Multi-mode driven generalized energy storage charging and discharging decision self-iteration optimization method and system

Abstract

The invention discloses a multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method and system. The method comprises the following steps: generating a reference path; when the mode is not determined, taking out one mode from the mode library; otherwise, constructing a current mode feasible path library; carrying out quantitative evaluation on the path, and selecting an optimal path for a corresponding mode; if in the mode selection state and the next mode exists, returning to the selection mode; otherwise, taking the optimal path in the current mode as the optimal path of the round, and then entering final judgment; comparing the target function value of the optimal path in each mode, and determining the mode with the minimum target function value as the mode of the subsequent round; judging whether a convergence condition is met; if not, entering the next round, and returning to mode judgment; otherwise, taking the optimal path of the round as an output result; compared with a traditional method, the method has the advantages that the safety and adequate guarantee capability of the power system is improved, and the method has good process traceability and result interpretability.

Description

Technical Field

[0001] The present invention relates to a generalized energy storage charging and discharging decision self-iterative optimization method and system, and in particular to a multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method and system, belonging to the technical field of power system operation and dispatching. Background Art

[0002] Building a new power system with new energy as the main body, the system is becoming increasingly complex and unstable under the high penetration rate of new energy. Conventional power sources must not only follow the load changes, but also balance the output fluctuations of new energy. Insufficient startup or regulation rate of coal-fired power will lead to the risk of power outages, and excessive startup will result in the waste of wind and solar resources, which brings huge challenges to the balance and coordination of the power system and the safe and stable operation.

[0003] The backup of traditional power systems mainly responds to the demand on the load side. The load fluctuations are regular and the random disturbances are small. The corresponding backup optimization configuration is relatively extensive but can still maintain the safe and economic operation of the power grid. After the grid connection of new energy, the system has changed from unilateral load fluctuations to fluctuations on both the power supply side and the load side. Due to the intermittent, random and volatile characteristics of new energy output, the fluctuations are less regular and the random disturbances are large. The traditional backup strategy system lacks the means of risk management and risk decision-making. In order to coordinate the prevention and control of wind and solar power abandonment and power outage risks, the existing backup system optimizes the up and down backup measures by exploring the advantages of starting and stopping thermal power units in the grid operation planning to minimize the risk cost. However, the existing backup system does not consider the synergy between energy storage and the start and stop of thermal power units after participating in the backup. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method and system that can efficiently utilize generalized energy storage resources to prevent and control grid operation risks while ensuring equipment safety.

[0005] Technical solution: The invention discloses a multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method, comprising:

[0006] Step 1: Generate a reference path;

[0007] Step 2: Determine the current optimization mode. If the mode has not been determined, define the current state as a mode selection state; otherwise, define the current state as a mode determination state and proceed to step 4.

[0008] Step 3: Take out a pattern from the pattern library;

[0009] Step 4: Construct a feasible path library of generalized energy storage devices or generalized energy storage device clusters under the current mode;

[0010] Step 5: Carry out quantitative evaluation on each path in the feasible path library and calculate the objective function value;

[0011] Step 6: Select the optimal path for the current mode by comparing the multi-path objective function values;

[0012] Step 7: If it is in the mode selection state, it is necessary to determine whether there is a next mode; if so, return to step 3; if not, proceed to the next step; if it is not in the mode selection state, take the preferred path in the current mode as the best path for this round, and then proceed to step 9;

[0013] Step 8: By comparing the objective function values ​​of the optimal paths in each mode, the mode with the smallest objective function value is determined as the mode for the subsequent rounds;

[0014] Step 9: Determine whether the convergence condition is met; if not, proceed to the next round and return to step 2; otherwise, take the optimal path of this round as the output result.

[0015] Furthermore, the path is a trajectory of the evolution of the power of a generalized energy storage device or a generalized energy storage device cluster over time on a plane with time and energy storage power as coordinates;

[0016] The generalized energy storage device is a power-adjustable resource in which the accumulated power generation / consumption in the power system is greater than or equal to a given value within a given scheduling cycle, and has interruptible and shiftable properties, and / or an energy storage upper limit is less than a given value.

[0017] Furthermore, the step 1 comprises:

[0018] Prepare the data required for path generation, the data including the power feasible domain and the power guiding area, specifically: obtain the current state and final state requirements of each generalized energy storage device, calculate the power feasible domain of the generalized energy storage device at each time step based on the physical and economic constraints of the generalized energy storage device; then set the power guiding area according to the power feasible domain, and adjust the guiding area parameters to control the characteristics of the generated path;

[0019] Starting from the starting point of a single flat start path, a path is randomly generated. The single flat start path is a path in which the energy storage output is set to zero throughout the entire scheduling period. Specifically: at each section, the power feasible domain and the power guidance area are first updated, where the power guidance value does not exceed the net outflow of the remaining period; then the power of the generalized energy storage device or the generalized energy storage device cluster at any section is used as the generation object, and the updated power feasible domain is used as the upper and lower bounds to complete the random power generation; all the generation results are collected to construct a path;

[0020] Repeat the steps of randomly generating paths, generate multiple paths, establish a path library, and store each generated path in the path library; use the risk cost as the objective function value, quantitatively evaluate each path, calculate the risk cost of all paths under different modes, and select the path with the smallest objective function value as the benchmark path.

[0021] Furthermore, the mode described in step 2 is a set of charging and discharging decision constraints of a generalized energy storage device or a generalized energy storage device cluster that satisfies the rule;

[0022] The rule R includes but is not limited to a polarization rule and an anti-polarization rule; the polarization rule allows the current charge and discharge power to be increased but not decreased, which is represented by R=1; the anti-polarization rule allows the current charge and discharge power to be decreased but not increased, which is represented by R=-1;

[0023] The modes include but are not limited to polarization mode, reverse polarization mode, consumption mode and supply guarantee mode; the polarization mode refers to strengthening the current charge / discharge intensity within the power feasible domain, charging more when charging, or discharging more when discharging, thereby releasing the power regulation potential of energy storage; the reverse polarization mode refers to weakening the current charge / discharge intensity within the power feasible domain, charging less when charging, and discharging less when discharging, thereby suppressing the power regulation potential of energy storage; the consumption mode refers to applying the polarization rule in the discharge state and the reverse polarization rule in the charge state; the supply guarantee mode refers to using the polarization rule in the charge state and the reverse polarization rule in the discharge state;

[0024] The entering into the mode selection state refers to the process of testing different modes in sequence when the mode is not determined, and selecting the best mode according to the testing results.

[0025] Furthermore, the step 4 comprises:

[0026] Step 4.1: Locate the starting point of a new path and establish a path library corresponding to the current mode;

[0027] Step 4.2: Reset the total period outflow of the generalized energy storage device or its cluster;

[0028] Step 4.3: Take out a scheduling window;

[0029] Step 4.4: Determine the charge and discharge status of the current window based on the previous value;

[0030] Step 4.5: Based on the total outflow power constraint of the period, generate the total power added value of the generalized energy storage device or its cluster in the scheduling window in the current mode according to the current rules, and perform rolling update of the path;

[0031] Step 4.6: Deduct the window power obtained by the total power added integral in step 4.5 from the total outflow power of the cycle;

[0032] Step 4.7: Determine whether there is a next window; if satisfied, return to step 4.3; otherwise, proceed to the next step;

[0033] Step 4.8: Add the updated path to the path library corresponding to the current mode;

[0034] Step 4.9: Determine whether the maximum number of paths is reached; if so, proceed to step 4.10; otherwise, return to step 4.1;

[0035] Step 4.10: Output the pattern path library.

[0036] Furthermore, the previous value described in step 4.4 is the power value obtained by the current window at the end of the previous round of optimization; the charging and discharging state P, when P=1, indicates the discharging state, and when P=-1, indicates the charging state; the charging and discharging state of the current window is determined according to the previous value, specifically, when the current value is greater than 0, the energy storage is in the discharging state, at which time P=1, otherwise it is in the charging state, P=-1.

[0037] Further, the step 4.5 comprises:

[0038] Step 4.5.1: Determine the activation region of the bootstrap distribution based on the previous generation value:

[0039] According to the combination of the preset optimization rules and the charge and discharge state of the generalized energy storage device or the generalized energy storage device cluster, the activation area is modified according to the following rules, specifically: determine whether the product of P and R is equal to 1; if satisfied, modify the lower limit of the guided distribution activation area to the previous generation value; if equal to -1, modify the upper limit of the guided distribution activation area to the previous generation value;

[0040] The guided distribution is a probability distribution space that guides the power decision of the generalized energy storage device or its cluster into a feasible domain according to the risk preference setting of the user, and randomly explores the optimization direction in the feasible domain;

[0041] The activation region is a sub-region of the guide distribution where power value extraction is actually performed after the mode is considered during the optimization process;

[0042] Step 4.5.2: Redistribute the probability distribution of the activation area of ​​the guided distribution;

[0043] The probability distribution refers to the probability law of the power value randomly drawn in the activation area of ​​the guided distribution; wherein the probability density is the proportion of the number of power values ​​drawn by setting random numbers in the corresponding area to the total number of times drawn, and the sum of the probability densities of the guided distribution is 1; the process of reallocating the probability distribution of the activation area of ​​the guided distribution refers to re-dividing the inner and outer guided sub-distributions of the activation area in step 4.5.2 according to the different probability densities of the sub-areas of the guided distribution, so that the sum of the probability densities of the activation area is still 1;

[0044] Step 4.5.3: Randomly extract the generalized energy storage device or the total power of its cluster based on the guided distribution activation area;

[0045] Step 4.5.4: Allocate the total power to each measure in the cluster.

[0046] Furthermore, the step 5 includes: taking the risk cost as the objective function value, calculating the risk cost of all paths under different modes;

[0047] The quantitative assessment refers to optimizing thermal power reserve and evaluating capacity risk cost after inputting the transmission power curve of the relevant interconnection lines in the target area and optimizing the generalized energy storage equipment or its cluster path for each feasible path. The total risk cost refers to the superposition of capacity risk cost and electricity risk cost.

[0048] Furthermore, step 6 includes: selecting a path with the smallest objective function value as the best path in the current mode.

[0049] Furthermore, step 7 includes, if it is in the mode selection state, traversing the modes in the mode library and performing optimization under the corresponding mode; if it is not in the mode selection state, the optimization mode of this round and subsequent rounds has been determined, then the preferred path under the current mode is used as the optimal path of this round.

[0050] Furthermore, the convergence condition described in step 9 is specifically that if in the mode selection state, the quantitative evaluation result of the best path selected in step 8 is improved relative to the quantitative evaluation result of the path selected in the previous round, and the objective function value is reduced, then the mode selected in step 8 is used as the given mode for the subsequent round of optimization, and the best path selected in the current round is used as the reference path for the next cycle, and the next round of optimization is performed to achieve self-iterative optimization; if it is not in the mode selection state, then when the quantitative evaluation result of the best path selected in step 6 is improved relative to the quantitative evaluation result of the path selected in the previous round, and the objective function value is reduced, the next round of self-iterative optimization is continued.

[0051] Based on the same inventive concept, the present invention also provides a multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization system, comprising:

[0052] Generate benchmark path module, in the first round of optimization, generate a massive path library from a single flat start point and select a path;

[0053] Construct a mode path module, and use the previous generation value as the starting point for generating the next generation charge and discharge path; then, divide the subsequent rounds of search into polarization rules and anti-polarization rules according to the same or opposite direction as the previous generation solution, thereby deriving multiple modes, and improving the generation of charge and discharge paths for multiple modes. By comparing and optimizing the generated paths under different modes, the optimization direction of the subsequent generation is determined;

[0054] The self-iterative optimization module selects the best mode as the optimization mode specified in the subsequent rounds, and uses the previous generation solution as the initial value for the next generation path generation. After multiple generations of cycles, the self-iterative evolution of the generalized energy storage charging and discharging strategy is achieved.

[0055] Beneficial effects: Compared with the prior art, the present invention overcomes the inefficiency, repetitiveness and conservatism of generating a massive path library from a single flat start point by introducing orthogonal rules and patterns. While significantly accelerating the decision convergence process, it avoids the problem of accuracy reduction caused by conservative decisions, high-frequency oscillations, directional hesitation, etc., and realizes efficient generation of pattern paths. The present invention can efficiently utilize generalized energy storage resources to prevent and control grid operation risks and enhance the safety and sufficiency guarantee capability of the power system under the premise of ensuring equipment safety, and has good process traceability and result interpretability. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0057] Figure 2 A schematic diagram of guide distribution setting according to an embodiment of the present invention;

[0058] Figure 3 A schematic diagram of the guide distribution and activation area thereof according to an embodiment of the present invention;

[0059] Figure 4 A flow chart of a mode power path library under a specified rule for constructing an embodiment of the present invention;

[0060] Figure 5 The generated power of the embodiment of the present invention is reserved for the total window power in this mode and allocated to the measure flow chart;

[0061] Figure 6 A power allocation flow chart of an embodiment of the present invention;

[0062] Figure 7 This is a schematic diagram of optimal configuration of thermal power reserve resources according to an embodiment of the present invention;

[0063] Figure 8This is a flow chart of dynamic optimization of multi-window thermal power reserve according to an embodiment of the present invention. DETAILED DESCRIPTION

[0064] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0065] As attached Figure 1 As shown, in view of the shortcomings of existing research, the multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method of this embodiment effectively improves the efficiency and accuracy of optimizing the energy storage power path, and seeks a better power reserve optimization configuration method, including:

[0066] Step 1: Generate a reference path;

[0067] Step 2: Determine the current optimization mode. If the mode has not been determined, define the current state as a mode selection state; otherwise, define the current state as a mode determination state and proceed to step 4.

[0068] Step 3: Take out a pattern from the pattern library;

[0069] Step 4: Construct a feasible path library of generalized energy storage devices or generalized energy storage device clusters under the current mode;

[0070] Step 5: Carry out quantitative evaluation on each path in the feasible path library and calculate the objective function value;

[0071] Step 6: Select the optimal path for the current mode by comparing the multi-path objective function values;

[0072] Step 7: If it is in the mode selection state, it is necessary to determine whether there is a next mode; if so, return to step 3; if not, proceed to the next step; if it is not in the mode selection state, take the preferred path in the current mode as the best path for this round, and then proceed to step 9;

[0073] Step 8: By comparing the objective function values ​​of the optimal paths in each mode, the mode with the smallest objective function value is determined as the mode for the subsequent rounds;

[0074] Step 9: Determine whether the convergence condition is met; if not, proceed to the next round and return to step 2; otherwise, take the optimal path of this round as the output result.

[0075] In order to achieve the purpose of the present invention, firstly, the operation constraints of the generalized energy storage device or its cluster are constructed. The constraints include objective operation constraints, subjective operation constraints and implicit operation constraints, etc. Specifically, they are as follows:

[0076] a) Objective operational constraints

[0077] The path generation process is subject to the performance limitations of the equipment itself, such as energy storage output constraints and energy storage power constraints.

[0078] Energy storage output constraints, that is, the range constraints of the power difference between each time step:

[0079] P s.min ≤P s ≤P s.max

[0080] Among them, P s is the actual output of energy storage, P s.max is the upper limit of energy storage output, P s.min It is the lower limit of energy storage output.

[0081] The energy storage capacity constraint requires that the capacity of each measure be taken into account during the evolution of the path, and that the capacity should neither be lower than the minimum capacity nor exceed the set maximum capacity:

[0082] E s.min ≤E s ≤E s.max

[0083] In the formula, E s is the energy storage capacity, E s.min 、E s.max are the minimum and maximum energy storage capacities respectively.

[0084] b) Subjective operational constraints

[0085] In addition to the inherent output constraints and power constraints, each measure in the measure set will also have independent operation requirements:

[0086] E s.end.min ≤E s.end ≤E s.end.max

[0087] Among them, E s.end is the electricity consumption at the end of the measure period, E s.end.max is the upper limit of electricity consumption at the end of the measure period, E s.end.min is the lower limit of the end-of-period electricity consumption of this measure. If it is not specified in the actual algorithm, the end-of-period electricity consumption constraint is the energy storage electricity consumption constraint.

[0088] In actual operation, routine maintenance of measures may result in not being connected to the network throughout the entire cycle. According to the overall arrangement, the following situations may occur:

[0089]

[0090] Among them, P s is the actual output of energy storage, P s.max is the upper limit of energy storage output, Ps.min is the lower limit of energy storage output, t is the time step number at the current moment, t start is the time for the measure to enter the network, t end The off-network time of this measure. If it is not specified in the actual algorithm, it is assumed that this measure is always in the network state.

[0091] In addition, there is an upper limit on the number of charging times, that is, for the number of charging times that limit this measure within a cycle, from the charging state to the discharging state and then to the charging state, the number of charging times of this measure is counted plus one:

[0092] n charge ≤n charge.max

[0093] Among them, n charge is the current number of charges, n charge.max It is the upper limit of the number of charging times in the measure cycle.

[0094] c) Implicit operational constraints

[0095] Since the system gives a total outflow value for the period, that is, the end-of-period electricity is a given value. There are infinite paths from the initial value to the end-of-period value. However, due to the maximum and minimum output limits of the generator, the path will eventually become infeasible due to excessive / small slopes. A feasible path must be within the boundary, but not all paths within the boundary are feasible paths.

[0096] Therefore, the intersection of the allowable output ranges of each constraint is taken as the feasible region of the power of the measure, referred to as the feasible region. For a single measure, under the premise of the end-of-period power, it is subject to implicit upper and lower bounds of the feasible region.

[0097] The implicit upper bound formula is as follows:

[0098] E u.j =E s.end -(N w *N T -j)*P s.max / η discharge

[0099] In the formula, E u.j is the upper limit of energy storage capacity at the jth time step in the cycle, E s.end is the final value of the measured power, N w is the number of windows in one cycle, N T is the number of time steps in a window, P s.max is the upper limit of energy storage output, η discharge is the discharge efficiency of energy storage;

[0100] The implicit lower bound formula is as follows:

[0101] E l.j=E s.end -(N w *N T -j)*P s.max *η charge

[0102] In the formula, E l.j is the lower limit of energy storage capacity at the jth time step in the cycle, E s.end is the final value of the measured power, N w is the number of windows in one cycle, N T is the number of time steps in a window, P s.max is the upper limit of energy storage output, η charge The charging efficiency of energy storage.

[0103] Next, the random generation algorithm of the benchmark path is introduced:

[0104] 1. Input the overall target outflow of regional energy storage;

[0105] The overall target outflow of the generalized energy storage device or its cluster refers to the net outflow of all energy storage measures in the area as a whole within a scheduling cycle.

[0106] 2. Construct the boundaries of the overall energy storage path planning. Specifically include:

[0107] First, the intersection of the allowable output ranges of each constraint of the generalized energy storage device or its cluster is taken as the power feasible domain of the power backup measure; then the feasible domain boundaries of all energy storage devices are accumulated to obtain the total upper and lower envelopes as the boundaries of the overall energy storage path planning.

[0108] 3. Divide the entire scheduling cycle into several scheduling windows, and each scheduling window is divided into the same number of time steps.

[0109] The scheduling window refers to a part of the continuous time period that constitutes the scheduling cycle.

[0110] 4. Take out a dispatch window and randomly generate the overall power of energy storage in the dispatch window. Specifically include:

[0111] 4.1 Divide the bootstrap distribution into positive and negative deviation regions according to the system bootstrap value;

[0112] The system guide value refers to the total amount of electricity to be discharged from the system divided by the number of remaining windows in the cycle; the guide distribution refers to the value range of the overall energy storage power in the window. The area in the guide distribution where the value is greater than the system guide value is the positive deviation area, and the area in the guide distribution where the value is less than the system guide value is the negative deviation area.

[0113] 4.2 Draw random numbers to activate a specific power deviation area; and divide the deviation area into inner and outer guide distributions;

[0114] like Figure 2 , the inner and outer guide distributions refer to the inner and outer regions of the guide distribution in the deviation region. The division formula of the inner and outer guide distributions is as follows:

[0115] p bound =p exp +a*v*|p max -p exp |

[0116] In the formula, p bound is the dividing line between the inner and outer guide distributions, p exp is the system boot value, p max is the maximum power of the guide sub-distribution, a takes the value of 1 in the positive deviation area and takes the value of -1 in the negative deviation area, v is the guide distribution ratio, which refers to the ratio of the data width of the guide distribution to the data width of the total feasible domain, where the total feasible domain is the intersection of the allowable output range under the constraint;

[0117] 4.3 Draw random numbers and activate specific guide distributions;

[0118] The method of activating a specific guide sub-distribution is to extract a random number between 0 and 1. If it is less than the guide distribution probability p, the inner guide sub-distribution is activated; otherwise, the outer guide sub-distribution is activated. The guide distribution probability p refers to the ratio of the expected number of times the randomly extracted power is within the guide distribution to the total number of times the generated power is generated, that is, the probability that the generated power falls within the guide distribution. The guide distribution and its activation area are shown in Figure 3 shown.

[0119] 4.4 Draw random numbers and extract power within a specific guide distribution;

[0120] 4.5 Output the extracted power, which is the power of the overall energy storage in this window.

[0121] By modifying the input guided distribution probability p and guided distribution proportion v, adjusting the guided distribution probability and guided distribution proportion, the relevant characteristics of the generated power path, such as target deviation and path divergence, can be affected. The target deviation is the ratio of the error between the actual outflow power of the path and the target outflow power to the target outflow power value, and the path divergence is the ratio of the average variance of the generated path at this time step to the average variance of the generated path at the most extreme time step.

[0122] 5. Allocate the total energy storage power to various energy storage measures and add them to the path;

[0123] First, exclude off-grid measures from all measure sets, and filter measures that can interact with electricity into the effective measure set.

[0124] Then construct the sorting index T d , sort the measures in the effective measure set according to, and the expression is as follows:

[0125] T d =|E s.end -E s | / [(N w -n w +1)*P s.max ]

[0126] Among them, E s.end is the electricity value at the end of the measure period, E s is the actual power value in the current window, N w is the number of remaining windows in the cycle, n w is the serial number of the current window, P s.max is the absolute value of the maximum power of this measure. Sorting index T d The meaning of is the minimum number of windows required to meet the end-of-period power value. In order to maximize the end-of-period power requirements of each measure, T d The smaller the measure, the higher the priority and the higher the ranking. It is worth noting that the number of remaining windows in a cycle is not necessarily equal to the number of all windows in the cycle minus the current window number, because the impact of the off-network time cannot be ignored. The number of remaining windows in a cycle should be equal to the number of remaining windows in the network state.

[0127] Finally, the window power is allocated one by one according to the order. When the remaining power to be allocated is higher than the maximum power of the current measure, the measure is output at the maximum power and enters the next measure. When the remaining power to be allocated is less than or equal to the maximum power of the current measure, the measure is output at the power to be allocated, the power allocation is terminated, and the power and power changes of each measure at this time step are recorded and stored.

[0128] 6. Determine whether the next window exists; if so, return to step 2.4; otherwise, proceed to the next step;

[0129] 7. Add the path to the feasible path library of energy storage charging and discharging strategy;

[0130] 8. Determine whether the maximum number of paths has been reached; if not, return to step 2.2; otherwise, proceed to the next step;

[0131] 9. Output a feasible path library for regional energy storage charging and discharging strategies.

[0132] In this process, the influence of charge and discharge efficiency needs to be considered.

[0133] The power of the defined measure is negative when charging and positive when discharging. Under the same window power, the charging and discharging efficiency of the measure is different, and the change in its power is also different. When the power is positive, that is, when the measure is discharging, the formula for the change in window power is as follows:

[0134] delta E =-power*delta t *n T / η discharge

[0135] Among them, delta E is the amount of change in the power measurement, power is the window power, delta t is the step size for each time step, N T is the number of time steps in a window, η discharge is a measure of discharge efficiency.

[0136] When the power is negative, that is, when charging, the formula for the window power change is as follows:

[0137] delta E =-power*delta t *n T *η charge

[0138] Among them, delta E is the amount of change in the power measurement, power is the window power, delta t is the step size for each time step, N T is the number of time steps in a window, η charge To measure charging efficiency.

[0139] Ideally, the charging and discharging efficiency is 1, but in practical applications, the loss of power reserve charging and discharging needs to be taken into account, and the charging and discharging efficiency is less than 1. Different types of power reserve have different losses, so the charging and discharging efficiency is also different. For example, the efficiency of pumped storage is different from that of electrochemical energy storage. The efficiency of electrochemical energy storage varies according to its type, and is also affected by external factors such as environment and temperature.

[0140] Further, step 1: generate a reference path, specifically:

[0141] The process of generating the reference path is to first prepare the key data required for path generation, including the power feasible domain and the power guiding area. The specific steps are as follows: obtaining the current state and final state requirements of each generalized energy storage device, and calculating the power feasible domain of each generalized energy storage device at each time step in accordance with the principle of not violating the physical and economic constraints of the device itself; further, setting the power guiding area according to the power feasible domain, and controlling the divergence and other characteristics of the generated path by adjusting the guiding area parameters to cope with diversified generation needs;

[0142] Then, starting from the starting point of a single flat start path (i.e., a path where the energy storage output is set to zero throughout the entire scheduling cycle), a path is randomly generated: at each section, the power feasible domain and the power guidance area must be updated first, where the power guidance value should not exceed the net outflow of the remaining period; then, the power of the generalized energy storage device or its cluster at any section is used as the generation object, and the updated power feasible domain is used as the upper and lower bounds to complete the random power generation; all the generated results are collected to construct a path;

[0143] Repeat the path generation steps to generate multiple paths and put them into the path library; evaluate each path in the path library and select the best path as the benchmark path.

[0144] Step 2: If the mode has not been determined, set it to the mode selection state; otherwise, set it to the state where the mode has been determined and jump directly to step 4;

[0145] The mode refers to a set of charging and discharging decision constraints of generalized energy storage or its cluster that meets certain rules;

[0146] The rules (indicated by R) include but are not limited to polarization and anti-polarization rules: the polarization rule (indicated by R=1) means that the current charge and discharge power is allowed to be increased but not decreased, and the anti-polarization rule (indicated by R=-1) is the opposite of the polarization rule, which means that the current charge and discharge power is allowed to be decreased but not increased;

[0147] The modes include but are not limited to polarization mode, reverse polarization mode, consumption mode and supply guarantee mode, etc.: polarization mode refers to strengthening the current charge / discharge intensity within the power feasible domain, that is, charging more when charging, or discharging more when discharging, thereby releasing the power regulation potential of energy storage; reverse polarization mode refers to weakening the current charge / discharge intensity within the power feasible domain, that is, charging less when charging, and discharging less when discharging, thereby suppressing the power regulation potential of energy storage; consumption mode refers to applying polarization rules in the discharge state and reverse polarization rules in the charging state, and is named because it is conducive to the consumption of new energy; supply guarantee mode refers to using polarization rules in the charging state and reverse polarization rules in the discharge state, and is named because it is conducive to the supply guarantee during the tight balance period;

[0148] The entering into the mode selection state refers to the process of testing different modes in sequence when the mode is not determined, and selecting the best mode according to the testing results.

[0149] Step 3: Take a pattern from the pattern library;

[0150] Step 4: Construct a feasible path library of generalized energy storage devices or their clusters under a given mode;

[0151] like Figure 4As shown, the construction of the mode power path library of the specified mode includes the following steps:

[0152] 4.1 Locate the starting point of a new path and establish a path library corresponding to the current mode;

[0153] 4.2 Reset the total periodic outflow of the generalized energy storage device or its cluster;

[0154] 4.3 Take out a scheduling window;

[0155] 4.4 Determine the charge and discharge status of the current window based on the previous value;

[0156] The previous value refers to the power value obtained by the current window at the end of the previous round of optimization; the charging and discharging state is represented by the symbol P, when P is equal to 1, it is the discharging state, and when P is equal to -1, it is the charging state; judging the charging and discharging state of the current window according to the previous value means that when the current value is greater than 0, the energy storage is in the discharging state, at which time P is equal to 1, otherwise it is in the charging state, at which time P is equal to -1.

[0157] 4.5 Under the premise of not violating the total outflow power constraint of the period, generate the total power added value of the generalized energy storage device or its cluster in the dispatch window under the specified mode according to the current rules;

[0158] The total power of the power reserve window in the specified mode is generated, such as Figure 5 As shown, the following steps are included:

[0159] 4.5.1 Determine the activation region of the guide distribution based on the previous generation value;

[0160] The guided distribution refers to a probability distribution space that can guide the power decision of a generalized energy storage device or its cluster into a feasible domain according to the user's risk preference setting, and has the function of randomly exploring the optimization direction within the feasible domain; the activated area refers to a sub-area of ​​the guided distribution where the power value is actually extracted after considering the mode during the optimization process.

[0161] 4.5.2 Redistribute the probability distribution of the activation area of ​​the guided distribution;

[0162] The probability distribution refers to the probability law of the power value randomly drawn in the activation area of ​​the guided distribution; wherein the probability density refers to the proportion of the number of power values ​​drawn by setting random numbers in the corresponding area to the total number of times drawn, and the sum of the probability densities of the guided distribution is 1; the process of reallocating the probability distribution of the activation area of ​​the guided distribution refers to re-dividing the inner and outer guided sub-distributions of the activation area in step 4.5.2 according to the different probability densities of the sub-areas of the guided distribution, so that the sum of the probability densities of the activation area is still 1;

[0163] 4.5.3 Randomly extract the total power of generalized energy storage devices or their clusters based on the guided distribution activation area;

[0164] 4.5.4 Allocate the total power to each measure within the cluster, such as Figure 6 shown.

[0165] First, exclude off-grid measures from all measures, and filter measures that can interact with electricity into the effective measures set. Then construct the sorting index T d , sort the measures in the effective measure set according to, and the expression is as follows:

[0166] T d =|E s.end -E s | / [(N w -n w +1)*P s.max ]

[0167] Among them, E s.end is the electricity value at the end of the measure period, E s is the actual power value in the current window, N w is the number of remaining windows in the cycle, n w is the serial number of the current window, P s.max is the absolute value of the maximum power of this measure. Sorting index T d The meaning of is the minimum number of windows required to meet the end-of-period power value. In order to maximize the end-of-period power requirements of each measure, T d The smaller the measure, the higher the priority and the higher the ranking. It is worth noting that the number of remaining windows in a cycle is not necessarily equal to the number of all windows in the cycle minus the current window number, because the impact of the off-network time cannot be ignored. The number of remaining windows in a cycle should be equal to the number of remaining windows in the network state.

[0168] Finally, the window power is allocated one by one according to the sorting results. When the remaining power to be allocated is higher than the maximum power of the current measure, the measure is output at the maximum power and enters the next measure. When the remaining power to be allocated is less than or equal to the maximum power of the current measure, the measure is output at the power to be allocated, the power allocation is terminated, and the power and power changes of each measure at this time step are recorded and stored.

[0169] 4.6 Deduct the window power obtained by the total power added integral in step 4.5 from the total outflow power of the cycle;

[0170] The purpose of deducting the window power from the total period outflow is to meet the total period outflow of the set regional power reserve, that is, the final power is the same as the initial power. After the power generation of the current window is completed, this power value should be subtracted from the total period outflow of the set regional power reserve.

[0171] 4.7 Determine whether there is a next window; if satisfied, return to step 4.3; otherwise, proceed to the next step;

[0172] 4.8 Add the path to the pattern path library;

[0173] 4.9 Determine whether the maximum number of paths has been reached; if so, proceed to the next step; otherwise, return to step 4.1;

[0174] 4.10 Output mode power path library.

[0175] Step 5: Carry out quantitative evaluation on each path in the power path library and calculate the objective function value;

[0176] The quantitative evaluation of the paths in the library refers to taking the risk cost as the objective function value, calculating the risk cost of all paths under different modes, and thus quantifying the optimization effect under the corresponding mode. The quantitative evaluation refers to optimizing the thermal power reserve and evaluating the capacity risk cost after optimizing the generalized energy storage device or its cluster path for each feasible path by inputting the transmission power curve of the relevant interconnection line in the target area. The total risk cost refers to the superposition of the capacity risk cost and the electricity risk cost.

[0177] The multi-window thermal power reserve dynamic optimization method specifically includes:

[0178] A day or a cycle is divided into multiple windows, and the same number of time steps is set in each window. For the capacity backup resources, an initial sorting table is first generated according to the empirical sequence method, and then the cost-performance ratio index is calculated for secondary sorting to select the optimal backup measure to achieve and verify the goal of minimizing the total risk cost in the region.

[0179] Since backup resources involve various situations, including but not limited to start-up and shutdown costs, start-up and shutdown times, prices, types, output limits, output costs, whether to participate in deep regulation, and deep regulation costs, it is necessary to establish a specific unit economic cost indicator to achieve the most economical backup resource allocation.

[0180] To solve the above problems, the CPR index, i.e., the ratio of the risk control cost of the standby unit to the effective power it can provide under the risk scenario, is increased to consider the deep adjustment cost to improve the index accuracy. The improved CPR formula is:

[0181]

[0182] Among them, C cap,i To control the capacity cost of i, C st,i To control the start-stop cost of i, C risk,i,s,t To control the risk electricity cost of i, ΔR i is the alternative cost of removing the control object i.

[0183] The replacement cost is to evaluate the performance of the control object by using the alternative control cost after removing the control object. The formula is:

[0184] ΔR i =R up,i -(R dn,P_min -R dn,P_min-P_min,i )

[0185] Among them, R up,i To remove the upward adjustment cost of the control object i, R dn,P_min The minimum output P before removing the measures min Calculate the downgrade cost, R dn,P_min-P_min,i The downward adjustment cost is calculated by subtracting the minimum output after removing the measure i from the minimum output after removing the measure.

[0186]

[0187] Among them, u up,i is the expected load rate of control measure i, P max,i To control the maximum output of i, T is the window length, price up,i Increase the risk cost for the unit that controls i.

[0188]

[0189] Among them, R dn,i is the downward adjustment cost of control measure i, P min,i is the minimum output of control measure i, P min is the minimum output sum of all selected measures in the previous round, R dn represents the average total cost of downward adjustment risk in the previous round, I dn,i It represents the benefit obtained by controlling object i from participating in deep adjustment.

[0190] like Figure 7 As shown in the figure, at each time step, the flexible units are classified according to whether they have been selected. The selected units can be removed by calculating the CPR index, and the unselected units can be turned on by calculating the CPR index. In the same window, all time steps use the backup resource startup plan table generated by the decision of the window. The start and stop time of the unit should be considered between time steps, and finally the optimal configuration of backup resources is completed.

[0191] In the thermal power reserve resource optimization configuration algorithm, the optimization process is divided into rounds, such as Figure 8As shown in the figure, each large cycle is divided into several windows, with the first window as the entry point and the rounds determined. If it is the first round of optimization, the expected output line to be regulated in this scenario is generated, and the backup measures are sorted according to the empirical sorting method: the available measure set of this window is screened, and sorted from small to large according to the scheduling order. If the scheduling order is the same, it is sorted from large to small according to the maximum output. After sorting, the sum of the capacity of all measures can cover the expected output line to be regulated.

[0192] If it is not the first round of optimization, the CPR prioritization method is used after updating the local and remote action sets, and the backup resources are arranged according to the CPR indicators.

[0193] When a round faces the first window, the control start and stop state selects the state corresponding to the same window in the previous round. When the window is greater than 1, the start and stop state of the previous window is introduced and the capacity reserve is considered to increase or decrease measures. The process is as follows: calculate the CPR index of all unselected available measures at the specified power level. If there are measures with a CPR value less than 1, update them to the new power level, and loop until there are no measures with a CPR value less than 1 under the new power standard. If there are measures with a CPR value greater than 1, remove the measure and update the power level, and loop until there are no measures with a CPR value greater than 1 in the selected units.

[0194] After completing the set of control measures, the deep peak-shaving cost will be evaluated when any output scenario to be regulated is lower than the lower output limit within the window. The deep peak-shaving evaluation includes the evaluation of the total risk cost of deep regulation and the deep regulation benefits of each deep peak-shaving measure. When the deep peak-shaving measures cannot meet the demand, the reference abandoned wind and solar costs are used for estimation. The window is then moved back, and the control measures in the window are re-prioritized in turn until the last window is prioritized, and the backup resource configuration plan for this round of optimization is obtained. Enter the next round of optimization until multiple rounds of iterative optimization meet the convergence conditions, and finally obtain the resource configuration plan for thermal power reserve.

[0195] Step 6: Select the optimal path for a given mode by comparing the values ​​of the multi-path objective function;

[0196] Selecting the optimal path for a given mode refers to selecting a path with the smallest objective function value, and this path is the best path in the corresponding mode.

[0197] Step 7: If it is in the mode selection state, determine whether there is a next mode; if so, return to step 3; if not, proceed to the next step; if it is not in the mode selection state, take the preferred path under the given mode as the optimal one in this round, and then proceed to step 9;

[0198] The judgment condition means that if it is in the mode selection state, the modes in the mode library are traversed and optimization is performed under the corresponding mode; if it is not in the mode selection state, that is, the optimization mode of this round and subsequent rounds has been determined, the preferred path under the given mode is used as the optimal path of this round.

[0199] Step 8: By comparing the objective function values ​​of the optimal paths in each mode, the mode with the best objective function value is determined as the mode for the subsequent rounds;

[0200] Step 9: Determine whether the convergence condition is met; if so, proceed to the next round and return to step 2; otherwise, take the optimal path of this round as the output result.

[0201] The convergence condition means that if the quantitative evaluation result of the optimal power path selected in step 8 is improved relative to the quantitative evaluation result of the power path selected in the previous round when in the mode selection state, then the best mode selected in step 8 is used as the given mode for the subsequent round of optimization, and the optimal power path selected in the current round is used as the reference path for the next cycle, and the next round of optimization is performed, thereby forming a self-iterative optimization mechanism; if it is not in the mode selection state, then when the quantitative evaluation result of the optimal power path selected in step 6 is improved relative to the quantitative evaluation result of the power path selected in the previous round, the next round of self-iterative optimization is continued.

[0202] Based on the same inventive concept, the present invention also provides a multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method and system, including:

[0203] Generate benchmark path module, in the first round of optimization, generate a massive path library from a single flat start point and select a path;

[0204] Construct a mode path module, and use the previous generation value as the starting point for generating the next generation charge and discharge path; then, divide the subsequent rounds of search into polarization rules and anti-polarization rules according to the same or opposite direction as the previous generation solution, thereby deriving multiple modes, and improving the generation of charge and discharge paths for multiple modes. By comparing and optimizing the generated paths under different modes, the optimization direction of the subsequent generation is determined;

[0205] The self-iterative optimization module selects the best mode as the optimization mode specified in the subsequent rounds, and uses the previous generation solution as the initial value for the next generation path generation. After multiple generations of cycles, the self-iterative evolution of the generalized energy storage charging and discharging strategy is achieved.

Claims

1. A self-iterative optimization method for generalized energy storage charging and discharging decision-making driven by multiple modes, characterized in that: include: Step 1: Generate a reference path; Step 2: Determine the current optimization mode. If the mode has not been determined, define the current state as a mode selection state; otherwise, define the current state as a mode determination state and proceed to step 4. Step 3: Take out a pattern from the pattern library; Step 4: Construct a feasible path library of generalized energy storage devices or generalized energy storage device clusters under the current mode; Step 5: Carry out quantitative evaluation on each path in the feasible path library and calculate the objective function value; Step 6: Select the optimal path for the current mode by comparing the multi-path objective function values; Step 7: If in the mode selection state, it is necessary to determine whether there is a next mode; if so, return to step 3; If it does not exist, go to the next step; if it is not in the mode selection state, take the preferred path in the current mode as the best path for this round, and then go to step 9; Step 8: By comparing the objective function values ​​of the optimal paths in each mode, the mode with the smallest objective function value is determined as the mode for the subsequent rounds; Step 9: Determine whether the convergence condition is met; if not, proceed to the next round and return to step 2; otherwise, take the optimal path of this round as the output result.

2. The multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method according to claim 1 is characterized in that: The path is a trajectory of the evolution of the amount of energy of a generalized energy storage device or a generalized energy storage device cluster over time on a plane with time and energy storage amount as coordinates; The generalized energy storage device is a power-adjustable resource in which the accumulated power generation / consumption in the power system is greater than or equal to a given value within a given scheduling cycle, and has interruptible and shiftable properties, and / or an energy storage upper limit is less than a given value.

3. The multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method according to claim 1 is characterized in that: The step 1 comprises: Prepare the data required for path generation, the data including the power feasible domain and the power guiding area, specifically: obtain the current state and final state requirements of each generalized energy storage device, calculate the power feasible domain of the generalized energy storage device at each time step based on the physical and economic constraints of the generalized energy storage device; then set the power guiding area according to the power feasible domain, and adjust the guiding area parameters to control the characteristics of the generated path; Starting from the starting point of a single flat start path, a path is randomly generated. The single flat start path is a path in which the energy storage output is set to zero throughout the entire scheduling period. Specifically: at each section, the power feasible domain and the power guidance area are first updated, where the power guidance value does not exceed the net outflow of the remaining period; then the power of the generalized energy storage device or the generalized energy storage device cluster at any section is used as the generation object, and the updated power feasible domain is used as the upper and lower bounds to complete the random power generation; all the generation results are collected to construct a path; Repeat the steps of randomly generating paths, generate multiple paths, establish a path library, and store each generated path in the path library; use the risk cost as the objective function value, quantitatively evaluate each path, calculate the risk cost of all paths under different modes, and select the path with the smallest objective function value as the benchmark path.

4. The multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method according to claim 1 is characterized in that: The mode described in step 2 is a set of charging and discharging decision constraints of a generalized energy storage device or a generalized energy storage device cluster that satisfies the rule; The rule R includes but is not limited to a polarization rule and an anti-polarization rule; the polarization rule allows the current charge and discharge power to be increased but not decreased, which is represented by R=1; the anti-polarization rule allows the current charge and discharge power to be decreased but not increased, which is represented by R=-1; The modes include but are not limited to polarization mode, reverse polarization mode, consumption mode and supply guarantee mode; the polarization mode refers to strengthening the current charge / discharge intensity within the power feasible domain, charging more during charging, or discharging more during discharging, thereby releasing the power regulation potential of energy storage; The reverse polarization mode refers to weakening the current charge / discharge intensity within the power feasible domain, charging less during charging and discharging less during discharging, thereby suppressing the power regulation potential of energy storage; the consumption mode refers to applying the polarization rule in the discharge state and the reverse polarization rule in the charge state; the supply guarantee mode refers to using the polarization rule in the charge state and the reverse polarization rule in the discharge state; The entering into the mode selection state refers to the process of testing different modes in sequence when the mode is not determined, and selecting the best mode according to the testing results.

5. The multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method according to claim 1 is characterized in that: The step 4 comprises: Step 4.1: Locate the starting point of a new path and establish a path library corresponding to the current mode; Step 4.2: Reset the total period outflow of the generalized energy storage device or its cluster; Step 4.3: Take out a scheduling window; Step 4.4: Determine the charge and discharge status of the current window based on the previous value; Step 4.5: Based on the total outflow power constraint of the period, generate the total power added value of the generalized energy storage device or its cluster in the scheduling window in the current mode according to the current rules, and perform rolling update of the path; Step 4.6: Deduct the window power obtained by the total power added integral in step 4.5 from the total outflow power of the cycle; Step 4.7: Determine whether there is a next window; if satisfied, return to step 4.3; otherwise, proceed to the next step; Step 4.8: Add the updated path to the path library corresponding to the current mode; Step 4.9: Determine whether the maximum number of paths is reached; if so, proceed to step 4.10; otherwise, return to step 4.1; Step 4.10: Output the pattern path library.

6. The multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method according to claim 5 is characterized in that: The previous value described in step 4.4 is the power value obtained by the current window at the end of the previous round of optimization; the charging and discharging state P, when P=1, it indicates the discharging state, and when P=-1, it indicates the charging state; the charging and discharging state of the current window is determined according to the previous value, specifically, when the current value is greater than 0, the energy storage is in the discharging state, at which time P=1, otherwise it is in the charging state, P=-1.

7. The multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method according to claim 5, characterized in that: The step 4.5 comprises: Step 4.5.1: Determine the activation region of the bootstrap distribution based on the previous generation value: According to the combination of the preset optimization rules and the charge and discharge state of the generalized energy storage device or the generalized energy storage device cluster, the activation area is modified according to the following rules, specifically: determine whether the product of P and R is equal to 1; if satisfied, modify the lower limit of the guided distribution activation area to the previous generation value; if equal to -1, modify the upper limit of the guided distribution activation area to the previous generation value; The guided distribution is a probability distribution space that guides the power decision of the generalized energy storage device or its cluster into a feasible domain according to the risk preference setting of the user, and randomly explores the optimization direction in the feasible domain; The activation region is a sub-region of the guide distribution where power value extraction is actually performed after the mode is considered during the optimization process; Step 4.5.2: Redistribute the probability distribution of the activation area of ​​the guided distribution; The probability distribution refers to the probability law of the power value randomly drawn in the activation area of ​​the guided distribution; wherein the probability density is the proportion of the number of power values ​​drawn by setting random numbers in the corresponding area to the total number of times drawn, and the sum of the probability densities of the guided distribution is 1; the process of reallocating the probability distribution of the activation area of ​​the guided distribution refers to re-dividing the inner and outer guided sub-distributions of the activation area in step 4.5.2 according to the different probability densities of the sub-areas of the guided distribution, so that the sum of the probability densities of the activation area is still 1; Step 4.5.3: Randomly extract the generalized energy storage device or the total power of its cluster based on the guided distribution activation area; Step 4.5.4: Allocate the total power to each measure in the cluster.

8. The multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method according to claim 1, characterized in that: The step 5 includes: taking the risk cost as the objective function value, calculating the risk cost of all paths under different modes; The quantitative assessment refers to optimizing thermal power reserve and evaluating capacity risk cost after inputting the transmission power curve of the relevant interconnection lines in the target area and optimizing the generalized energy storage equipment or its cluster path for each feasible path. The total risk cost refers to the superposition of capacity risk cost and electricity risk cost.

9. The multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method according to claim 1, characterized in that: The step 6 includes: selecting a path with the smallest objective function value as the best path in the current mode.

10. The multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method according to claim 1, characterized in that: The step 7 includes, if it is in the mode selection state, traversing the modes in the mode library and optimizing under the corresponding mode; if it is not in the mode selection state, the optimization mode of this round and subsequent rounds has been determined, then taking the preferred path under the current mode as the optimal path of this round.

11. The multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization method according to claim 1, characterized in that: The convergence condition described in step 9 is specifically that if in the mode selection state, the quantitative evaluation result of the best path selected in step 8 is improved relative to the quantitative evaluation result of the path selected in the previous round, and the objective function value is reduced, then the mode selected in step 8 is used as the given mode for the subsequent round of optimization, and the best path selected in the current round is used as the reference path for the next cycle, and the next round of optimization is performed to achieve self-iterative optimization; if it is not in the mode selection state, then when the quantitative evaluation result of the best path selected in step 6 is improved relative to the quantitative evaluation result of the path selected in the previous round, and the objective function value is reduced, the next round of self-iterative optimization is continued.

12. A multi-mode driven generalized energy storage charging and discharging decision self-iterative optimization system, characterized in that: include: Generate benchmark path module, in the first round of optimization, generate a massive path library from a single flat start point and select the best path; Construct a mode path module, and use the previous generation value as the starting point for generating the next generation charge and discharge path; then, divide the subsequent rounds of search into polarization rules and anti-polarization rules according to the same or opposite direction as the previous generation solution, thereby deriving multiple modes, and improving the generation of charge and discharge paths for multiple modes. By comparing and optimizing the generated paths under different modes, the optimization direction of the subsequent generation is determined; The self-iterative optimization module selects the best mode as the optimization mode specified in the subsequent rounds, and uses the previous generation solution as the initial value for the next generation path generation. After multiple generations of cycles, the self-iterative evolution of the generalized energy storage charging and discharging strategy is achieved.

Images