ optimal charging temperature.
[0139] The reason for using the above calculation method is that the optimal charging temperature may differ among different molten salt thermal storage systems. If the temperature deviation is used directly for calculation, it is difficult to unify the evaluation standard between different systems. By normalizing using the optimal charging temperature, the degree of deviation can be converted into a dimensionless parameter, improving the applicability among different thermal storage systems.
[0140] Subsequently, the squared value of the deviation is calculated based on the dimensionless relative temperature deviation. That is: Deviation square value = dimensionless temperature relative deviation².
[0141] The reason for using the squared treatment is that when the temperature deviation is small, the degree of deviation is appropriately smoothed to make the change in thermal storage efficiency more gradual; while when the temperature deviation gradually increases, the growth rate of the squared term accelerates, which can enhance the sensitivity to larger temperature deviations and make the subsequent decay process more consistent with the actual operating characteristics of the thermal storage system.
[0142] Furthermore, a nonlinear transformation is performed on the squared deviation value to obtain the initial attenuation factor. Wherein, Initial attenuation factor = 1 ÷ (1 + deviation squared value).
[0143] After adopting the above-mentioned square inverse function, when the working fluid temperature is close to the optimal charging temperature, the deviation square value approaches zero, and the initial attenuation factor approaches 1. As the temperature deviation gradually increases, the initial attenuation factor changes continuously in a monotonically decreasing manner without abrupt changes, which is beneficial to improving the stability of thermal power regulation during the rolling optimization process.
[0144] Subsequently, the initial attenuation factor is constrained within a range to obtain the thermal storage efficiency attenuation coefficient. Since the initial attenuation factor theoretically ranges from 0 to 1, to improve the robustness of the algorithm under abnormal data input, sensor failure, or numerical calculation errors, when the calculated initial attenuation factor exceeds the range of 0 to 1, it is truncated to the corresponding boundary value, thus obtaining the thermal storage efficiency attenuation coefficient.
[0145] For example, the optimal charging temperature of a molten salt thermal storage system is 565℃. Step S300 calculates the real-time average working fluid temperature to be 520℃. Therefore, the working fluid temperature deviation is 520 - 565 = -45℃. Further calculation shows the dimensionless relative temperature deviation is 45 ÷ 565 ≈ 0.0796. The squared deviation value is 0.0796² ≈ 0.00634. The initial attenuation factor is 1 ÷ (1 + 0.00634) ≈ 0.9937. Since the calculation result is between 0 and 1, no interval correction is needed, and the final thermal storage efficiency attenuation coefficient is taken as 0.9937. In subsequent inner-layer optimization, this thermal storage efficiency attenuation coefficient is applied to the thermal storage power term, allowing the optimization model to comprehensively consider the impact of the current thermal state of the thermal storage system on the actual charging efficiency, thereby obtaining thermal storage optimization results that better conform to actual operating conditions.
[0146] It should be noted that the above values are for illustrative purposes only and do not constitute a limitation on the present invention.
[0147] In summary, this step involves obtaining the real-time average working fluid temperature, remaining usable volume ratio, and equipment design parameters of the thermal storage tank. It then calculates the working fluid temperature deviation, thermal storage efficiency decay coefficient, and maximum storable heat capacity. The thermal storage efficiency decay coefficient is used to correct the thermal storage power term in the inner objective function. Simultaneously, the maximum storable heat capacity is used as an upper limit constraint on the cumulative heat storage. This allows the two-layer optimization model to comprehensively consider the current thermal state, charging efficiency, and remaining thermal storage capacity of the thermal storage system. Under the premise of ensuring the safe operation of the thermal storage equipment, this improves the effective charging capacity of the thermal storage process and the engineering feasibility of the optimization results.
[0148] S400: Solve the modified bi-level optimization model under joint constraints, output the current cycle's thermal storage working fluid pump speed sequence and generate the first step speed command; Step S400 in the method provided in this embodiment of the invention includes: From the optimal thermal storage working fluid pump speed sequence obtained by solving the modified bi-layer optimization model, the pump speed value corresponding to the first sampling time is extracted as the initial speed command; The actual rotational speed of the thermal storage medium pump is collected at the current moment, and the difference between the initial speed command and the actual speed is calculated to obtain the speed deviation. Based on the upper limit of the speed regulation rate of the thermal storage working fluid pump and the sampling period determined by the discrete sampling time, the speed deviation is limited within the maximum allowable adjustment amount in a single step to obtain the feasible speed increment; The actual speed is added to the feasible speed increment to obtain the dynamic feasible speed command; The dynamic feasible speed command is compared with the upper and lower limits of the adjustable range of the thermal storage working fluid pump speed. If the dynamic feasible speed command exceeds the adjustable range, it is truncated to the nearest range boundary to obtain the first speed command.
[0149] Specifically, after solving the modified two-layer optimization model, the optimal thermal storage medium pump speed sequence covering the entire prediction window can be obtained. Since rolling optimization control only executes the control quantity of the current control cycle, and subsequent sampling times will be re-optimized based on the latest operating state, this embodiment only extracts the pump speed value corresponding to the first sampling time in the optimal thermal storage medium pump speed sequence as the initial speed command.
[0150] Subsequently, the actual operating speed of the thermal storage medium pump is collected in real time, and the difference between the initial speed command and the actual speed is calculated to determine the speed adjustment range required for the current control cycle. Speed deviation = Initial speed command - Actual speed.
[0151] When the speed deviation is positive, it indicates that the speed of the thermal storage medium pump needs to be increased; when the speed deviation is negative, it indicates that the speed of the thermal storage medium pump needs to be decreased.
[0152] Considering that the thermal storage medium pump and its variable frequency drive device both have a maximum allowable speed adjustment rate, in order to avoid drastic fluctuations in molten salt flow, mechanical impact on the pump body and increased thermal stress in the heat exchanger due to excessively rapid speed changes, this embodiment limits the speed adjustment amount based on the upper limit of the allowable speed adjustment rate of the equipment.
[0153] The upper limit of the speed adjustment rate is determined by the operating procedures of the thermal storage working fluid pump equipment or the performance parameters of the frequency converter, and can be expressed as the maximum speed that can be adjusted per unit time, for example, 40 to 60 r / min·min.
[0154] Subsequently, the maximum permissible speed adjustment per step is calculated based on the rolling optimization sampling period. Among these, The maximum permissible speed adjustment in a single step = the upper limit of the speed adjustment rate × the sampling period.
[0155] The sampling period adopts the rolling optimization control period corresponding to the two-layer optimization model, such as 15 minutes.
[0156] Furthermore, the speed deviation is compared with the maximum permissible speed adjustment in a single step.
[0157] When the absolute value of the speed deviation is greater than the maximum permissible speed adjustment amount in a single step, the maximum permissible speed adjustment amount in a single step is used as the feasible speed increment, and the direction of the speed deviation remains unchanged; when the absolute value of the speed deviation is less than or equal to the maximum permissible speed adjustment amount in a single step, the speed deviation is directly used as the feasible speed increment.
[0158] Subsequently, a dynamically feasible speed command is calculated based on the actual operating speed and the feasible speed increment. Among these, Dynamic feasible speed command = actual speed + feasible speed increment.
[0159] Furthermore, to ensure that the thermal storage medium pump always operates within the allowable range of the equipment, the minimum stable operating speed and the maximum continuous operating speed of the thermal storage medium pump are obtained, and an adjustable speed range is constructed.
[0160] When the dynamically feasible speed command is higher than the upper limit of the speed adjustable range, the dynamically feasible speed command is corrected to the upper limit value; when the dynamically feasible speed command is lower than the lower limit of the speed adjustable range, the dynamically feasible speed command is corrected to the lower limit value; when the dynamically feasible speed command is within the speed adjustable range, its value remains unchanged, thus obtaining the first speed command.
[0161] Subsequently, the initial speed command is sent to the frequency converter of the thermal storage medium pump. The frequency converter adjusts the output frequency according to the initial speed command, so that the thermal storage medium pump gradually runs to the corresponding target speed.
[0162] Preferably, after the current control cycle ends, real-time data such as unit power output, average working fluid temperature of the thermal storage tank, remaining available volume, and grid operating status are collected again, and steps S100 to S400 are executed again to generate the first speed command corresponding to the next control cycle, thereby realizing rolling optimization closed-loop control.
[0163] Furthermore, when the deviation between the actual feedback speed of the thermal storage working fluid pump and the initial speed command continuously exceeds the preset deviation threshold, the current control command is immediately terminated, and the two-layer optimization model is restarted to solve the problem, so as to ensure that the thermal storage system is always optimized and controlled according to the latest operating state.
[0164] For example, at a certain moment, the bi-level optimization model finds the optimal thermal storage pump speed at the first sampling moment to be 1450 r / min. The current actual operating speed of the thermal storage pump is 1320 r / min. Therefore, the speed deviation = 1450 - 1320 = 130 r / min. Assuming the upper limit of the allowable speed adjustment rate of the thermal storage pump is 4 r / min·s, and the rolling optimization sampling period is 15 min (900 s), the maximum allowable speed adjustment in a single step = 4 × 900 = 3600 r / min. Since 130 < 3600, the feasible speed increment = 130 r / min. Further, the dynamic feasible speed command = 1320 + 130 = 1450 r / min is obtained. Assuming the allowable operating speed range of the thermal storage pump is 600–1500 r / min, 1450 r / min is within the allowable operating range, therefore no range correction is needed, and the final initial speed command = 1450 r / min is obtained. The frequency converter adjusts the output frequency according to the initial speed command, so that the thermal storage medium pump gradually increases to 1440 r / min, and recalculates the new control command in the next rolling optimization cycle.
[0165] It should be noted that the above values are for illustrative purposes only and do not constitute a limitation on the present invention.
[0166] In summary, this step uses the optimal rotational speed at the first sampling moment output by the two-layer optimization model as the control target. Combined with the speed regulation rate constraint and operating range constraint of the thermal storage working fluid pump, the optimization results are corrected for engineering feasibility. A rolling optimization closed-loop control method is adopted to update the initial rotational speed command in real time. Under the premise of ensuring the safe and stable operation of the equipment, the thermal storage working fluid pump can accurately track the optimized peak-shaving strategy, thereby improving the control accuracy of the thermal storage system and the reliability of engineering applications.
[0167] S500: Executes the first speed command, corrects the two-layer optimization model based on the deviation between the actual and theoretical remaining available volume ratio, and uses the current sequence remaining value as the initial solution for the next window, and solves the problem in a rolling manner until the cumulative deviation converges to the preset tolerance.
[0168] Step S500 in the method provided in this embodiment of the invention includes: The initial speed command is encoded according to the communication protocol format of the thermal storage working fluid pump actuator and then sent to the thermal storage working fluid pump actuator to execute the initial speed command. After the first speed command is executed, the current molten salt volume of the thermal storage tank is collected in real time through the tank's continuous liquid level gauge. The actual remaining usable volume ratio is calculated by combining the tank's rated usable volume and then the difference between the actual remaining usable volume ratio and the ratio before execution is obtained. Obtain the change in the theoretical remaining usable volume ratio corresponding to the first step speed command, which is synchronously output by the corrected two-layer optimization model during the solution process, and calculate the single-step deviation between the actual change in the remaining usable volume ratio and the theoretical change in the remaining usable volume ratio. The single-step deviation is superimposed with the historical deviation accumulated in the current scrolling window to obtain the cumulative deviation. It is then determined whether the cumulative deviation converges to the preset tolerance. The preset tolerance is determined based on the ratio of the minimum resolvable volume of the tank continuous level gauge to the rated usable volume of the tank. If the cumulative deviation does not converge to the preset tolerance, the single-step deviation will be fed back to the corrected two-layer optimization model to offset the correction coefficient of the thermal power term in the objective function of the next inner window. The remaining pump speed sequence values in the current sequence, excluding the first speed command that has already been executed, are used as the initial solution for the inner layer optimization problem of the next window. The optimization window is then pushed forward by one sampling period, and the rolling solution process is repeated.
[0169] like Figure 3 As shown, specifically, after solving the corrected two-layer optimization model, the optimal thermal storage medium pump speed sequence covering the entire prediction window can be obtained. Since the rolling optimization control only executes the control quantity corresponding to the current control cycle, and subsequent sampling times will be re-optimized based on the latest operating state, this embodiment only extracts the pump speed value corresponding to the first sampling time in the optimal thermal storage medium pump speed sequence as the first step speed command.
[0170] Subsequently, the initial speed command is encoded according to the industrial communication protocol corresponding to the thermal storage medium pump actuator and sent to the thermal storage medium pump frequency converter for execution. Preferably, the industrial communication protocol can be an industrial control protocol such as Modbus TCP, PROFINET, or IEC 61850. After receiving the initial speed command, the frequency converter adjusts the output frequency to drive the thermal storage medium pump to the corresponding target speed, thereby changing the thermal storage medium flow rate and the actual thermal storage power.
[0171] After the initial rotation speed command is executed for one sampling cycle, the current molten salt level is measured in real time using a continuous level gauge installed on the thermal storage tank. This measurement, combined with the tank's geometric dimensions, is used to calculate the current molten salt volume, and further calculates the percentage of actual remaining usable volume after execution. Actual remaining usable volume percentage = (rated usable volume of tank - current molten salt volume) ÷ rated usable volume of tank.
[0172] Subsequently, the actual remaining available volume percentage after execution is compared with the remaining available volume percentage recorded before executing the first speed command to obtain the change in the actual remaining available volume percentage. Among these, Change in actual remaining available volume percentage = Actual remaining available volume percentage after execution - Remaining available volume percentage before execution
[0173] Meanwhile, when solving for the first-step speed command, the modified two-layer optimization model simultaneously calculates the theoretical cumulative heat storage based on the thermal power, sampling period, and physical properties such as thermal working fluid density and specific heat capacity corresponding to the first-step speed. Based on the theoretical cumulative heat storage, it calculates the change in the theoretical remaining usable volume ratio corresponding to the first-step speed command, which is used to characterize the change in the remaining usable volume of the thermal storage tank under ideal operating conditions.
[0174] Subsequently, the one-step deviation between the actual change in the percentage of remaining usable floor space and the theoretical change in the percentage of remaining usable floor space is calculated. Among these, Single-step deviation = Change in actual remaining usable volume percentage - Change in theoretical remaining usable volume percentage.
[0175] The single-step deviation is used to reflect the difference between the model's prediction results and the actual operating results. When the single-step deviation is small, it indicates that the model's prediction results are basically consistent with the actual operating state; when the single-step deviation continues to increase, it indicates that the actual thermal storage process is affected by factors such as changes in heat exchange efficiency, changes in molten salt properties, and equipment execution errors, and the model needs to be corrected and optimized in a timely manner.
[0176] Furthermore, the current single-step deviation is accumulated with the deviations generated in each historical sampling period within the rolling window to obtain the cumulative deviation. Among these, Cumulative deviation = Historical cumulative deviation + Current single-step deviation.
[0177] Subsequently, it is determined whether the cumulative deviation has converged to the preset tolerance range. The preset tolerance is determined based on the measurement accuracy of the continuous level gauge. Specifically, Preset tolerance = minimum resolvable volume of level gauge ÷ rated usable volume of tank.
[0178] The reason for adopting the above method is that the continuous level gauge itself has a minimum resolution capability. When the cumulative deviation is less than the minimum volume ratio that the level gauge can resolve, the deviation can no longer reflect the actual change in the thermal storage state. Therefore, it can be considered that the model prediction result is consistent with the actual operation result.
[0179] Preferably, the minimum resolvable volume of the level gauge is derived from the level gauge product manual or metrological verification report. For example, if a thermal storage tank has a rated usable volume of 12,000 m³ and the minimum resolvable volume of the continuous level gauge is 6 m³, then: preset tolerance = 6 ÷ 12,000 = 0.0005.
[0180] When the absolute value of the cumulative deviation is no greater than 0.0005, the cumulative deviation is considered to have converged.
[0181] If the rolling cumulative deviation does not converge to the preset tolerance range, the current single-step deviation is fed back to the thermal storage efficiency attenuation coefficient calculated in step S300, and bias compensation is performed on the thermal storage efficiency attenuation coefficient. The correction factor for the thermal storage power term is equal to the original correction factor plus the offset compensation.
[0182] Preferably, the offset compensation amount is proportional to the single-step deviation, and its proportionality coefficient can be obtained offline based on historical operating data. The proportionality coefficient is preferably between 0.05 and 0.20, derived from parameter identification of historical thermal storage operating data, ensuring that the cumulative deviation within multiple consecutive rolling windows converges quickly without significant system oscillations. In this embodiment, a value of 0.10 is preferred.
[0183] For example, when the current single-step deviation is -0.003, the bias compensation amount = 0.10 × (-0.003) = -0.0003.
[0184] Subsequently, the corrected thermal storage efficiency decay coefficient is substituted back into the inner objective function of the next rolling window to correct the thermal storage power term, thereby reducing model prediction error and improving the accuracy of subsequent optimization.
[0185] Furthermore, to improve the efficiency of rolling optimization, this embodiment uses the remaining optimal pump speed sequence in the current optimization sequence, excluding the first step speed command that has been executed, as the initial solution for the next rolling window optimization calculation.
[0186] The reason for adopting the above method is that the system operating state changes little between adjacent scrolling windows. Using the remaining optimization results of the previous scrolling window as the initial solution of the next window can effectively reduce the number of optimization iterations and improve the speed of online optimization solution.
[0187] Subsequently, the optimization window is advanced one sampling cycle along the time axis to reacquire the latest real-time data such as the unit operating status, the average temperature of the working fluid in the thermal storage tank, the percentage of remaining available capacity, and the grid operating status. Steps S200 to S500 are then re-executed to form the optimization result corresponding to the next rolling window.
[0188] When the absolute value of the rolling cumulative deviation is not greater than the preset tolerance, the offset compensation amount is stopped from being updated, and the control continues to be executed according to the first speed command calculated by the rolling optimization model. The rolling cumulative deviation is judged again to see if it meets the convergence condition only after the latest running status is collected again in the next rolling window.
[0189] For example, at the start of a certain rolling window, the remaining usable volume percentage of the thermal storage tank is 0.225. After executing the first rotation speed command, the continuous level gauge measures the current remaining usable volume percentage as 0.220. Therefore, the actual change in the remaining usable volume percentage is 0.220 - 0.225 = -0.005. The corrected two-layer optimization model synchronously predicts a theoretical change in the remaining usable volume percentage of -0.004. Therefore, the single-step deviation is -0.005 - (-0.004) = -0.001. Assuming the cumulative deviation of the previous rolling window is 0.0004, the cumulative deviation is 0.0004 + (-0.001) = -0.0006. If the minimum resolvable volume of the level gauge is 6 m³ and the rated usable volume is 12000 m³, then the preset tolerance is preset tolerance = 6 ÷ 12000 = 0.0005. Since |-0.0006| > 0.0005, the rolling cumulative deviation has not yet converged. In this embodiment, the current single-step deviation is fed back to the thermal storage efficiency attenuation coefficient in step S300 for bias compensation. At the same time, the pump speed values at each time point in the current optimization sequence, excluding the first speed command, are used as the initial solution for the next rolling window optimization, and the optimization window is advanced by one sampling period to continue rolling optimization. When the absolute value of the subsequent rolling cumulative deviation is not greater than 0.0005, the model prediction accuracy is considered to meet the engineering requirements, the bias compensation amount is stopped, and the thermal storage working fluid pump control continues to be executed according to the rolling optimization results.
[0190] It should be noted that the above values are for illustrative purposes only and do not constitute a limitation on the present invention.
[0191] In summary, this step executes the initial rotation speed command and collects the actual operating status of the thermal storage system in real time. It compares the actual remaining available volume change with the model prediction results, uses the rolling cumulative deviation to compensate the thermal storage efficiency decay coefficient online, and uses the remaining optimization sequence of the previous rolling window as the initial solution for the next window optimization. This achieves online self-correction and rolling iteration of the modified bi-layer optimization model, improving the model prediction accuracy while taking into account the online optimization solution efficiency and the operational stability of the thermal storage system.
[0192] It should be noted that the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some implementations, multitasking and parallel processing are possible or may be advantageous.
[0193] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
[0194] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.
Claims
1. An optimization method for collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units, characterized in that, The method includes: The system obtains the photovoltaic power forecast sequence, grid load forecast sequence and current power output of the generating units during the noon period, constructs the net load forecast curve, and generates the peak-shaving power deficit based on the maximum power reduction demand during the off-peak period and the adjustable capacity margin between the current power output and the minimum technical output. A two-layer optimization model is constructed, and the peak power deficit is set as the endpoint constraint of the power output reduction depth of the outer layer at the valley time. The outer layer of the two-layer optimization model takes maximizing the cumulative amount of the power output reduction depth of the unit as the optimization objective, and the inner layer takes maximizing the net heat storage of the thermal storage system as the optimization objective. The outer layer has a higher priority than the inner layer. Based on the real-time average working fluid temperature and the percentage of remaining usable volume in the thermal storage tank, the thermal storage efficiency attenuation coefficient when the working fluid temperature deviates from the optimal charging temperature is calculated, and the thermal storage power term in the inner objective function is corrected. Solve the modified bi-level optimization model under joint constraints, output the current cycle's thermal storage working fluid pump speed sequence, and generate the first step speed command; The first speed command is executed, and the two-layer optimization model is corrected based on the deviation between the actual and theoretical remaining available volume ratio. The remaining value of the current sequence is used as the initial solution for the next window, and the solution is rolled until the cumulative deviation converges to the preset tolerance.
2. The method for optimizing the collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units according to claim 1, characterized in that, The steps for determining the noon time period include: Obtain the all-weather irradiance prediction sequence output by the numerical weather prediction system, as well as the longitude and latitude of the location of the photovoltaic power station; Based on the longitude and latitude, the solar hour angle and solar declination angle of the day are calculated in combination with the date information, and the clock time corresponding to the noon solar hour of the day is determined. Obtain the sunrise and sunset times for the day, calculate the time expansion width, and expand the time expansion width to both sides of the clock time corresponding to the noon solar time to form an initial noon window; Extract the irradiance prediction value corresponding to the initial noon window from the all-weather irradiance prediction sequence, calculate the average value of the irradiance prediction value within the window, and use a preset proportion of the average value as the irradiance threshold. From the irradiance prediction values at each time point within the initial noon window, the continuous period during which the irradiance prediction value is consistently higher than the irradiance threshold is determined as the noon period.
3. The method for optimizing the collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units according to claim 1, characterized in that, Obtain the photovoltaic power forecast sequence, grid load forecast sequence, and current power output of generating units during the noon period, and construct a net load forecast curve, including: The predicted irradiance values for each moment during the noon period are obtained, and the corresponding predicted photovoltaic power values are obtained after irradiance-power conversion. These values are then arranged by time index to form the photovoltaic power prediction sequence for the noon period. Based on the ultra-short-term load forecast data released by the power grid dispatching system, load forecast values that are consistent with the time index of the photovoltaic power forecast sequence are extracted and arranged in chronological order to form the power grid load forecast sequence. The active power output value of the unit is collected in real time using the power transmitter on the output side of the unit, which is used as the current power output of the unit. The photovoltaic power prediction sequence and the power grid load prediction sequence are aligned point by point according to the time index. At each time point, the power grid load prediction value is subtracted from the photovoltaic power prediction value to calculate the net load prediction value at each time point. The net load prediction curve is generated by combining the values at each time point.
4. The method for optimizing the collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units according to claim 3, characterized in that, The predicted irradiance values for each time period during noon are obtained, and the corresponding predicted photovoltaic power values are obtained after irradiance-power conversion, including: The predicted irradiance values at various times during the noon period were obtained, and the standard irradiance under standard test conditions and the installed capacity of the photovoltaic power station were used as the conversion benchmark. Divide the predicted irradiance value at each time point by the standard irradiance to obtain the irradiance ratio, and combine it with the installed capacity to obtain the initial DC power value; The rated operating temperature and temperature power coefficient of the photovoltaic module are obtained, and the predicted irradiance value at each time moment is substituted into the definition of the rated operating temperature to calculate the module operating temperature at each time moment. Calculate the difference between the component's operating temperature and the standard temperature under standard test conditions, and combine it with the temperature power coefficient to obtain the temperature correction coefficient; Multiply the initial DC power value by the temperature correction factor, and then combine it with the comprehensive efficiency coefficient of the photovoltaic power station to obtain the photovoltaic power prediction value at each time. The comprehensive efficiency coefficient is the comprehensive reduction coefficient of the energy loss along the entire path from the DC side output of the photovoltaic power station to the AC grid connection point. The energy loss along the entire path includes at least the DC-AC conversion loss, the collector line loss, the inverter loss, and the dust shading loss on the surface of the photovoltaic modules.
5. The method for optimizing the collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units according to claim 1, characterized in that, Peak-shaving power deficits are generated based on the maximum power demand during off-peak hours and the adjustable capacity margin between the current power output and the minimum technical output, including: The low-end period is located from the net load forecast curve, and the time corresponding to the minimum net load forecast value within the low-end period is identified as the low-end time. The low-end period is determined based on the changing trend of the net load forecast value within the noon period of the net load forecast curve. Extract the predicted net load value at the off-peak time as the off-peak net load value, and calculate the difference between the off-peak net load value and the current power output of the unit as the maximum power reduction demand during the off-peak period; Obtain the minimum technical output of the unit, calculate the difference between the current electrical output of the unit and the minimum technical output, and use it as the adjustable capacity margin; The maximum power demand during the off-peak period is compared with the adjustable capacity margin, and the smaller of the two values is taken as the peak-shaving power deficit.
6. The method for optimizing the collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units according to claim 1, characterized in that, The construction steps of the two-level optimization model include: Using the power output reduction depth of each discrete sampling time during the noon period as the outer decision variable, and setting the peak power deficit as the lower limit constraint of the reduction depth at the valley time, an outer objective function is established to maximize the cumulative power output reduction depth during the entire noon period. Using the pump speed of the thermal storage medium at each sampling time as the inner layer decision variable, and based on the inherent characteristic relationship between the pump speed of the thermal storage medium and the thermal storage power, as well as the thermal storage efficiency decay coefficient, an inner layer objective function is established to maximize the cumulative amount of thermal storage power after correction by the thermal storage efficiency decay coefficient during the noon period. Establish a coupled inequality constraint that the power reduction of the unit's electrical output at any given moment is not less than the charging power of the thermal storage system; Under the combined constraint conditions of outer constraints, inner constraints, and the coupled inequality constraints, a layered solution mechanism prioritizing outer constraints is adopted for solving the problem, including: Solve the outer optimization problem that satisfies the lower limit constraint of the endpoint to obtain the optimal power output reduction depth sequence, and take the power reduction value at each time in the optimal power output reduction depth sequence as the upper limit value of the thermal storage power. Solve the inner-layer optimization problem with the pump speed of the thermal storage medium as the variable. Under the premise that the thermal storage power does not exceed the corresponding upper limit value at each time, obtain and output the optimal pump speed sequence that maximizes the cumulative amount of the corrected thermal storage power.
7. The method for optimizing the collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units according to claim 1, characterized in that, Based on the real-time average working fluid temperature and the percentage of remaining usable volume in the thermal storage tank, the thermal storage efficiency attenuation coefficient is calculated when the working fluid temperature deviates from the optimal charging temperature. The thermal storage power term in the inner objective function is then corrected, including: Obtain the optimal charging temperature and the maximum allowable operating temperature of the thermal storage tank, and combine the real-time average working fluid temperature and the percentage of remaining usable volume in the thermal storage tank to calculate the deviation of the current working fluid temperature from the optimal charging temperature. The real-time average working fluid temperature of the thermal storage tank is obtained by collecting and weighting the data from temperature measuring points deployed at different heights of the tank. The remaining usable volume ratio is obtained by the ratio of the current molten salt volume measured by the tank's continuous level gauge to the tank's rated usable volume. Based on the ratio of the absolute value of the deviation to the optimal charging temperature, and combined with the nonlinear decay relationship between the deviation and the charging efficiency, a thermal storage efficiency decay coefficient is generated, wherein the thermal storage efficiency decay coefficient is a positive number not greater than 1. Multiply the thermal storage power term in the inner objective function by the thermal storage efficiency decay coefficient to obtain the effective thermal storage power term after temperature correction, and substitute it into the inner objective function to replace the thermal storage power term before correction. The remaining available volume of the thermal storage tank is determined by the percentage of the remaining available volume, and the maximum heat storage capacity corresponding to the current remaining available volume is used as the upper limit constraint of the integral value of the thermal storage power term in the inner objective function.
8. The method for optimizing the collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units according to claim 7, characterized in that, Based on the ratio of the absolute value of the deviation to the optimal charging temperature, and combined with the nonlinear decay relationship between the deviation and the charging efficiency, a thermal storage efficiency decay coefficient is generated, including: Calculate the ratio of the absolute value of the deviation to the optimal charging temperature to obtain the dimensionless relative temperature deviation. Based on the dimensionless relative temperature deviation, calculate the squared value of the deviation. An initial attenuation factor is obtained by performing a nonlinear transformation based on the squared deviation value, wherein the nonlinear transformation causes the initial attenuation factor to decrease monotonically as the squared deviation value increases. The initial attenuation factor is subject to upper and lower limit truncation constraints to obtain the thermal storage efficiency attenuation coefficient, wherein the thermal storage efficiency attenuation coefficient is a positive number that is not less than 0 and not greater than 1.
9. The method for optimizing the collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units according to claim 1, characterized in that, The steps for generating the initial speed command include: From the optimal thermal storage working fluid pump speed sequence obtained by solving the modified bi-layer optimization model, the pump speed value corresponding to the first sampling time is extracted as the initial speed command; The actual rotational speed of the thermal storage working fluid pump at the current moment is collected, and the difference between the initial rotational speed command and the actual rotational speed is calculated to obtain the rotational speed deviation; Based on the upper limit of the speed regulation rate of the thermal storage working fluid pump and the sampling period determined by the discrete sampling time, the speed deviation is limited within the maximum allowable adjustment amount in a single step to obtain the feasible speed increment; The actual rotational speed is added to the feasible rotational speed increment to obtain the dynamic feasible rotational speed command; The dynamic feasible speed command is compared with the upper and lower limits of the adjustable range of the thermal storage working fluid pump speed. If the dynamic feasible speed command exceeds the adjustable range, it is truncated to the nearest range boundary to obtain the first speed command.
10. The method for optimizing the collaborative peak-shaving strategy of molten salt thermal storage coupled coal-fired power units according to claim 1, characterized in that, Execute the initial rotation speed command, and correct the two-layer optimization model based on the deviation between the actual and theoretical remaining available volume ratio. Use the current sequence remaining value as the initial solution for the next window, and continuously solve until the cumulative deviation converges to the preset tolerance, including: The initial speed command is encoded according to the communication protocol format of the thermal storage working fluid pump actuator and then sent to the thermal storage working fluid pump actuator to execute the initial speed command; After the first speed command is executed, the current molten salt volume of the thermal storage tank is collected in real time by the tank continuous level gauge. The actual remaining usable volume ratio is calculated by combining the rated usable volume of the tank and the difference between the actual remaining usable volume ratio and the ratio before execution. The change in the actual remaining usable volume ratio is obtained. Obtain the change in the theoretical remaining usable volume ratio corresponding to the first step speed command, which is synchronously output by the modified bi-layer optimization model during the solution process; and calculate the single-step deviation between the actual change in the remaining usable volume ratio and the theoretical change in the remaining usable volume ratio. The single-step deviation is superimposed with the historical deviation accumulated in the current scrolling window to obtain the cumulative deviation. It is then determined whether the cumulative deviation converges to a preset tolerance, wherein the preset tolerance is determined based on the ratio of the minimum resolvable volume of the tank continuous level gauge to the rated usable volume of the tank. If the cumulative deviation does not converge to the preset tolerance, the single-step deviation is fed back to the corrected two-layer optimization model to offset the correction coefficient of the thermal power term in the objective function of the next window. The remaining pump speed sequence values in the current sequence, excluding the first speed command that has already been executed, are used as the initial solution for the inner layer optimization problem of the next window. The optimization window is then pushed forward by one sampling period, and the rolling solution process is repeated.
