A photovoltaic power balancing method and system for distribution station area based on double-layer architecture

Abstract

The present invention discloses a photovoltaic power balancing method for distribution substations based on a double-layer architecture, including predicting and analyzing the power supply and demand of the distribution substation, calculating the off-grid load and on-grid power of the distribution substation based on real-time data, and conducting a game for the lower photovoltaic power of the distribution substation to obtain the optimal solution for the user energy consumption and energy storage charging strategy; for the upper photovoltaic power balance of the distribution substation, a static game is conducted between PV power generation, energy storage charging and discharging, and user load to solve the optimal economic strategy for PV balance in the distribution substation. Finally, a photovoltaic power generation control strategy and an energy storage charging and discharging control strategy for the distribution substation are generated, and power control of the distribution substation is performed to achieve an economic balance of photovoltaic power in the distribution substation. The present invention adopts a double-layer architecture joint control method to calculate the optimal PV absorption target, reduce the proportion of PV abandoned light, and has good practicality.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power supply and demand balancing, and specifically relates to a photovoltaic power balancing method and system for a distribution station area based on a double-layer architecture. Background Art

[0002] Currently, distributed photovoltaic (PV) systems are being integrated into distribution networks on a large scale. These networks are evolving from passive, one-way, radiating networks to bidirectional networks deeply coupled with the power grid. This integration of distributed PV systems into distribution networks presents a temporal and spatial imbalance in power supply and demand. Therefore, exploring PV power balance within distribution networks is crucial for addressing this imbalance.

[0003] Many scholars have conducted extensive research on photovoltaic power balance in distribution substations. PV balance is mainly divided into two categories: low-voltage flexible DC energy circulation method and optimized absorption method. Among them, the low-voltage flexible DC energy circulation method uses low-voltage DC technology to interconnect multiple distribution substations, thereby establishing a new low-voltage topology structure for distribution substations. According to the distributed photovoltaic power situation of different distribution substations, this method uses energy routers to distribute photovoltaic power to the load or energy storage in the distribution substation, thereby achieving the problem of photovoltaic power balance in the distribution substation. Low-voltage flexible DC energy circulation can effectively solve the problem of photovoltaic power balance in distribution substations, but the cost of building a low-voltage flexible DC energy system is high and cannot meet the requirements of large-scale promotion.

[0004] The optimized consumption method combines information models with electrical control models to control energy storage in distribution stations, while limiting the amount of PV power fed back to the distribution station, thereby achieving PV power balance. While the optimized scheduling and consumption method can address PV power balance in distribution stations from a model control perspective, it does not consider optimal operation of distribution stations. Under the constraints of PV power fed back to the distribution station, if the energy storage capacity is configured too low, photovoltaic power will fully fill the energy storage equipment, leading to curtailment. However, if the energy storage capacity is configured too high, the cost will be high, resulting in wasted resources. Summary of the Invention

[0005] The present invention aims to provide a method and system for balancing photovoltaic power in distribution substations based on a two-tier architecture, aiming to address the issue of poor energy storage operation performance in optimizing photovoltaic power consumption in distribution substations. In the lower layer, the present invention addresses the PV power balance issue within distribution substations. In the upper layer, a dynamic game theory approach is employed to determine the optimal consumption value between PV aggregators, energy storage aggregators, and power users.

[0006] The present invention is mainly achieved through the following technical solutions:

[0007] A photovoltaic power balancing method for a distribution station area based on a double-layer architecture includes the following steps:

[0008] Step S100: forecasting the photovoltaic power and load of the distribution substation area respectively, and conducting forecast analysis on the power supply and demand of the distribution substation area;

[0009] Step S200: Obtain the real-time changes in the power supply and demand of the current distribution area and obtain the energy storage charging power; when the sum of the user load of the distribution area predicted in step S100 and the energy storage charging power is less than the photovoltaic power predicted in step S100, calculate the distribution area grid power p UDA ; If the power of the distribution area is p UDA Greater than the power limit value Δp in the distribution area NL , then PV power generation is abandoned and the process goes to step S300; otherwise, the process goes directly to step S300;

[0010] Step S300: The lower photovoltaic power game of the distribution area is used to obtain the optimal solution of the user energy consumption and energy storage charging strategy; the lower photovoltaic power game model p USG for:

[0011]

[0012] in: is the SG game function;

[0013] p DDA Unload grid load for distribution substation area;

[0014] The photovoltaic power of the distribution area predicted in step S100;

[0015] Provides energy storage and charging power for distribution station areas;

[0016] P ALLREM It is the sum of the loads of the electricity meters of users in the distribution station area;

[0017] Step S400: For the upper photovoltaic power balance of the distribution area, a static game is conducted between PV power generation, energy storage charging and discharging, and user load to solve the optimal economic strategy for PV balance in the distribution area, so as to achieve the optimal benefits for the four entities of the distribution area: power supply company, power user, PV aggregator, and energy storage aggregator; the upper photovoltaic power balance game model p of the distribution area DSG for:

[0018]

[0019] Where: p DDA (t) and v DDA They are the power supply quantity and power supply price within the time period of the distribution station area;

[0020] and v REAPV are the photovoltaic power generation and power generation price in the distribution station area;

[0021] and v ES are the charging and discharging capacity and charging and discharging price of energy storage in the distribution station area respectively;

[0022] p ALLREM (t) and v ALLREM They are the electricity consumption and electricity price of users in the distribution station area;

[0023] Step S500: Generate a photovoltaic power generation control strategy and an energy storage charge and discharge control strategy for the distribution station area based on the photovoltaic power value and the energy storage discharge power value curve respectively; then, perform power control on the distribution station area to achieve an economic balance of photovoltaic power in the distribution station area.

[0024] In order to better implement the present invention, further, in step S200, the power p of the distribution area is UDA The calculation formula is as follows:

[0025]

[0026] in: The load of different user meters;

[0027] Power limit value Δp for the distribution area NL The calculation formula is as follows:

[0028]

[0029] Where: n REA The number of distribution stations within the distribution dispatch area;

[0030] It is the minimum load of different distribution station areas.

[0031] In order to better implement the present invention, further, in step S300, when the sum of the user load of the distribution station area predicted in step S100 and the energy storage charging power is greater than or equal to the photovoltaic power predicted in step S100, the off-grid load p of the distribution station area is calculated. DDA ;

[0032] When the energy storage is in the charging state, the off-grid load p DDA for:

[0033]

[0034] Where: n DA The number of household meters in the distribution area;

[0035] The load of different user meters;

[0036] Provides energy storage and charging power for distribution station areas;

[0037] The photovoltaic power of the distribution station area;

[0038] When the energy storage is in the discharge state, the load p DDA for:

[0039]

[0040] in: Discharge power for distribution station area.

[0041] In order to better implement the present invention, further, in step S100, an LSTM network is used in combination with meteorological data to predict the photovoltaic power of the distribution station area; in the LSTM network structure, the input gate receives numerical weather forecast data, photovoltaic power data at the previous moment, and user load data respectively; the forgetting gate selectively memorizes the numerical weather forecast data, photovoltaic power data at the previous moment, and user load data, and superimposes the user load and photovoltaic power data at the current moment to form a memory state; and the output gate outputs the photovoltaic power data and user load data at the next moment respectively.

[0042] In order to better implement the present invention, further, in step S100, when predicting the photovoltaic power, the photovoltaic power at the current moment retained by the forget gate for:

[0043]

[0044] Where: δ() is the activation function in the LSTM for photovoltaic power prediction;

[0045] w F is the historical weight of the photovoltaic power in the distribution station area;

[0046] s WF Numerical weather forecast data provided to meteorological service providers;

[0047] The photovoltaic power data of the distribution area at the previous moment;

[0048] The photovoltaic voltage data of the distribution area at the previous moment;

[0049] l FGPV Function bias for the LSTM forget gate.

[0050] In order to better implement the present invention, further, in step S100, when predicting photovoltaic power, the state q of the LSTM network unit NET for:

[0051]

[0052] Where: qNET t-1 is the photovoltaic power state variable of the distribution station area at the previous moment;

[0053] Enter the photovoltaic power data of the gate distribution station area at the current moment;

[0054] The photovoltaic power data of the distribution area of ​​the coupling unit in LSTM;

[0055] LSTM distribution station photovoltaic power prediction output gate data for:

[0056]

[0057] Where: q NET is the photovoltaic power state variable of the distribution area at the current moment;

[0058] w OUT The LSTM output gate weights for the photovoltaic power of the distribution station area;

[0059] The photovoltaic power data of the distribution area at the current moment;

[0060] Tanh() is the activation function in the LSTM network.

[0061] In order to better implement the present invention, the step S400 further includes the following constraints:

[0062] Energy storage device power configuration of energy storage aggregator p NLEV for:

[0063]

[0064] Where: p NLTF The rated power of the distribution transformer in the distribution station area;

[0065] is the average power generation predicted by PV;

[0066] is the predicted maximum user load;

[0067] Energy storage capacity configuration h NLEV for:

[0068]

[0069] Where: h MDA The hourly average power consumption during the morning peak in the distribution area;

[0070] h NDA It is the hourly average electricity consumption during the evening peak in the distribution station area.

[0071] A photovoltaic power balancing system for a distribution station area based on a double-layer architecture includes a memory and a processor; the memory stores a computer program; the processor is used to execute the computer program in the memory to implement the above method.

[0072] The beneficial effects of the present invention are as follows:

[0073] This invention achieves optimal photovoltaic power balance by leveraging the power of the lower PV layer within the distribution substation, while satisfying energy conservation and the grid company's restrictions on photovoltaic power access within the distribution substation. This invention optimizes the benefits of the four entities within the distribution substation: the power supply company, electricity users, PV aggregators, and energy storage aggregators. This invention utilizes a dual-layer architecture for joint control to calculate the optimal PV absorption target, reducing the proportion of PV curtailment and demonstrating good practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 This is an overall flow chart of a photovoltaic power balancing method for a distribution station area based on a double-layer architecture according to the present invention. DETAILED DESCRIPTION

[0075] Example 1:

[0076] A photovoltaic power balancing method for distribution area based on a two-layer architecture, such as Figure 1 As shown, the following steps are included:

[0077] 1.1 Analysis of power supply and demand in distribution station area:

[0078] Collect historical electrical data of distribution substations such as current, voltage, and power;

[0079] Then, the PV power of the distribution substation area is predicted in combination with meteorological data to obtain the power supply forecast of the distribution substation area;

[0080] Forecast the load of the distribution substation area to obtain the power demand forecast of the distribution substation area.

[0081] 1.2 Power Game in the Lower Layer of the Distribution Area:

[0082] Collect real-time data of the power distribution area to obtain the real-time changes in power supply and demand in the current power distribution area;

[0083] Based on the predicted value of 1.1, the difference between the predicted PV power generation in the distribution area and the predicted load and the measured energy storage charging is analyzed to determine whether the PV power generation can be absorbed by the load and energy storage charging in the distribution area. UDA If the calculated photovoltaic power of the distribution substation is greater than the grid-connected limit, the distributed PV power generation will be abandoned, and then the power of the lower layer of the distribution substation will be gambled to solve the optimal user energy consumption and energy storage charging strategy.

[0084] If the power of the distribution area is p UDA If it is less than or equal to the access limit, the lower-level power of the distribution station area is directly priced to solve the optimal user energy consumption and energy storage charging strategy.

[0085] 1.2 Upper power balance of distribution station area:

[0086] A static game is conducted among PV power generation, energy storage charging and discharging, and user load to solve the optimal economic strategy for PV balance in the distribution station area.

[0087] Secondly, the PV power generation control strategy and energy storage charging and discharging control strategy of the distribution station area are generated respectively.

[0088] Finally, power control is performed in the distribution area to achieve economical balance of PV power in the distribution area.

[0089] Preferably, the specific steps of power supply and demand analysis in the distribution station area in step 1.1 are as follows:

[0090] Distributed PV generation in distribution areas is highly intermittent and random, severely impacting power balance and safe and stable operation within the distribution area. When distributed PV generation exceeds user load, power backflow can occur, compromising the safe operation of the 10kV distribution network. When distributed PV generation is low and local distribution network power supply is limited, it can be unable to meet user load demands. Therefore, accurate analysis of power supply and demand within distribution areas is crucial for balancing PV power.

[0091] In the distribution substation area, meteorological data can be combined to predict distributed PV and user loads, thereby obtaining accurate power supply and demand conditions in the distribution substation area.

[0092] Long Short-Term Memory (LSTM) is a temporally recurrent neural network designed to address the long-term dependency issues of recurrent neural networks. LSTM employs a forget, input, and output gate structure and is often used to address correlation issues in multi-timescale forecasting, thereby improving forecast accuracy. Therefore, LSTM is used to forecast photovoltaic power and load in distribution substations.

[0093] In the LSTM network structure, the input gate receives numerical weather forecast data, photovoltaic power data at the previous moment, and user load data respectively; the forget gate selectively memorizes the above data and superimposes the user load and photovoltaic power data at the current moment to form a memory state; the output gate outputs the photovoltaic power data and user load data at the next moment respectively.

[0094] The current photovoltaic power data pFG c retained by the forget gate is:

[0095]

[0096] Where: δ is the activation function in PV prediction LSTM; w F is the historical weight of photovoltaic power in the distribution station area; WF Numerical weather forecast data provided to meteorological service providers; The photovoltaic power data of the distribution area at the previous moment; is the PV voltage data of the distribution station area at the previous moment; FGPV Function bias for the LSTM forget gate.

[0097] The state of the LSTM network unit is composed of the photovoltaic power state of the distribution area at the previous moment and the photovoltaic power input of the distribution area at the current moment. The state q of the LSTM network unit is NET for:

[0098]

[0099] in: is the photovoltaic power state variable of the distribution station area at the previous moment; Enter the photovoltaic power data of the gate distribution station area at the current moment; It is the photovoltaic power data of the distribution station area of ​​the coupling unit in LSTM.

[0100] The LSTM output gate determines the proportion of the sub-unit output photovoltaic power results of the distribution substation, and the LSTM distribution substation photovoltaic power prediction output gate data for:

[0101]

[0102] Where: w OUT The LSTM output gate weights for the photovoltaic power of the distribution station area; is the photovoltaic power data of the distribution station area at the current moment; tanh is the activation function in the LSTM network.

[0103] The use of LSTM for distribution station area load forecasting is similar to photovoltaic power forecasting. Due to space limitations, this article will not elaborate on this.

[0104] Preferably, the specific steps of the photovoltaic power game in the lower layer of the distribution station area in step 1.2 are as follows:

[0105] The purpose of the PV power game in the lower layer of the distribution substation is to balance the constraints at the physical level of electric energy, that is, to meet the energy conservation and the grid company's restrictions on the PV power access to the distribution substation.

[0106] During the real-time data collection phase of the distribution substation, the intelligent fusion terminal collects electrical data from the main meter, PV, household meters, energy storage, and other devices. The off-grid load and on-grid power of the distribution substation are calculated based on the real-time data.

[0107] When the sum of the predicted user load in the distribution area and the measured energy storage charging power is greater than or equal to the predicted photovoltaic power, there is a grid load in the distribution area. DDA for:

[0108]

[0109] Where: n DA The number of household meters in the distribution area; The load of different user meters; Charging power for distribution area; is the PV power generation power.

[0110] When the energy storage is in the discharge state, the load p DDA for:

[0111]

[0112] in: Discharge power for distribution station area.

[0113] When the sum of the predicted user load in the distribution area and the measured energy storage charging power is less than the predicted photovoltaic power, there is grid-connected power in the distribution area. UDA for:

[0114]

[0115] Constraints:

[0116] p UDA ≤Δp NL (7)

[0117] Where: Δp NL The power limit value for the distribution area access provided by the power supply company.

[0118] The power limit for the distribution area is:

[0119]

[0120] Where: n REA The number of distribution stations within the distribution dispatch area; It is the minimum load of different distribution station areas.

[0121] When the power supply of the distribution area is p UDA Greater than the power limit value Δp in the distribution area NL In order to ensure the stable operation of the distribution network, PV power generation is abandoned.

[0122] Under the condition of meeting the basic operation of PV in the distribution station area, the lower-level photovoltaic power game is carried out to improve the photovoltaic power absorption rate.

[0123] A static game (SG) is a type of leader-based game model. In an SG model, a dominant player and several followers form a game strategy. The dominant player first formulates a strategy, and the followers then adjust their strategies based on the dominant player's strategy and their own. In this paper, the game is led by the power supply operator of the distribution substation area. Therefore, the power supply operator is the dominant player, while users, PV aggregators, and energy storage aggregators are followers.

[0124] Lower photovoltaic power game model p USG for.

[0125]

[0126] in: is the SG game function; P ALLREM The sum of the meter loads of users in the distribution station area.

[0127] Through the lower-level energy game, the optimal solution for user energy consumption and energy storage charging strategy can be obtained.

[0128] Preferably, the specific steps for balancing the photovoltaic power in the upper layer of the distribution station area in step 1.3 are as follows:

[0129] In the photovoltaic power balance at the upper level of the distribution network, a static game theory is used among the four value chains of power supply companies, power users, PV aggregators, and energy storage aggregators to determine the optimal value scenario for each entity. The power supply company aims to reduce forward and reverse overload in the distribution network to avoid the investment required for distribution network renovation, while also hoping to sell more electricity to users. PV aggregators aim to generate more electricity, sell it at a higher price, and thus obtain higher profits. Energy storage aggregators aim to reduce energy storage investment while purchasing electricity when prices are low and selling it when prices are high, thereby obtaining higher profits. Power users aim to purchase the cheapest electricity to reduce their energy investment.

[0130] The calculation basis for energy storage aggregators to configure the device power is the rated power of the distribution transformer plus the average power generation power predicted by PV minus the predicted maximum user load. The power configuration of the energy storage device is p NLEV for.

[0131]

[0132] Where: p NLTF The rated power of the distribution transformer in the distribution station area; is the average power generation predicted by PV; is the predicted maximum user load.

[0133] The calculation basis of energy storage device power is the average user power consumption during the predicted peak period. NLEV for:

[0134]

[0135] Where: h MDA is the hourly average power consumption during the morning peak in the distribution station area; NDA It is the hourly average electricity consumption during the evening peak in the distribution station area.

[0136] The upper photovoltaic power balance game model of the distribution substation area DSG for:

[0137]

[0138] Where: p DDA (t) and v DDA They are the power supply quantity and power supply price within the time period of the distribution station area; and v REAPV are the power generation and power generation price of PV in the distribution station area respectively; and v ES are the charging and discharging capacity and charging and discharging price of energy storage in the distribution station area respectively; p ALLREM (t) and v ALLREM They are the electricity consumption and electricity price of users in the distribution station area respectively.

[0139] Through the PV double-layer game in the distribution station area, the PV power generation balance under the optimal value can be solved.

[0140] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the scope of protection of the present invention.

Claims

1. A photovoltaic power balancing method for a distribution area based on a double-layer architecture, characterized in that: The following steps are involved: Step S100: forecasting the photovoltaic power and load of the distribution substation area respectively, and conducting forecast analysis on the power supply and demand of the distribution substation area; Step S200: Obtain the real-time changes in the power supply and demand of the current distribution area and obtain the energy storage charging power; when the sum of the user load of the distribution area predicted in step S100 and the energy storage charging power is less than the photovoltaic power predicted in step S100, calculate the distribution area grid power p UDA ; If the power of the distribution area is p UDA Greater than the power limit value Δp in the distribution area NL , then PV power generation is abandoned and the process goes to step S300; otherwise, the process goes directly to step S300; Step S300: The lower photovoltaic power game of the distribution area is used to obtain the optimal solution of the user energy consumption and energy storage charging strategy; the lower photovoltaic power game model p USG for: in: is the SG game function; p DDA Unload grid load for distribution substation area; The photovoltaic power of the distribution area predicted in step S100; Provides energy storage and charging power for distribution station areas; P ALLREM It is the sum of the loads of the electricity meters of users in the distribution station area; Step S400: For the upper photovoltaic power balance of the distribution area, a static game is conducted between PV power generation, energy storage charging and discharging, and user load to solve the optimal economic strategy for PV balance in the distribution area, so as to achieve the optimal benefits for the four entities of the distribution area: power supply company, power user, PV aggregator, and energy storage aggregator; the upper photovoltaic power balance game model p of the distribution area DSG for: Where: p DDA (t) and v DDA They are the power supply quantity and power supply price within the time period of the distribution station area; and v REAPV are the photovoltaic power generation and power generation price in the distribution station area; and v ES are the charging and discharging capacity and charging and discharging price of energy storage in the distribution station area respectively; p ALLREM (t) and v ALLREM They are the electricity consumption and electricity price of users in the distribution station area; Step S500: Generate a photovoltaic power generation control strategy and an energy storage charge and discharge control strategy for the distribution station area based on the photovoltaic power value and the energy storage discharge power value curve respectively; then, perform power control on the distribution station area to achieve an economic balance of photovoltaic power in the distribution station area.

2. The photovoltaic power balancing method for a distribution station area based on a double-layer architecture according to claim 1 is characterized in that: In step S200, the power p UDA The calculation formula is as follows: in: The load of different user meters; Power limit value Δp for the distribution area NL The calculation formula is as follows: Where: n REA The number of distribution stations within the distribution dispatch area; It is the minimum load of different distribution station areas.

3. The photovoltaic power balancing method for a distribution station area based on a double-layer architecture according to claim 1 is characterized in that: In step S300, when the sum of the user load of the distribution station area predicted in step S100 and the energy storage charging power is greater than or equal to the photovoltaic power predicted in step S100, the off-grid load p of the distribution station area is calculated. DDA ; When the energy storage is in the charging state, the off-grid load p DDA for: Where: n DA The number of household meters in the distribution area; The load of different user meters; Provides energy storage and charging power for distribution station areas; The photovoltaic power of the distribution station area; When the energy storage is in the discharge state, the load p DDA for: in: The discharge power of the distribution station area.

4. A photovoltaic power balancing method for a distribution station area based on a double-layer architecture according to any one of claims 1 to 3, characterized in that: In step S100, an LSTM network is used in combination with meteorological data to predict the photovoltaic power of the distribution station area. In the LSTM network structure, the input gate receives numerical weather forecast data, photovoltaic power data at the previous moment, and user load data respectively; the forget gate selectively memorizes the numerical weather forecast data, photovoltaic power data at the previous moment, and user load data, and superimposes the user load and photovoltaic power data at the current moment to form a memory state; and the output gate outputs the photovoltaic power data and user load data at the next moment respectively.

5. The photovoltaic power balancing method for a distribution station area based on a double-layer architecture according to claim 4 is characterized in that: In step S100, when predicting the photovoltaic power, the photovoltaic power at the current moment retained by the forget gate for: Where: δ() is the activation function in the LSTM for photovoltaic power prediction; w F is the historical weight of the photovoltaic power in the distribution station area; s WF Numerical weather forecast data provided to meteorological service providers; The photovoltaic power data of the distribution area at the previous moment; The photovoltaic voltage data of the distribution area at the previous moment; l FGPV Function bias for the LSTM forget gate.

6. The photovoltaic power balancing method for a distribution station area based on a double-layer architecture according to claim 5, characterized in that: In step S100, when predicting photovoltaic power, the state q of the LSTM network unit NET for: in: is the photovoltaic power state variable of the distribution station area at the previous moment; Enter the photovoltaic power data of the gate distribution station area at the current moment; The photovoltaic power data of the distribution area of ​​the coupling unit in LSTM; LSTM distribution station photovoltaic power prediction output gate data for: Where: q NET is the photovoltaic power state variable of the distribution area at the current moment; w OUT The LSTM output gate weights for the photovoltaic power of the distribution station area; The photovoltaic power data of the distribution area at the current moment; Tanh() is the activation function in the LSTM network.

7. The photovoltaic power balancing method for a distribution station area based on a double-layer architecture according to claim 1, characterized in that: The step S400 includes the following constraints: Energy storage device power configuration of energy storage aggregator p NLEV for: Where: p NLTF The rated power of the distribution transformer in the distribution station area; The average power generated by PV is predicted; is the predicted maximum user load; Energy storage capacity configuration h NLEV for: Where: h MDA The hourly average power consumption during the morning peak in the distribution area; h NDA It is the hourly average electricity consumption during the evening peak in the distribution station area.

8. A photovoltaic power balancing system for distribution stations based on a double-layer architecture, characterized in that: The invention comprises a memory and a processor; the memory stores a computer program; the processor is configured to execute the computer program in the memory to implement the method according to any one of claims 1 to 6.

Images