Shortage of electricity may be worsened by inefficiencies in the end-use system. The inefficiencies in the end-use system are caused by obsolete technology of industrial procedures and old equipment, lack of awareness of emerging energy practices in the industry and inadequate policy drivers. Handling peak consumption only from supply side may lead to negative environmental consequences. In order to meet the peak demand and avoid blackouts, fossil fuels are most commonly used [3] . Demand side management (DSM) has emerged as an alternative method for energy management in order to maintain the balance while focusing on the consumer side [4] . DSM has different meanings for different categories of stakeholders. For utility companies, DSM means avoiding or delaying the need to build new generating capacity by reducing or shifting consumers’ energy use duration. For consumers, DSM presents an opportunity to save money by reduction of their electricity bill by taking advantage of financial incentives provided by the utility company. Introduction of Time of Use tariff enables consumers to further reduce their energy costs through DSM [5] . DSM programs are also thought to be an effective strategy to reduce both economic and environmental impacts due to the increasing electricity demand in the near future [6] .

Peak Demand Management (PDM) involves curtailing of energy consumption during periods of system peak thereby lowering capacity charges, hence reduced overall power costs. Peak demand reduction is a key element in demand management programs targeting the stabilization of the electricity grid [3] .

High demand instances cause utilities challenges since they must have capacity to meet the peak which is only required for a fraction of a 24 hour period, in addition to operations and maintenance costs, and power system losses. Traditionally, the peak load problem is often solved by capacity additions whereby a percentage of the total installed capacity must be reserved for managing peak loads, with the rest being dedicated to base loads [7] . Recent years have seen momentum build in the energy industry in favor of introducing more cost-reflective electricity tariffs with energy retailers and distributors, as well as market regulators and policymakers, increasingly calling for more dynamic pricing [8] . In Kenya, the Regulator announced time of use tariff [9] . It may encourage consumers to shift some of their operations to off peak hours. The tariff for industrial consumers has a component of peak demand charges for every billing cycle thus the need for peak demand management practices.

[10] indicates that producing scientific and reasonable maintenance plans helps to increase the reliability of power distribution system operation. In practice, maintenance scheduling is organized artificially based on the experience of the power department that is monitored by the operations team to ensure stability of power distribution system. The arrangement only focuses on the security of distribution system while neglecting the economic efficiency. The authors further contend that artificial scheduling and maintenance scheduling ought to be an optimized procedure based on scientific and effective mathematical model that can avoid the subjectivity and randomness. In order to meet the needs and benefit from the potential underlying in the energy system, planning processes from production to energy management need to operate seamlessly and with real time data. The energy management system needs to be integrated with production management and mill control systems in order to achieve the desired results [11] . For industrial load, decision makers need to find the optimal scheduling of the industrial load [12] . A study scheduling on a single machine with an energy consumption limit proposed mathematical algorithms [13] that are complex, used on a single machine and small instances; and may not be very applicable in real complex industrial cases. There is still a gap between industrial needs and academic research [14] . This is despite the fact that integration of energy awareness into production scheduling is getting more attention [15] .

Peak shrinking deals with the installation of equipment that uses less energy. Using more efficient equipment reduces the base load of energy used at all times effectively shrinking the entire demand curve in a downward direction. Flattening out daily peak and critical peak demand can improve the cost-effectiveness of the energy infrastructure but requires consumers to alter their behavior to achieve it. In a study on reducing household electricity consumption during evening peak demand times [3] it was noted that there was the strongest behavioral change in consumers who received monetary incentives, with only weak evidence of changes in the non-monetary treatment groups.

Peak shaving involves the leveling out of peaks in electricity use by consumers. The fundamental idea of peak shaving is to limit the amount of power drawn from the grid to some level, and substitute for power demand higher than that level from local on-site power sources. A common practice to reduce these peaks is to introduce energy storage systems into the main electrical system. The energy storage system will be charged during off-peak operation hours and will discharge during peak operation hours thus reducing the peak power drawn from the grid [16] .

A key benefit of generating own energy is that it helps to reduce peak demand charges from the utility. Using forms of alternative energy, e.g. solar panels, to generate a portion of the total energy needed is a great way to reduce the power draw from the grid. The challenge with peak shaving is to design a control scheme that detects the peaks on time and fully exploits the capacity of the distribution system [1] . Most of the control schemes found in literature suggest using a predefined shave level depending on the maximum load or how the load looks like. Since load forecasting is quite difficult to achieve, if hard limits are applied there is a good chance to miss the peak or to discharge the battery in smaller peaks leaving the biggest peak intact.

As part of the study, historical electricity consumption data of the Steel firms

under study was analyzed based on monthly billing data. The monthly consolidated electricity costs for each firm were compared with monthly energy consumption and cost index computed as per Equation (2). For each firm, Monthly Load factor, which is an indicator of the amount of energy used in a time period versus amount that would have been used, was computed as per Equation Number (3).

Cost   Index = Monthlybill ( ksh ) Monthlyenergyconsumption ( kwh ) (2)

Load   Factor = Monthlyenergyconsumption Peakdemand × 30   days × 24   hours × pf (3)

For the purposes of the study, the formula assumes unity power factor (pf).

From the historical data analyzed, there was a direct relationship between cost index and peak demand as per Figure 2 thus a higher cost index for the billing cycles when the peak demand was high and vice versa. The Pearson correlation was found to be 0.557 with p-value of 0.044, thus statistically significant at significance of less than 0.05.

A questionnaire designed to identify the extent of use of the three peak demand management strategies namely demand scheduling, peak shrinking and peak shaving in the firms under study was issued. The respondents were asked to indicate the extent to which they agreed on the items of the statements relating to demand management strategies. Each item had a 5-point Likert-type scale ranging from “no extent” (1) to “very large extent” (5).

This section began by first analyzing the descriptive statistics and then inferential statistics. The descriptive statistics was used to summarize features of

central tendency of the analyzed data while the inferential statistics was used for correlations and regression analysis.