CN109088470B - Battery-super capacitor hybrid energy storage independent photovoltaic system optimization control method - Google Patents

Abstract

The invention discloses an optimization control method for a hybrid energy storage independent photovoltaic system with a battery and a super capacitor. The method comprises the following steps: 1. establishing a battery-super capacitor-based hybrid energy storage independent photovoltaic system model; 2. reducing dynamic stress of the battery by using a moving average filter; 3. introducing power allocation algorithm, using two thresholds

And

controlling the charging and discharging of the super capacitor; 4. optimizing power distribution algorithm parameters based on power demand and the charge state of the super capacitor in a short time by utilizing a self-organizing map and particle swarm combined algorithm; 5. and carrying out charge and discharge protection on the super capacitor. The invention provides higher flexibility for the battery-super capacitor energy storage system by introducing the charging threshold and the discharging threshold. The optimization method comprises the steps of optimizing power distribution algorithm parameters based on the predicted power demand and the charge state of the super capacitor by self-organizing mapping and a particle swarm algorithm, and relieving the peak demand and the transient charge-discharge cycle of the battery.

Description

Battery-super capacitor hybrid energy storage independent photovoltaic system optimization control method

The technical field is as follows:

the invention belongs to the technical field of solar photovoltaic control, and particularly relates to an optimal control method for a hybrid energy storage independent photovoltaic system with a battery and a super capacitor.

Background art:

the battery-super capacitor hybrid energy storage system scheme is an effective method in a photovoltaic power generation system, the size and the stress level of a battery can be greatly reduced, and the total investment cost of an independent photovoltaic system is reduced. The control strategy is essentially an algorithm that determines and controls the operation of the battery-supercapacitor HESS based on the state of the system. Optimal control strategies can significantly improve the performance and economic viability of the overall system.

The invention content is as follows:

the invention provides a novel optimization control method for an independent photovoltaic system with a battery-super capacitor HESS. Unlike conventional rule-based controllers and fuzzy logic controllers, the present invention proposes a power distribution algorithm that proposes two thresholds, namely a charge threshold and a discharge threshold (T)_cAnd T_d) Flexible operation is provided for the battery-supercapacitor HESS, where the HESS can be charged and discharged without limitation. In addition, a new optimization method is proposed, including self-organizing map (SOM) and Particle Swarm Optimization (PSO) to optimize T_cAnd T_d. The idea of the method is to use the SOM to narrow the search space of the PSO to speed up the optimization process of the PSO. With this approach, the control strategy can be based on the predicted power demand and state of charge SOC of the supercapacitor_SCOptimization was performed every minute.

In order to reduce the battery dynamic stress and peak power requirements in the battery-super capacitor HESS independent photovoltaic system, the battery-super capacitor HESS operates in the optimal state, and the technical scheme of the invention is as follows:

an optimization control method for a hybrid energy storage independent photovoltaic system with a battery-super capacitor comprises the following steps:

step 1, establishing a battery-super capacitor-based hybrid energy storage independent photovoltaic system model;

step 2, reducing the dynamic stress of the battery by using a moving average filter;

step 3, introducing a power distribution algorithm, and using two threshold values T respectively representing charging_cAnd threshold value T of discharge_dControlling the charging and discharging of the super capacitor;

step 4, optimizing power distribution algorithm parameters based on power demand and super capacitor charge state in a short time by using a self-organizing map and particle swarm combined algorithm;

and 5, performing charge and discharge protection on the super capacitor.

The battery-super capacitor hybrid energy storage independent photovoltaic system model in the step 1 comprises a photovoltaic array, a load, a charge controller and a hybrid energy storage system HESS; the HESS comprises a battery, a super capacitor and a bidirectional direct current converter, and the self-organizing mapping is combined with a particle swarm algorithm to carry out threshold T_cAnd T_dAnd (6) optimizing.

The dynamic power model of the independent photovoltaic system is represented by the following formula:

P_PV-P_batt-P_SC′-P_load＝0 (1)

wherein P is_PVOutput power for a photovoltaic array; p_battFor battery flow power, P_SC‘Refers to the power, P, of the super capacitor after power conversion_loadIs the load demand power; the relationship between the power deficit dP between the photovoltaic array power and the load and the sum of the super-capacitor and battery power is obtained by converting the following equation:

dP＝P_PV-P_load＝P_batt+P_SC’ (2)

in the step 2, a filter controller-based moving average filter is used to eliminate the fast transient power in the battery demand, so as to prevent the battery from providing the high-frequency power demand in the hybrid energy storage system HESS; wherein, the dP low frequency power in the hybrid energy storage system HESS is represented as:

wherein T is_sIs the mobile sampling time, with low frequency power, the high frequency power requirement is expressed as:

dP_HF＝dP-dP_LF (4)

the power distribution algorithm in the step 3 provides higher flexibility for the hybrid energy storage system HESS of the battery-super capacitor by allowing the super capacitor to be charged and discharged under any condition; the input to the proposed power allocation algorithm is dP_LFAnd state of charge SOC of super capacitor_SCThe output is the reference power P of the super capacitor_SCPDA(ii) a Introduction of T_cAnd T_dTwo thresholds control the charging and discharging operations of the supercapacitor, only one threshold being active at a time; t is_cIs always greater than T_d(ii) a Super capacitor at dP_LF<Provide power to the system at 0 and at dP_LF>And (3) charging at 0, and outputting the reference power of the super capacitor according to the following three different conditions:

(1)dP_LF>T_c,P_{SC PDA}＝dP_LF-T_c

(2)dP_LF<T_d,P_{SC PDA}＝dP_LF-T_d

(3)T_c>dP_LF>T_d,P_{SC PDA}＝dP_LF。

the self-organizing map and the particle swarm algorithm in the step 4 are ranges T for providing an optimization scheme for the particle swarm algorithm by introducing the self-organizing map_c' and T_d′，T_c' and T_d' represents threshold ranges for charging and discharging, respectively, after self-organizing map training, such that the initial population of particles is at T_c' and T_d' initialization in; in a self-organizing map neural network, by using multiple one-hour dP's in the system over a period of time_LFTo train the self-organizing map; self-organizing mapping high-order input signal dP_{LF prediction}Converting into two-dimensional discrete mapping with input vector defined as V_iN represents time, SOM is composed of two-dimensional neuron array M ═ M₁,m₂,…,m_QWhere Q is the total number of neurons arranged in a hexagonal network topology, the weight vector m of the neurons_i＝[m_i1,…,m_in]The self-organizing mapping neural network training comprises the following steps:

6.1. initializing weight vectors, initializing m_iE, taking M as a random value before the training of the SOM;

6.2. random selection of input vector V from input space_i；

6.3. Finding out dominant neuron m_cIn the process of searching for dominant neurons, m is obtained by applying the following formula_iAnd V_iThe shortest euclidean distance between them, i.e. m_c＝argmin||V_i-m_i||m_i∈M

6.4. Updating weight vectors of all neurons near the winning neuron;

6.5. repeating steps 6.2-6.4 for each input vector;

6.6. updating the learning rate and the neighborhood range;

6.7. and 6.2-6.6 are repeated until the learning rate is attenuated to 0, and the training of the two-dimensional SOM neural network is completed.

T after self-organizing mapping training and particle swarm optimization_cAnd T_dApplies two fitness functions f₁() And f₂() To evaluate its fitness, the fitness function is as follows:

wherein P is_{batt_max}And P_{batt_min}Refer to the maximum and minimum points of battery power, respectively, for different dP_{LF prediction}The case takes different optimization schemes. When P is present_PV>P_loadWhen is, dP_{LF prediction}When the maximum value and the minimum value are both greater than 0, T is not required to be powered by the battery_dSet to 0 to prevent the battery from switching from a charged to a discharged state, SOM-PSO using f only₁(x) For T_cOptimizing; when P is present_PV<P_loadWhen is, dP_{LF prediction}When the maximum value of f is positive and negative, the HESS supplies power to the system by using f₂(x) To T_cAnd T_dAnd (6) optimizing.

After optimizing the threshold of the power distribution algorithm, the reference power P of the super capacitor obtained by the power distribution algorithm_{SC PDA}With high frequency component dP_HFThe sum is given to P_{SC ref′}In the introduction of P_{SC ref′}Before sending to the controller of the bidirectional DC converter, performing overcharge/discharge protection to prevent the super capacitor from exceeding a specified working voltage range; SOC of super capacitor_SCThe working range is regulated to be between 25% and 100%, a multiplier beta is introduced to realize overcharge/discharge protection, beta is 1 or 0, and the specific values of beta are as follows:

the final input to the controller of the bi-directional dc converter is determined by the following equation:

P_{SC ref}＝β*P_{SC ref}′ (7)。

the invention has the beneficial effects that:

the invention provides higher flexibility for the battery-super capacitor energy storage system by introducing the charging threshold and the discharging threshold. The optimization method comprises the steps of optimizing power distribution algorithm parameters based on the predicted power demand and the charge state of the super capacitor by self-organizing mapping and a particle swarm algorithm, and relieving the peak demand and the transient charge-discharge cycle of the battery. The proposed optimization method enables faster convergence than conventional optimization methods.

Description of the drawings:

FIG. 1 is a diagram of a battery-supercapacitor based HESS standalone photovoltaic system;

FIG. 2 is a diagram of a proposed control strategy architecture;

fig. 3 is a flow chart for the proposed threshold optimization.

The specific implementation mode is as follows:

the invention is further described with reference to the following drawings and detailed description:

an optimization control method for a hybrid energy storage independent photovoltaic system with a battery-super capacitor comprises the following steps:

step 1, establishing a battery-super capacitor-based hybrid energy storage independent photovoltaic system model;

step 2, reducing the dynamic stress of the battery by using a moving average filter;

step 3, introducing a power distribution algorithm, and using two threshold values T respectively representing charging_cAnd threshold value T of discharge_dControlling the charging and discharging of the super capacitor;

step 4, optimizing power distribution algorithm parameters based on power demand and super capacitor charge state in a short time by using a self-organizing map and particle swarm combined algorithm;

and 5, performing charge and discharge protection on the super capacitor.

As shown in fig. 1, the battery-super capacitor hybrid energy storage independent photovoltaic system model in step 1 includes a photovoltaic array, a load, a charge controller, and a hybrid energy storage system HESS; the HESS comprises a battery, a super capacitor and a bidirectional direct current converter, and the self-organizing mapping is combined with a particle swarm algorithm to carry out threshold T_cAnd T_dAnd (6) optimizing.

The dynamic power model of the independent photovoltaic system is represented by the following formula:

P_PV-P_b-P_SC′-P_load＝0 (1)

wherein P is_PVOutput power for a photovoltaic array; p_battFor battery flow power, P_SC‘Refers to the power, P, of the super capacitor after power conversion_loadIs the load demand power; the relationship between the power deficit dP between the photovoltaic array power and the load and the sum of the super-capacitor and battery power is obtained by converting the following equation:

dP＝P_PV-P_load＝P_batt+P_SC’ (2)

in the step 2, a filter controller-based moving average filter is used to eliminate the fast transient power in the battery demand, so as to prevent the battery from providing the high-frequency power demand in the hybrid energy storage system HESS; wherein, the dP low frequency power in the hybrid energy storage system HESS is represented as:

wherein T is_sIs the mobile sampling time, with low frequency power, the high frequency power requirement is expressed as:

dP_OF＝dP-dP_LF (4)

the power allocation algorithm in step 3 above provides higher flexibility for the hybrid energy storage system HESS of the battery-super capacitor by allowing the super capacitor to charge and discharge under any condition, as shown in fig. 2; the input to the proposed power allocation algorithm is dP_LFAnd state of charge SOC of super capacitor_SThe output is the reference power P of the super capacitor_SCPDA(ii) a Introduction of T_cAnd T_dTwo thresholds control the charging and discharging operations of the supercapacitor, only one threshold being active at a time; t is_cIs always greater than T_d(ii) a Super capacitor at dP_LF<Provide power to the system at 0 and at dP_LF>And (3) charging at 0, and outputting the reference power of the super capacitor according to the following three different conditions:

(1)dP_LF>T_c,P_{SC PDA}＝dP_LF-R_c

(2)dP_LF<T_d,P_{SC PDA}＝dP_LF-T_d

(3)T_c>dP_LF>T_d,P_{SC PDA}＝dP_LF。

the self-organizing map and the particle swarm algorithm in the step 4 are ranges T for providing an optimization scheme for the particle swarm algorithm by introducing the self-organizing map_c' and T_d′，T_c' and T_d' represents threshold ranges for charging and discharging, respectively, after self-organizing map training, such that the initial population of particles is at T_c' and T_d' initialization in; in a self-organizing map neural network, by using multiple one-hour dP's in the system over a period of time_LFTo train the self-organizing map; self-organizing mapping high-order input signal dP_{LF prediction}Converting into two-dimensional discrete mapping with input vector defined as V_iN represents time, SOM is composed of two-dimensional neuron array M ═ M₁,m₂,…,m_QWhere Q is the total number of neurons arranged in a hexagonal network topology, the weight vector m of the neurons_i＝[m_i1,…,m_in]The self-organizing mapping neural network training comprises the following steps:

6.1. initializing weight vectors, initializing m_iE, taking M as a random value before the training of the SOM;

6.2. random selection of input vector V from input space_i；

6.3. Finding out dominant neuron m_cIn the process of searching for dominant neurons, m is obtained by applying the following formula_iAnd V_iThe shortest euclidean distance between them, i.e. m_c＝argmin||V_i-m_i||m_i∈M

6.4. Updating weight vectors of all neurons near the winning neuron;

find out the dominant neuron m_cIts neighborhood is then determined by the following equation, where the area of the neighborhood shrinks with time by the decay function:

wherein σ₀Is represented at time t₀The field size of time, μ represents the time constant, and t represents the current time step. In determining m_cAfter the neighborhood area, m in the area is updated using the following equation_i：

m_i(t+1)＝m_i(t)+θ(t)L(t)(V_i(t)-m_i(t)) (6)

Wherein, L (t) and V_i(t) are the learning rate and the input vector at time t, respectively. In fact, the learning effect should be consistent with the selected m_iAnd m_cAnd therefore the learning rate influence at time t, θ (t), is also included, which can be calculated using the following formula:

θ (t) decays over time based on a Gaussian decay function.

6.5. At update m_iThen repeating steps 6.2-6.4 for each input vector;

6.6. updating the learning rate and the neighborhood range using equations (5) and (7) above;

6.7. and 6.2-6.6 are repeated until the learning rate is attenuated to 0, and the training of the two-dimensional SOM neural network is completed. Each M in the set M_iIndicating a dP that is not tagged with a response_LF，SOC_SCIn the range of 25% -100%. The SOM is constructed based on a hexagonal network topology, with each master neuron correlated with six other neurons. dP to be predicted_LFAnd SOC_SCInput to a pre-trained SOM, a set of neighboring neurons including a main neuron is denoted as { s₀,…,s₆For a given SOC_SCT in the range_cAnd T_dCan be represented as T_cK＝{T_c-S0…T_c-S6And T_dK＝{T_d-s0…T_d-S6Get the training result T'_c＝[max(T_ck),min(T_cK)]T′_d＝[max(T_dk),min(T_dk)]。

T after self-organizing mapping training and particle swarm optimization_cAnd T_dApplies two fitness functions f₁(x) And f₂(x) To evaluate its fitness, the fitness function is as follows:

as shown in fig. 3, different optimization methods are adopted for three different situations. For case 1, dP_LFIs positive, i.e. P_PV>P_loadNo battery power is required. Will T_dSet to 0 to prevent the battery from switching from a charged state to a discharged state to prevent an unpredictable sudden drop in photovoltaic array output power or sudden peak load demand that could cause dP_LFShort term is negative. And when SOC is_SCAt low time, T_cSet to a smaller value to let the supercapacitor distribute more charging power, on the other hand, when the SOC is_SCAt a high time, T_cSet to a larger value. In this case, only f is used₁(x) Optimizing T_cSo that P is_{batt_max}>The case of 0W is minimized to minimize the battery peak charging power.

Second case, dP_LFIs negative, i.e. P_load>P_PVAt that time, the HESS supplies power to the system. When SOC is reached_SCAt a high time, T_dPut at 0W to prevent short charge-discharge cycles while T_dPut at 0W to provide load demand. When SOC is reached_SCAt lower time, T_cAt a lower value, T_dSet to the same value. T is_dIs to limit/adjust P_{batt_min}Threshold value of (1), T_cIs used to reduce the peak load of the battery. The SOM-PSO method now uses f₂(x) To perform the optimization. In some cases, T_cNot necessarily the minimization of the peak load of the battery, and at this time, whether T needs to be evaluated according to the fitness_cOptimizing when T is not matched_cOptimization with varying fitness means that T is required_cTo minimize battery peak load and use f₁(x) Optimizing the same; on the contrary, only need to be rightT_dOptimization is carried out, at this time T_cSet to 0W to prevent short charge-discharge cycles of the battery.

Third case, dP_LFThe values of (A) have positive and negative values, and the optimization method using the SOM-PSO is similar to that of case 2. T is_cAnd T_dUsing f simultaneously₂(x) And (6) optimizing.

Finally, after optimizing the threshold of the power distribution algorithm, the reference power P of the super capacitor obtained by the power distribution algorithm_{SC PDA}With high frequency component dP_HFThe sum is given to P_{SC ref′}In the introduction of P_{SC ref′}Before being sent to the controller of the bidirectional DC-DC converter, the super capacitor is prevented from exceeding a specified working voltage range by implementing over-charge/discharge protection. SOC of super capacitor in the invention_SCThe working range is regulated to be between 25% and 100%, a multiplier beta is introduced to realize overcharge/discharge protection, beta is 1 or 0, and the specific values of beta are as follows:

the final input to the controller of the bi-directional dc converter is determined by the following equation:

P_{SC ref}＝β*P_{SC ref′} (10)。

Claims (4)

1. a battery-super capacitor hybrid energy storage independent photovoltaic system optimization control method is characterized by comprising the following steps:

step 1, establishing a battery-super capacitor-based hybrid energy storage independent photovoltaic system model;

the dynamic power model of the independent photovoltaic system is represented by the following equation:

P_PV-P_batt-P_SC′-P_load＝0 (1)

wherein P is_PVOutput power for a photovoltaic array; p_battFor battery flow power, P_SC‘Is through a power supplyConverted power of super capacitor, P_loadIs the load demand power; the relationship between the power difference dP between the photovoltaic array power and the load and the sum of the super-capacitor and battery power is obtained by converting the following equation:

dP＝P_PV-P_load＝P_batt+P_SC’ (2)

step 2, reducing the dynamic stress of the battery by using a moving average filter;

using a filter controller based moving average filter to eliminate fast transient power in battery demand, preventing the battery from providing high frequency power demand in the hybrid energy storage system HESS; wherein, the dP low frequency power in the hybrid energy storage system HESS is represented as:

wherein T is_sIs the mobile sampling time, with low frequency power, the high frequency power requirement is expressed as:

dP_HF＝dP-dP_LF (4)

step 3, introducing a power distribution algorithm, and using two threshold values T respectively representing charging_cAnd threshold value T of discharge_dControlling the charging and discharging of the super capacitor;

the power distribution algorithm, by allowing the super capacitor to charge and discharge under any condition, has the input dP_LFAnd state of charge SOC of super capacitor_SCThe output is the reference power P of the super capacitor_{SC PDA}(ii) a Introduction of T_cAnd T_dTwo thresholds control the charging and discharging operations of the supercapacitor, only one threshold being active at a time; t is_cIs always greater than T_d(ii) a Super capacitor at dP_LF<Provide power to the system at 0 and at dP_LF>And (3) charging at 0, and outputting the reference power of the super capacitor according to the following three different conditions:

(1)dP_LF>T_c,P_{SC PDA}＝dP_LF-T_c

(2)dP_LF<T_d,P_{SC PDA}＝dP_LF-T_d

(3)T_c>dP_LF>T_d,P_{SC PDA}＝dP_LF

step 4, optimizing power distribution algorithm parameters based on power demand and super capacitor charge state in a short time by using a self-organizing map and particle swarm combined algorithm;

the self-organizing map and the particle swarm optimization algorithm are a range T which is used for providing an optimization scheme for the particle swarm optimization algorithm by introducing the self-organizing map_c' and T_d′，T_c' and T_d' represents threshold ranges for charging and discharging, respectively, after self-organizing map training, such that the initial population of particles is at T_c' and T_d' initialization in; in a self-organizing map neural network, by using multiple one-hour dP's in the system over a period of time_LFTo train the self-organizing map; self-organizing mapping high-order input signal dP_{LF prediction}Converting into two-dimensional discrete mapping with input vector defined as V_iN represents time, and the self-organizing map is formed by a two-dimensional neuron array M ═ M₁,m₂,…,m_QWhere Q is the total number of neurons arranged in a hexagonal network topology, the weight vector m of the neurons_i＝[m_i1,…,m_in]The self-organizing mapping neural network training comprises the following steps:

6.1. initializing weight vectors, initializing m_iE, taking the E M as a random value before training the self-organizing mapping;

6.2. random selection of input vector V from input space_i；

6.3. Finding out dominant neuron m_cIn the process of searching for dominant neurons, m is obtained by applying the following formula_iAnd V_iThe shortest euclidean distance between them, i.e. m_c＝argmin||V_i-m_i||m_i∈M

6.4. Updating weight vectors of all neurons near the winning neuron;

6.5. repeating steps 6.2-6.4 for each input vector;

6.6. updating the learning rate and the neighborhood range;

6.7. repeating the steps 6.2-6.6 until the learning rate is attenuated to 0, and finishing the training of the two-dimensional SOM neural network;

and 5, performing charge and discharge protection on the super capacitor.

2. The optimal control method of the battery-super capacitor hybrid energy storage independent photovoltaic system according to claim 1, characterized in that: the battery-super capacitor hybrid energy storage independent photovoltaic system model in the step 1 comprises a photovoltaic array, a load, a charge controller and a hybrid energy storage system HESS; the HESS comprises a battery, a super capacitor and a bidirectional direct current converter, and the self-organizing mapping is combined with a particle swarm algorithm to carry out threshold T_cAnd T_dAnd (6) optimizing.

3. The optimal control method of the battery-super capacitor hybrid energy storage independent photovoltaic system according to claim 1, characterized in that: t after self-organizing mapping training and particle swarm optimization_cAnd T_dApplies two fitness functions f₁(x) And f₂(x) To evaluate its fitness, the fitness function is as follows:

wherein P is_{batt_max}And

refer to the maximum and minimum points of battery power, respectively, for different dP_{LF prediction}The case adopts different optimization schemes; when in useP_PV>P_loadWhen is, dP_{LF prediction}When the maximum value and the minimum value are both greater than 0, T is not required to be powered by the battery_dSet to 0 to prevent battery switching from charged to discharged state, self-organizing map-particle swarm algorithm with f only₁(x) For T_cOptimizing; when P is present_PV<P_loadWhen is, dP_{LF prediction}When the maximum value of f is positive and negative, the hybrid energy storage system HESS supplies power to the system by using f₂(x) To T_cAnd T_dAnd (6) optimizing.

4. The optimal control method of the battery-super capacitor hybrid energy storage independent photovoltaic system according to claim 1, characterized in that: after optimizing the threshold of the power distribution algorithm, the reference power P of the super capacitor obtained by the power distribution algorithm_{SC PDA}With high frequency component dP_HFThe sum is given to P_{SC ref′}In the introduction of P_{SC ref′}Before sending to the controller of the bidirectional DC converter, performing overcharge/discharge protection to prevent the super capacitor from exceeding a specified working voltage range; SOC of super capacitor_SCThe working range is regulated to be between 25% and 100%, a multiplier beta is introduced to realize overcharge/discharge protection, beta is 1 or 0, and the specific values of beta are as follows:

the final input to the controller of the bi-directional dc converter is determined by the following equation:

P_{SC ref}＝β*P_SCref′ (7)。

Images