Construction of a framework for real-time evaluation of regulation effect of distributed photovoltaic users based on multi-source data fusion

With global warming, energy demand is increasing day by day, and in recent years, there have been frequent “power shortages”, China’s energy structure is irrational, more than three-quarters of its energy supply comes from coal, China’s large-scale mining and burning of coal, and the excessive dependence of the power industry on fossil fuels have not only caused serious damage to China’s natural ecological environment, but also affected social and economic development [1–3]. The development of clean energy has been an inevitable trend. Vigorously developing renewable energy technologies such as solar energy, wind power, hydropower, nuclear energy, etc. is the inevitable choice to guarantee the security of China’s energy supply and to promote the sustainable development of economy and society [4–5].

Nowadays, in the central and eastern provinces of China, which are densely populated, with high electricity prices and strong demand for electricity, distributed photovoltaic (PV) has been equipped for large-scale application due to its good economics, by virtue of its geographical environmental advantages and economic demand for electricity. Among various renewable energy sources, solar energy is riding on the renewable energy track with its rapid development due to its outstanding advantages of inexhaustible source, safety and cleanliness [6–7].

Distributed photovoltaic is an important part of photovoltaic power generation, which is quite popular in the energy market because of the natural advantages of both distributed power generation and clean power generation, which are flexible, efficient and environmentally friendly [8–9]. In recent years, the overall growth rate of electricity consumption in society has been on a downward trend, and the personalized and detailed demand on the user side has been gradually improved, which provides a good opportunity and environment for the development of distributed PV power generation and further promotes the rapid development of distributed PV power generation, and distributed PV power generation has become an important part of China’s electric power grid at present [10–11].

In the development mode of photovoltaic power generation industry, changing the traditional rough operation and maintenance and management mode, improving the level of operation and maintenance management, improving the safe and efficient operation of photovoltaic power plants, and trying to control the risk, is the basis of the high-quality development of photovoltaic power generation [12–13]. Operation and maintenance and technical supervision work itself around the equipment, can improve the reliability of equipment and facilities, improve the level of management and power plant technology, can ensure the safe and efficient operation of photovoltaic power plants [14–15].

With the continuous progress and development of society, the user’s requirements for a reliable power grid have been raised, the government has also continuously strengthened the supervision of power supply services, and the general public is also always concerned about the power grid, which puts forward higher requirements for the quality service of electric power enterprises [16–17]. For photovoltaic power generation system it has the characteristics of simple system structure, large amount of equipment, failure prone and difficult to locate the problematic equipment. On the other hand, distributed photovoltaic grid-connected increases the difficulty of load forecasting in its area, changes the existing load growth pattern, leads to changes in the voltage distribution on the feeder, is prone to harmonic pollution, affecting the quality of power, and access to a large number of distributed power sources makes the transformation and management of the distribution network more complex [18–19]. In the face of these problems, equipment standardization, normalization and refinement of management can be strengthened, especially to explore lean production and realize real-time regulation and management of users. At present, the total installed capacity of distributed PV up to and the number of grid-connected users is climbing year by year, and there is an urgent need to form a set of new management mode that can be updated in real time to effectively control the user side of the power supply, optimize the mode of operation, and achieve cost reduction and efficiency [20–21].

The article introduces the output and cost models of distributed PV power generation systems and proposes a regulation model for distributed PV systems based on the economic assessment process. The model considers the influence of multi-source data on PV energy dispatch, constructs the economy objective function, time comfort objective function and temperature comfort objective function respectively, and considers the constraints such as charging and discharging power, state of charge and number of charging and discharging times. Meanwhile, for the advantage of particle swarm algorithms in solving continuous problems, an improved discrete binary particle swarm algorithm is used to solve the objective function. Different residential adjustable load participation cases are simulated and analyzed respectively, and the load response analysis is used to verify the deployment effect of this paper’s method on the user’s electricity load. At the same time, this paper also analyzes the difference between the scheduling effect of the solution obtained in this paper and other solutions under different weather conditions.

2

Distributed photovoltaic power generation system model

Grid-connected photovoltaic systems include dispatchable and non-dispatchable systems. Non-dispatchable systems mainly include PV arrays and grid-connected inverters. The dispatchable distributed PV-storage power generation system mainly consists of a PV array, a single converter, a battery bank, a bidirectional converter, and a grid-connected inverter [22].

In the non-dispatchable PV power generation system, the output power from the PV array is directly converted to AC power by the inverter, which is supplied to the user. When the output power of the PV array is insufficient to meet the user’s demand, the user needs to utilize grid power to supply electricity.

In the dispatchable PV power generation system, when the power generated by the PV power generation system is greater than the load demand, the excess power can be fed into the grid, or the battery can be used to store the excess power. In the current situation where the feed-in tariff is lower than the electricity tariff, it is more economical to prioritize the use of batteries to store the excess power. Therefore, in the analysis process of this paper, when the PV output power is greater than the load demand, the priority is to charge the battery. When the PV output is less than the load demand, the battery is discharged to meet the load demand.

2.1

Modeling of PV-storage system

2.1.1

PV module output characteristics

The most basic electricity generating unit in a photovoltaic power generation system is the photovoltaic cell, which is capable of photovoltaic conversion under sunlight. It is a semiconductor P–N junction device, the use of light-excited minority carriers through the P–N junction power generation, in the power generation process, there is no chemical reaction, with no noise, no smell and no pollution of the environment.

The output power of a single cell is small, so in order to obtain a higher output power and output voltage, it is usually necessary to encapsulate more than one solar cell unit after the composition, according to the number of solar cell units connected in series and parallel is different, the solar photovoltaic module can obtain different output power.

The power size of solar crystalline silicon photovoltaic module output is greatly affected by the temperature and light intensity of the environment, the greater the light intensity, the greater the output power of the module, and the lower the temperature, the higher the output power of the module. When the temperature is constant, the maximum output power of the module is proportional to the light intensity, while at a specific solar irradiation intensity, the maximum output power of the PV module shows a negative proportionality with the temperature.

2.1.2

PV module output model

Based on the above analysis, the output power of the PV module can be expressed as a function of module area, light intensity, module conversion efficiency and temperature. Let the area of the solar PV module be S, the light intensity at a certain moment i be Φ(i), the conversion efficiency of the PV module be η_p, the outdoor temperature be T, and the outdoor temperature difference be ΔT(i). At moment i, the output power of the PV module can be expressed as: 1 $P_{v} (i) = S \cdot ϕ (i) \cdot η_{p} [1 - 0.005 (T + Δ T (i)) - 25]$

Where the light intensity unit is W/m², the area of the solar module unit is m², the PV output power unit is W. From the above equation, it can be seen that in different seasons, by the temperature and the intensity of solar radiation, the intensity of solar radiation will be converted to the total efficiency of electrical energy is different. Therefore, the above equation can be expressed as a function of irradiation intensity, total conversion efficiency and module area, at this time the PV module output power model: 2 $P_{v} (i) = S \cdot ϕ (i) \cdot η (i)$

Where, η(i) denotes the conversion efficiency of the PV module as well as the total conversion efficiency affected by temperature, which is mainly affected by the weather temperature. In this paper, different total conversion efficiencies are considered in the summer, winter and spring/fall seasons, so as to calculate the output power of PV modules per unit of installed capacity.

Solar modules generate DC power, which needs to be converted into AC power by the inverter and connected to the AC power grid, at this time, the power input to the grid from PV modules needs to take into account the efficiency of the inverter, so that the power converted by the inverter and delivered to the AC power grid is modeled as: 3 $P_{g} (i) = P_{v} (i) \cdot η_{c}$

Where, P(i) is the power input to the grid from the inverter after inverting the PV power, and η_c is the inverter conversion efficiency.

2.2

Whole Life Cycle Cost Modeling

In this paper, the annual repair and maintenance costs are not considered in the operation of the PV system, only the initial investment cost is considered. In this paper dispatchable PV power generation system, the initial investment cost mainly consists of 3 parts, which are PV module cost, energy storage device cost (battery composition cost) and inverter cost.

The PV module cost considered in this paper is considered to be related to the installed capacity of the PV module, and is proportional to the relationship, where the PV module cost, refers to the one-way converter cost of the PV module and PV batteries as a whole system for costing, as a unit of installed capacity of PV module prices for calculation. The cost of the battery in this paper will also include the cost of the bidirectional converter as a unified whole for consideration.

In this paper, we set the operating life of PV module as n, and the initial investment cost model of PV module is: 4 $C_{P V} = W C_{0 P V}$

Where, C_pv is the initial investment cost of the PV module, W is the installed capacity of the PV module, and C_OPV is the cost of installing a unit capacity of the PV module.

The cost model of the energy storage device is: 5 $C_{E S S} = C_{E Q} Q_{Σ}$ 6 $C_{E 0} = \frac{C_{E}}{Q \times η_{E} \times n_{E}}$

Where, C_ESS is the annual operating cost of the energy storage system, C_E0 is the cost of charging and discharging per unit of electricity, Q_Σ is the total amount of charging and discharging of the battery in a year, C_E is the cost of purchasing the battery cells, Q is the individual capacity of the battery, η_E is the charging efficiency of the battery, Λ is the maximum discharge depth of the battery, and n_E is the number of times that the battery cycles the battery at this discharge depth.

The inverter cost model is: 7 $C_{C} = C_{C 0} Q_{C}$

Where, C_C is the inverter investment cost, C_CO is the cost per unit power inverter, and Q_C is the inverter installation capacity.

In the dispatchable PV power system, the system cost model is: 8 $C = C_{P V} + C_{E S S} \times n + C_{C}$

In a non-dispatchable PV system, the system cost is modeled as: 9 $C = C_{P V} + C_{C}$ where C is the total initial investment cost of the system.

2.3

Economic assessment process

After installing the PV power generation system, the user can get benefits from 3 aspects. Firstly, according to the national power subsidy policy, as long as the PV system has power output, the user can get the state subsidy according to the amount of power generated; secondly, when the PV output is larger than the user’s demand, the excess power is fed into the Internet through the inverter, and the proceeds from the sale of this part of the power can be used as part of the user’s income; lastly, the user’s part of the load demand can be met by the PV power generation, which reduces the need to buy power from the grid, which is equivalent to obtaining indirect income. It is equivalent to obtaining indirect income.

In order to analyze the net benefits of users under different PV installation capacities, this paper defines the PV capacity/load electricity ratio, which is expressed as the ratio of the sum of the annual power generation of the PV system and the annual electricity consumption of the electricity user after the actual installation of the PV system, and indirectly indicates the PV installation capacity. And set to K, the calculation formula is: 10 $K = \frac{\sum_{i = 1}^{365} P V (i) h_{i}}{\sum_{i = 1}^{365} L o a d (i) h_{i}}$

The installed capacity K_p of the PV system can be calculated for a value of K for the capacity ratio: 11 $K_{p} = \frac{K \cdot \sum_{i = 1}^{365} L o a d (i) h_{i}}{\sum_{i = 1}^{365} P V_{0} (i) h_{i}}$ where PV_o is the generation power of the PV system per unit capacity.

In the dispatchable system, an energy storage battery of capacity Q is set up for the user: 12 $P = \frac{Q}{\sum_{i = 1}^{365} P V (i) h_{i} / 365}$

Then: 13 $Q = \frac{P \cdot \sum_{i = 1}^{365} P V (i) h_{i}}{365} = P \cdot K \cdot \sum_{i = 1}^{365} L o a d (i) h_{i}$ where P denotes the relationship between the storage battery capacity and the average daily generation of the PV system. The following calculations analyze the net benefits to the user at different values of P.

3

Distributed PV regulation model with multi-source data fusion

3.1

Optimal control model for home energy management system

3.1.1

Objective function

1)

Economic objective function

The economy objective function is the total cost of household electricity consumption, expressed as the cost of purchasing electricity minus the benefit of surplus electricity going online: 14 $F_{1} = \sum_{t = 1}^{n} (f_{g} (t) \cdot p_{g r i d} (t) \cdot \frac{24}{n})$

Where F₁ is the total cost of electricity consumption by households in a planning cycle, n is the total number of time periods in the cycle, f_g(t) is the price of electricity purchased by users in time period t or the price of electricity sold to the grid, which is taken as the peak time or valley time price depending on the time period, yuan/kW-h, p_grid(t) is the power exchanged between the household and the grid during the t time period, when p_grid>0 the household absorbs power from the grid, and when p_grid(t)<0 the household transmits power kW to the grid.

2)

Time comfort objective function

The optimization objective in terms of time comfort is to make the appliance work schedule as consistent as possible with the user’s original power consumption habits, i.e., the change in the load working time is as small as possible. The objective function is expressed as: 15 $S_{i} (t) = {\begin{array}{l} 1 & t \in [t_{i}^{s t a r t}, t_{i}^{e n d}] A n d t \notin [T_{i}^{s t a r t}, T_{i}^{e n d}] \\ 0 & O t h e r \end{array}$ $${S_i}\left( t \right) = \{ \matrix{ 1 \hfill \unicode {x00A0} {t \in \left[ {t_i^{start},t_i^{end}} \right]\,And\,t \notin \left[ {T_i^{start},T_i^{end}} \right]} \hfill \cr 0 \hfill \unicode {x00A0} {Other} \hfill \cr } $$ 16 $T_{i}^{L} = T_{i}^{s a r t e} - T_{i}^{s t a r t s}$ 17 $F_{2 i} = \frac{\sum_{i = 1}^{n} S_{i} (t) \cdot Δ t}{T_{i}^{L}} (i = 1, 2 \dots K)$ where F_2i is the time comfort of load i, K is the total number of dispatchable loads in the household, Δt is the duration of the unit time period in the planning cycle, $T_{i}^{L}$ is the duration of the feasible startup period of load i, $t_{i}^{s t a r t}$ and $t_{i}^{e n d}$ are the actual startup and stopping times of load i after scheduling, $T_{i}^{s t a r t}$ and $T_{i}^{e n d}$ are the optimal startup and stopping times of load i, and S_i(t) is the state parameter of the load comfort, S_i(t) = 1 indicates that the load i is on during the t time period and is in a different switching state than it would be under the optimum work schedule.

3)

Temperature comfort objective function

The optimization objective in terms of temperature comfort is to keep the room temperature as close as possible to the user’s set value. The home energy management system will improve the efficiency of electricity consumption by adjusting the upper and lower limits of the room temperature control setpoints, and care should be taken to satisfy the user’s demand for ambient temperature comfort when adjusting the setpoints. The evaluation index of comfort is related to the room temperature and the upper and lower limits of the temperature setpoint.

Within the range of the set temperature, the greater the percentage of room temperature deviation from the optimum temperature, the lower the user comfort. In addition, the sensitivity of the human body to the sensation of temperature change is variable, and the closer the room temperature is to the human body’s comfort temperature, the lower the human body’s sensitivity to temperature change. That is, when the temperature difference between body temperature and room temperature is small, the temperature feeling sensitivity is low, and when the temperature difference is large, the temperature feeling sensitivity is high. The user’s comfort is exponentially related to the difference between the current temperature and the optimal set temperature, and the comfort objective function is expressed as: 18 $u_{A C} (t) = \frac{e^{l_{A C (t)}} - 1}{e - 1}$ 19 $F_{3} = \frac{\sum_{t = 1}^{n} u_{A C} (t) \cdot k_{A C} (t)}{\sum_{t = 1}^{n} k_{A C} (t)}$

Where, k_AC(t) is the air conditioning working state parameter, k_AC(t) = 1 when t belongs to the air conditioning working hours, and k_AC(t) = 0 when t does not belong to the air conditioning working hours.

3.1.2

Constraints

The constraints of the optimal control problem of the court energy management system include both system power balance constraints and energy storage system constraints. 1)

Power balance constraints

The home load power, photovoltaic power, energy storage device charging and discharging power, and home and grid interaction power are in equilibrium at each time period. The power balance constraint in the system is: 20 $P_{l o a d} (t) - P_{P V} (t) - P_{b} (t) = P_{g r i d} (t)$

Where, P_lood(t) is the total power of the household load, kW, P_b(t) is the discharge power of the energy storage device, kW, P_pV(t) is the PV output power, kW, P_grid(t) the power flowing from the grid to the user’s household, kW.

2)

Energy storage system constraints

The energy storage system constraints include three parts: equation (21) is the power and discharging power constraints, equation (22) is the charging state constraints, and equation (23) (24) is the charging and discharging time constraints. The restrictions on the number of charging and discharging times are designed to reduce the frequent charging and discharging processes caused by fluctuations in the PV output, and to extend the service life of the battery. During the daytime when the PV output value is high, the battery is charged, and at night when the PV output value is low, the battery is discharged. The battery’s charging and discharging process is restricted to one charging and one discharging per day based on the size of the PV output value. The constraints of the energy storage system are: 21 $- P_{b \min} \leq P_{b} (t) \leq P_{b \max}$ 22 $S O C_{\min} \leq S O C (t) \leq S O C_{\max}$ 23 $P_{b} (t) (P_{P V} (t) - P_{P V}^{L}) \leq 0$ 24 $P_{b} (t) (P_{P V} (t) - P_{P V}^{H}) \leq 0$

Where p_bmax is the maximum charging power of the battery, kW, p_bmin is the maximum discharging power of the battery, kW, SOC(t) is the charging state of the battery, SOC_max, SOC_min are the maximum and minimum values of the charging state of the battery, $P_{P V}^{H}$ , $P_{P V}^{L}$ is the battery charging and discharging threshold, when the photovoltaic power is higher than $P_{P V}^{H}$ when the battery is charging, lower than $P_{P V}^{I}$ when the battery is discharging, kW.

3.2

Particle swarm optimization scheduling algorithm

3.2.1

Optimizing the scheduling algorithm

In the choice of algorithms for optimal energy scheduling, genetic algorithms or particle swarm optimization algorithms are generally chosen. In comparison, the genetic algorithm is more suitable for solving discrete problems, while the particle swarm algorithm is suitable for solving continuous problems. Therefore, the particle swarm algorithm is more in line with the requirements of this paper.

Particle swarm algorithm (PSO) is a more commonly used global stochastic search algorithm. It seeks the optimal solution in a continuous space through multiple iterations and evaluates the results of each iteration using the fitness index. The position and velocity of each particle are updated variables, and in each iteration, the global optimal solution is compared to the historical optimal solution of the current particle. However, the particle swarm algorithm does not perform selection, crossover, or mutation operations on individual particles; its main advantage is its fast computational speed and better global search capability [23].

If we use 0 to denote the stop working state of the load and 1 to denote the power supply operating state of the load, the optimization objective is ultimately in finding the optimal set of states that contains all the load states, so that the objective function reaches the optimal extreme value. Therefore, it is a nonlinear 0-1 overall planning problem in nature. In order to fit the problem of this paper more closely, we use the discrete binary particle swarm algorithm (DBPsO) [24] here.

DBPSO is developed on the basis of PSO, and its velocity update formula has been the same as PSO as: 25 $v_{i}^{k + 1} = ω^{k} \cdot v_{i}^{k} + c_{1} \cdot r a n d (p b e s t_{i}^{k} - x_{i}^{k}) + c_{2} \cdot r a n d \cdot (g b e s t_{i}^{k} - x_{i}^{k})$ where $v_{i}^{k}$ , $x_{i}^{k}$ , ω^k are the velocity, position and inertia weights of i at the k th iteration, and

$p b e s t_{i}^{k}$ is the historical optimal solution of particle i at the k th iteration, and

$g b e s t_{i}^{k}$ is the historical optimal solution of particle i at the k th iteration of the swarm, and

c₁, c₂ are the learning factors.

In solving the particle position, the sigmoid function is used to map the velocity value into the probability between [0,1], and then the particle position is limited to one of 0 and 1 according to the velocity. Since the value of the mapping function of the particle is 0.5 when the particle velocity is 0, which is not conducive to the convergence of the algorithm, the mapping function can be changed to: 26 $S_{i}^{k + 1} = | \frac{2}{1 + \exp (- v_{i}^{k + 1})} - 1 |$

Then the position of the particle can be expressed as: 27 $x_{i}^{k + 1} = {\begin{array}{l} 1, & i f r a n d () \leq S_{i}^{k + 1} \\ 0, & o t h e r w i s e \end{array}$

To improve the local search capability, we iteratively update the inertia weight ωeach time as well: 28 $ω^{k} = ω_{\max} + \frac{k - 1}{K - 1} \cdot (ω_{\min} - ω_{\max})$ where ω_max is the preset maximum inertia weight, and

ω_min is the preset maximum inertia weight, and

K is the preset maximum number of iterations.

3.2.2

Algorithm realization flow

We represent the operating state of all controllable loads in a day by the set matrix X : 29 $X = {[X_{1}, \dots X_{c}, \dots X_{n}]}^{T} = [\begin{matrix} x_{1}^{1} & x_{1}^{2} & \dots & x_{1}^{48} \\ ⋮ & ⋮ & ⋮ & ⋮ \\ x_{c}^{1} & x_{c}^{2} & \dots & x_{c}^{48} \\ ⋮ & ⋮ & ⋮ & ⋮ \\ x_{n}^{1} & x_{n}^{2} & \dots & x_{n}^{48} \end{matrix}]$

In the matrix, n is the total number of controllable loads in the electrical load. $x_{c}^{i}$ denotes the operating state of the device numbered c at the i rd time period. We use 0 to denote the state of stopped operation and 1 to denote the state of power supply operation. We convert the above equation to vector form and considering the working set state matrix of controllable loads at all times of the day, that is X, as a particle, the i th particle X_i can be expressed as: 30 $X_{i} = [x_{i}^{1}, \dots, x_{i}^{48}, \dots, x_{i}^{48 (c - 1) + 1}, \dots, x_{i}^{48 c}, \dots, x_{i}^{48 (n - 1) + 1}, \dots, x_{i}^{48 n}]$

The particle swarm position matrix X_PS can be expressed as: 31 $X_{P S} = {[X_{1}, \dots X_{i}, \dots X_{N}]}^{T} = [\begin{matrix} x_{1}^{1} & x_{1}^{2} & \dots & x_{1}^{48 c} & \dots & x_{1}^{48 n} \\ ⋮ & ⋮ & ⋮ & ⋮ & ⋮ & ⋮ \\ x_{i}^{1} & x_{i}^{2} & \dots & x_{i}^{48 c} & \dots & x_{i}^{48 n} \\ ⋮ & ⋮ & ⋮ & ⋮ & ⋮ \\ x_{N}^{1} & x_{N}^{2} & \dots & x_{N}^{48 c} & \dots & x_{N}^{48 n} \end{matrix}]$

The particle swarm velocity matrix V_PS can be expressed as: 32 $V_{P S} = {[V_{1}, \dots V_{i}, \dots V_{N}]}^{T} = [\begin{matrix} v_{1}^{1} & v_{1}^{2} & \dots & v_{1}^{48 c} & \dots & v_{1}^{48 n} \\ ⋮ & ⋮ & ⋮ & ⋮ & ⋮ & ⋮ \\ v_{i}^{1} & v_{i}^{2} & \dots & v_{i}^{48 c} & \dots & v_{i}^{48 n} \\ ⋮ & ⋮ & ⋮ & ⋮ & ⋮ \\ v_{N}^{1} & v_{N}^{2} & \dots & v_{N}^{48 c} & \dots & v_{N}^{48 n} \end{matrix}]$ where N is the number of particle swarms and n is the number of flexible loads.

The optimal scheduling strategy for distributed PV generation systems without energy storage is essentially a reconstruction of the household load profile using demand-side management tools to reduce electricity consumption costs. However, due to the existence of load operating time present value and intermittent PV power generation, the savings in electricity consumption costs by relying on load shifting alone are extremely limited. So energy storage devices need to be introduced for regulation. The optimal scheduling process of energy storage devices is as follows:

The self-generation and self-consumption ratio of distributed PV power generation can be derived: 33 $μ = 1 - \frac{\sum P_{s e l l}^{t}}{P_{P V - a l l}}$

4

Example analysis and validation

In this section, the example of Village G in County L is analyzed for simulation. Two hundred households in the region are selected as samples, in which the average PV installation capacity of each residential house is 3.7 kW, i.e., the total installed capacity of residential PV in the region is 780 kW.At the same time, it is assumed that each household owns an electric vehicle with a battery capacity of 15 kWh, and all of them can participate in the scheduling of the residential demand response strategy. The charging and discharging process of EVs and the power generation process of PVs are compared to the constant power operation process in this analysis. In addition to EVs, residential loads in the region include air conditioners, washing machines, kitchen appliances, water heaters, etc. The load parameters of specific residential power equipment are shown in Table 1.

Table 1.

Household load parameters

Category	Device	Rated power/kw
Adjustable load	Electric vehicle	3.5
	Rice cooker	0.7
	Electric hot water heater	3
	Washing machine	0.5
	Air conditioner	2.8
Rigid load	TV	0.6*3
Rigid load	Others	2

4.1

Simulation of user-regulated loads in multiple contexts

4.1.1

Residential adjustable loads not participating in demand response simulation

When the regional residential adjustable loads do not participate in demand response, based on the load probability matrix, Monte Carlo simulation is applied to generate a 300-household electricity load data sample, and the distributed power outlets, the residential set of gross loads and their net loads in the region are shown in Figure 1.

From the graph analysis, the maximum net residential load is reached at time period 42 (20:30~21:00) with a value of about 1444.97kW. The net load reaches its minimum at time slot 19 (9:00 to 9:30) with a value of about -267.2kW. During the three time periods 18~19 (8:30~9:30), 26~28 (12:30~14:00) and 33 (16:00~16:30) the net residential load is ≤0, i.e., the distributed power sources are not fully consumed during the time periods here, and the remaining load is delivered to the grid in the reverse direction. In addition, during time periods 34 and 48, the net load peaks are formed mainly from EV charging and electric water heaters. Therefore, EVs, washing machines, and electric water heaters are considered as adjustable loads for day-ahead dispatch to promote local consumption of distributed power sources.

4.1.2

Residential Adjustable Load Participation in Demand Response Simulation

According to the load characteristics of electric water heaters, washing machines and electric cars, they can be flexibly mobilized within a certain time range on the premise of meeting the electricity demand of users. Different scenarios are simulated: the adjustable load participating in day-ahead scheduling in Scenario 1 consists of electric water heaters and washing machines, and the adjustable load participating in day-ahead scheduling in Scenario 2 consists of electric cars, electric water heaters and washing machines.

The load response curve for Scenario 1 is shown in Figure 2. The corresponding net loads of - 174.3kW and -267.2kW change to -53.6kW and -27.6kW, respectively, for the hours 18~19 (8:30~9:30) when the net loads are less than zero. The original net loads of -112.7kW, -101.1kW and -183.8kW corresponding to the time periods 26~28 (12:30~14:00) became 98.9kW, -33.1kW and 37.4kW respectively. The net load corresponding to the original 33 (16:00~16:30) time period changed from the original -46kW to 6.7kW. The rise in net loads in time periods 28 to 30 is mainly due to residential customers responding to demand management by advancing the execution of electricity consumption plans for electric water heaters and washing machines that were originally scheduled for time periods 38 to 44. The maximum value of residential net load at time period 42 (20:30~21:00) changed from 1444.97kW to 1046.4kW, and the time period was also postponed from the original 42 to 44. After the participation of electric water heaters and washing machines in demand response, the net load distribution of the users changed due to the differences in the load dispatch of electric water heaters and washing machines in their homes as a result of the differences in the electricity consumption habits and consumption behaviors of each residential user.

The load response curve for Scenario 2 is shown in Figure 3. After EVs participate in demand response together with electric water heaters and washing machines, the net loads corresponding to - 174.3kW and -267.2kW in time periods 18~19 (8:30~9:30) change to -63.1kW and 30.4kW, respectively; however, compared with Scenario 1 before EV participation, the net loads corresponding to time periods 18~19 under Scenario 2 change very little, which is due to the fact that the EVs in this time period are are almost always not in the home parking state. The consumption of distributed power in time periods 26~28 (12:30~14:00) and 33 (16:00~16:30) changed significantly, where the net load is less than zero in time periods 26~28, whose corresponding net loads become 26.5kW, 176.5kW and 134.6kW, respectively. The net load in time slot 33 (16:00~16:30) has changed from the original -46kW to 15.6kW. Meanwhile, the maximum peak-to-valley difference of net load during the daytime (time period 10~40) changes from 1712.18kW before elimination to 1214.87kW after elimination, which improves the load profile to some extent.

4.1.3

Load response analysis

The maximum and minimum values of the net load, the maximum peak-to-valley difference, and the results of the unabated distributed power generation for the three scenarios before the implementation of DR, the participation of electric water heaters and washing machines in DR, and the participation of EVs along with electric water heaters and washing machines in DR are shown in Table 2.

Table 2.

Analysis of experimental results in multiple scenarios

Scenario		Minimum net load/kw	Maximum net load/kw	Maximum peak valley difference/kw	The power of the undenied/kWh
Before adjustable load involvement		-305.9	1484.6	1790.0	-530.5
After adjustable load involvement	Scenario1	-85.4	1185.3	1271.2	-120.8
After adjustable load involvement	Scenario2	-81.6	1154.0	1235.7	-56.0

As analyzed in Table 2, after the participation of adjustable loads in the response, the maximum value of the regional residential net load is reduced by 21.21% due to the fact that the electric water heaters and washing machines as well as the electric vehicles during the peak hours are shifted to the hours when the distributed power sources are not consumed. At the same time, after the adjustable load participates in the DR, the minimum value of the net load of residents shows a significant downward trend, and the peak-to-valley difference is further reduced, which makes the regional residential electricity load more gentle, which is conducive to the stable operation of the distribution network, in addition, the electricity of the distributed power generation that has not been absorbed locally is reduced from the original 530.5kW to 120.8kW (Scenario 1) and 56.0kW (Scenario 2) respectively, and the growth rate of local consumption is 77.2% and 89.4% respectively. To summarize, when the normal life of residents is not affected, the local consumption of distributed power significantly increases after the participation of adjustable loads in their homes in demand response.

4.2

Simulation analysis of scheduling scheme considering weather data

4.2.1

Simulation analysis of different scheduling schemes

Scheduling simulation analysis is carried out using the improved optimal scheduling model, assuming that the scheduling cycle is 7 days, and the scheduling cycle contains three types of weather at the same time, which are rainy, sunny, and cloudy, and assuming that the time of each scheduling day is from 8:00 a.m. to the next day’s 8:00 a.m., and treating every three hours of the time period as a phase.

In order to provide a better comparative analysis of the various scheduling models used in distributed PV power systems, this paper compares this scheme with three other scheduling schemes. Scenario 1: Optimal scheduling of distributed PV power system using traditional scheduling strategy. Scheme 2: Optimal scheduling of distributed PV power systems using the global optimal scheduling model without considering weather factors. Scheme 3: Optimal scheduling of distributed PV power system using energy storage direct formulation optimal scheduling model. Scheme of this paper: optimal scheduling of distributed PV power system using optimal scheduling model. Matlab software is used to simulate and analyze the four different scheduling schemes, and the solution algorithm is the improved particle swarm algorithm mentioned above.

The scheduling simulation results of this paper are given in this paper, in which Fig. 4 shows the scheduling optimization simulation results for each day of the scheduling cycle, and Fig. 5 shows the allocation results for each day of the scheduling cycle.

Analysis of the simulation results shows that when the next day of the scheduling cycle is rainy or cloudy, such as the second day, the fourth day and the sixth day of the scheduling cycle selected in this paper, the PV arrays in these three days almost do not produce electricity, from the simulation results, it can be seen that the power supplied to the user in the three days is almost all from the power released by the battery, and less power is purchased from the grid, and the power stored in the battery in the end of the day is able to ensure that the next day scheduling. The power stored in the battery at the end of the previous day can ensure that the next day’s scheduling needs, can effectively deal with the impact of various weather conditions, as far as possible, so that the user to obtain the maximum economic benefits.

4.2.2

Comparative analysis of economic benefits of different scheduling options

In the scheduling cycle, the economic benefits obtained by each scheduling scheme are compared and analyzed. Table 3 shows the economic benefits obtained by each scheduling scheme for each scheduling day in the cycle. The analysis shows that the economic benefits under the four different scheduling strategies are ranked as follows: this paper scheme > scheme 3 > scheme 2 > scheme 1, of which the economic benefit of this paper scheme is 113.215 yuan, an increase of 12.593% compared with scheme 1. Comparing Scheme I with Scheme II, it can be found that the economic benefit of Scheme II is obviously better than Scheme I when the weather is sunny, but when the weather is rainy the economic benefit under both schemes is about -14.5 yuan, and the optimization effect of Scheme II is not obvious. Comparing Scheme II and Scheme III, it can be found that the economic benefit of Scheme III is lower than that of Scheme II when the weather is sunny, but the total economic benefit is higher than that of Scheme II. This is because Scheme III must consider that the next day is rainy or sunny to store energy in advance, and to buy a large amount of electricity from the grid in the valley tariff hours, so as to reduce the cost of electricity in the high tariff hours and to ensure that the overall benefit is maximized. The advantages of the scheme in this paper are mainly reflected in cloudy or rainy days, which indicates that the proposed dispatch scheme plays an effective role, thus achieving the overall optimal economic benefits.

Table 3.

The comparison of economic benefits of the four schemes

Date	Weather	Scheme1 (yuan)	Scheme2 (yuan)	Scheme3 (yuan)	Ours (yuan)
9/1	Sunny	30.975	32.753	27.730	27.074
9/2	Rainy	-14.587	-14.457	-8.092	-7.007
9/3	Sunny	30.975	32.770	30.428	27.075
9/4	Cloudy	5.843	6.596	8.627	13.537
9/5	Sunny	30.976	32.784	27.728	27.074
9/6	Rainy	-14.586	-14.463	-8.082	-7.006
9/7	Sunny	30.956	32.794	32.455	32.468
Sum.		100.552	108.777	110.794	113.215
Revenue growth			8.180%	10.186%	12.593%

In summary, compared with scheme I, scheme II and scheme III, the best effect is achieved when using the scheme of this paper to optimize the scheduling of the distributed photovoltaic power generation system, especially in the case of rainy and cloudy days, and the storage capacity of the battery obtained by using the method of assessment takes full account of the effects of various types of weather, and has a strong scientific and objective nature.

5

Conclusion

In this study, economic factors, electricity consumption habits, indoor temperature and other data are fused and modeled, and the model is solved using a discrete binary particle swarm algorithm to achieve real-time assessment of the effect of user regulation.

The user regulation simulation results show that the formation of net load peaks mainly comes from charging electric vehicles and electric water heaters, and that differences in electricity consumption habits and consumer behavior have an impact on the load dispatch situation. After solving the objective function and obtaining the regulation scheme, it is found that the application of this paper’s scheme can significantly improve the net load situation of the customer’s electricity consumption, and the growth rate of the in situ consumption under the two scenarios is 77.2% and 89.4%, respectively. This paper’s scheme is able to effectively meet the degree demand when unfavorable weather for PV power generation occurs, and its benefit grows by 12.593% compared to the initial deployment scheme, which is more obvious than the economic benefits realized by other schemes.

Therefore, it is sufficient to demonstrate that the work in this paper is relevant and can improve the regulation of distributed PV power generation.

Język:: Angielski

Częstotliwość wydawania:: 1 razy w roku
Dziedziny czasopisma:: Nauki biologiczne, Nauki biologiczne, inne, Matematyka, Matematyka stosowana, Matematyka ogólna, Fizyka, Fizyka, inne

Kanał RSS czasopisma

Construction of a framework for real-time evaluation of regulation effect of distributed photovoltaic users based on multi-source data fusion

Bo Feng

Junpeng Zhao

Yifeng Wang

Yongliang Jia

Data publikacji: 19 mar 2025

Otrzymano: 24 paź 2024

Przyjęty: 10 lut 2025

DOI: https://doi.org/10.2478/amns-2025-0473

Słowa kluczoweDistributed photovoltaic power generation, User regulation assessment, Multi-source data fusion, Particle swarm algorithm

© 2025 Bo Feng et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Słowa kluczowe
Distributed photovoltaic power generation, User regulation assessment, Multi-source data fusion, Particle swarm algorithm