Control method and system of binary hybrid energy storage system based on multi-energy power supply platform

Abstract

The application relates to the technical field of new energy power supply, and particularly provides a control method and system of a binary mixed energy storage system based on a multi-energy power supply platform, the method comprising the following steps: performing short-term power prediction of the multi-energy power supply platform through a CNN+LSTM hybrid neural network; combining a short-term power prediction result with a complete ensemble empirical mode decomposition (EEMD)+Hilbert transform (HT)+adaptive filter algorithm to perform power decomposition, so as to obtain low-frequency and high-frequency reference power; and in order to realize DC bus voltage stabilization, combining a power distribution method based on a state of charge (SOC) value of an energy storage device and a finite set model predictive control (FCS-MPC) control method to stabilize unbalanced power in the system, so as to solve the problem of strong instability of wind, light and waves in time and space when the multi-energy power supply platform is used on the sea.

Description

Technical Field

[0001] This invention relates to the field of electrical engineering, and in particular to a control method and system for a binary hybrid energy storage system based on a multi-energy power supply platform. Background Technology

[0002] DC microgrids have become a research hotspot both domestically and internationally because they do not require consideration of factors such as frequency and reactive power. Due to the randomness of weather changes affecting the output of renewable energy generation platforms, renewable energy generation exhibits volatility and uncertainty. Its direct grid connection can impact the safe and stable operation of the distribution network. Therefore, energy storage systems are needed to smooth out power imbalances in DC microgrid systems connected to distributed power sources. Traditional single-cell batteries struggle to address power density and energy density issues and have slow response times, leading to frequent charging and discharging, thus affecting battery lifespan.

[0003] The smoothing effect of an energy storage system depends on the given reference value and the control strategy. Since the multi-energy power supply platform is used at sea, wind, solar, and wave conditions exhibit strong instability in time and space, resulting in nonlinear and non-stationary output power. The first-order filter has poor power cutoff characteristics at its cutoff frequency, leading to integral action when responding to power commands at high frequencies. This results in charging or discharging capacity accumulation in the energy storage elements responding to high-frequency power. Furthermore, the complex sea conditions at sea exacerbate aliasing for a single filter. Simultaneously, traditional PI controllers and droop control in the control strategy have poor dynamic response capabilities due to fixed parameters, leading to increased DC bus voltage fluctuations under complex environmental conditions. Summary of the Invention

[0004] To address the shortcomings mentioned in the background art, the present invention aims to provide a control method and system for a binary hybrid energy storage system based on a multi-energy power supply platform, which can adapt to complex sea conditions and achieve unbalanced power suppression and DC bus voltage stabilization of the multi-energy power supply platform.

[0005] The technical solution to achieve the purpose of this invention is as follows:

[0006] A control method for a binary hybrid energy storage system based on a multi-energy power supply platform includes:

[0007] Using historical weather data as input, the power of distributed power sources is predicted through a hybrid neural network of CNN and LSTM.

[0008] By combining the predicted distributed power with the actual output power and load power, the power required by the hybrid energy storage system under the stable condition of the system bus voltage is determined through a fully ensemble empirical mode decomposition, Hilbert transform and adaptive filter algorithm.

[0009] Based on the SOC value of the energy storage system, the required power is determined by a secondary allocation strategy and allocated to the hybrid energy storage system of battery and supercapacitor.

[0010] Based on the power requirements allocated by the battery and the supercapacitor, the FCS-MPC strategy is adopted to construct a discrete mathematical model and objective function for the bidirectional DC / DC converter in the system, determine the switching signals in the bidirectional DC-DC converter in the hybrid energy storage system, and realize the control of the binary hybrid energy storage system.

[0011] Furthermore, the CNN includes convolutional layers, pooling layers, fully connected layers, and an output layer; its model is as follows:

[0012] Y = X * W + b

[0013]

[0014] p t =max{z t ,z t+1 ,…,z t+k-1}

[0015] y = W·x + b

[0016]

[0017] In the formula, Y represents the data after convolution, X is the input feature matrix, W is the convolution kernel, b is the bias term, f(x) is the activation function, x is the original value input to the function, α is the negative slope parameter in the activation function f(x), y is the output of the fully connected layer, and P(y) = ... i ) is the probability of category i. It is the exponent sum of the output values ​​of all categories, y i This represents the activation value for each neuron in the output layer.

[0018] Furthermore, the LSTM incorporates a forget gate, an input gate, an output gate, and cell states, and its model is as follows:

[0019] f t =σ(W f ·[h t-1 ,x t-1 ]+b f )

[0020]

[0021] In the formula, W f It is the weight matrix of the forget gate, i t For input gate output, For candidate cell states, C t For the final cell state, bf b c b o It is a bias term, x t It is the input for the current time step, h t It is the hidden state at the current time step, h t-1 It is the hidden state of the previous time step, and the σ function maps the input value to a probability value between 0 and 1.

[0022] Furthermore, the specific steps of obtaining the power required by the hybrid energy storage system under stable bus voltage conditions through complete set empirical mode decomposition, Hilbert transform, and adaptive filter algorithms include:

[0023] Power decomposition of the predicted distributed power source is performed using complete set empirical mode decomposition and Hilbert transform.

[0024] The adaptive filter algorithm combines the current SOC value of the energy storage system to give the corresponding cutoff frequency and performs power decomposition on the predicted power error.

[0025] By combining the two power decomposition data, the low-frequency reference power P used for real-time control is obtained. low With high-frequency reference power P high .

[0026] Furthermore, by performing power decomposition on the predicted distributed power source through complete set empirical mode decomposition and Hilbert transform, the final components are obtained as follows:

[0027]

[0028]

[0029] in, They are respectively, Let represent the i-th IMF component obtained through forward and reverse noise-assisted analysis, respectively; k represents the number of times the ensemble average is applied, i.e., the number of repetitions of noise addition; i represents the index of the IMF component; and j represents the iteration index of the ensemble average.

[0030] Furthermore, the transfer function of the adaptive filter algorithm is designed as follows:

[0031]

[0032] In the formula T L is the time constant.

[0033] Furthermore, the time constant T L The setting principle is:

[0034] 1) When the SOC value of the supercapacitor is in the normal operating range, T LThe default time constant T0 is set.

[0035] 2) When the hybrid energy storage output power P hess When the SOC value is greater than 0, if the supercapacitor's SOC value is within the discharge warning zone, then the time constant T is reduced. L If the supercapacitor's SOC value is within the charging warning zone at this time, then the time constant T is increased. L ;

[0036] 3) When the hybrid energy storage output power P hess When the SOC value is less than 0, if the supercapacitor's SOC value is within the discharge warning zone, then the time constant T is increased. L If the supercapacitor's SOC value is within the charging warning zone at this time, then the time constant T should be reduced. L .

[0037] Furthermore, a SOC power allocation strategy is adopted to perform a secondary allocation judgment on the required power and distribute it to the hybrid energy storage system of batteries and supercapacitors, specifically including:

[0038] 1) When the battery is in its normal operating range, and the supercapacitor is in its normal operating range or warning range:

[0039]

[0040] In the formula P batref P is the reference power for the battery. low For low-frequency reference power, P scref For the super electric reference power, P high This is the high-frequency reference power;

[0041] 2) When the battery's SOC value is greater than SOC bmax In the discharge region and P hess <0 or SOC value is less than SOC bmin In the charging area and P hess >0, while the supercapacitor is in the normal or warning zone; or when the supercapacitor's SOC value is greater than SOC scmax In the discharge region and P hess >0 or SOC value less than SOC scmin In the charging area and P hess When the value is less than 0, and the battery is in its normal operating range, then:

[0042]

[0043] Among them, SOC bmax This refers to the upper limit of the battery's SOC value. bmin This is the lower limit of the battery's SOC value. scmax This represents the upper limit of the SOC value of a supercapacitor. scminThis represents the lower limit of the SOC value for supercapacitors.

[0044] 3) When the battery's SOC value is greater than SOC bmax In the discharge region and P hess >0 or SOC value less than SOC bmin In the charging area and P hess <0; or when the SOC value of the supercapacitor is greater than SOC scmax In the discharge region and P hess <0 or SOC value is less than SOC scmin In the charging area and P hess If the value is greater than 0, and the battery is in its normal operating range, then:

[0045]

[0046] 4) When all supercapacitor batteries are in the discharge zone and P hess When the value is less than 0, the battery and supercapacitor stop charging, and a command is issued to stop the maximum power control of the distributed power source.

[0047] 5) When all supercapacitor batteries are in the charging zone and P hess When the value is greater than 0, the battery and supercapacitor stop outputting and the DC bus load is disconnected according to its importance.

[0048] Furthermore, the discretized mathematical model and objective function of the converter on the battery side of the system are as follows:

[0049]

[0050] In the formula, i b (k) represents the battery output current at time k, i b (k+1) represents the output current of the energy storage device at time k+1, u o (k) represents the high-voltage side voltage at time k, u o (k) represents the high-voltage side voltage at time k+1, u b Where L is the low-voltage side voltage, R is the inductance value, C is the load voltage regulator capacitor, and R is the resistance value. L This is the load resistance value. P batref The low-frequency power that the battery needs to suppress.

[0051] Furthermore, the discretized mathematical model and objective function of the converter on the supercapacitor side of the system are as follows:

[0052]

[0053] In the formula, i sc (k) represents the supercapacitor output current at time k, i sc(k+1) represents the supercapacitor output current at time k+1, i c (k) represents the current flowing through capacitor C, α and β are the voltage and power weighting coefficients of the supercapacitor, respectively, and the weighting coefficients are related to the allowable voltage fluctuation range. scref This refers to the high-frequency power that needs to be smoothed for supercapacitors.

[0054] A control system for a binary hybrid energy storage system based on a multi-energy power supply platform includes:

[0055] The distributed power prediction unit uses historical weather data as input and predicts the power of distributed power sources through a hybrid neural network of CNN and LSTM.

[0056] The power requirement determination unit for the hybrid energy storage system combines the predicted distributed power with the actual output power and load power, and determines the power required by the hybrid energy storage system under the stable condition of the system bus voltage through a fully ensemble empirical mode decomposition, Hilbert transform and adaptive filter algorithm.

[0057] The power distribution unit, based on the SOC value of the energy storage system, performs a secondary distribution judgment on the required power and allocates the required power to the hybrid energy storage system of batteries and supercapacitors.

[0058] The switching signal control unit, based on the power demand allocated by the battery and the power demand allocated by the supercapacitor, adopts the FCS-MPC strategy. Through the discrete mathematical model of the bidirectional DC / DC converter in the system and the objective function, it determines the switching signal in the bidirectional DC-DC converter of the hybrid energy storage system, thereby realizing the control of the binary hybrid energy storage system.

[0059] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention uses a CNN+LSTM hybrid neural network to perform short-term power prediction for a multi-energy power supply platform; the short-term power prediction results are combined with a complete set empirical mode decomposition + Hilbert transform + adaptive filter algorithm to perform power decomposition, obtaining low-frequency and high-frequency reference power; secondly, in order to achieve DC bus voltage stabilization, a power allocation method based on the state of charge (SOC) value of energy storage devices and an FCS-MPC control method are combined to smooth unbalanced power in the system; this invention can achieve real-time power decomposition and, through optimized control strategies, fully utilize the characteristics of battery and supercapacitor energy storage devices to smooth unbalanced power in the system, improve system stability, stabilize DC bus voltage, and extend the service life of energy storage devices. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a flowchart of the method of the present invention;

[0062] Figure 2 This is a diagram of the SOC control strategy of this invention;

[0063] Figure 3 This is a flowchart of the control system of the present invention;

[0064] Figure 4 This is a simplified equivalent model diagram of the supercapacitor of this invention;

[0065] Figure 5 This is the power training diagram of the photovoltaic part of the neural network in this invention;

[0066] Figure 6 This is a diagram showing the power decomposition results of this invention;

[0067] Figure 7 This is a simulation experiment diagram of the battery output power of this invention;

[0068] Figure 8 This is a simulation experiment diagram of the supercapacitor output power of this invention;

[0069] Figure 9 This is a simulation experiment diagram of the DC bus voltage of this invention. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Example 1

[0072] like Figure 1 and Figure 4 As shown, a control method for a binary hybrid energy storage system based on a multi-energy power supply platform includes the following steps:

[0073] CNN+LSTM hybrid neural network power prediction: Using historical weather data as input, it performs short-term and medium-short-term distributed power generation (photovoltaic, wind power, wave energy) output power prediction to obtain approximate power curves for future times for CEEMD (Complete Ensemble Empirical Mode Decomposition) + HT (Hilbert Transform) power decomposition preprocessing.

[0074] CEEMD+HT+Adaptive Filter Algorithm: The CEEMD+HT algorithm is used for predictive power decomposition. The adaptive filter algorithm combines the current SOC value of the energy storage system to give the corresponding cutoff frequency to decompose the predicted power error. Combining the two parts of power decomposition data, the reference low-frequency power and reference high-frequency power used for real-time control are obtained.

[0075] Based on the SOC value of the energy storage system, a SOC power allocation strategy is adopted to perform secondary allocation judgment on the required power and allocate it to the hybrid energy storage system of batteries and supercapacitors; (upper-level power allocation strategy)

[0076] Based on the power demand allocated by the battery and the power demand allocated by the supercapacitor, the FCS-MPC strategy (Power Allocation Method and Finite Set Model Predictive Control Strategy) determines the switching signal at the next moment through the discrete mathematical model of the bidirectional DC / DC converter in the system and the objective function.

[0077] It should be further explained that, in the specific implementation process, the binary hybrid energy storage system composed of the battery and the supercapacitor: the battery is used to smooth the low-frequency power in the unbalanced power of the system, and the supercapacitor is used to smooth the high-frequency power in the unbalanced power of the system and to stabilize the DC bus voltage; since the battery is an energy-type device with high energy density, low power density, and slow response time, frequent switching between charging and discharging will lead to damage to the battery's lifespan, and it is not suitable for smoothing the high-frequency power in the system for a long time. Therefore, the supercapacitor, as a power-type device, is used to smooth the high-frequency power of the unbalanced power in the system. Compared with the energy storage system of a single battery, the hybrid energy storage system extends the lifespan of the energy storage devices.

[0078] It should be further explained that, in the specific implementation process, the CNN+LSTM hybrid neural network power prediction uses historical weather data as input to perform short-term and medium-short-term power prediction, obtaining approximate power curves for future times, which are then used for power decomposition preprocessing in CEEMD+HT. In the CNN+LSTM power prediction, the CNN is used for input feature dimensionality reduction. The CNN mainly consists of convolutional layers, pooling layers, fully connected layers, and an output layer, and its mathematical expression is as follows:

[0079] Y = X * W + b

[0080]

[0081] p t =max{z t ,z t+1 ,…,z t+k-1}

[0082] y = W·x + b

[0083]

[0084] In the formula, Y is the data after convolution, X is the input feature matrix, W is the convolution kernel, b is the bias term, f(x) is the activation function, y is the output of the fully connected layer, and P(y) is the output of the fully connected layer. i ) is the probability of category i. It is the exponent sum of the output values ​​of all categories, y i This represents the activation value for each neuron in the output layer.

[0085] LSTM is a special type of recurrent neural network specifically designed for processing and predicting time-series data. LSTM can capture long-term dependencies in time-series data and effectively avoids the vanishing and exploding gradient problems encountered by traditional RNNs in long-term data. LSTM introduces the concepts of forget gate, input gate, output gate, and cell state, and its mathematical expression is as follows:

[0086] f t =σ(W f ·[h t-1 ,x t-1 ]+b f )

[0087]

[0088] In the formula W f This is the weight matrix of the forget gate, b f b c b o It is a bias term, x t It is the input for the current time step, h t-1 It is the hidden state of the previous time step. The σ function maps the input value to a probability value between 0 and 1.

[0089] It should be further explained that, in the specific implementation process, reference should be made to... Figure 5 The CEEMD+HT+adaptive filter algorithm, which combines CEEMD+HT with power prediction, solves at least 85% of the unbalanced power. Its CEEMD mathematical expression is as follows: First, for a given signal x, CEEMD... k (t), multiple additions of positive and negative white noise to the group.

[0090] x k + (t)=x(t)+ηk + (t)

[0091] x k - (t)=x(t)+η k - (t)

[0092] In the formula, x(t) is the original signal, and η k + (t) is the positive white noise added for the kth time, η k - (t) is the negative white noise added for the kth time, x k + (t) is the signal processed in the kth iteration of the white noise group, x k - (t) The signal processed in the kth iteration of the negative white noise group. Then, for x... k + (t) and x k - (t) Perform EMD decomposition to obtain the intrinsic mode function (IMF) and residuals:

[0093]

[0094] In the formula Let r be the i-th intrinsic mode function of the k-th group. k (t) represents the residual of the k-th group. Then, the ensemble average of the IMF components within the group is performed:

[0095]

[0096] Finally, for all The final IMF component is obtained by averaging the component sets:

[0097]

[0098] The remaining power is the power error between the actual power and the actual power. Frequency division is performed using an adaptive filtering algorithm, and its transfer function is:

[0099]

[0100] In the formula T L T is the time constant. L The setting principle is:

[0101] 1) When the SOC value of the supercapacitor is in the normal operating range, T L The default time constant is T0.

[0102] 2) When the hybrid energy storage output power P hessWhen the SOC value is greater than 0, the energy storage system needs to absorb power. If the SOC value of the supercapacitor is in the discharge warning zone at this time, the time constant is reduced, that is, the power absorbed by the supercapacitor is reduced to avoid overcharging. If the SOC value of the supercapacitor is in the charging warning zone at this time, the time constant is increased, that is, the output power of the supercapacitor is increased to avoid over-discharging.

[0103] 3) When P hess When the SOC value is less than 0, the energy storage system needs to output power. If the SOC value of the supercapacitor is in the discharge warning zone at this time, the time constant is increased, that is, the absorption power of the supercapacitor is increased to avoid overcharging of the supercapacitor. If the SOC value of the supercapacitor is in the charging warning zone at this time, the time constant is decreased, that is, the output power of the supercapacitor is reduced to avoid over-discharging of the supercapacitor.

[0104] It should be further explained that, in the specific implementation process, reference should be made to... Figure 2 , Figures 5-9 The required power is secondary-allocated using a SOC power allocation strategy and distributed to the hybrid energy storage system consisting of batteries and supercapacitors. The low-frequency and high-frequency power obtained from the power decomposition algorithm are reconstructed as follows:

[0105] 1) When the battery is in its normal operating range, and the supercapacitor is in its normal operating range or warning range:

[0106]

[0107] In the formula P batref P is the reference power for the battery. low For low-frequency reference power, P scref For the super electric reference power, P high This is the high-frequency reference power.

[0108] 2) When the battery's SOC value is greater than SOC bmax In the discharge region and P hess <0 or SOC value is less than SOC bmin In the charging area and P hess >0, while the supercapacitor is in the normal or warning zone; or when the supercapacitor's SOC value is greater than SOC scmax In the discharge region and P hess >0 or SOC value less than SOC scmin In the charging area and P hess <0, when the battery is in its normal operating range:

[0109]

[0110] When the battery SOC value is greater than SOC bmax In the discharge region and P hess >0 or SOC value less than SOCbmin In the charging area and P hess <0; or when the SOC value of the supercapacitor is greater than SOC scmax In the discharge region and P hess <0 or SOC value is less than SOC scmin In the charging area and P hess >0, while the battery is in its normal operating range:

[0111]

[0112] 4) When all supercapacitor batteries are in the discharge zone and P hess When the value is less than 0, the battery and supercapacitor stop charging and issue a command to stop the maximum power control of the distributed power source, thereby reducing power generation.

[0113] 5) When all supercapacitor batteries are in the charging zone and P hess When the value is greater than 0, the battery and supercapacitor stop outputting power and disconnect the DC bus load according to its importance to reduce power consumption.

[0114] To achieve reference power and reference DC bus voltage tracking, the reference values, along with the current output current and output voltage, are used as inputs to the FCS-MPC. The switching signal of the bidirectional DC / DC converter at the next moment is obtained through the objective function. The discretized mathematical model of the battery-side converter and the objective function are constructed as follows:

[0115]

[0116] In the formula i b (k) represents the battery output current at time k, i b (k+1) represents the output current of the energy storage device at time k+1, u o (k) represents the high-voltage side voltage at time k, u o (k) represents the high-voltage side voltage at time k+1, u b Where L is the low-voltage side voltage, R is the inductance value, C is the load voltage regulator capacitor, and R is the resistance value. L This is the load resistance value. P batref The low-frequency power that needs to be smoothed for the battery. A simplified equivalent model of a supercapacitor is shown below. Figure 3 As shown.

[0117] Referring to the simplified equivalent model of a supercapacitor, the discretized mathematical model and objective function of the supercapacitor-side converter are constructed as follows:

[0118]

[0119] In the formula i sc (k) represents the supercapacitor output current at time k, isc (k+1) represents the supercapacitor output current at time k+1, i c (k) represents the current flowing through capacitor C, α and β are the voltage and power weighting coefficients of the supercapacitor, respectively, and the weighting coefficients are related to the allowable voltage fluctuation range. scref This refers to the high-frequency power that needs to be smoothed for supercapacitors.

[0120] A control system for a binary hybrid energy storage system based on a multi-energy power supply platform includes:

[0121] The distributed power prediction unit uses historical weather data as input and predicts the power of distributed power sources through a hybrid neural network of CNN and LSTM.

[0122] The power requirement determination unit for the hybrid energy storage system combines the predicted distributed power with the actual output power and load power, and determines the power required by the hybrid energy storage system under the stable condition of the system bus voltage through a fully ensemble empirical mode decomposition, Hilbert transform and adaptive filter algorithm.

[0123] The power distribution unit, based on the SOC value of the energy storage system, performs a secondary distribution judgment on the required power and allocates the required power to the hybrid energy storage system of batteries and supercapacitors.

[0124] The switching signal control unit, based on the power demand allocated by the battery and the power demand allocated by the supercapacitor, adopts the FCS-MPC strategy. Through the discrete mathematical model of the bidirectional DC / DC converter in the system and the objective function, it determines the switching signal in the bidirectional DC-DC converter of the hybrid energy storage system, thereby realizing the control of the binary hybrid energy storage system.

[0125] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0126] The foregoing has shown and described the basic principles, main features, and advantages of this disclosure. Those skilled in the art should understand that this disclosure is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this disclosure. Various changes and modifications can be made to this disclosure without departing from its spirit and scope, and all such changes and modifications fall within the scope of this disclosure as claimed.

Claims

1. A control method for a binary hybrid energy storage system based on a multi-energy power supply platform, characterized in that, include: Using historical weather data as input, the power of distributed power sources is predicted through a hybrid neural network of CNN and LSTM. By combining the predicted distributed power with the actual output power and load power, the power required by the hybrid energy storage system under the stable condition of the system bus voltage is determined through a fully ensemble empirical mode decomposition, Hilbert transform and adaptive filter algorithm. Based on the SOC value of the energy storage system, the required power is secondary allocated and determined using the SOC power allocation strategy, and then allocated to the hybrid energy storage system of battery and supercapacitor. Based on the power demand of battery allocation and supercapacitor allocation, the FCS-MPC strategy is adopted to construct the discrete mathematical model and objective function of the bidirectional DC / DC converter in the system, determine the switching signal in the bidirectional DC-DC converter in the hybrid energy storage system, and realize the control of the binary hybrid energy storage system. The discretized mathematical model and objective function of the converter on the battery side of the system are as follows: In the formula, Let k be the battery output current. for The energy storage device outputs current at all times. Let k be the voltage on the high-voltage side. for High voltage side voltage at all times This is the low-voltage side voltage. This is the inductance value. The inductance is very small, and C is the load voltage regulator capacitor. This is the load resistance value. , The low-frequency power that the battery needs to suppress; The discretized mathematical model and objective function of the converter on the supercapacitor side of the system are as follows: In the formula, Let k be the supercapacitor output current. for The supercapacitor outputs current at all times. The current flowing through capacitor C, , These are the supercapacitor voltage and power weighting coefficients, respectively, and the weighting coefficients are related to the allowable voltage fluctuation range. This refers to the high-frequency power that needs to be smoothed for supercapacitors.

2. The control method for a binary hybrid energy storage system based on a multi-energy power supply platform according to claim 1, characterized in that, The CNN includes convolutional layers, pooling layers, fully connected layers, and an output layer; its model is as follows: In the formula, Data after convolution, For the input feature matrix, is the convolution kernel, and b is the bias term. For activation function, The original value input to this function. It is an activation function The negative slope parameter in For the output of a fully connected layer, It is the probability of category i. It is the sum of the exponents of all category output values. This represents the activation value for each neuron in the output layer.

3. The control method for a binary hybrid energy storage system based on a multi-energy power supply platform according to claim 1, characterized in that, The LSTM incorporates a forget gate, an input gate, an output gate, and cell states; its model is as follows: In the formula, It is the weight matrix of the forget gate. For input gate output, Candidate cell state, This is the final cell state. It is a bias term. This is the input for the current time step. It is the hidden state at the current time step. It is the hidden state of the previous time step. The function maps input values ​​to probability values ​​between 0 and 1.

4. The control method for a binary hybrid energy storage system based on a multi-energy power supply platform according to claim 1, characterized in that, The required power of the hybrid energy storage system under stable bus voltage conditions is obtained through complete ensemble empirical mode decomposition, Hilbert transform, and adaptive filter algorithms. Specifically, this includes: Power decomposition of the predicted distributed power source is performed using complete set empirical mode decomposition and Hilbert transform. The adaptive filter algorithm combines the current SOC value of the energy storage system to give the corresponding cutoff frequency and performs power decomposition on the predicted power error. By combining the two power decomposition data, the low-frequency reference power used for real-time control is obtained. With high-frequency reference power .

5. The control method for a binary hybrid energy storage system based on a multi-energy power supply platform according to claim 4, characterized in that, By performing power decomposition on the predicted distributed power using complete set empirical mode decomposition and Hilbert transform, the final components are obtained as follows: in, , Let represent the i-th IMF component obtained through forward and reverse noise-assisted analysis, respectively; k represents the number of times the ensemble average is applied, i.e., the number of repetitions of noise addition; i represents the index of the IMF component; and j represents the iteration index of the ensemble average.

6. The control method for a binary hybrid energy storage system based on a multi-energy power supply platform according to claim 4, characterized in that, The transfer function of the adaptive filter algorithm is designed as follows: In the formula, It is an adjustable time constant.

7. The control method for a binary hybrid energy storage system based on a multi-energy power supply platform according to claim 6, characterized in that, time constant The adjustment method is as follows: 1) When the SOC value of the supercapacitor is in the normal operating range, The default time constant is set. ; 2) When the hybrid energy storage output power P hess When the SOC value is greater than 0, if the supercapacitor's SOC value is within the discharge warning zone, then the time constant should be reduced. If the supercapacitor's SOC value is within the charging warning zone at this time, the time constant is increased. ; 3) When the hybrid energy storage output power P hess When the SOC value is less than 0, if the supercapacitor's SOC value is within the discharge warning zone, then the time constant is increased. If the supercapacitor's SOC value is within the charging warning zone at this time, then the time constant should be reduced. .

8. The control method for a binary hybrid energy storage system based on a multi-energy power supply platform according to claim 5, characterized in that, The SOC power allocation strategy is adopted to perform secondary allocation judgment on the required power and allocate it to the hybrid energy storage system of batteries and supercapacitors, specifically including: 1) When the battery is in its normal operating range, and the supercapacitor is in its normal operating range or warning range: In the formula This is the reference power for the battery. For low-frequency reference power, For super electric reference power, This is the high-frequency reference power; 2) When the battery's SOC value is greater than In the discharge region and P hess <0 or SOC value less than In the charging area and P hess >0, while the supercapacitor is in the normal or warning zone; or when the supercapacitor's SOC value is greater than In the discharge region and P hess >0 or SOC value less than In the charging area and P hess When the value is less than 0, and the battery is in its normal operating range, then: in, This represents the upper limit of the battery's SOC value. This is the lower limit of the battery's SOC value. This represents the upper limit of the SOC value of a supercapacitor. This represents the lower limit of the SOC value for supercapacitors. 3) When the battery's SOC value is greater than In the discharge region and P hess >0 or SOC value less than In the charging area and P hess <0; or when the SOC value of the supercapacitor is greater than 0; In the discharge region and P hess <0 or SOC value less than In the charging area and P hess If the value is greater than 0, and the battery is in its normal operating range, then: 4) When all supercapacitors and batteries are in the discharge zone and P hess When the value is less than 0, the battery and supercapacitor stop charging, and a command is issued to stop the maximum power control of the distributed power source. 5) When all supercapacitors and batteries are in the charging zone and P hess When the value is greater than 0, the battery and supercapacitor stop outputting and the DC bus load is disconnected according to its importance.

9. A control system for a binary hybrid energy storage system based on a multi-energy power supply platform that implements the method of any one of claims 1-8, characterized in that, include: The distributed power prediction unit uses historical weather data as input and predicts the power of distributed power sources through a hybrid neural network of CNN and LSTM. The power requirement determination unit for the hybrid energy storage system combines the predicted distributed power with the actual output power and load power, and determines the power required by the hybrid energy storage system under the stable condition of the system bus voltage through a fully ensemble empirical mode decomposition, Hilbert transform and adaptive filter algorithm. The power distribution unit, based on the SOC value of the energy storage system, performs a secondary distribution judgment on the required power and allocates the required power to the hybrid energy storage system of batteries and supercapacitors. The switching signal control unit, based on the power demand allocated by the battery and the power demand allocated by the supercapacitor, adopts the FCS-MPC strategy. Through the discrete mathematical model of the bidirectional DC / DC converter in the system and the objective function, it determines the switching signal in the bidirectional DC-DC converter of the hybrid energy storage system, thereby realizing the control of the binary hybrid energy storage system.

Images