Independent energy storage multi-time scale transaction optimization method based on power spot market

Abstract

The application belongs to the technical field of electric power market, and particularly relates to an independent energy storage multi-time scale transaction optimization method based on an electric power spot market. First, a first power scheduling instruction sequence of an independent energy storage system under a day-ahead scale and a second power scheduling instruction sequence at a start time of an intra-day rolling optimization cycle are acquired. The two power scheduling instruction sequences are time-aligned in an overlapping interval. Whether a power instruction difference at a time boundary exceeds a preset jump threshold is detected. If the power instruction difference exceeds the preset jump threshold, a virtual buffer period is set before and after the time boundary to construct a power smooth transition curve. The independent energy storage system is controlled to perform power transition. An energy deviation amount caused by power smoothing in the virtual buffer period is taken as a cost correction item and fed back to an intra-day rolling optimization target in the next round to adjust a subsequently generated power scheduling instruction sequence. The strategy consistency and execution robustness of the independent energy storage in the multi-time scale market are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power market, and specifically relates to an independent energy storage multi-time scale transaction optimization method based on a power spot market. BACKGROUND

[0002] Under the framework of the power spot market, various market subjects need to formulate transaction strategies on different time scales according to price signals and load forecasts to maximize economic benefits. Independent energy storage, as a flexible resource with fast response and bidirectional regulation capability, is no longer limited to a single dispatch instruction in the market, but needs to be cooperatively decided on multiple time scales such as day-ahead, intra-day and real-time to cope with the dual challenges of market price fluctuations and system balance requirements.

[0003] Currently, there are two typical technical solutions in the industry to handle the multi-time scale transaction optimization of independent energy storage: one is single-scale independent optimization, that is, independent decision is made on each time scale, and the plan of the latter scale completely covers or replaces that of the former scale.

[0004] The other is multi-scale static connection, that is, fixed rules such as linear interpolation are used to decompose coarse-scale instructions to fine-scale.

[0005] The above existing technical solutions have the following defects in actual application: 1. Although the existing technology has introduced a rolling optimization mechanism to avoid the serious disconnection between day-ahead planning and real-time execution, the new and old instruction sequences are prone to sharp power instruction jumps at the time boundary. The existing method lacks effective detection and suppression strategies, and such jumps may exceed the physical regulation capacity of the energy storage device, resulting in the inability to execute the instructions or even damage the safety and service life of the device.

[0006] 2. The existing connection rules are statically preset and cannot be adaptively adjusted according to the real-time state of the energy storage system and market dynamic conditions, resulting in a transition process that is either too conservative to lose benefits or too aggressive to be risky.

[0007] 3. The instruction smoothing operation to suppress power jumps will inevitably cause a deviation between the actual executed electric quantity and the market bid quantity, thereby bringing additional economic costs or loss of benefits. The existing method generally ignores the economic value of this deviation and does not feed it back to the subsequent optimization, causing the contradiction between the economic theory of the overall strategy and the loss after execution. SUMMARY

[0008] In order to overcome the shortcomings in the background art, the embodiments of the present application provide an independent energy storage multi-time scale transaction optimization method based on a power spot market, which can effectively solve the problems involved in the above background art.

[0009] The object of the present application can be achieved by the following technical solution: an independent energy storage multi-time scale transaction optimization method based on a power spot market, comprising: obtaining a first power scheduling instruction sequence of an independent energy storage system at a day-ahead scale, and a second power scheduling instruction sequence at a start time of an intra-day rolling optimization period, wherein the second power scheduling instruction sequence overlaps with the first power scheduling instruction sequence in time.

[0010] The first power scheduling instruction sequence and the second power scheduling instruction sequence are time-aligned in the overlapping interval, and whether a power instruction difference at a time boundary exceeds a preset jump threshold is detected.

[0011] If the preset jump threshold is exceeded, a virtual buffer period is set before and after the time boundary, and a power smoothing transition curve is constructed in the virtual buffer period to control the independent energy storage system to perform power transition.

[0012] An energy deviation amount caused by power smoothing in the virtual buffer period is taken as a cost correction term and fed back to a next round of intra-day rolling optimization target to adjust a subsequently generated power scheduling instruction sequence.

[0013] Compared with the prior art, the embodiments of the present application have at least the following advantages or beneficial effects: (1) The present application can automatically identify and smooth a power instruction mutation that may exceed the physical bearing capacity of the device when the multi-time scale instructions are connected, which helps to avoid problems such as device overload and accelerated battery degradation caused by abrupt changes in planned instructions, and improves the safety and reliability of the independent energy storage system in actual execution.

[0014] (2) The connection strategy of the present application is not static or empirical, but can dynamically adjust the jump judgment threshold, the buffer period length and the smoothing transition curve according to the real-time state of charge, the maximum power capacity and other parameters of the independent energy storage system. This adaptive mechanism ensures that the instruction transition process is always within the safe and feasible range of the device, and can make optimal adjustments when the system state changes, taking into account safety constraints and economic needs.

[0015] (3) The present application quantifies the energy deviation amount caused by smoothing transition as an economic cost correction term and feeds it back to the subsequent rolling optimization objective function, which can perceive the real economic impact of instruction connection, so as to actively prefer to select a smoother and less execution deviation strategy when making subsequent plans, which helps to maximize the economic efficiency of the whole cycle transaction. BRIEF DESCRIPTION OF DRAWINGS

[0016] The application is further described with the accompanying drawings, but the embodiments in the drawings do not constitute any limitation to the application, and other embodiments can be obtained by those of ordinary skill in the art without creative labor on the basis of the following drawings.

[0017] Figure 1 The method of the application implements a step flowchart.

[0018] Figure 2 The dynamic jump threshold setting flowchart of the application.

[0019] Figure 3 The power smooth transition curve construction flowchart of the application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the protection scope of the application.

[0021] Referring to Figure 1 As shown in the figure, the application provides an independent energy storage multi-time scale transaction optimization method based on a power spot market, which comprises the following steps: S1. obtaining a first power scheduling instruction sequence of an independent energy storage system at a day-ahead scale and a second power scheduling instruction sequence at a start time of a day-ahead rolling optimization cycle, wherein the second power scheduling instruction sequence and the first power scheduling instruction sequence overlap in time.

[0022] Based on the operation mechanism of the power spot market multi-time scale transaction, the output plan of the independent energy storage system at the day-ahead market stage needs to be corrected and executed according to the updated market information and system state at the day-ahead and real-time stages. Therefore, at the start of each day-ahead rolling optimization cycle, the newly generated day-ahead scale plan will coexist with the day-ahead scale plan that has not been executed in the future time period, thereby forming an overlapping time interval in time. The overlapping time interval specifically refers to a continuous time period from the start time of the current day-ahead rolling optimization cycle to a certain time in the future covered by the day-ahead plan. The interval is the coexistence period of the day-ahead plan that has not been executed and the day-ahead new plan that has started to take effect, and is also the area where different scale instructions may conflict and need to be coordinated.

[0023] Specifically, from the day-ahead market clearing result file issued by the power dispatching institution or the power trading platform, the first power dispatching instruction sequence corresponding to the independent energy storage system is parsed and read. At the same time, from the output result of the local intra-day rolling optimization platform, the second power dispatching instruction sequence newly generated at the starting moment of the current period is directly called. The acquisition process is automatically completed through a predefined data interface.

[0024] S2. Time align the first power dispatching instruction sequence and the second power dispatching instruction sequence in the overlapping interval, and detect whether the power instruction difference at the time boundary exceeds the preset jump threshold.

[0025] Since the first power dispatching instruction sequence is derived from the day-ahead market, its time granularity is usually hourly, representing the average power commitment for each complete hour. The second power dispatching instruction sequence is generated by intra-day rolling optimization, with a finer time granularity, such as 15 minutes.

[0026] To compare the power settings of the two at the same physical time, they must be mapped to the same time reference. The time boundary specifically refers to the starting moment of the overlapping interval, i.e., the moment when the current intra-day rolling optimization period begins. This point is the junction of day-ahead plan execution and intra-day new plan effectiveness, and is also the location with the highest risk of power instruction discontinuity.

[0027] The time alignment implementation process is as follows: determine the time range of the overlapping interval, and establish a unified time axis covering the overlapping interval based on the time granularity of the second power dispatching instruction sequence.

[0028] For each time node on the unified time axis, it is attributed to the corresponding specific hour time period in the first power dispatching instruction sequence, and the original power setting value corresponding to the hour time period is directly determined as the first aligned power value corresponding to the time node.

[0029] The original power setting value corresponding to each time granularity in the second power dispatching instruction sequence is directly assigned to the corresponding time node on the unified time axis with the same timestamp as the second aligned power value.

[0030] Through this process, two columns of power sequences are obtained, which are strictly aligned on the unified time axis and can be directly compared point by point.

[0031] On the basis of time alignment, to quantitatively evaluate the smoothness of instruction junction, it is necessary to detect whether the power instruction difference at the time boundary exceeds the preset jump threshold. The specific implementation process is as follows: determine the last instruction value of the first power dispatching instruction sequence at the starting moment of the intra-day rolling optimization period, and the initial instruction value of the second power dispatching instruction sequence at the same starting moment.

[0032] calculating an absolute difference between the final instruction value and the initial instruction value.

[0033] comparing the absolute difference with a preset jump threshold, if the absolute difference is greater than the preset jump threshold, determining that there is a power instruction jump, and synchronously recording a time point at which the power instruction jump occurs and a corresponding instruction value.

[0034] Referring to Figure 2 The preset jump threshold is dynamically set according to the physical characteristic parameters of the independent energy storage system, and the implementation process is as follows: obtaining the maximum allowable power change rate of the independent energy storage system and the battery cycle life attenuation sensitive interval, wherein the maximum allowable power change rate is an inherent technical parameter obtained from the converter technical specification or the equipment nameplate parameter of the independent energy storage system, and the battery cycle life attenuation sensitive interval is a state of charge range predefined and stored by the battery management system of the independent energy storage system based on the cycle life test data or the electrochemical model provided by the battery manufacturer.

[0035] multiplying the maximum allowable power change rate by the current trading cycle length of the power spot market, obtaining the maximum power step amount allowed in a unit control cycle, and multiplying the step amount by a preset safety margin coefficient to obtain a basic jump threshold.

[0036] It should be noted that the preset safety margin coefficient is used to reserve a certain buffer space before the theoretical limit to prevent the device action from exceeding the safety boundary due to errors or instantaneous overload, and is determined based on historical operation statistical data of the independent energy storage system, and the determination process is as follows: extracting the operation data of multiple complete trading days in the past from the historical operation database of the independent energy storage system. For each trading day, the following data in each market trading cycle are obtained: the planned power instruction value received at the start of the cycle, and the average actual output power value actually measured and recorded in the cycle through the sensor.

[0037] For each trading cycle in the historical operation statistical data set, the absolute value of the difference between the planned power instruction value and the actual output power value is calculated. The absolute value is divided by the absolute value of the corresponding planned power instruction value to obtain an instruction execution deviation rate. This deviation rate quantifies the deviation degree of the historical execution effect relative to the instruction. In addition, when the planned instruction value is zero, a very small positive number is used instead of the denominator to avoid division by zero error.

[0038] The instruction execution deviation rates calculated for all historical trading cycles are statistically analyzed. The preset high percentile is calculated as a statistical characteristic value, for example, the 95th percentile. The characteristic value represents the upper limit of the accuracy that the actual execution capability of the system can achieve relative to the theoretical requirements of the instruction in most cases, which can be defined by the technician.

[0039] A mapping relationship between the statistical characteristic value and the safety margin coefficient is established. The mapping relationship is constructed as a reverse correlation, that is, the larger the statistical characteristic value, the greater the historical execution deviation and the higher the system uncertainty, and the smaller the safety margin coefficient obtained by mapping should be.

[0040] As an exemplary mapping function, the safety margin coefficient can be the difference between 1 and the statistical characteristic value, and to ensure the effectiveness of the coefficient, a lower limit value and an upper limit value need to be set. If the calculated value is lower than the lower limit, the lower limit is taken, and if it is higher than the upper limit, the upper limit is taken. Finally, the preset safety margin coefficient for the current threshold calculation is obtained.

[0041] To protect the battery life, the base jump threshold needs to be further adjusted according to the real-time state. It is determined whether the current state of charge of the independent energy storage system is in the battery cycle life attenuation sensitive interval, if so, the Euclidean distance between the current state of charge and the midpoint of the battery cycle life attenuation sensitive interval is calculated, and a reduction coefficient is determined according to a predefined negative correlation mapping relationship between the distance and the reduction coefficient. The reduction coefficient is multiplied by the base jump threshold to obtain the final dynamic jump threshold.

[0042] The predefined negative correlation mapping relationship between the distance and the reduction coefficient can be exemplarily referred to the following process: the Euclidean distance between the current state of charge and the midpoint of the battery cycle life attenuation sensitive interval is compared with half of the span of the battery cycle life attenuation sensitive interval to obtain the normalized distance.

[0043] The normalized distance is substituted into a standard negative exponential function to output the reduction coefficient, and the reduction coefficient is compared with a predefined reduction coefficient specification interval. If the output value is lower than the lower limit, the lower limit is taken, and if it is higher than the upper limit, the upper limit is taken to obtain the final reduction coefficient.

[0044] It should be noted that the standard negative exponential function is only an evolution algorithm of a negative mapping relationship provided in this embodiment. As other implementation manners, implementers can also use other feasible algorithms in the prior art as long as the purpose of realizing the negative mapping relationship is achieved, and the present application does not limit this.

[0045] If the current state of charge is not in the battery cycle life attenuation sensitive interval, the base jump threshold is directly used as the final dynamic jump threshold. This dynamic setting process is automatically executed when each daily rolling optimization period starts, ensuring that the threshold always matches the current safety operation boundary of the system.

[0046] S3. If the preset jump threshold is exceeded, a virtual buffer period is set before and after the time boundary, and a power smooth transition curve is constructed in the virtual buffer period to control the independent energy storage system to perform power transition.

[0047] When the power instruction difference value is detected to exceed the dynamic jump threshold, it indicates that there is a significant power step at the junction point between the day-ahead plan and the intra-day new plan. In order to ensure the safe operation of power electronic devices such as energy storage converters, avoid the instantaneous sharp change of power causing overcurrent protection or electrical stress damage, and at the same time make the actual output curve of the energy storage system smoother to meet the grid dispatching requirements, a transition period needs to be introduced at the instruction switching point. This transition period is not the actual market clearing period, so it is called a virtual buffer period, and its role is to provide a time window for the smooth change of power.

[0048] The virtual buffer period setting process includes: first, determining the basic time required for transition according to the physical limits of the device. The absolute value of the power instruction difference value is divided by the maximum allowed power change rate of the independent energy storage system to obtain the minimum transition time required to complete the power instruction jump.

[0049] However, only meeting the theoretical minimum time may not be enough to cope with certain real-time constraints. Therefore, the real-time state of the system at the transition starting point needs to be further obtained, including the real-time state of charge and the real-time maximum charge and discharge power limit under the current working condition.

[0050] Based on the sign of the power instruction difference value and the real-time state of charge, it is determined whether there is a power output capability limitation before and after the power transition. The specific determination process is: from the battery technical specification book or cycle life test report provided by the battery manufacturer matched with the independent energy storage system, the recommended state of charge long-term operation window corresponding to the relatively flat battery capacity decay rate is extracted.

[0051] From the historical operation database of the independent energy storage system, the actual running state of charge data of the battery in the past period of time is extracted, a statistical distribution histogram of the state of charge is drawn and its cumulative distribution function is calculated, according to the preset coverage probability target of the state of charge range in which the battery is running most of the time, the state of charge that makes the cumulative distribution function value equal to the complement of half of the coverage probability target is identified as the lower fractile, and the state of charge that makes the cumulative distribution function value equal to the coverage probability target itself is identified as the upper fractile, and the actual running experience window of the state of charge is determined according to the upper and lower fractiles.

[0052] The recommended state of charge long-term operation window and the actual running experience window are intersected to obtain a comprehensive safe operation window. The upper limit value of this comprehensive window is taken as the first state of charge high threshold, and the lower limit value is taken as the first state of charge low threshold.

[0053] When the power instruction difference value sign is positive, it indicates that the instruction jump direction is to increase the output power. When it is negative, it indicates that the instruction jump direction is to reduce the output power. In combination with the real-time state of charge, the limited state logic is judged: if the current state of charge is greater than the first state of charge high threshold and the power instruction difference value is positive, or the current state of charge is less than the first state of charge low threshold and negative, it is determined that the starting side adjustment tension is high, otherwise it is determined that the starting side adjustment tension is low.

[0054] If the current state of charge is greater than the first state of charge high threshold and the power instruction difference value is negative, or the current state of charge is less than the second state of charge low threshold and the power instruction difference value is positive, it is determined that the target side adjustment tension is high, otherwise it is determined that the target side adjustment tension is low.

[0055] When it is determined that the starting side or the target side adjustment tension is high, it indicates that there is a situation of limited power output capability before and after the power transition.

[0056] Based on the adjustment tension judgment results of the starting side and the target side, on the basis of the minimum theoretical transition time, the time lengths of the forward buffer sub-period and the backward buffer sub-period are dynamically allocated to form the virtual buffer period, wherein if there is a situation of limited power output capability, a longer buffer sub-period is allocated for the limited side.

[0057] Specifically, if the starting side and the target side adjustment tension are both low, equal lengths are allocated for the forward and backward buffer sub-periods, specifically each being half of the total length of the virtual buffer period, and the allocation is the base value of the forward and backward buffer sub-periods.

[0058] If the starting side adjustment tension is high and the target side adjustment tension is low, the backward buffer sub-period maintains the base value, and the forward buffer sub-period is allocated more than half of the total length of the virtual buffer period on the basis of the base value.

[0059] If the target side adjustment tension is high and the starting side adjustment tension is low, the forward buffer sub-period maintains the base value, and the backward buffer sub-period is allocated more than half of the total length of the virtual buffer period on the basis of the base value.

[0060] The embodiment of the application can automatically identify and smooth the power instruction mutation that may exceed the physical bearing capacity of the device when the instructions of multiple time scales are connected, which helps to avoid problems such as device overload and battery accelerated attenuation caused by abrupt planned instructions, and improves the safety and reliability of the independent energy storage system in actual execution.

[0061] The virtual buffer period defines the time window of the transition, but how the power should change in this window needs to construct a power smooth transition curve. The goal of the construction of this curve is to complete the power change in a given time while minimizing the mechanical and electrical impact on the device as much as possible and strictly following all real-time operation constraints. Based on this, referring to Figure 3 The power smooth transition curve construction process includes calculating the ratio of the absolute value of the power command difference to the total length of the virtual buffer period, defined as the required average power change rate.

[0062] Compare the required average power change rate with the maximum allowed power change rate, and select the function type of the transition curve according to the comparison result: if the required average power change rate is less than or equal to the first proportion threshold of the maximum allowed power change rate, it indicates that the transition demand is gentle and the impact on the device stress is small. At this time, select a linear ramp function with high efficiency and direct response as the construction function. Its implementation is to directly take the starting power value and the target power value as the end points, and perform linear interpolation in the virtual buffer period to generate a preliminary power smooth transition curve.

[0063] If it is greater than the first proportion threshold of the maximum allowed power change rate, it indicates that the transition demand is urgent, and directly using a linear ramp may cause excessive instantaneous stress on the device. At this time, a nonlinear function is selected to provide a smoother connection at the end points. Specifically, a polynomial function of the third order or higher is selected as the construction function. Its implementation and constraint method is to calculate the normalized excess amplitude of the required average power change rate exceeding the first proportion threshold.

[0064] Set the order of the polynomial function and construct the polynomial expression with time as the variable.

[0065] Take the starting power value, target power value, and start and end time of the virtual buffer period as the boundary conditions. According to the normalized excess amplitude, impose a derivative constraint: when the excess amplitude is less than half of the first proportion threshold, additionally constrain the first derivative of the polynomial function at the start and end points to be zero. When the excess amplitude is greater than or equal to half of the first proportion threshold, then additionally constrain the first and second derivatives of the polynomial function at the start and end points to be zero.

[0066] Substitute the initial command value, the final command value, and the start and end time of the virtual buffer period into the selected construction function to solve and generate a preliminary power smooth transition curve.

[0067] It should be noted that the first proportion threshold is obtained by: in the typical working interval of the independent energy storage system converter, taking the power change rate starting from zero and increasing in steps as the control command to drive the converter to perform a power step response test.

[0068] Synchronously collect and record the average switching loss rate of the power module of the converter and the temperature rise rate of the key heat dissipation point in each test to form a loss rate sequence and a temperature rise rate sequence corresponding to the power change rate instruction. The key heat dissipation point refers to the junction temperature monitoring point or the heat sink substrate temperature monitoring point of the power device which is most sensitive to temperature or most likely to fail due to overheating.

[0069] Calculate the change between adjacent data points in the loss rate sequence and the temperature rise rate sequence, respectively. Identify the data point in each sequence where the change first exceeds the predetermined multiple of the previous average value, and mark it as the loss inflection point and the thermal stress inflection point, respectively. Record the corresponding power change rate test value.

[0070] Divide the power change rate test value corresponding to the loss inflection point and the thermal stress inflection point, respectively, by the maximum allowed power change rate to obtain two candidate ratio values.

[0071] Select the smaller of the two candidate ratio values as the basic criterion ratio. Multiply the basic criterion ratio by a pre-set engineering margin coefficient less than 1 to obtain the final ratio criterion used for subsequent comparison.

[0072] Example implementation process: The loss inflection point is 70% of the maximum allowed value, and the thermal stress inflection point is 65% of the maximum allowed value. Select the smaller 65% as the basic criterion ratio, and multiply it by the pre-set engineering margin coefficient of 0.85 to obtain the final ratio criterion for comparison, which is 55.25%. In the subsequent step, when the required average power change rate is less than or equal to 55.25% of the maximum allowed power change rate, a linear ramp function is selected. When the ratio exceeds this ratio, a nonlinear function is selected.

[0073] The generated preliminary power smooth transition curve must be checked for feasibility to ensure its physical executability. The specific process is as follows: obtain the real-time maximum charge and discharge power limit of the independent energy storage system within the virtual buffer period.

[0074] Check whether all power values on the preliminary power smooth transition curve exceed the real-time maximum charge and discharge power limit.

[0075] If the check fails, identify all points on the preliminary power smooth transition curve where the power value exceeds the real-time maximum charge and discharge power limit, and calculate the maximum absolute amplitude of the limit value.

[0076] Divide the power limit amplitude by the maximum allowed power change rate of the independent energy storage system to obtain the minimum additional transition time required based on the current limit severity, which is used as the total length adjustment of the virtual buffer period in this iteration.

[0077] The current virtual buffer period total length is increased by the adjustment amount, the time lengths of the forward and backward buffer sub-periods are re-allocated based on the new total length, and the power smoothing transition curve construction process is returned to generate a new preliminary curve and perform a feasibility check, until the check passes, but when the virtual buffer period total length is extended to exceed a preset maximum allowed buffer length, the iteration is terminated and a safety fault tolerance mechanism is triggered, which requires abandoning the single smoothing transition and decomposing the original jump into multiple smaller sub-jumps within the power limit, which are completed step by step in multiple subsequent control periods.

[0078] The adaptation strategy of the embodiments of the present application is not static or empirical, but can dynamically adjust the jump judgment threshold, the buffer period length and the smoothing transition curve according to the real-time state of charge, maximum power capability and other parameters of the independent energy storage system. The adaptive mechanism ensures that the instruction transition process is always within the safe and feasible domain of the device, and can make optimal adjustments when the system state changes, taking into account both safety constraints and economic demands.

[0079] S4. The energy deviation amount caused by power smoothing in the virtual buffer period is taken as a cost correction term and fed back to the next round of daily rolling optimization target to adjust the subsequent generated power scheduling instruction sequence.

[0080] Based on the rules of the electricity spot market for settlement of actual delivered energy, the market revenue of the independent energy storage system is directly determined by the physical execution of the energy curve. The power smoothing operation in the virtual buffer period causes a deviation between the actual charging and discharging energy of the system and the energy planned by the daily rolling optimization. If the economic value change corresponding to this deviation is not taken into account, the daily rolling optimization process will continue to generate power instruction sequences with pure market revenue as the target, ignoring the risk of high-cost smoothing operation that may be triggered when the sequence is physically executed. This will lead to a cycle of planned jump-forced smoothing-revenue loss. Therefore, the energy deviation amount caused by smoothing needs to be converted into a cost correction term and fed back as an input data to the objective function solving process of the next round of rolling optimization. By internalizing this cost, the subsequent power scheduling instruction sequence obtained by solving can balance the power jump that may bring high smoothing cost while pursuing market revenue, ultimately achieving the unity of operation safety and economy.

[0081] The energy deviation amount caused by power smoothing in the virtual buffering period is taken as a cost correction term. The implementation process is as follows: first, in the virtual buffering period, the difference between the power value of the power smoothing transition curve and the corresponding power value of the second power scheduling instruction sequence at each time node is calculated based on the unified time axis. The power difference value sequence is numerically integrated with the time interval of the unified time axis as the step size to obtain the total energy deviation amount in kilowatt-hours.

[0082] Secondly, the total energy deviation amount is converted into economic cost. Since the node price of the electricity spot market changes over time, the corresponding real-time price sequence in the virtual buffering period is obtained. The weighted value of the total energy deviation amount and the real-time price sequence is calculated.

[0083] A specific embodiment is that if the price in the virtual buffering period is constant, the cost correction term is the product of the total energy deviation amount and the constant price. If the price changes, the energy deviation component in each different price sub-period needs to be calculated based on the power difference value curve, and then each energy deviation component is multiplied by the corresponding sub-period price and added to obtain the cost correction term.

[0084] The feedback to the next round of daily rolling optimization target is as follows: when solving the objective function of the next round of daily rolling optimization, the cost correction term is taken as a penalty term in the objective function.

[0085] The sign of the penalty term is set as follows: when the energy deviation amount leads to a decrease in expected income, the penalty term is negative. When the energy deviation amount leads to an increase in expected cost, the penalty term is positive.

[0086] By optimizing and solving the objective function containing the penalty term, a new second power scheduling instruction sequence is generated, and the power instruction jump at the time boundary of the new second power scheduling instruction sequence is inhibited by the penalty term.

[0087] It should be noted that the cost correction term is a historical or immediate economic deviation calculated based on the virtual buffering period that has occurred or has been planned. Feedback to the subsequent optimization period aims to affect the newly generated power scheduling instruction sequence in the subsequent period, so that it is smoother at the time boundary in the future, forming a continuous improvement cycle of evaluation-compensation-prevention, and does not constitute a retroactive adjustment of the executed instructions.

[0088] The method further comprises a learning optimization step: after each time the virtual buffering and power smoothing operation is completed, the feature data of this time is recorded as a sample. The recorded data at least includes: the power instruction difference value at the time of triggering the jump, the dynamic jump threshold value used, the total length of the virtual buffering period and the allocation ratio of its preceding and subsequent sub-periods, and the absolute value of the finally calculated cost correction term.

[0089] After a certain number of historical samples are accumulated, the ratio of the power instruction difference value and the dynamic jump threshold value is used as the independent variable to reflect the relative severity of the jump, and the cost correction term is used as the dependent variable. A prediction model from the independent variable to the dependent variable is established using linear regression or nonlinear fitting method.

[0090] At the start of each new daily rolling optimization, when the dynamic jump threshold value is initially calculated, based on the current market forecast and system state, the typical power instruction difference value that is likely to occur is estimated. This is substituted into the prediction model to obtain a predicted smoothing cost. If the predicted smoothing cost exceeds the preset tolerance cost, it is determined that the current threshold setting is likely to be too loose, allowing high-cost jumps to occur. At this time, the initial value of the dynamic jump threshold value to be used for detection in the current period is pre-tightened by proportionally reducing the calculation result of the basic jump threshold value. Through this feedforward adjustment based on historical experience, the probability of triggering high-cost smoothing events is actively reduced, realizing online learning and continuous improvement of the system.

[0091] The energy deviation caused by the smooth transition is quantified as an economic cost correction term in the embodiments of the present application, and is fed back to the subsequent rolling optimization objective function. The real economic impact brought by the instruction connection can be perceived, so as to actively prefer to select a smoother strategy with smaller execution deviation when formulating the subsequent plan, which helps to maximize the economic efficiency of the whole period transaction.

[0092] It should be additionally pointed out that the instruction smoothing connection method provided by the present application is not limited to the application between the two specific time scales of day-ahead and intra-day. The method can be cascaded and applied to the optimization instruction connection link between any two adjacent time scales from coarse to fine, such as between day-ahead, intra-day, real-time and second-level control.

[0093] In specific implementation, a structured instruction coordination unit can be deployed between the output interfaces of the optimization modules of each pair of adjacent levels. Each unit independently runs and performs instruction alignment, jump detection and virtual buffering smoothing operation for the level it is in. When a new instruction sequence is generated by the lower module representing a finer time granularity, the unit performs time alignment between the new instruction sequence and the instruction sequence of the upper module representing a coarser time granularity that has been issued and is within the overlapping period, and detects the power instruction difference value at all connection time points.

[0094] For each detected power jump point exceeding the corresponding level dynamic jump threshold, the unit independently constructs a local virtual buffer period for it and generates a local power smooth transition curve. Subsequently, all generated local smooth curves are spliced with the original instruction fragments that do not occur jump in time sequence, thereby forming a global continuous modified power instruction sequence suitable for execution between the two levels.

[0095] The modified instruction sequence will be used as the input reference of the next finer time scale optimization module, or as the reference for feedback execution constraints to the upper coarser time scale module. Through such step-by-step deployment and connection between multiple levels, the present application can systematically guarantee the continuity and executability of the entire instruction chain from long-term planning to instantaneous control, and improve the overall robustness of the complex time scale architecture.

[0096] The above is only an example and description of the structure of the present application. Those skilled in the art can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, as long as they do not deviate from the structure of the present application or exceed the scope defined by the present application.

Claims

1. An optimization method for independent energy storage trading across multiple time scales based on the electricity spot market, characterized in that: include: Obtain the first power dispatch command sequence of the independent energy storage system on the day-ahead scale, and the second power dispatch command sequence when the intraday rolling optimization cycle starts, wherein the second power dispatch command sequence and the first power dispatch command sequence have an overlapping period in time; The first power scheduling instruction sequence and the second power scheduling instruction sequence are time-aligned within the overlapping interval, and the power instruction difference between the two at the time boundary is detected to exceed a preset jump threshold. If the preset jump threshold is exceeded, a virtual buffer period is set before and after the time boundary, and a power smooth transition curve is constructed within the virtual buffer period to control the independent energy storage system to perform power transition. The energy deviation caused by power smoothing during the virtual buffer period is used as a cost correction item and fed back into the next round of intraday rolling optimization targets to adjust the subsequently generated power scheduling instruction sequence.

2. The method for optimizing independent energy storage multi-timescale trading based on the electricity spot market according to claim 1, characterized in that, The time alignment implementation process is as follows: The time range of the overlapping interval is determined, and a unified time axis covering the overlapping interval is established based on the time granularity of the second power scheduling instruction sequence. For each time node on the unified time axis, it is assigned to the specific hourly time period corresponding to the first power scheduling instruction sequence, and the original power setting value corresponding to the hourly time period is directly determined as the first aligned power value corresponding to the time node. The original power setting value corresponding to each time granularity in the second power scheduling instruction sequence is directly assigned to the corresponding time node with the same timestamp on the same time axis, as the second aligned power value.

3. The method for optimizing independent energy storage multi-timescale trading based on the electricity spot market according to claim 1, characterized in that, The process for detecting whether the power command difference between the two at the time boundary exceeds the preset transition threshold is as follows: Determine the last instruction value of the first power scheduling instruction sequence at the start of the intraday rolling optimization cycle, and the initial instruction value of the second power scheduling instruction sequence at the same start time; Calculate the absolute difference between the final instruction value and the initial instruction value; The absolute difference is compared with a preset transition threshold. If the absolute difference is greater than the preset transition threshold, it is determined that there is a power command transition, and the time point of the power command transition and the corresponding command value are recorded synchronously.

4. The method for optimizing independent energy storage multi-timescale trading based on the electricity spot market according to claim 3, characterized in that, The preset switching threshold is dynamically set based on the physical characteristic parameters of the independent energy storage system, and the implementation process is as follows: Obtain the maximum permissible power change rate and the battery cycle life degradation sensitive range of the independent energy storage system; Multiply the maximum allowable power change rate by the current clearing cycle length of the electricity spot market to obtain the maximum allowable power step within a unit control cycle, and then map the result to the basic jump threshold using a preset safety margin coefficient. Determine whether the current state of charge of the independent energy storage system is within the battery cycle life degradation sensitive range. If so, calculate the distance between the current state of charge and the midpoint of the battery cycle life degradation sensitive range, and use this distance to plan a reduction factor to correct the basic jump threshold, thus obtaining the final dynamic jump threshold. Otherwise, the basic transition threshold is directly used as the final dynamic transition threshold; This process is repeated at the start of each intraday rolling optimization cycle to update the dynamic jump threshold.

5. The method for optimizing multi-timescale trading of independent energy storage based on the electricity spot market according to claim 4, characterized in that, The virtual buffer period setting process includes: Calculate the minimum transition time required to complete the power command transition based on the maximum permissible power change rate of the independent energy storage system; Obtain the real-time state of charge and maximum charge / discharge power limit of the independent energy storage system before the time boundary; Based on the sign direction of the power command difference and the real-time state of charge, it is determined whether the system has limited power output capability before and after the power transition; Based on the judgment results, the time lengths of the forward buffer sub-period and the backward buffer sub-period are dynamically allocated on the basis of the minimum theoretical transition time to form the virtual buffer period. If there is a situation where the power output capability is limited, a longer buffer sub-period is allocated to the limited side.

6. The method for optimizing multi-timescale trading of independent energy storage based on the electricity spot market according to claim 5, characterized in that, The process of constructing the power smooth transition curve includes: The ratio of the absolute value of the power command difference to the total length of the virtual buffer period is defined as the required average power change rate. Compare the required average power change rate with the maximum allowable power change rate, and select the function type of the transition curve based on the comparison result: If the required average power change rate is less than or equal to the first proportional threshold of the maximum allowable power change rate, then a linear ramp function is selected as the constructor. If the value exceeds the first proportional threshold of the maximum allowable power change rate, a nonlinear function is selected as the constructor, and the first or second derivative of the function is constrained to zero at the start and end points of the virtual buffer period based on the magnitude of the required average power change rate exceeding the first proportional threshold. The initial instruction value, the final instruction value, and the start and end times of the virtual buffer period are used as boundary conditions and substituted into the selected constructor to solve for the initial power smooth transition curve.

7. The method for optimizing independent energy storage multi-timescale trading based on the electricity spot market according to claim 6, characterized in that, The process of constructing the power smoothing transition curve also includes a feasibility verification of the preliminary power smoothing transition curve: Obtain the real-time maximum charge and discharge power limit of the independent energy storage system during the virtual buffer period; Verify that all power values ​​on the preliminary power smoothing transition curve do not exceed the real-time maximum charge / discharge power limit. If the verification fails, the total length of the virtual buffer period is dynamically adjusted according to the preset rules, and the curve is reconstructed until the verification passes, generating the final power smooth transition curve.

8. The method for optimizing independent energy storage multi-timescale trading based on the electricity spot market according to claim 1, characterized in that, The energy deviation caused by power smoothing during the virtual buffer period is used as a cost correction item, and the implementation process is as follows: During the virtual buffer period, the energy deviation is calculated by time-granular integration based on the power difference between the constructed power smooth transition curve and the original second power scheduling instruction sequence. The energy deviation is converted into an equivalent economic cost value by weighted calculation based on the current spot electricity market price signal, and this value is used as a cost correction item.

9. The method for optimizing independent energy storage multi-timescale trading based on the electricity spot market according to claim 1, characterized in that, The feedback is incorporated into the next round of intraday rolling optimization targets, and the implementation process is as follows: When solving its objective function in the next round of intraday rolling optimization, the cost adjustment term will be included as a penalty term in the objective function; The sign of the penalty term is set as follows: when the energy deviation leads to a decrease in expected revenue, the penalty term is negative; when the energy deviation leads to an increase in expected cost, the penalty term is positive. A new second power scheduling instruction sequence is generated by optimizing the objective function that includes the penalty term. The power instruction jump tendency of the new second power scheduling instruction sequence at the time boundary is suppressed by the penalty term.

10. The method for optimizing independent energy storage multi-timescale trading based on the electricity spot market according to claim 1, characterized in that, The method also includes a learning optimization step: Record the setting parameters, power smoothing transition curves, and corresponding cost correction items for each virtual buffer period; Based on historical data, establish the correlation between the preset jump threshold, the power command difference, and the size of the cost correction item; At the start of each intraday rolling optimization, the initial value of the preset jump threshold for the current period is pre-adjusted using the aforementioned correlation to reduce the probability of triggering the virtual buffer and generating cost correction items in the future.

Images