Electric bus non-electricity index evaluation and transaction method

Abstract

The invention discloses an electric bus non-electricity index evaluation and transaction method. The method comprises the following steps: firstly, acquiring information such as planned operation time, battery capacity, maximum charging rate, station passenger flow volume and the like; based on the passenger flow information and the extreme weather information of different stations, the time and the residual electric quantity of the electric bus returning to the charging station are corrected, a Bi-LSTM-based prediction model of the passenger flow and the extreme weather on the running time parameters of the electric bus is provided, and the passenger flow influence parameters and the extreme weather influence parameters are predicted. Based on the electric bus load, non-electric indexes such as the economic regulation capability, the green power regulation capability and the environment regulation capability of the electric bus are evaluated, the charging strategy of the electric bus is optimized through reinforcement learning, and intelligent and low-carbon charging scheduling is achieved. Settlement, including power support transaction settlement, green electricity transaction settlement and the like, is carried out when the electric bus participates in non-electricity index transaction, effective evaluation is carried out on the non-electricity indexes of the electric bus, and the load side adjusting capacity is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system regulation and control, and particularly to a method for evaluating and trading non-electric indicators of electric buses. Background Art

[0002] Evaluating the non-electric indicators of electric buses and constructing a trading mechanism can accurately evaluate non-electric parameters such as the driving mileage, on-vehicle load, and carbon emission reduction contribution of buses, improve the market value of electric buses, increase the operation revenue, and at the same time contribute to the green and low-carbon development of urban transportation, enhance the flexibility of power grid dispatching, and achieve the dual optimization of bus energy utilization and environmental protection benefits.

[0003] However, the existing methods have the following deficiencies: 1) The real-time performance and integrity of load information collection are insufficient. The existing load data collection system responds slowly and fails to track the load changes of buses in real time. Especially during peak hours or extreme weather, the load fluctuates violently, which is difficult to capture by traditional methods. Traditional load collection mainly focuses on electrical parameters (such as SOC, charging power, etc.), while paying less attention to non-electric parameters such as passenger flow, driving trajectory, and traffic conditions, resulting in inaccurate load prediction. 2) The accuracy of charging load prediction is insufficient. The current charging load prediction methods are mainly based on historical power data and do not fully consider the impact of passenger flow changes on charging demand. Weather factors (such as heavy rain, cold snap, heat wave, etc.) will affect the bus driving route and load demand, but the existing models are difficult to dynamically adjust the prediction results. Traditional linear models are difficult to capture the non-linear change characteristics of charging load, resulting in large prediction errors. 3) The evaluation method of load non-electric indicators is not perfect. The existing methods mainly focus on power consumption and lack a systematic evaluation of non-electric indicators such as vehicle utilization rate, carbon emission reduction contribution, and passenger flow carrying rate. Making full use of deep reinforcement learning (such as DQN) to optimize the evaluation of the adjustment ability of non-electric indicators results in insufficient intelligence in bus operation scheduling. 4) The non-electric indicator trading mechanism is insufficient. Currently, there is no effective market mechanism for carbon credit trading and load regulation trading of buses, which cannot give full play to the load regulation ability and carbon emission reduction value of electric buses. The existing trading mechanism is difficult to reasonably evaluate the load regulation value of buses and lacks a dynamic pricing model based on comprehensive factors such as passenger flow, driving time, and environmental impact.

[0004] For the above reasons, the present invention designs a method for evaluating and trading non-electric indicators of electric buses to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art, and to propose a method for evaluating and trading non-electric indicators of electric buses. This method formulates an evaluation and trading mechanism for non-electric indicators of electric buses based on Bi-LSTM and DQN, and proposes a bus load information collection technology to improve data integrity and real-time performance.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] An evaluation and trading method for non - electrical indicators of electric buses, comprising the following steps:

[0008] Step (1), collecting the load information of electric buses;

[0009] Step (2), constructing a correction model for the charging load of electric buses based on passenger flow and extreme weather, correcting the time when the electric bus returns to the charging station and the remaining power, and outputting the corrected load; proposing a prediction model for the operating time parameters of electric buses to predict the passenger flow influence coefficient and the extreme weather influence coefficient;

[0010] Step (3), based on non - electrical indicators, adopting a deep - network algorithm to optimize the charging strategy of electric buses through reinforcement learning, and evaluating its economic regulation ability, green - electricity regulation ability and environmental regulation ability;

[0011] Step (4), settling the non - electrical indicator trading of electric buses, dynamically adjusting the charging strategy according to the power support price, electricity price and carbon emission factor to maximize the market revenue.

[0012] As a further preferred solution of the present invention, in step (1), when collecting the load information of electric buses, the collected data is expressed by formula (1):

[0013]

[0014] Where: T base (i) is the planned operation time, SOC init (i) is the initial SOC, E cap (i) is the battery capacity, P charge (i) is the maximum charging rate, SOC target is the target SOC, P s (j,t) is the passenger flow at the station, P(i) is the total bus passenger flow influence, α pass is the influence coefficient of passenger flow on operation time, γ pass is the influence coefficient of passenger flow on energy consumption, W(i) is the weather influence factor, β weather is the influence coefficient of weather on operation time, δ weather is the influence coefficient of weather on energy consumption, P max is the maximum capacity of the charging station, T charge (i) is the residence time at the charging station, C(t) is the real - time electricity price, P DR (t) is the demand response signal.

[0015] As a further preferred solution of the present invention, in step (2), a correction model for the charging load of electric buses based on passenger flow and extreme weather is constructed to correct the return time and remaining power of the electric buses to the charging station, and the process of outputting the corrected load is as follows:

[0016] Calculate the actual running time of the bus as shown in formula (2):

[0017]

[0018] In the formula: T run (i) is the actual running time of bus i, T base (i) is the planned running time of bus i under normal conditions, ΔT pass (i) is the additional running time caused by the change in passenger flow, ΔT weather (i) is the additional running time caused by extreme weather, P(i) is the weighted sum of the passenger flows at each station passed by bus i, W(i) is the weather impact factor during the operation of bus i, α pass , β weather are empirical parameters to measure the impact of flow and weather on the running time;

[0019] Calculate the remaining power when the bus arrives at the station, as shown in formula (3). Under different running times, the remaining power of the bus when it arrives at the station is affected by the running time, air conditioner use and load conditions, and the correction is as follows:

[0020]

[0021] In the formula: SOC arr (i) is the remaining battery power of bus i when it arrives at the charging station, SOC init (i) is the battery capacity of bus i when it leaves the station, E base (i) is the theoretical energy consumption of bus i under normal conditions, ΔE pass (i) is the additional energy consumption due to the increase in passenger flow, ΔE weather (i) is the additional energy consumption due to extreme weather such as low temperature and high temperature, E cap (i) is the total battery capacity of bus i, γ pass is the coefficient of the impact of passenger flow on unit energy consumption, δ weather is the coefficient of the impact of weather on unit energy consumption;

[0022] Calculate the charging demand, as shown in formula (4). The charging demand is to ensure that the battery is charged to the target SOC, and the corrected charging power is calculated as follows:

[0023] E charge (i) = max(SOC target ·E cap(i) - State of Charge (SOC) arr (i)·E cap (i),0) (4)

[0024] Where: SOC target is the target SOC, and E charge (i) is the charging demand of bus i;

[0025] Calculate the corrected charging load. As shown in formula (5), assuming that the charging power of the bus is limited by the capacity of the charging station, the corrected charging load is calculated as follows:

[0026]

[0027] Where: P load (t) is the total charging load of the charging station at time t, N(t) is the set of buses in the charging state at time t, and P charge (i) is the charging power of bus i, and T charge(i) is the residence time of bus i at the charging station.

[0028] As a further preferred solution of the present invention, a prediction model of the influence of passenger flow and extreme weather on the operation time parameters of electric buses based on Bi-LSTM is proposed to predict the passenger flow influence coefficient and the extreme weather influence coefficient, including the following process:

[0029] Taking the time step t as the input window, the passenger flow data of the past T time steps, as shown in formula (6), the weather factor data of the past T time steps as shown in formula (7), and the historical bus operation time data as shown in formula (8):

[0030] Ρ t =[P(t - T), P(t - T + 1),..., P(t - 1)] (6)

[0031] W t =[W(t - T), W(t - T + 1),..., W(t - 1)] (7)

[0032] Τ run =[T run (t - T), T run (t - T + 1),..., T run (t - 1)] (8)

[0033] Where: P(t) is the total passenger flow at time t, P(t - T) is the total passenger flow in the past T time steps, P(t - T + 1) is the total passenger flow in the past T - 1 time steps, P(t - 1) is the total passenger flow in the past 1 time step, W(t) is the weather impact factor at time t, W(t - T) is the weather impact factor in the past T time steps, W(t - T + 1) is the weather impact factor in the past T - 1 time steps, W(t - 1) is the weather impact factor in the past 1 time step, and T run (t) is the bus running time at time t; T run (t - T) is the bus running time in the past T time steps, T run (t - T + 1) is the bus running time in the past T - 1 time steps, T run (t - 1) is the bus running time in the past 1 time step;

[0034] Construct the input matrix of Bi-LSTM. The input matrix is shown in Equation (9):

[0035]

[0036] Where: Dimension: X t ∈ T×3 , P(t) is the total passenger flow at time t, W(t) is the weather impact factor at time t, and T run (t) is the bus running time at time t;

[0037] Bi-LSTM is composed of a forward LSTM and a backward LSTM, as shown in Equation (10):

[0038]

[0039] Where: is the hidden state of the forward LSTM, is the hidden state of the backward LSTM, and H t is the combined hidden state of the bidirectional LSTM. Dimension: X t ∈ T×3 ;

[0040] Predict the passenger flow impact coefficient α pass (t) and the extreme weather impact coefficient β weather (t) through the fully connected layer, as shown in Equation (11):

[0041]

[0042] Where: W α 、W β are weight matrices, and b α 、b β are bias terms, and Ht It is the combined hidden state of the bidirectional LSTM.

[0043] As a further preferred solution of the present invention, in step (3), based on the load of the electric bus, evaluate its non-electric indicators of economic regulation ability, green power regulation ability and environmental regulation ability, adopt a deep network algorithm, and optimize the charging strategy of the electric bus through reinforcement learning. The process of evaluating its economic regulation ability, green power regulation ability and environmental regulation ability is as follows:

[0044] Step 31), define the state space. At time step t, the system state S t is composed of the following variables, as shown in formula (12):

[0045] S t ={P grid (t), P RE (t), C(t), E CO2 (t), SOC bus (t), P charge (t)} (12);

[0046] Step 32), define the action space. The reinforcement learning agent selects a charging power adjustment strategy at each time step t, as shown in formula (13):

[0047]

[0048] Step 33), use the reward function based on DQN to guide the agent to learn the optimal adjustment strategy, as shown in formula (14):

[0049]

[0050] Step 34), use Q-learning based on DQN for training, update the Q value, and through continuous iteration, make the Q value converge to find the optimal charging adjustment strategy, as shown in formula (15):

[0051]

[0052] Step 35), calculate the non-electric indicator adjustment ability. After the training is completed, use the optimal strategy predicted by DQN to evaluate the non-electric indicator adjustment ability, and evaluate the cost reduced by the bus through intelligent charging, the carbon emission reduction during charging, and the proportion of green power used through the economic regulation ability, green power regulation ability, and environmental regulation ability, as shown in formula (16):

[0053]

[0054] In the above formula: S t is the system state, P grid(t) is the power supplied by the power grid, P RE (t) is the power supplied by renewable energy, C(t) is the real-time electricity price, is the carbon emission factor per unit of electricity at a point, SOC bus (t) is the SOC of the current bus, P charge (t) is the charging power of the current bus, A t is the reinforcement learning agent, P charge (t) takes values in the range [P min , P max , with a step size of ΔP, a t ∈{-1, 0, 1}, R econ (t) is the economic regulation ability, Δt is the time step, R green (t) is the green electricity regulation ability, R env (t) is the environmental regulation ability, R t is the comprehensive reward function, ω1, ω2, ω3 are the weight coefficients of different regulation abilities, α is the learning rate, γ is the discount factor, Q(S t , A t ) is the current state-action value function, is the future optimal Q value, EAC is the economic regulation ability, GAC is the green electricity regulation ability, and EVC is the environmental regulation ability.

[0055] As a further preferred solution of the present invention, in step (4), when settling the non-electricity index trading of electric buses, the charging strategy is dynamically adjusted according to the power support price, electricity price and carbon emission factor to maximize the market revenue, including the following steps:

[0056] Step (41), establish a non-electricity index trading market for electric buses, define the electric bus operator as the seller, providing: power support, adjusting the charging power to respond to the power grid demand; green electricity trading, using green electricity for charging to improve the new energy consumption rate, environmental regulation support, reducing carbon emissions, and providing low-carbon charging services for the power grid; the power grid operator, renewable energy power generation company, and government department act as buyers and pay corresponding fees;

[0057] Step (42), power support trading settlement, the bus participates in the demand response of the power grid by adjusting the charging power, providing peak shaving and valley filling support, and the power support revenue is as shown in formula (17):

[0058] R power (t) = λ power (t)·|P baseline (t) - P charge (t)| (17);

[0059] Step (43): Green electricity settlement transaction. The bus optimizes the charging time period according to the green electricity price to maximize the green electricity trading revenue. The revenue calculation is shown in formula (18):

[0060] R green (t) = λ green (t)·P RE (t) (18);

[0061] Step (44): Environmental regulation support settlement. The bus reduces carbon emissions by optimizing the charging strategy. The revenue calculation is shown in formula (19):

[0062]

[0063] Step (45): Total transaction revenue calculation. The total revenue of the bus operator consists of three parts. The bus company can dynamically adjust the charging strategy according to the power support price, electricity price, and carbon emission factor to maximize the market revenue, as shown in formula (20):

[0064]

[0065] In the above formula: R power (t) is the power support revenue, λ power (t) is the power support subsidy price provided by the power grid, P baseline (t) is the charging power of the electric bus before adjustment, P charge (t) is the actual charging power of the electric bus, R green (t) is the green electricity trading revenue, λ green (t) is the green electricity market settlement price, P RE (t) is the green electricity power used by the bus at time t, R env (t) is the environmental regulation revenue, λ env (t) is the carbon emission reduction subsidy price, is the carbon emission of the bus when using ordinary grid electricity, is the carbon emission of the bus after adopting the optimized charging strategy, R total is the total revenue of the bus during the entire trading cycle.

[0066] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention proposes a bus load information acquisition technology to improve data integrity and real-time performance; establishes a correction model for passenger flow plus extreme weather load to enhance the prediction accuracy of charging load; uses DQN to optimize the evaluation of non-electric index adjustment ability to improve the intelligence of bus V2G scheduling; constructs a non-electric index trading mechanism to promote the commercial application of electric buses in the carbon credit and load regulation markets. Thus, the technical solution of the present invention can improve the prediction accuracy of the load of electric buses, realize the quantitative evaluation of the non-electric indexes of electric buses, including regulation capabilities such as economy, green electricity, and environment, and enhance the effectiveness of non-electric index trading of electric buses. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a schematic flow chart of a method for evaluating and trading non-electric indexes of an electric bus proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0069] Figure 1 It is a schematic flow chart of a method. In this embodiment, a method for evaluating and trading non-electric indexes of an electric bus based on a bidirectional long short-term memory recurrent neural network (Bi-LSTM) and a reinforcement learning algorithm (Deep Q Network, DQN) is proposed. The method mainly includes the following steps:

[0070] (1) Propose an electric bus load information acquisition technology.

[0071] Electric vehicle charging information acquisition: The collected information mainly includes planned operation time, initial SOC, battery capacity, maximum charging rate, target SOC, passenger flow at the station, total bus passenger flow impact, passenger flow impact coefficient on operation time, passenger flow impact coefficient on energy consumption, weather impact factor, weather impact coefficient on operation time, weather impact coefficient on energy consumption, charging station capacity, charging station stay time, real-time electricity price, demand response signal.

[0072] The specific data collected is shown in formula (1):

[0073]

[0074] Where: T base (i) is the planned operation time (in minutes), SOC init (i) is the initial SOC, within the range of (0 - 1); Ecap (i) is the battery capacity (unit: kWh), P charge (i) is the maximum charging rate (unit: kWh), SOC target is the target SOC, also within the range of (0 - 1); P s (j,t) is the passenger flow of the station (persons), P(i) is the total bus passenger flow impact (persons), α pass is the influence coefficient of passenger flow on running time (minutes / person), γ pass is the influence coefficient of passenger flow on energy consumption (kWh / person), W(i) is the weather influence factor, β weather is the influence coefficient of weather on running time (minutes / weather factor), δ weather is the influence coefficient of weather on energy consumption (kWh / weather factor), P max is the maximum capacity of the charging station (kW), T charge (i) is the residence time at the charging station (hours), C(t) is the real-time electricity price (yuan / kWh), P DR (t) is the demand response signal (kW).

[0075] (2) Construct a correction model for the charging load of electric buses based on passenger flow and extreme weather.

[0076] Based on the passenger flow information and extreme weather information of different stations, correct the return time and remaining power of electric buses to the charging station, and then realize the correction of the charging load and output the corrected load. A prediction model based on Bi-LSTM for the influence of passenger flow and extreme weather on the running time parameters of electric buses is proposed to predict the passenger flow influence coefficient (the influence coefficient of passenger flow on running time) and the extreme weather influence coefficient (the influence coefficient of weather on running time).

[0077] Calculate the actual running time of the bus as shown in formula (2). Since the passenger flow and weather affect the arrival time of the bus, correct the running time of the electric bus based on the above information.

[0078]

[0079] In the formula: T run (i) is the actual running time of bus i, T base (i) is the planned running time of bus i under normal conditions, ΔT pass (i) is the additional running time caused by the change in passenger flow, ΔT weather (i) is the additional running time caused by extreme weather (rain, snow, strong wind, etc.), P(i) is the total bus passenger flow impact (persons), W(i) is the weather influence factor during the operation of bus i (dimensionless, weighted after normalization of wind speed, temperature, etc.), α pass , β weatherAre empirical parameters that respectively measure the impacts of traffic flow and weather on the running time.

[0080] Calculate the remaining battery power when the bus arrives at the station. As shown in formula (3), under different running times, the remaining battery power SOC when the bus arrives at the station arr (i) Affected by the running time, air conditioner usage (affected by weather), and load conditions, it is corrected as follows:

[0081]

[0082] In the formula: SOC arr (i) Is the remaining battery power of bus i when it arrives at the charging station, SOC init (i) Is the battery capacity of bus i when it departs from the station, E base (i) Is the theoretical energy consumption of bus i under normal conditions, ΔE pass (i) Is the additional energy consumption due to the increase in passenger flow, ΔE weather (i) Is (the additional energy consumption due to extreme weather such as low temperature and high temperature), E cap (i) Is the total battery capacity of bus i, γ pass Is the influence coefficient of passenger flow on energy consumption, δ weather Is the influence coefficient of weather on energy consumption.

[0083] Calculate the charging demand. As shown in formula (4), the charging demand is to ensure that the battery is charged to the target SOC (such as 90%). The corrected charging power E charge (i) Is calculated as follows:

[0084] E charge (i) = max(SOC target ·E cap (i) - SOC arr (i)·E cap (i), 0) (4)

[0085] In the formula: SOC target Is the target SOC (such as 0.9), E charge (i) Is the charging demand of bus i.

[0086] Calculate the corrected charging load. As shown in formula (5), assuming that the charging power of the bus is limited by the capacity of the charging station, the corrected charging load is calculated as follows:

[0087]

[0088] In the formula: P load (t) Is the total charging load of the charging station at time t, N(t) is the set of buses in the charging state at time t, P charge(i) is the charging power of bus i, and T charge(i) is the residence time of bus i at the charging station.

[0089] A prediction model for the operation time parameters of electric buses based on Bi-LSTM considering passenger flow and extreme weather is proposed to predict the passenger flow influence parameters and extreme weather influence parameters.

[0090] Taking the time step t as the input window, the passenger flow data of the past T time steps is shown in Equation (6). The weather factor data of the past T time steps is shown in Equation (7). The historical bus operation time data is shown in Equation (8).

[0091] Ρ t = [P(t - T), P(t - T + 1),..., P(t - 1)] (6)

[0092] W t = [W(t - T), W(t - T + 1),..., W(t - 1)] (7)

[0093] Τ run = [T run (t - T), T run (t - T + 1),..., T run (t - 1)] (8)

[0094] Where: P(t) is the total passenger flow at time t, P(t - T) is the total passenger flow data of the past T time steps, P(t - T + 1) is the total passenger flow data of the past T - 1 time steps, P(t - 1) is the total passenger flow data of the past 1 time step, W(t) is the weather influence factor at time t, W(t - T) is the weather influence factor of the past T time steps, W(t - T + 1) is the weather influence factor of the past T - 1 time steps, W(t - 1) is the weather influence factor of the past 1 time step, T run (t) is the bus operation time at time t (unit: minute); T run (t - T) is the bus operation time of the past T time steps, T run (t - T + 1) is the bus operation time of the past T - 1 time steps, T run (t - 1) is the bus operation time of the past 1 time step.

[0095] The Bi-LSTM input matrix is shown in Equation (9).

[0096]

[0097] Where: dimension X t ∈ T×3, P(t) is the total passenger flow at time t (unit: person), W(t) is the weather influence factor (dimensionless), and T run (t) is the running time of the bus at time t (unit: minute).

[0098] Bi-LSTM consists of a forward LSTM and a backward LSTM, as shown in formula (10).

[0099]

[0100] In the formula: is the forward LSTM hidden state, is the backward LSTM hidden state, and H t is the combined hidden state of the bidirectional LSTM, with dimension X t ∈ T×3 .

[0101] The passenger flow influence coefficient and extreme weather influence coefficient are predicted through a fully connected layer (FC), as shown in formula (11):

[0102]

[0103] In the formula: W α , W β are weight matrices, b α , b β are bias terms, α pass (t) is the coefficient of the passenger flow affecting the running time of the bus (passenger flow influence coefficient), (unit: minute / person), and β weather (t) is the coefficient of the weather factor affecting the running time of the bus (extreme weather influence coefficient), (unit: minute / weather factor), and H t is the combined hidden state of the bidirectional LSTM.

[0104] (3) A method for evaluating the non-electric index regulation ability of electric buses based on DQN is proposed.

[0105] Based on the load of electric buses, evaluate their non-electric indexes such as economic regulation ability, green power regulation ability, and environmental regulation ability. This method uses a deep Q-network (DQN) to optimize the charging strategy of electric buses through reinforcement learning, and evaluate the regulation ability of non-electric indexes such as economic regulation ability (EAC), green power regulation ability (GAC), and environmental regulation ability (EVC) to achieve intelligent and low-carbon charging scheduling.

[0106] Method for evaluating the non-electric index regulation ability of electric buses based on DQN:[[]]

[0107] Step 31), state space definition, time step t and system state St It consists of the following variables, as shown in formula (12):

[0108]

[0109] In step 32), the action space is defined. The reinforcement learning agent selects a charging power adjustment strategy at each time step t, as shown in formula (13):

[0110]

[0111] In step 33), the reward function is designed. DQN uses the reward function to guide the agent to learn the optimal adjustment strategy. The rewards include economic adjustment ability, green power adjustment ability, and environmental condition ability, as shown in formula (14):

[0112]

[0113] In step 34), the DQN training process. DQN uses Q-learning for training to update the Q value. Through continuous iteration, DQN makes the Q value converge to find the optimal charging adjustment strategy, as shown in formula (15):

[0114]

[0115] In step 35), calculate the non-electric index adjustment ability. After the training is completed, the optimal strategy predicted by DQN can evaluate the non-electric index adjustment ability. The cost reduced by the bus through intelligent charging, the carbon emissions reduced during charging, the proportion of green power used, etc. are evaluated through economic adjustment ability, green power adjustment ability, and environmental adjustment ability, as shown in formula (16).

[0116]

[0117] In the above formula: S t is the system state, P grid (t) is the grid power supply (kW), P RE (t) is the renewable energy (green power) supply (kW), C(t) is the real-time electricity price (yuan / kWh), is the carbon emission factor of electricity at the location (kg CO2 / kWh), SOC bus (t) is the SOC of the current bus, and its value range is (0 - 1), A t is the reinforcement learning agent, P charge (t) is the charging power (kW) of the current bus, P charge (t) has a value range of [P min , P max , P max is the maximum capacity (kW) of the charging station, Pmin is the minimum capacity (kW) of the charging station, and the step size ΔP (e.g., 10 kW)a t ∈{-1, 0, 1} (decrease, remain unchanged, increase the charging power), R econ (t) is the economic adjustment capacity (EAC), and Δt is the time step (hours), R green (t) is the green power trading revenue (yuan), R env (t) is the environmental adjustment capacity (EVC), R t is the comprehensive reward function, ω1, ω2, ω3 are the weight coefficients of different adjustment capabilities, α is the learning rate, γ is the discount factor, which measures the importance of future seizure management, Q(S t ,A t ) is the current state-action value function, is the future optimal Q value, EAC is the economic adjustment capacity, GAC is the green power adjustment capacity, and EVC is the environmental adjustment capacity.

[0118] (4) Propose a non-electric index trading mechanism for electric buses.

[0119] Settle the participation of electric buses in non-electric index trading, including power support trading settlement, green power trading settlement, and environmental adjustment support settlement. The non-electric index trading mechanism for electric buses enables bus operators to participate in the power market, provide power support, green power trading, and environmental adjustment support, and obtain corresponding economic benefits while meeting their own charging needs through market-based means.

[0120] For the non-electric index trading mechanism of electric buses, first of all, bus operators, as the sellers in the trading market, provide adjustable charging loads to level the power grid peaks and valleys, while preferentially using renewable energy for charging and reducing carbon emissions through optimized charging strategies. Grid operators, renewable energy power generators, and government agencies, as the buyers, pay corresponding subsidies to encourage buses to improve their load adjustment capabilities, promote the consumption of new energy, and reduce carbon emissions. The trading mechanism includes three core links. Power support trading allows buses to respond to grid demands by adjusting their charging powers and thus obtain power support revenues, which are calculated based on the adjusted power and market subsidy prices. Green power trading encourages buses to preferentially use renewable energy for charging and obtain revenues according to the amount of green power used. Environmental adjustment support trading is based on the reduction of carbon emissions, and bus companies can obtain subsidies from the government or the carbon market. Finally, bus operators calculate the total trading revenue based on the revenues of each trading link and optimize their charging strategies to maximize economic benefits, improve the utilization rate of green power, and reduce carbon emissions at the same time, contributing to the development of intelligent and low-carbon transportation.

[0121] Step 41), establish a non-electric index trading market for electric buses, define electric bus operators (bus companies) as sellers, and provide: power support, adjusting the charging power to respond to grid demands. Green power trading, using green power for charging to increase the consumption rate of new energy. Environmental regulation support, reducing carbon emissions and providing low-carbon charging services for the grid. Grid operators, renewable energy generators, and government departments act as buyers and pay corresponding fees.

[0122] Step 42), power support trading settlement. Buses can participate in the grid's demand response by adjusting the charging power P charge (t), providing peak shaving and valley filling support. Buses can adjust the charging strategy to reduce peak loads or fill valley loads, increasing the power trading revenue. The power support revenue is shown in formula (17):

[0123] R power (t) = λ power (t) · |P baseline (t) - P charge (t)| (17)

[0124] Step 43), green power settlement trading. The higher the proportion of green power used by buses during charging, the greater the green power trading revenue. Buses can optimize the charging time according to the green power price to maximize the green power trading revenue. The green power trading revenue calculation is shown in formula (18):

[0125] R green (t) = λ green (t) · P RE (t) (18)

[0126] Step 44), environmental regulation support settlement. Buses reduce carbon emissions by optimizing the charging strategy, and the government or carbon market provides subsidies. Buses can obtain higher carbon trading revenue if they use low-carbon electricity or avoid high-emission periods during charging. The revenue calculation is shown in formula (19):

[0127]

[0128] Step 45), total trading revenue calculation. The total revenue of bus operators consists of three parts. Bus companies can dynamically adjust the charging strategy according to the power support price, electricity price, carbon emission factor, etc. to maximize the market revenue. As shown in formula (20):

[0129]

[0130] In the formula: R power (t) is the power support revenue (yuan), λ power (t) is the power support subsidy price provided by the grid (yuan / kW), P baseline(t) is the charging power (kW) of the electric bus before adjustment, P charge (t) is the actual charging power (kW) of the electric bus, R green (t) is the green power trading revenue (yuan), λ green (t) is the green power market settlement price (yuan / kWh), P RE (t) is the green power consumed by the bus at time t (kW), R env (t) is the environmental regulation revenue (yuan), λ env (t) is the carbon emission reduction subsidy price (yuan / kgCO2), is the carbon emission (kgCO2) of the bus when using ordinary grid power, is the carbon emission (kgCO2) of the bus after adopting the optimized charging strategy, R total is the total revenue (yuan) of the bus during the entire trading cycle.

[0131] The above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered within the protection scope of the present invention.

Claims

1. An evaluation and trading method for non - electrical indicators of electric buses, characterized in that, It includes the following steps: Step (1), collecting the load information of electric buses; Step (2), constructing a correction model for the charging load of electric buses based on passenger flow and extreme weather, correcting the return time and remaining power of electric buses to the charging station, and outputting the corrected load; proposing a prediction model for the operating time parameters of electric buses to predict the passenger flow influence coefficient and extreme weather influence coefficient; Step (3), based on non-electric indicators, using a deep network algorithm to optimize the charging strategy of electric buses through reinforcement learning, and evaluating its economic regulation ability, green power regulation ability and environmental regulation ability; Step (4), settling the participation of electric buses in non-electric indicator transactions, dynamically adjusting the charging strategy according to the power support price, electricity price and carbon emission factor to maximize the market revenue.

2. The non-electric index evaluation and trading method for an electric bus according to claim 1, characterized in that, In step (1), when collecting the load information of electric buses, the collected data is expressed by formula (1): Where: T base (i) is the planned operation time, SOC init (i) is the initial SOC, E cap (i) is the battery capacity, P charge (i) is the maximum charging rate, SOC target is the target SOC, P s (j,t) is the passenger flow at the station, P(i) is the impact of the total bus passenger flow, α pass is the coefficient of the impact of passenger flow on operation time, γ pass is the coefficient of the impact of passenger flow on energy consumption, W(i) is the weather impact factor, β weather is the coefficient of the impact of weather on operation time, δ weather is the coefficient of the impact of weather on energy consumption, P max is the maximum capacity of the charging station, T charge (i) is the residence time at the charging station, C(t) is the real-time electricity price, P DR (t) is the demand response signal.

3. The non-electric index evaluation and trading method for an electric bus according to claim 1, characterized in that In step (2), the process of constructing a correction model for the charging load of electric buses based on passenger flow and extreme weather, correcting the return time and remaining power of electric buses to the charging station, and outputting the corrected load is as follows: Calculating the actual operating time of the bus as shown in formula (2): Where: T run (i) is the actual running time of bus i, T base (i) is the planned running time of bus i under normal conditions, ΔT pass (i) is the additional running time caused by the change in passenger flow, ΔT weather (i) is the additional running time caused by extreme weather, P(i) is the weighted sum of the passenger flows at each stop passed by bus i, W(i) is the weather impact factor during the running of bus i, α pass , β weather are empirical parameters that measure the impact of flow and weather on the running time; Calculating the remaining power when the bus arrives at the station, as shown in formula (3). At different operating times, the remaining power of the bus when it arrives at the station is affected by the operating time, air conditioner use and load conditions, and the correction is as follows: Where: SOC arr (i) is the remaining battery power of bus i when it arrives at the charging station, SOC init (i) is the battery capacity of bus i when it departs from the station, E base (i) is the theoretical energy consumption of bus i under normal conditions, ΔE pass (i) is the additional energy consumption caused by the increase in passenger flow, ΔE weather (i) is the additional energy consumption caused by extreme weather such as low temperature and high temperature, E cap (i) is the total battery capacity of bus i, γ pass is the coefficient of the influence of passenger flow on unit energy consumption, δ weather is the coefficient of the influence of weather on unit energy consumption; Calculating the charging demand, as shown in formula (4). The charging demand is to ensure that the battery is charged to the target SOC, and the corrected charging power is calculated as follows: E charge (i)=max(SOC target ·E cap (i)-SOC arr (i)·E cap (i),0) (4) Where: SOC target is the target SOC, and E charge (i) is the charging demand of bus i; Calculating the corrected charging load, as shown in formula (5). Assuming that the charging power of the bus is limited by the charging station capacity, the corrected charging load is calculated as follows: Where: P load (t) is the total charging load of the charging station at time t, N(t) is the set of buses in the charging state at time t, P charge (i) is the charging power of bus i, T charge(i) is the residence time of bus i at the charging station.

4. The method for evaluating and trading non-electrical indicators of electric buses according to claim 3, characterized in that: Proposing a prediction model for the influence of passenger flow and extreme weather on the operating time parameters of electric buses based on Bi-LSTM to predict the passenger flow influence coefficient and extreme weather influence coefficient, including the following process: Taking the time step t as the input window, the passenger flow data of the past T time steps, as shown in formula (6), the weather factor data of the past T time steps, as shown in formula (7), and the historical bus operating time data, as shown in formula (8): Ρ t =[P(t-T),P(t-T+1),...,P(t-1)] (6) W t = [W(t - T), W(t - T + 1),..., W(t - 1)] (7) Τ run = [T run (t - T), T run (t - T + 1),..., T run (t - 1)] (8) Where: P(t) is the total passenger flow at time t, P(tT) is the total passenger flow in the past T time steps, P(t-T+1) is the total passenger flow in the past T-1 time steps, P(t-1) is the total passenger flow in the past 1 time step, W(t) is the weather impact factor at time t, W(tT) is the weather impact factor in the past T time steps, W(t-T+1) is the weather impact factor in the past T-1 time steps, W(t-1) is the weather impact factor in the past 1 time step, T run (t) is the bus running time at time t; T run (tT) is the bus running time in the past T time steps, T run (t-T+1) is the bus running time in the past T-1 time steps, T run (t-1) is the bus running time in the past time step; Bi-LSTM input matrix, constructing the input matrix as shown in formula (9): Where: Dimension: X t ∈ T×3 , P(t) is the total passenger flow at time t, W(t) is the weather influence factor at time t, and T run (t) is the bus running time at time t; Bi-LSTM is composed of a forward LSTM and a backward LSTM, as shown in formula (10): In the formula: is the forward LSTM hidden state, is the backward LSTM hidden state, H t is the combined hidden state of the bidirectional LSTM, dimension: X t ∈ T×3 ; Predict passenger flow impact coefficient α through the fully connected layer pass (t) and extreme weather impact coefficient β weather (t), as shown in formula (11): Where: W α and W β are weight matrices, b α and b β are bias terms, and H t is the combined hidden state of the bidirectional LSTM.

5. The non - electrical index evaluation and trading method for an electric bus according to claim 1, characterized in that, In step (3), based on the load of electric buses, evaluating its non-electric indicators of economic regulation ability, green power regulation ability and environmental regulation ability, using a deep network algorithm to optimize the charging strategy of electric buses through reinforcement learning, and the process of evaluating its economic regulation ability, green power regulation ability and environmental regulation ability is as follows: Step 31), state space definition. At time step t, the system state S t is composed of the following variables, as shown in Equation (12): Step 32), defining the action space, and the reinforcement learning agent selects a charging power adjustment strategy at each time step t, as shown in formula (13): Step 33), using a reward function based on DQN to guide the agent to learn the optimal regulation strategy, as shown in formula (14): Step 34), based on DQN, Q-learning is used for training to update the Q value. Through continuous iteration, the Q value converges to find the optimal charging regulation strategy, as shown in Equation (15): Step 35), calculate the non-electric index regulation ability. After the training is completed, the optimal strategy predicted by DQN is used to evaluate the non-electric index regulation ability. The cost reduction of the bus through intelligent charging, the carbon emission reduction during charging, and the proportion of green electricity used are evaluated through the economic regulation ability, green electricity regulation ability, and environmental regulation ability, as shown in Equation (16): In the above formula: S t is the system state, P grid (t) is the power supplied by the power grid, P RE (t) is the power supplied by renewable energy, C(t) is the real-time electricity price, is the carbon emission factor per unit of electricity at a point, SOC bus (t) is the SOC of the current bus, P charge (t) is the charging power of the current bus, A t is the reinforcement learning agent, P charge (t) takes values in the range [P min , P max , the step size is ΔP, a t ∈{-1, 0, 1}, R econ (t) is the economic regulation ability, Δt is the time step, R green (t) is the green electricity regulation ability, R env (t) is the environmental regulation ability, R t is the comprehensive reward function, ω1, ω2, ω3 are the weight coefficients of different regulation abilities, α is the learning rate, γ is the discount factor, Q(S t , A t ) is the current state-action value function, is the future optimal Q value, EAC is the economic regulation ability, GAC is the green electricity regulation ability, and EVC is the environmental regulation ability.

6. The method for evaluating and trading non-electrical indicators of electric buses according to claim 1, characterized in that: In step (4), the settlement of the non-electric index trading of electric buses is carried out. According to the power support price, electricity price, and carbon emission factor, the charging strategy is dynamically adjusted to maximize the market revenue, including the following steps: Step (41), establish a non-electric index trading market for electric buses. Define the electric bus operator as the seller, providing: power support, adjusting the charging power to respond to the grid demand; green electricity trading, using green electricity for charging to improve the new energy consumption rate; environmental regulation support, reducing carbon emissions, and providing low-carbon charging services for the grid; the grid operator, renewable energy power generation company, and government department as the buyers, paying the corresponding fees; Step (42), power support trading settlement. The bus participates in the demand response of the grid by adjusting the charging power, providing peak shaving and valley filling support, and the power support revenue, as shown in Equation (17): R power R(t) = λ power R(t) · |P baseline (t) - P charge (t)| (17); Step (43), green electricity settlement trading. The bus optimizes the charging period according to the green electricity price to maximize the green electricity trading revenue, and the revenue calculation is as shown in Equation (18): R green λ(t) = green P(t) · RE (t) (18); Step (44), environmental regulation support settlement. The bus reduces carbon emissions by optimizing the charging strategy, and the revenue calculation is as shown in Equation (19): Step (45), total trading revenue calculation. The total revenue of the bus operator consists of three parts. The bus company can dynamically adjust the charging strategy according to the power support price, electricity price, and carbon emission factor to maximize the market revenue, as shown in Equation (20): In the above formula: R power (t) is the power support revenue, and λ power (t) is the power support subsidy price provided by the power grid, P baseline (t) is the charging power of the electric bus before adjustment, P charge (t) is the actual charging power of the electric bus, R green (t) is the green electricity trading revenue, and λ green (t) is the green electricity market settlement price, P RE (t) is the green electricity power consumed by the bus at time t, R env (t) is the environmental regulation revenue, and λ env (t) is the carbon emission reduction subsidy price, is the carbon emission of the bus when using ordinary grid power, is the carbon emission of the bus after adopting the optimized charging strategy, R total is the total revenue of the bus over the entire trading cycle.

Images