A wind storage combined optimization configuration method based on power spectral density

Abstract

The application discloses a wind storage joint optimization configuration method based on power spectrum density, and belongs to the technical field of wind power generation dispatching. The method is a storage energy configuration optimization algorithm based on power spectrum density and used for suppressing source side wind power fluctuation. The power spectrum density of wind farm output is fitted and divided by using a hinge model, a suitable low frequency decomposition frequency is found, and then the optimal grid-connected power curve and the storage energy system power curve are determined by using fast Fourier transform decomposition. The grid-connected output satisfies the 1min and 10min limit value constraints proposed in the Technical Regulation for Access to Power System of Wind Farm. In addition, by considering the charge and discharge loss coefficient and single-day energy balance in the operation of the storage energy system, the rated power and the rated capacity of the storage energy system are determined, so that the SOC of the storage energy device is at the optimal position at the beginning of each operation cycle. The application suppresses wind power output fluctuation, improves the wind power output consumption proportion and reduces the source end storage energy cost.

Description

Technical Field

[0001] This invention belongs to the field of wind power generation dispatching technology, and specifically relates to a wind-storage joint optimization configuration method based on power spectral density. Background Technology

[0002] Promoting an energy revolution, continuously increasing the proportion of non-fossil energy in total energy consumption, building a clean, low-carbon, safe, and efficient energy system, and improving energy supply security are the future directions of energy development. Driven by policy support and ever-increasing energy demand, clean energy, primarily wind power, has developed on a large scale in my country, with installed capacity continuously increasing. However, the inherent volatility, intermittency, and anti-peak-shaving characteristics of wind power, as well as its harmonic generation, lead to numerous problems when it participates in large-scale grid connection and system dispatch. Significantly increasing the wind power output absorption ratio while maintaining grid safety and stability is crucial for controlling source-end power generation costs, and the rational application of advanced energy storage technologies can effectively solve this problem.

[0003] The proper combination of wind power and energy storage systems can not only smooth power generation output and increase the proportion of wind power consumption, but also improve the flexibility and resilience of the entire power grid system. Currently, the main configuration methods include filtering control algorithms, fuzzy control algorithms, moving average algorithms, and model predictive control algorithms. Among these, filtering control algorithms are the simplest and most widely used for smoothing power fluctuations. Whether it's wavelet packet decomposition or empirical mode decomposition, the core of filtering control algorithms is the reasonable selection of filtering parameters to obtain a relatively smooth output curve. While wavelet packet decomposition can quickly handle short-term power fluctuations, the algorithm is relatively complex, and the selection of wavelet basis functions and the number of decomposition levels is cumbersome. Empirical mode decomposition decomposes signals based on the time-scale characteristics of the data itself, without the need to pre-set any basis functions, but it is prone to mode aliasing.

[0004] Since the power system has a limited capacity to accept fluctuating energy, in order to ensure that the wind power output meets the limit requirements for active power variation proposed in the "Technical Regulations for Wind Farm Access to Power System", the fast Fourier transform decomposition method can be used to decompose the wind power, thereby determining the optimal grid-connected power and energy storage power, so as to improve the wind power output absorption ratio and reduce the source-end power generation cost. Summary of the Invention

[0005] The purpose of this invention is to propose a wind-storage joint optimization configuration method based on power spectral density. The method is characterized by being a power spectral density-based energy storage configuration optimization algorithm for mitigating source-side wind power fluctuations. This energy storage optimization configuration method includes using power spectral density to find the low-frequency decomposition frequency, then using the Fast Fourier Transform decomposition method to determine the optimal grid-connected power and energy storage power. Finally, it considers energy loss and daily energy balance to determine the rated power and capacity of the energy storage system, ensuring that the energy storage system (SOC) is in an optimal position at the beginning of each operating cycle. Specifically, it includes:

[0006] Step 1: Based on Passevar's theorem, plot the power spectral density image of the wind farm based on its active power output;

[0007] Step 2: Use the hinge model to segment the power spectral density (PSD) image and fit it into several affine functions. Then find two fitted straight lines whose slopes are close to the Kolmogorov slope of "-5 / 3" and take the frequency values ​​corresponding to their nodes as the low-pass decomposition frequencies.

[0008] Step 3: Using the low-pass decomposition frequency as a benchmark, the wind power output is decomposed using the fast Fourier transform decomposition method; the low-frequency part is the power directly connected to the grid, and the medium and high frequency part is the power of the energy storage system. Verify whether the grid-connected power meets the fluctuation requirements of 1 minute and 10 minutes.

[0009] Step 4: Define the upper and lower limits of the state of charge of the energy storage system, and calculate the rated power and rated capacity of the energy storage system considering daily energy balance and charging and discharging losses;

[0010] Step 5: Calculate the grid-connected utilization conversion rate of wind power generation and the configuration cost of the energy storage system.

[0011] In step 1, for a discrete signal x(n), n = 0, 1, 2, ..., N-1, the mathematical model of its power spectral density S(k) is as follows:

[0012]

[0013] In the formula, X(k) is the Fourier transform form of the discrete signal x(n), k = 0, 1, 2, ..., N-1. j is the imaginary number.

[0014] The windowed average periodogram method used is as follows:

[0015]

[0016] In the formula, M is the number of segments in the random sequence, L is the length of each segment, and ω(n) is the window function.

[0017] The hinge model in step 2 is as follows:

[0018]

[0019] Its continuity condition is:

[0020] H j-1 (k j )=H j (k j )=H j+1 (k j )

[0021] In the formula, k j The x-axis coordinate interval is [k1, k N The x-coordinate of the node within ], H j For k j For the hinge model of the intermediate node, H l,j and H r,j H are hinges respectively j The left and right halves of H j-1 and H j+1 They are adjacent to H respectively j Two hinge models, one on the left and one on the right;

[0022] The fitted radiometric function is in the form of:

[0023] P x (f)≈αf -β +γ

[0024] In the formula, f is the frequency, α is a scale, β is a power value, and γ is a constant; if we assume that γ is 0, then taking the logarithm of both sides of the equation yields:

[0025] logP x (f)≈logα-βlogf

[0026] In logarithmic coordinates, β represents the slope and logα represents the y-intercept; the piecewise slope of the power spectral density may vary depending on the model order and the geographical location of the wind farm.

[0027] In step 3, for a continuous time-domain signal x(t), the discrete signals corresponding to its N points are taken as x(n), then its discrete Fourier transform signal is:

[0028]

[0029]

[0030] In the formula, f(n) is the corresponding frequency of x(n), f s T is the sampling frequency. s The sampling period.

[0031] For the original output power P of the wind farm wind N sampling points P wind (n) Performing a Fast Fourier Transform (FFT) yields the spectral analysis results:

[0032]

[0033] Subsequently, the obtained frequency band F was analyzed. P The frequencies are divided into F based on high, medium, and low frequencies. PH F PM F PL The corresponding frequency band amplitudes are X PH X PM X PL Using the low-frequency band X PL Example of calculation method:

[0034]

[0035] P L (n) = IFFT(X) PL (n),N)

[0036] In the formula, For F PL Regarding the frequency center point f s / 2 symmetrical frequency band, P L (n) is the low-frequency output power obtained through inverse fast Fourier transform, which is also the effective grid-connected power of wind power generation;

[0037] Grid connection constraints for wind power generation are:

[0038]

[0039] In the formula ΔP grid,1min and ΔP grid,10min These are the deviations of the maximum and minimum grid-connected power fluctuations within 1 minute and 10 minutes, respectively, P. r This refers to the nominal value of the installed capacity of the wind farm.

[0040] After considering the charging and discharging loss coefficient in step 4, the corrected power of the energy storage system is:

[0041]

[0042] In the formula, P E0 To obtain the power of the energy storage system from the decomposition, η d For discharge conversion efficiency, η c The charging conversion coefficient;

[0043] Considering daily energy balance, that is, assuming that the charging and discharging can be balanced within one working cycle, the final SOC value of the ESS on that day should be as close as possible to the initial SOC value of the energy storage system. Since the sampling period of the sampled data is the same, the energy balance can be converted into power balance:

[0044]

[0045] The formula for calculating the rated power of an energy storage system is:

[0046]

[0047] The calculation process for the rated capacity of an energy storage system is as follows:

[0048]

[0049]

[0050]

[0051]

[0052]

[0053] In the formula, E(n) is the cumulative charge / discharge amount of the ESS in the current state, and SOC0 is the initial SOC of the ESS. U and SOC L These are the upper and lower limits of SOC, E. rated This is the rated capacity of the ESS.

[0054] In step 5, the grid-connected portion of the wind farm's power output is:

[0055]

[0056] The grid-connected conversion rate of wind power is:

[0057]

[0058] The configuration cost of the energy storage system is:

[0059] C ESS =P rated C P +E rated C E

[0060] In the formula, C P Price per unit power of energy storage system, in RMB 10,000 / MW; C E Price per unit capacity of energy storage system, in RMB 10,000 / MWh.

[0061] The beneficial effect of this invention is that it uses the Fast Fourier Transform decomposition method to determine the optimal grid-connected power and energy storage power, and then considers the charging and discharging losses and daily energy balance to determine the rated power and capacity of the energy storage system, thereby reducing the cost of energy storage and solving the problem of wind power output power fluctuation. Attached Figure Description

[0062] Figure 1 Flowchart of wind farm energy storage configuration.

[0063] Figure 2 Wind farm output power spectral density and its fitting results;

[0064] Figure 3 Original output power and target grid-connected power of wind farm;

[0065] Figure 4 The power and SOC changes of the energy storage system; where a is the power curve of the energy storage system; b is the SOC change curve of the energy storage system; Detailed Implementation

[0066] This invention proposes a wind-storage joint optimization configuration method based on power spectral density. This method is an energy storage configuration optimization algorithm based on power spectral density for mitigating wind power fluctuations at the source side. The invention is described below with reference to the accompanying drawings.

[0067] like Figure 1 The diagram shows the wind farm energy storage configuration flowchart. The low-frequency decomposition frequency is found using the power spectral density, and then the optimal grid-connected power and energy storage power are determined using the Fast Fourier Transform (FFT) decomposition method. Then, considering energy loss and daily energy balance, the rated power and capacity of the energy storage system are determined to ensure that the energy storage system (SOC) is in an optimal position at the beginning of each operating cycle. The specific implementation steps are as follows:

[0068] 1) Based on Passevar's theorem, draw the power spectral density image of the wind farm based on its active power output.

[0069] For a discrete signal x(n), the mathematical model for its power spectral density is:

[0070]

[0071] The windowed average periodogram method used is as follows:

[0072]

[0073] In the formula, M is the number of segments in the random sequence, L is the length of each segment, and ω(n) is the window function.

[0074] 2) Use the hinge model to segment the power spectral density image and fit it into several affine functions. Then find two fitted straight lines with slopes close to the Kolmogorov slope of "-5 / 3", namely Part I and Part II. Take the frequency values ​​corresponding to their nodes as the low-pass decomposition frequencies.

[0075] The hinge model is as follows:

[0076]

[0077] Its continuity condition is:

[0078] H j-1 (k j )=H j (k j )=H j+1 (k j )

[0079] In the formula, k j The x-axis coordinate interval is [k1, k N The x-coordinate of the node within ], H j For k j For the hinge model of the intermediate node, H l,j and H r,j H are hinges respectively j The left and right halves of H j-1 and H j+1 They are adjacent to H respectively j Two hinge models, one on the left and one on the right.

[0080] The fitted radiometric function is in the form of:

[0081] P x (f)≈αf -β +γ

[0082] In the formula, f is the frequency, α is a scale, β is a power value, and γ is a constant. If we assume γ is 0, then taking the logarithm of both sides of the equation yields:

[0083] logP x (f)≈logα-βlogf

[0084] In logarithmic coordinates, β represents the slope, and logα represents the y-intercept. The piecewise slope of the power spectral density may vary depending on the model order and the geographical location of the wind farm.

[0085] 3) Using the low-pass decomposition frequency as a benchmark, the wind power output is decomposed using the Fast Fourier Transform (FFT) method. The low-frequency component represents the directly grid-connected power, while the mid-to-high-frequency component represents the energy storage system power. The grid-connected power is verified to meet the fluctuation requirements of 1 minute and 10 minutes.

[0086] For a continuous-time signal x(t), let x(n) be the discrete signal corresponding to N points. Then its discrete Fourier transform signal is:

[0087]

[0088]

[0089] In the formula, f(n) is the corresponding frequency of x(n), f s T is the sampling frequency. s The sampling period.

[0090] For the original output power P of the wind farm wind N sampling points P wind (n) Performing a Fast Fourier Transform (FFT) yields the spectral analysis results:

[0091]

[0092] Subsequently, the obtained frequency band F was analyzed. P The frequencies are divided into F based on high, medium, and low frequencies. PH F PM F PL The corresponding frequency band amplitudes are X PH X PM X PL Using the low-frequency band X PL Example of calculation method:

[0093]

[0094] P L (n) = IFFT(X) PL (n),N)

[0095] In the formula, For F PL Regarding the frequency center point f s / 2 symmetrical frequency band, P L (n) is the low-frequency output power obtained through inverse fast Fourier transform, which is also the effective grid-connected power of wind power generation.

[0096] Grid connection constraints for wind power generation are:

[0097]

[0098] In the formula ΔP grid,1min and ΔP grid,10min These are the deviations of the maximum and minimum grid-connected power fluctuations within 1 minute and 10 minutes, respectively, P. r This refers to the nominal value of the installed capacity of the wind farm.

[0099] 4) Define the upper and lower limits of the SOC of the energy storage system, and calculate the rated power and rated capacity of the energy storage system considering daily energy balance and charging and discharging losses.

[0100] After considering the charging and discharging loss coefficients, the corrected power of the energy storage system is:

[0101]

[0102] In the formula, P E0 To obtain the power of the energy storage system from the decomposition, η d For discharge conversion efficiency, η c This represents the charging conversion coefficient.

[0103] Considering daily energy balance, meaning assuming that the charge and discharge quantities can be balanced within one working cycle, the final SOC value of the ESS on that day should be as close as possible to the initial SOC value. Since the sampling period of the sampled data is the same, energy balance can be converted into power balance:

[0104]

[0105] The formula for calculating the rated power of an energy storage system is:

[0106]

[0107] The calculation process for the rated capacity of an energy storage system is as follows:

[0108]

[0109]

[0110]

[0111]

[0112]

[0113] In the formula, E(n) is the cumulative charge / discharge amount of the ESS in the current state, and SOC0 is the initial SOC of the ESS. U and SOC L These are the upper and lower limits of SOC, E. rated This is the rated capacity of the ESS.

[0114] 5) Calculate the grid-connected utilization conversion rate of wind power generation and the configuration cost of energy storage systems.

[0115] The grid-connected portion of the wind farm's output power is:

[0116]

[0117] The grid-connected conversion rate of wind power is:

[0118]

[0119] The configuration cost of the energy storage system is:

[0120] C ESS =P rated C P +E rated C E

[0121] In the formula, C P Price per unit power of energy storage system, in RMB 10,000 / MW; C E Price per unit capacity of energy storage system, in RMB 10,000 / MWh.

[0122] Taking a 30MW wind farm in Tianjin as an example, with a sampling frequency set to 0.1Hz, the corresponding sampling period is 10s; considering a daily period, the number of sampling points is 8640. According to Passevar's theorem, the power spectral density image and its fitting result based on the active power output of the wind farm are shown below. Figure 2 The slopes of the fitted lines for Part I and Part II are -1.9231 and -1.5873, respectively, both close to the -5 / 3 Kolmogorov slope, thus the frequency corresponding to the node between Part I and Part II is 0.005 Hz.

[0123] The obtained low-pass filter frequency was used to perform a fast Fourier transform decomposition on the wind power output power. The decomposed target grid-connected power is shown in [reference needed]. Figure 3 Solid line, see the original output curve of the wind farm. Figure 3 Dashed line. Its 1-minute and 10-minute volatility are calculated to be 9.44% and 31.68% respectively, which meet the grid connection requirements.

[0124] like Figure 4 As shown in Figure 'a', which represents the power curve of the energy storage system, by selecting the upper and lower limits of the SOC of the energy storage system as 0.8 and 0.2, respectively, we can obtain the following: Figure 4 In Figure b, the SOC variation curve of the energy storage system is shown. It can be seen that the initial SOC and the final SOC are the same in a single cycle, both being 0.575. The final rated power is determined to be 2.356MW and the rated capacity is 0.381MWh.

[0125] The grid-connected utilization conversion rate of wind power generation is calculated to be 96.03%. The configuration cost of the energy storage system is shown in Table 1. Finally, the energy storage configuration cost of this wind farm is calculated to be 5.1692 million yuan.

[0126] Table 1: Cost Price List of Energy Storage Systems

[0127] Energy storage unit 200 120

[0128] As the examples show, using a power spectral density-based energy storage configuration optimization algorithm to configure energy storage for wind farms, with the assistance of the energy storage system, ultimately meets the requirements for grid-connected power fluctuations, significantly improves the utilization efficiency of wind power generation, and also ensures a low energy storage configuration cost.

Claims

1. A wind-storage joint optimization configuration method based on power spectral density; characterized in that, This method is an energy storage configuration optimization algorithm based on power spectral density for mitigating source-side wind power fluctuations. The energy storage optimization configuration method includes finding the low-pass decomposition frequency using power spectral density, then determining the optimal grid-connected power and energy storage system power using fast Fourier transform decomposition, and finally determining the rated power and capacity of the energy storage system by considering charging and discharging losses and daily energy balance, ensuring that the energy storage system's state of charge (SOC) is at its optimal position at the beginning of each operating cycle. Specifically, it includes: Step 1: Based on Passevar's theorem, plot the power spectral density image of the wind farm based on its active power output; Step 2: Use the hinge model to segment the power spectral density (PSD) image and fit it into several affine functions. Then find two fitted straight lines whose slopes are close to the Kolmogorov "-5 / 3" slope, and take the frequency values ​​corresponding to their nodes as the low-pass decomposition frequencies. Step 3: Using the low-pass decomposition frequency as a benchmark, the wind power output is decomposed using the fast Fourier transform decomposition method; the low-frequency part is the power directly connected to the grid, and the medium and high frequency part is the power of the energy storage system. Verify whether the grid-connected power meets the fluctuation requirements of 1 minute and 10 minutes. Step 4: Define the upper and lower limits of the state of charge of the energy storage system, and calculate the rated power and rated capacity of the energy storage system considering daily energy balance and charging and discharging losses; Step 5: Calculate the grid-connected utilization conversion rate of wind power generation and the configuration cost of the energy storage system; The hinge model in step 2 is as follows: ； Its continuity condition is: ； In the formula, k j The x-axis coordinate interval is [k1, k N The x-coordinate of the node within ], H j For k j For the hinge model of the intermediate node, H l,j and H r,j H are hinges respectively j The left and right halves of H j-1 and H j+1 They are adjacent to H respectively j Two hinge models, one on the left and one on the right; The fitted affine function is in the form of: ； In the formula, f is the frequency, α is a scale, β is a power value, and γ is a constant; if we assume that γ is 0, then taking the logarithm of both sides of the equation yields: ； In logarithmic coordinates, β represents the slope and logα represents the y-intercept; the piecewise slope of the power spectral density may vary depending on the model order and the geographical location of the wind farm.

2. The wind-storage joint optimization configuration method based on power spectral density according to claim 1; characterized in that, In step 1, for a discrete signal x(n), n = 0, 1, 2, ..., N-1, the mathematical model of its power spectral density S(k) is as follows: ； In the formula, X(k) is the Fourier transform form of the discrete signal x(n), k = 0, 1, 2, ..., N-1; j is the imaginary sign; The windowed average periodogram method used is as follows: ； In the formula, M is the number of segments in the random sequence, L is the length of each segment, and ω(n) is the window function.

3. The wind-storage joint optimization configuration method based on power spectral density according to claim 1; characterized in that, In step 3, for a continuous time-domain signal x(t), the discrete signals corresponding to its N points are taken as x(n), then its discrete Fourier transform signal is: ； ； In the formula, f(n) is the corresponding frequency of x(n), f s T is the sampling frequency. s The sampling period; For the original output power P of the wind farm wind N sampling points P wind (n) Performing a Fast Fourier Transform (FFT) yields the spectral analysis results: ； Subsequently, the obtained frequency bands were analyzed. The frequencies are divided into F based on high, medium, and low frequencies. PH F PM F PL The corresponding frequency band amplitudes are X PH X PM X PL ; with low frequency band X PL Example of calculation method: ； ； In the formula, For F PL Regarding the frequency center point f s / 2 symmetrical frequency band, P L (n) is the low-frequency output power obtained through inverse fast Fourier transform, which is also the effective grid-connected power of wind power generation; Grid connection constraints for wind power generation are: ； In the formula ΔP grid,1 min and ΔP grid,10 min These are the deviations of the maximum and minimum grid-connected power fluctuations within 1 minute and 10 minutes, respectively, P. r This refers to the nominal value of the installed capacity of the wind farm.

4. The wind-storage joint optimization configuration method based on power spectral density according to claim 1; characterized in that, After considering the charging and discharging loss coefficient in step 4, the corrected power of the energy storage system is: ； In the formula, P E0 To obtain the power of the energy storage system from the decomposition, η d For discharge conversion efficiency, η c The charging conversion coefficient; Considering daily energy balance, that is, assuming that the charging and discharging can be balanced within one working cycle, the final SOC value of the ESS on that day should be as close as possible to the initial SOC value of the energy storage system. Since the sampling period of the sampled data is the same, the energy balance can be converted into power balance: ； The formula for calculating the rated power of an energy storage system is: ； The calculation process for the rated capacity of an energy storage system is as follows: ； ； ； ； ； In the formula, E(n) is the cumulative charge / discharge amount of the ESS in the current state, and SOC0 is the initial SOC of the ESS. U and SOC L These are the upper and lower limits of SOC, E rated This is the rated capacity of the ESS.

5. The wind-storage joint optimization configuration method based on power spectral density according to claim 1; characterized in that, In step 5, the grid-connected portion of the wind farm's power output is: ； The grid-connected conversion rate of wind power is: ； The configuration cost of the energy storage system is: ； In the formula, C P Price per unit power of energy storage system, in RMB 10,000 / MW; C E Price per unit capacity of energy storage system, in RMB 10,000 / MWh.

Images